William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

335
ANALYTICAL SOLIPSISM

description

William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Transcript of William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Page 1: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANALYTICAL SOLIPSISM

Page 2: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANALYTICAL SOLIPSISM

by

WILLIAM TODD

University of Cincinnati

• MARTINUS NI]HOFF / THE HAGUE / 1968

Page 3: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ISBN 978-94-011-8183-9 ISBN 978-94-011-8829-6 (eBook) 10.1007/978-94-011-8829-6

© 1968 by Martinus Nijhoff, The Hague, Netherlands. All rights reserved, including the right to translate or to

reproduce this book or parts thereof in any form.

DOI

Page 4: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

TABLE OF CONTENTS

Introduction vii

Acknowledgements ix

PART I: THE FOUNDATIONS OF SOLIPSISM 1

Chapter 1: Sensations and Images 3 Chapter 2: Language and Sensation 24 Chapter 3: Imagining Instances 46 Chapter 4: Comprehension 67 Chapter 5: Understanding and Synonymy 83 Chapter 6: Verifiability lO 1 Chapter 7: Objections 109

PART II: CAUSALITY 125

Chapter 8: Causes and Counterfactuals 127

PART III: SOLIPSISM PROPER 151

Chapter 9: A Quasi-Axiomatic Solipsistic System 153 Chapter 10: Alternatives to Solipsism 192 Chapter 11: Anti-Solipsism 216 Chapter 12: Further Development of the System: Phe-

nomenalism 253 Chapter 13: Statements about the Past 279 Chapter 14: Further Development of the System: Other

Minds 300 Chapter 15: Belief 310

Conclusion 317

Bibliography 319

Index 320

Page 5: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

INTRODUCTION

Philosophers usually have been anxious to avoid solipsism. A large number of good and great philosophers have tried to refute it. Of course, these philosophers have not always had the same target in mind and, like everything else, solipsism over the centuries has become increasingly elusive and subtle. In this book I undertake to state the position in its most modern and what I take to be its most plausible form.

At some points in the history of philosophy the solipsist has been one who denied the existence of everything except himself or even the existence of everything except his own present sensations. At other times, the solipsist instead of doubting these things has merely insisted that there could be no good reason for believing in the existence of anything beyond one's own present sensations. Roughly, this doubt is aimed at reasons rather than at things. A solipsist of this sort appears in Santayana's Scepticism and Animal Faith.

With the rise of analytic philosophy a new sort of solipsist has ap­peared. He no longer holds any of these older doubts, of either things or reasons. He does not deny that there are reasons for believing in the existence of the external world and the existence of other persons. In fact, he admits the truth of statements asserting the existence of these things. Rather, he is saying that one can admit all this and still not commit oneself ontologically to anything beyond the occurrence of one's own sensations and certain principles governing them. Thus, while the modern solipsist would never deny the existence ofthe world, his hypothesis is that we can assert its existence just by talking ulti­mately about our own sensations and making very complex assertions about them. His view is that we never have beliefs which commit us to anything more than this, even when shopping at the supermarket, kicking stones, and engaging in other mundane activities. The modern solipsist is nevertheless enough like the traditional one to deserve to be called by the same name. In a way his position is even more ex­treme. Instead of denying the existence of an irreducible physical world he tells us that we never believed in it anyway.

Page 6: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

viii INTRODUCTION

Our analytic solipsist will, unlike most of his predecessors, admit a principle of induction, and hence will be committed to the past and the future, but he will not be committed to irreducible physical ob­jects and to other minds. Further, as if to make up for this spurt of liberality about the past and the future, our solipsist will take a very dim view of such supposedly irreducible concepts as that of logical possibility, analyticity, and their related notions. These all will be reducible ultimately to empirical assertions about the mental phe­nomena of the subject.

Of course, this sort of position is not altogether new. Tendencies in this direction are to be found in Berkeley, Hume, and certain positi­vists such as A. J. Ayer. Much of the present account tries to adapt some of the basic principles of the British Empiricists to the sorts of difficulties which have arisen in the last two decades.

In all this, there are two main objects: the first is to show that, given appropriate premises, there is no logical difficulty in the solip­sist's position. He cannot be dismissed out of hand. I am urging that the solipsist presents a theory of considerable scope and complexity, and I argue that his theory compares favorably with other theories which attempt as much. The second main goal is to state modern solipsism thoroughly enough so that we can argue with it. The aim here is to see exactly what it is that we have to deny or add in order to go beyond it. It seems to me that a greater and wider understanding of solipsism would result in fewer people setting out to refute it. We could then expect to hear less philosophical conversation on the sub­ject.

Page 7: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ACKNOWLEDGEMENTS

Some chapters of this book contain material which has previously appeared in journal articles, but which has been re-worked for present purposes. For permission to use this material I am very grateful to Professor Richard Rudner of Philosophy of Science, Professor Soren Hallden of Theoria, and Professor G. P. Henderson of The Philosophical Quarterly. I am also indebted to the following publishers for giving me permission to quote from their books: Barnes & Noble, Cambridge University Press, The Clarendon Press, Cornell University Press, and John Wiley & Sons. I also wish to thank Mr. George Thompson of the University of Cincinnati who read the manuscript and offered many helpful comments.

Page 8: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

MARTINUS NIjHOFF - PUBLISHER - THE HAGUE

Adamczewski, Zygmunt, The tragic protest. 1963. VI and 282 pp. Gld. 27.-

Alexander, W. M., Johann Georg Hamann: philosophy, and faith. 1966. XII and 204 pp. Guilders 28.25

Atlas, Samuel, From critical to speculative idealism. The philoso­phy of Solomon Maimon. 1964. XI and 335 pp.

Cloth. Guilders 27.-Brunner, Constantin, Der endarvte Mensch. Herausgegeben und

eingeleitet von Lothar Bickel. 1951. XII and 205 pp. Guilders 9.-

la Charite, Raymond, The~ concept of judgment in Montaigne. 1968. IX and 149 pp. - Guilders 18.90

Darnoi, Dennis N. K., The unconscious and Eduard von Hart­mann. A historico-critical monograph. 1967. VII and 198 pp.

Guilders 23.40 Feibleman, James K., Foundations of empiricism. 1962. VIII and

389 pp. Cloth. Guilders. 27.-- Moral Strategy. An introduction to the ethics of confrontation.

1967. XI and 325 pp. Cloth. Guilders 37.25

Garelick, Herbert M., The anti-christianity of Kierkegaard. A study of "Concluding unscientific postscript." 1965. IX and 73 pp. Guilders 8.75

Hallett, H. F., Creation, emanation and salvation. A Spinozistic study. 1962. XI and 234 pp. Cloth. Guilders 25.-

Heelan, P. A., Quantum mechanics and objectivity. A study of the physical philosophy of Werner Heisenberg. 1965. XV and 207 pp.

Guilders 27.-Hildebrandt, Kurt, Leibniz und das Reich der Gnade. 1953. VIII

and 505 pp. Cloth. Guilders 26.50 Hocking, W. E., Philosophy, religion, and the coming world civili­

zation. Essays in honor of William Ernst Hocking. Edited by L. S. Rouner. 1966. XXII and 504 pp. With photograph.

Cloth. Guilders 54.­Iowa Publications in Phllosophy.

1. Essays in ontology. Edwin B. Allaire, May Brodbeck, Reinhardt Grossmann, Herbert Hochberg, Robert G. Turnbull. 1963. XI and 216 pp. Guilders 16.20

2. Moore and Ryle: two ontologists. Laird Addis and Douglas Lewis. 1965. VII and 184 pp. Guilders 16.20

Page 9: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

MARTINUS NUHOFF - PUBLISHER - THE HAGUE

3. Carnap and Goodman: Two formalists. Alan Hausman and Fred Wilson. 1967. IX and 225 pp. Guilders 21.60

Johnston, Willia:m M., The formative years of R. G. Collingwood. 1967. XIII and 167 pp. Guilders 25.20

Kattsoff, Louis 0., Logic and the nature of reality. 2nd edition. 1967. IX and 247 pp. Guilders 16.20

- Making moral decisions. An existential analysis. 1965. VIII and 279 pp. Guilders 28.75

- Physical science and physical reality. 1957. VIII and 311 pp. Guilders 17.75

MacDonald, Lauchlin D., John Grote. A critical estimate of his writings. 1966. XXI and 284 pp. Guilders 38.50

Rotenstreich, Nathan, Experience and its systematization. Studies in Kant. 1965. VII and 178 pp. Guilders 21.-

- Spirit and man. An essay on being and value. 1963. 257 pp. Guilders 20.75

Sa:minsky, Lazare, Physics and methaphysics of music and es­says on the philosophy of mathematics. (A green philosopher's peripeteia. Physics and metaphysics of music. The roots of arithmetic. Critique of new geometrical abstractions. The philo­sophical value of science). 1957. 151 pp. Cloth. Guilders lO.45

Spakovsky, Anatol von, Freedom, determinism, indeterminism. 1963. VII and 117 pp. Guilders 11.50

Vycinas, Vincent, Earth and Gods. An introduction to the philos­ophy of Martin Heidegger. 1961. XII and 323 pp.

Cloth. Guilders 26.50

- Greatness and philosophy. An inquiry into Western thought. 1966. XI and 294 pp. Guilders 34.75

Yolton, John W., The philosophy of science of A.S. Eddington. Preface par F. Gonseth. 1960. XV and 151 pp. Guilders 1l.50

Zabeeh, F., Hume, precursor of modern empiricism. An analysis of his opinions on meaning, metaphysics, logic and mathematics. 1960. 166 pp. Guilders 12.50

- Universals. A new look at an old problem. 1966. XII and 68 pp. Guilders 11.25

- What is in a name? 1968. VII and 74 pp. Guilders 10.-

One guilder = abo $ 0.28 = abo 2/4 sh. = env. Fr. 1.36 = ca. DMW 1.10 Obtainable through al!)' bookseller or directly from the publisher

Page 10: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

William Todd

ANALYTICAL

SOLIPSISM

. Martinus Nijhoff / The HaBue

Page 11: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

MARTINUS NIjHOFF - PUBLISHER - THE HAGUE

Analytic solipsism is defined as the view which says that all the state­

ments of ordinary language can be translated without loss of meaning

into a language whose primitive terms refer only to the mental phe­

nomena of the speaker. This view does not involve the doubting of any­

thing and, if correct, all the same things would exist that are ordinarily

supposed to exist. However, the theory does represent reductionism

carried to an extreme and, since the language of analysis is centered on

one person, it is the analogue of traditional solipsism within analytic

philosophy.

It is argued that the best way to do systematic philosophy is to state

all the competing theories as completely as possible; only then can we

be in a position to compare them in detail and to prefer some to others

on rational grounds. The object of this book is to state one of these

competing theories, analytical solipsism, as thoroughly as possible and to

- put it in the best possible light. It is then hoped that if solipsism is rejected

in favor of some other theory it will be rejected for the right reasons.

The first part of the book consists of a theory of meaning which is

considered to be part of the essential foundation for any sort of system­

atic philosophical theory. A discussion of causation and counterfactual

statements is then followed by a statement of a kind of phenomenalism.

This phenomenalism is unusual in that the analyses are always finite, as

is every analysis given in the system. The book then concludes with ana­

lyses of statements about other minds, the past, and belief which are con­

sistent with the very limited premises which have been chosen as the

basis of the system.

About the author: A.B., Harvard University (1955); Ph.D. University of Michigan (1960). Taught at Northwestern University (1960-62) and at the University of Cincinnati (1962-present ).

1968. IX and 329 pages. Guilders 42.75

Page 12: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PART I

THE FOUNDATIONS OF SOLIPSISM

Page 13: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 1

SENSATIONS AND IMAGES

This chapter deals with the fundamental building blocks of solips­ism - sensations and images. While no one denies the occurrence of these entities, there has been a great deal of controversy about their status. Since much of this controversy has concerned the sense-datum theory either explicitly or implicitly, we must try to isolate the confused as well as the correct elements in that theory.

I take G. E. Moore's article, "The Status of Sense-Data" to be one of the best and clearest statements of the theory in its original form.l The following are Moore's main theses:

(1) In all cases of sensory experience the entity which is experienced, whatever its nature may be, must be distinguished from the fact or event which consists in its being experienced. Moore calls the entity which is experienced the "sensible."

(2) When we see, hear, touch, etc., we have the same kind of re­lation to different kinds of sensibles. We even have the same kind of relation to an image of a sensible as we have to the sensible itself. The difference is in the kinds of objects.

(3) If two persons look at the same physical object they directly apprehend different sensibles. Since they are distinct, the surface of the object cannot be identical with both sensibles, and the physical object will be distinct from the sensible of it. Moore does not say that a physical object can never be identical with a sensible, but it would seem that he must say this in order to avoid absurdities.

(4) Acts of apprehension are in the mind in a very different way from that in which sensibles are in the mind. A sensible in the mind is an object of direct apprehension. An act of apprehension, may, how­ever, be in the mind without being directly apprehended. Moore seems to imply here that at times we can be conscious of our appre­hending and that in those cases the act of apprehension would also be the object of another act of apprehension. As a consequence, Moore deduces that there are at least two senses of the word "see,"

1 G. E. Moore, "The Status of Sense-Data," reprinted in Philosophual Studies.

Page 14: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

4 FOUNDATIONS OF SOLIPSISM

(a) the sense in which we can be said to see a physical object and (b) the sense in which "see" means merely to directly apprehend a visual sensible. These are always very different.

(5) A proposition of the form, "I see A," where A is a physical object, entails that there is a visual sensible, B, which I am directly apprehending.

The upshot of these assertions is that there are three distinct ele­ments in perception. First, there is the physical object which is per­ceived in one sense of "see," "hear," etc. This is always distinct from the sensible and is never a mental entity in any sense. In certain cases, such as dreams and hallucinations, there actually may be no physical object. Secondly, there is the sensible. This has certain characteristics in common with physical objects; it has internal spatial relations and has many of the same sorts of predicates applied to it as do physical objects. While a sensible is "in the mind" and is a mental rather than a physical entity, it may nevertheless be "from a place" in the way that a certain view of a famous building isfrom a certain place. In order to be consistent Moore probably would have to distinguish two senses of every predicate applied both to sensibles and physical objects.

Lastly, there is the act of apprehension of a sensible. This act is a purely mental event. It cannot be described in any of the ways in which we describe physical events. The special feature of these mental acts of apprehension is really that they have no content at all. Moore has put the content of these mental events into the sensibles which are apprehended, and there is no way of distinguishing acts of appre­hension from one another except by their objects.

The admitting of some entities besides physical objects in perception amounts simply to a denial of extreme behaviorism or materialism. But this issue will be taken up below. Granted that some sort of mental event is involved in a sensation, our present interest is in the assertion that a sensible and a distinct act of apprehension are both involved. Let us now look at some of the arguments Moore gave for this assertion, along with some others that he might have given.

It is often argued that in the case of a hallucination or a dream something is seen, but no physical object is there to be seen. Conse­quently, it must be something like one of Moore's sensibles which is seen and which is the object of consciousness at that moment. Simi­larly, when one sees a penny at an angle, there is a sense in which what is seen is elliptical. These arguments are not conclusive, and Moore may have realized this himself. No one would deny that when one sees

Page 15: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

SENSATIONS AND IMAGES 5

a penny at an angle one may have the same experience one would have in seeing an elliptical penny square on. Similarly, hallucinations and dreams may be very vivid. Their constitutive experiences may be just like experiences which turn out to be veridical. J. L. Austin has pointed out that there are many different kinds of hallucinations. Although he may be right about there being no sharp line between hallucinations and veridical experiences, here we need only the extreme cases.1 Austin's views on the subject will be discussed in detail in Chapter II, but even now we can safely say that some dream experiences are not sensations of any existing physical object, and that certain ordinary experiences are sensations of particular existing physical objects. Again, as Austin points out, the range of the term "physical object" may be question­able, but for this discussion it is sufficient if we can find some entity which is obviously a physical object and some entity that is obviously not a physical object. Again, we need only the extreme cases to get our discussion off the ground. Thus I do not think that Moore can be refuted because of the very language he uses.

Nevertheless, that there are at least one veridical experience and one non-veridical experience does not entail that there are sensibles distinct from our consciousness of them. We can merely conclude that there are a variety of mental events which we can easily distinguish from one another and that there is less than a perfect correlation be­tween these kinds of mental events and the physical environment of the perceiver. Of course, if sensations just differ from one another without having as a common element the consciousness of the other elements of the sensation, we cannot really speak of the object of a hallucination as opposed to its subject. Moore's use of sensibles and sense-data does always allow him to speak of objects. He seems to want such objects because one can say such things as "In my dream I really did see Caesar behead a Gaul," which is synonymous with, "I really did have a dream in which I saw Caesar behead a Gaul." In our ordinary speech we do say that certain things in dreams were seen even though everyone knows that they were not seen in the usual sense. The alternative is to say "I had an experience in my dream which was very like the experience I would have had if I had seen Caesar behead a Gaul." Here no object of the sensation is mentioned, and this sentence may be synonymous with the ones above. Nevertheless, many philoso­phers have been wary of hypothetical translations of categorical

1 J. L. Austin, Sense and Sensibilia, pp. 6-32. Austin's views will be discussed in much greater detail in a later chapter.

Page 16: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

6 FOUNDATIONS OF SOLIPSISM

sentences; a sufficiently wary person must, with Moore, postulate something like a sensible that is really seen. On the other hand, we can also say in our ordinary speech such things as, "I dreamed that I saw Caesar do certain things, but, of course, I really couldn't have seen Caesar at all, and I realized this when I woke up." We are willing to admit that there really was no object even though we want to insist that something happened to us. We can admit that, however, without postulating sensibles, so this argument does not seem to be decisive. Further, it now seems that someone who is determined to follow ordi­nary speech in such great detail will wind up both having real dream objects and not having them.

It should be remembered, nevertheless, that if we diverge from Moore in the postulation of sensibles we are, in certain cases, admitting hypo­thetical analyses of categorical sentences.

We must now turn to Moore's main argument for the distinction between sensibles and acts of apprehension which occurs in his "Refuta­tion of Idealism."! Moore here says that we must always distinguish in a sensation, say of blue, two parts - the consciousness on the one hand, and the sensible, blue, on the other hand. This is because the sensation of green and the sensation of blue have a common element, the con­sciousness, and they also differ. Consequently, each must have two parts so that one part is the same in both, and the remaining parts, the blue and the green, differ. Moore does not conclude that there can be consciousness which is not consciousness of anything, but he does con­clude that the sensibles, blue and green, are the sorts of things that could exist unperceived. This is also his main reason for saying that the act of apprehending a sensible is distinct from the sensible itself.

Let us now suppose that we have a red apple and a green apple. Here there is also a common element, the applehood, and a respect in which the two apples differ. If we used Moore's argument here, we would distinguish between the apples and their colors, and, of course, there is such a distinction. However, Moore's argument would also lead us to conclude that the properties of redness and greeness can exist independently. These happen to be sensible properties, so perhaps this consequence is not absurd from Moore's point of view. However, we could also choose examples involving very complex non-sensible properties, and Moore would then have to admit that these could exist independently as well. This seems somewhat odder, and makes one look for a fallacy in the argument.

1 G. E. Moore. "Refutation of Idealism," reprinted in Philosophical Studies.

Page 17: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

SENSATIONS AND IMAGES 7

The basic point is that in our language we can compare and contrast things or events in two different ways. First, we can point out similari­ties and differences between events by distinguishing the events them­selves from their properties. We may then find that the events share some properties, but not others. Secondly, we can compare two events by dividing each into parts and then pointing out that the events have some parts in common (or have some parts which are very similar) but have other parts which are dissimilar. When we distinguish between an event and its properties to account for differences, we need not suppose that the properties can exist apart from any events or things for them to characterize. But when we distinguish between the parts of a thing or event to account for differences, it ordinarily happens that the parts are separable and that one part can exist independently of anything like the other parts with which it happens to be in company in a particular whole. That is, the properties of an event are incomplete in themselves in a way that a part of an event is not.

In the case of the sensations of blue and green, the idealist is saying that blue is a property of one conscious event and that green is a pro­perty of the other conscious event. But Moore takes the distinction between the whole mental event and the blue to be a distinction be­tween a whole and one of its parts. Thus he says that the mistake of the idealist is "to identify a part either with the whole of which it is a part or else with the other part of the same whole."l Actually, Moore him­self has confused the thing-property distinction with the part-whole distinction, and this is why he arrives at the conclusion that the green is a separable sensible which could conceivably exist independently.

Consequently, there is no decisive argument for the existence of sensibles or sense-data in Moore's sense of the term. Whether one posits them will ultimately depend on the value one places on metaphysical economy as against logical simplicity. In this book, the latter will con­sistently be sacrificed to the former.

I still think, however, that there is something of value in the sense­datum theory. I will try to bring this out by making a distinction be­tween two kinds of sensation situations. I will then argue that Moore and the traditional sense-datum theorists made a very basic mistake in overlooking this distinction. Part of what they wanted to say is still true, but only if construed in a somewhat different way.

We must first show that having a sensation is a kind of conscious

1 Ibid., p. 18.

Page 18: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

8 FOUNDATIONS OF SOLIPSISM

state; thus given the usual sense of the word, it would be contradictory to assert that someone is having a sensation but is not conscious of it. It seems clear, further, that consciousness in general cannot be reduced to the behavioral correlates of consciousness or to any set of purely physical events. First, it is always possible to be conscious in some way of any physical event. Hence, if A's being conscious amounted to some set of physical events, he could also be conscious of those events. But under the behavioral account this consciousness itself would involve some further behavior on A's part, and he could be conscious of that, too. Hence A could be conscious of being conscious of being conscious, without limit, and each new consciousness of consciousness would in­volve further behavior. Of course, there may be a difference between consciousness and self-consciousness, but no higher order of conscious­ness adds anything new, contrary to what would be entailed by the extreme behaviorist position. Thus we cannot reduce conscious states in general to non-conscious events. Much the same could be said for the special case of sensation.

There are two different kinds of conscious situations now to be distinguished. First, there is the kind where we presuppose an object of consciousness, often a physical thing, distinct from the conscious state itself. An example of this would be my conscious state when I say, "I am conscious of someone following me." I will call sensation-situa­tions of this type "independent sensings," referring to the sensation it­self in this situation as an "independent sensation." In the other case, one does not suppose that there is any such object distinct from the conscious state itself. An example of this would be the conscious state I have when I say, "I am conscious of a dull pain in my finger." I will call sensations of this kind "dependent sensations" since it is never supposed that the pain has any sort of independent existence or that it will continue to exist when I am no longer conscious of it. In the second case sentences about consciousness assert only that a certain kind of mental event is taking place and the "I am conscious of .... " serves only to indicate that I, as opposed to someone else, am having the pain. It does not say that the consciousness occurs in addition to its object.

But does this distinction between independent and dependent sensa­tions rest on some more basic difference? Similar problems have some­times been discussed under the heading of the "intentionality" or the "referring quality" of sensation. It is clear that if sensations do refer to something else or to intended objects, this must be true only of in-

Page 19: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

SENSATIONS AND IMAGES 9

dependent sensations; in the case of a dependent sensation there is nothing for it to intend or refer to. Thus, in analyzing the independent vs. dependent distinction we will also be giving an account of the referring quality of sensation.

Given the distinction between independent and dependent con­sciousness and the corresponding distinction between independent and dependent sensations, it is important to notice that there can be a dependent sensation for every independent one, and vice versa. To a state such as that usually described by saying "I see an island over there" there corresponds the state describable by "I am having just the sort of sensation I would have if there were an island over there, but I know that it is just a mirage (hallucination) and that there isn't any­thing there." The content of these two sensations could bejust the same. The difference is only that in the one case we suppose that there is an object existing independently, and in the other case we think that there is no such object. Similarly, to a dependent sensation of the kind that might be described by "I feel a sharp pain" there will correspond a possible independent sensation, perhaps described by "A sharp pin is sticking into me," which might be exactly similar qualitatively. Thus, the distinction between the two kinds of sensation need not in general rest on a difference of quality between the content of the sensations themselves.

One might say that the difference here is only between a bodily sensation and a sensation of an external object. Thus a dependent sensation would be a sensation of one's body and an independent sensation would be of an external object. In this connnection it might be argued that when someone feels pain he is actually having a sensation of part of his own body. I do not think, however, that this is a good account of what happens. Ifl have a pain in my finger, the finger is not painful in the way that it is white. The pain is in the finger in the sense that the partial cause is there. Much the same thing happens when we say that a mirage is there on the horizon, but in this case the distinctive part of the cause is not even in the body. It would be even less plausible to say that when someone is seeing a mirage he is sensing his own eye­ball, yet this is what we should say in order to insist upon the distinc­tion's being only one of bodily sensations as opposed to those of external objects.

Further, the difference here is not the difference between a sensation and an image, even though there are some similarities between images and what I have called dependent sensations. In neither case do we

Page 20: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

10 FOUNDATIONS OF SOLIPSISM

suppose that there exists independently of consciousness an object of the conscious state. However, in other respects the dependent sensation is more like an independent sensation than an image. In particular, it is usually possible to produce an image or to rid oneself of an image without moving the body in any way, or by doing anything to affect the sense organs. Dependent sensations, on the other hand, cannot be produced, destroyed, or changed without such actions. One cannot make a pain disappear in the way that one can rid oneself of an image. One cannot see a mirage except by looking in the right direction, whereas one can produce an image of an island without looking in any direction.

It seems clear, therefore, that the only basic difference between the two kinds of sensation is a belief as to the existence or non-existence of certain independent objects, and that this difference is not the result of some other difference. Further, this belief is distinct from the sensa­tion itself. In fact, this belief does not seem to be at all active in the whole conscious state which includes the sensation. When I have a pain I do not consciously think to myself, "this pain will continue only until I am no longer conscious of it" just as I do not think to myself, "The pin which is causing the pain will continue to exist even when I no longer feel it." On the contrary, it never occurs to us to consider either of these questions. Thus the difference between dependent and independent sensations need not, and usually does not, imply any differ­ence between the conscious states which include the sensations. Con­sequently, it is better not to speak of the referring quality of sensations since the conscious state itself doesn't refer to or intend anything. The connection with the object is exterior to the sensation and has to do with the context in which it occurs.

There are many cases where we can be said to have certain beliefs without ever having been conscious of them or having considered the issues involved. If I say that I have time to get to the bank before it closes, I am presupposing, among other things, that a flood will not interfere with my progress towards the bank. Clearly there are a great many things which we presuppose in this way simply because it has never occurred to us to doubt them, and these can also be called beliefs. In our expectations, our thinking, and our behavior we operate on the basis of such presuppositions being true just as much as we do with any other beliefS. Similarly, in certain sorts of sensation-situations we have a not-necessarily conscious belief to the effect that there is an object being sensed, and in other sorts of sensation-situations the opposite

Page 21: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

SENSATIONS AND IMAGES 11

belief is presupposed. There are, of course, rather rare borderline cases where one might wonder whether the island one sees is a mirage or not, or where one might wonder whether a persistent mirage should be treated as a physical object. Even here we can switch from one hypo­thesis to the other without altering the sensation in any way. Once we make a decision as to the reality of the island the consequent belief will then become unconscious in subsequent viewings of the island.

We must now allow for the possibility that beliefs influence percep­tion. It might be that when one sees a distant mountain top and believes that it is covered with snow, one's sensation is different from what it would have been if one had believed the mountain top to be covered with white canvas. The difference would have to do with a perception of cold in the snow and not in the canvas. Could it be that when one has a visual sensation of something which is believed to be cold, the sensation has what would normally be a tactual quality? That is, might it be similar to the sensation one would have upon looking at and touching a similar object which is cold? If so, we must consider the possibility of whether a belief in the existence of the object sensed could make a systematic difference in the sensation itself and thus contradict what we have said about the qualitative identity of independent and dependent sensations.

It is, of course, possible that, in a particular case, a sudden realiza­tion that the sensation is hallucinatory might cause a qualitative change in the sensation. However, this says little because practically anything that might happen could cause such a change. The important question is whether such a belief could make a systematic and regular difference to the sensation. While it is possible that some beliefs might alter sensa­tions systematically, this does not seem possible in the case of a belief about the existence of the object of the sensation. If the belief in the non-existence of the object did alter the sensation it would have to alter certain qualities of it; this would not make the sensation a sensa­tion of a non-existent object, but it would rather make it a sensation of a merely different object. It might be supposed, for instance, that a belief in the non-existence of an island which is being viewed would make the outlines of the island in the visual field become vague and indistinct. But then one could have exactly the same sensation if one were looking at the island on a foggy day, and in this case one might believe in the existence of the island. One need not even suppose a change in the intervening medium; a momentary abnormality of vision of which the perceiver is aware could produce the same effect that the

Page 22: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

12 FOUNDATIONS OF SOLIPSISM

belief in the non-existence of the island is supposed to produce. Thus, there is no quality which a sensation can have which always accom­panies dependent sensations and never accompanies independent sen­sations, or vice versa; we can always imagine an independent (depen­dent) sensation which would have the same quality. Hence, the belief in the existence or non-existence of the object does not make any general difference to the sensation.

It is important to notice now that there is a very close relationship between independent and dependent sensations. In fact, whenever an independent sensation occurs it takes only something which is distinct from the sensation itself, a change in conscious or unconscious belief, to convert it into a dependent sensation, and vice versa. It will then be true that whenever someone has an independent sensation he is having something which could be a dependent sensation in the context of different beliefs.

Returning to the sense-datum theory, two theses are most important. The first is that in sensing there is always an object (the sense-datum or sensible) which is sensed and which is distinct from the conscious state itself. We have already rejected this thesis; in dependent sensation there is no object at all, and in independent sensation there is believed to be an object, but this object is a physical thing rather than a sense­datum. The second thesis is that whenever we see anything we have sense-data. We must also reject this insofar as the term "sense-datum" is defined by the first thesis. However, the heart of the second thesis is that whenever we see anything, something is happening which is simi­lar to what might be happening if there were no object, and further, that the same thing might be happening even while there was believed to be no object. The sense-datum theory says that in all these cases we are having sense-data, and that these sense-data might be the same. While sense-datum theorists have not been very clear on this point, I think that one of their basic attempts is to group together cases such as "feeling pains caused by pins" and "feeling pins which cause pains." My argument that there need be no conscious difference between dependent and independent sensations would, if correct, substantiate this second thesis insofar as it is separable from the rest of the sense­datum theory.

The next question that traditionally arises concerns the privacy of sensations. Some issues connected with this question will be discussed in later chapters, but many will never have to be raised at all. Al­though a statement about one's own sensations and a statement about

Page 23: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

SENSATIONS AND IMAGES 13

someone else's sensations will differ formally in ways to be specified under solipsism, statements about the sensations of others ultimately, of course, will be reduced to statements about one's own sensations. We need not trouble greatly about the privacy of sensation at this point, but can go on to discuss the second of the solipsistic building blocks.

The theory of meaning to be developed will depend a great deal on the concept of an image. Much of what has been said about sensations will also be true of images, but it will be useful here to compare and contrast images and sensations. In making this distinction we will depart somewhat from the ordinary usage of the terms "image" and "sensation" or "sense experience." In the first place, the ordinary meaning of these terms is not entirely clear and it is doubtful that they can be clarified without ignoring certain aspects of their use. In the second place, we are not attempting to give an analysis of these terms, but rather to use them in the analysis of other notions. For this reason it is better to clarify these terms as much as possible, even if it does in­volve a certain distortion of their ordinary meaning. We will therefore try to give criteria for deciding whether a given mental event is an image or a sense experience at the time that it occurs; insofar as this is possible we will approximate the ordinary meaning of these terms.

The term "sense-experience" or "sensation" will be used in such a way as to exclude images, and the term "phenomenon" will be used to include either. Phenomena such as my experience of the table in front of me clearly belong to the class of sensations, which my present image of a centaur does not. But there are other cases which are not so clear. Consider, for instance, the hallucinations of pink rats which drunks are alleged to have. Although a candidate for "image," this sort of phe­nomenon is better classed with sense experiences simply because it is impossible to distinguish it from non-hallucinatory sense experiences at the time the experience occurs. Phenomenally, the two must there­fore be similar. The same can be said for certain dream experiences since they are, in themselves, indistinguishable from waking experiences. These examples have been of independent sensations, but the same will hold for the corresponding dependent sensations, as will be seen later.

We have a different situation with my image of the centaur. I am immediately able to distinguish it from my sensation of the table even at the time it occurs; thus there must be some phenomenal difference between the two which we must make more explicit. It is sometimes

Page 24: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

14 FOUNDATIONS OF SOLIPSISM

suggested that the distinction rests on certain relations between the phenomenon and the world. For instance, it might be that a phenom­enon is a sense experience if and only if it is believed to represent some part of the physical universe (whatever this may mean). We have al­ready seen that this is not true. We are often confronted with mirages and one's being told that what had first been taken to be an island is in fact a mirage does not seem to affect one's experience of the mirage itself. Even if it did, this would not constitute grounds for saying that what started out as a sensation became an image. We still look at the mirage as if it were "out there" even though we know that there is actually no entity in the world which it represents in the way that one's view of an island represents the island. In general, if the "representa­tive function" of phenomena is construed as a belief or attitude con­cerning the phenomenon, it does not have the sort of effect on the phenomenon itself which would lead us to move it from one classifica­tion to the other. Thus the difference between a sensation and an image is not the difference between an independent sensation and a dependent sensation. I could look at a mirage of a centaur knowing that it was not real, and this would be different from having an image of the centaur. We must therefore look further for this difference.

But before doing this it should be pointed out that there is a complete qualitative correspondence between images and sensations of either kind. An image may have any quality in any degree that any sense experience has, and vice versa. The distinction is thus not a qualitative one. It also seems quite clear that the distinction is not, as Hume thought, one of intensity or vividness. This is seen most clearly when we realize that we can have auditory sense experiences which are just loud enough to be noticeable while, on the contrary, we can imagine very loud noises. The qualities of sensations and images are comparable in this way, even though, as we shall see, sensations and images do not have qualities in exactly the same sense. We can specify the qualities of images only by mentioning the qualities of sensations or physical objects. Even taking vividness to transcend mere intensity, it still seems that we can have phenomena which are indisputably images and which are more vivid in aTl:Y sense than certain sensations.

It might also be that the distinction rests on the presence or absence of certain kinaesthetic phenomena, such as accompany focusing the eyes, reaching out the hand, etc. Sensations of objects are often accom­panied by these kinaesthetic sensations, whereas images are generally not accompanied by the corresponding kinaesthetic images. However,

Page 25: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

SENSATIONS AND IMAGES 15

it is possible to imagine the kinaesthetic sensations along with the ima­ges of the object, while, at the same time, it is not clear that all our sensations are accompanied by the kinaesthetic sensations. In any case this would not give us a criterion for distinguishing between kinaesthe­tic sensations and kinaesthetic images.

Another possibility is one of an emotive tone that goes along with sensations and is absent from imagery. This can be called the "tone of sensation," and it is irreducible to intensity or any other quality of the phenomenon. Rather it is something sometimes added to the phenom­enon without affecting it in any way. Other than this, we can give no description of the tone of sensation; to say that it accompanies one phenomenon and not another that is identical qualitatively is just to say that there is a phenomenal difference in the way the phenomena occur, and that the one accompanied by the tone is in some sense "richer" than the other.

There are two reasons that might be given for thinking that such a tone at least sometimes accompanies what we call sense experiences, as opposed to images. First, there would be personal experience and introspection. Secondly, there is the fact that we seem to be able to differentiate images and sensations on the basis of the phenomena alone. Thus if there is no other qualitative difference, some such tone might be plausible. This could also be explained by positing an in­definable tone which accompanies images, but perhaps we generally feel that, if anything, there simply is more to sensation than to imagination.

However, even if this feeling does accompany sensation, it will not preserve the usual distinction between sensation and imagination. This is because it is possible to imagine a situation in exactly the way it might be experienced. That is, it is possible, though more difficult, to complement an image with the feeling that ordinarily goes with sensa­tion. One way to do it is to imagine a whole series of events rather than a single event or object. For instance, one might read a detailed account of a battle (and perhaps even see a movie of it); one then sets out to imagine the whole battle. Mter several practice runs one can generally put oneselfinto the situation so completely that things begin to happen by themselves and although we are imagining the whole thing our images become indistinguishable from sensations in themselves. Fur­ther, it does seem possible to sometimes have sensations which do not have any such tone of sensation. If one puts oneself into a meditative mood and gazes out at the countryside as if it were in a dream, the resulting sensations phenomenally are rather like images. Thus, it does

Page 26: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

16 FOUNDATIONS OF SOLIPSISM

not matter for our purposes whether or not there is such an irreducible emotive tone that goes with sensations.

From this we can see that some other distinction must be involved, and it may have to do with the fact that images are often preceded by a conscious resolve. That is, one purposely sets out to imagine something and then the image occurs. It sometimes happens, of course, that we have images that are not intended in this way; what we call images just pop up, perhaps because they are related in some way to something that we are sensing or thinking about. And, of course, one could also resolve to have a sensation and then have it. In such a case, however, it seems that one has to do something besides deciding to have the sensation in order to actually have it. Such an additional required action might be opening the eyes, looking in a particular direction, sniffing with the nose, etc. Nevertheless, the difference does not seem to be just the presence or absence of a conscious resolve, although this is involved. Rather, it seems that a phenomenon which we ordinarily classify as an image is constantly dependent on our intention, but not on our actions. Even if the image pops up without our intending it, we can alter it in all sorts of ways just because we want to do so without doing anything at all with our bodies. There may be certain kinds of images (obsessions, preoccupations, etc.) which we cannot always rid ourselves of so easily, but if they are images, we can at least make them change perhaps with psychiatric advice, in ways that a sensation will not change. Such a phenomenon as MacBeth's vision of the dagger may be very close to the dividing line between a persistent image and a dependent, or even independent, sensation. If MacBeth eliminated the latter possibility by assuring us that he didn't actually suppose a dagger to be in front of him, we would decide the remaining question by asking such questions as "Can you make the dagger spin around or get larger, or does it behave the way it wants to, just as if there were a real dagger there?" The answers to such questions would be decisive.

We may still have failed to eliminate all the middleground between sensations and images, and to some extent we may have departed from the ordinary use of these terms. But it now should be fairly clear what will be meant by the terms involved. It is hoped that something like the ordinary meaning has been refined.

Having distinguished images from sensations, let us remember that they share two important properties. First, both images and sensations are mental events in the sense that they are processes of consciousness which we ourselves distinguish from the physical things and events

Page 27: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

SENSATIONS AND IMAGES 17

which are the objects of consciousness. Secondly, there is no inherent qualitative difference between images and sensations.

The theory of imagination presented here contrasts with that of Gilbert Ryle.1 I have re-ordered his main assertions in order to make them clearer. The first of these is that there are many ways of being imaginative; having images is only one of these. We do often say things of the form, "I imagine that .... " without implying that images actually are occurring. This can be roughly equivalent to "I think that ... " Chiefly, Ryle wants to deny that the imaginary has any other-worldly reality or that images are non-physical pictures which are contem­plated. It does seem incorrect to treat images as if they were contem­plated pictures existing in some special realm. I have said nothing which implies this. As was pointed out in the corresponding case of sensation, we cannot divorce in this way the person who is conscious of the image from the image and then treat the former as if he were contemplating the latter. But Ryle denies that images are mental enti­ties that are irreducible to the physical universe. Thus he would also be denying the view held here that images denote certain conscious processes, in our non-behavioral sense of "conscious." It is not clear whether he intends his arguments to apply just to the theory of "special non-physical pictures" or to our sort of theory as well. For instance, he says, "It is part of this doctrine [the one he is arguing against] that the picture that I see is not, as snapshots are, in front of my face; on the contrary, it has to be not in physical space, but in a space of another kind. The child, then, who imagines her wax-doll smiling is seeing a picture of a smile. But the picture of the smile is not where the doll's lips are, since they are in front of the child's face. So the imagined smile is not on the doll's lips at all. Yet this is absurd. No one can imagine an unattached smile, and no doll-owner would be satisfied with an un­smiling doll plus a separate and impossible simulacrum of a smile suspended somewhere else."2 But we do not want to imply with Ry1e that the image of the smile has any spatial location. This reluctance in no way commits us to the view that one can imagine a smile without also imagining the lips along with it. On the contrary, we are merely holding that one can imagine any concrete chunk of the universe without imagining the neighboring chunks.

In another place Ryle says: "and to say that someone pictures a dragon is to say, not that he dimly sees a dragon of a peculiar kind or

1 Gilbert Ryle, The Concept of Mind, pp. 245-279 . • Ibid., p. 248.

Page 28: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

18 FOUNDATIONS OF SOLIPSISM

something else very like a dragon, but that he does not see a dragon, or anything dragon-like at all."! We have already seen that we should not speak of seeing sense experiences or sense-data, and this also holds of images. But it is still the case that images have a great deal in com­mon with sensations; having an image of an object will be very like having a sensation of it, in fact, like seeing it. It does not follow that we are seeing a dragon dimly or that we are seeing something else which is like a dragon, but just that we are going through a conscious process similar to that which would occur if we did see a dragon except in the respects indicated above. Thus, the mental picture language is justified to a certain extent so long as this is all we mean by it - that is, so long as we do not suppose that there are two entities involved, the image and the process of being conscious of it.

Ryle's second main contention is that when we see and "see" (image) we are doing things which are totally different in kind. Here he is denying what we have just asserted. He further says that we can see and hear only what is there to be seen and heard; we only imagine or image what is not really there. He admits that we can be mistaken about whether we are seeing or only "seeing," but says that this does not erase the distinction. However, if the two processes were totally different in kind, as Ryle asserts, one would suppose that we could never mistake one for the other. Further, Ryle has to include under "seeing" not only visual images but visual dreams and hallucinations and the seeing of mirages, because the object is not really present in any of these cases. It then seems particularly strange, if there is no similarity in kind between seeing and "seeing," that there is no phenomenal difference between seeing an island and "seeing" an island when we are looking at a mirage. In fact the similarity may be so strong that we can tell the difference only at a later time by consulting someone else or by sailing through the mirage, or by something of that sort.

Ryle also points out that we talk about ordinary sensations different­ly from the way that we talk about hallucinations and images. We can say that the latter are vivid, lifelike, and faithful; we cannot say this of the former. But this does not show that there is any great intrinsic difference between sensations and images. It is just an indication that we consider the former more informative as to the nature of material objects. Thus, to say of an image that it is any of these things may be just to say that it is very similar to the sensations which are cha­racteristic of the object. This difference in language merely shows

I Ibid., p. 251.

Page 29: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

SENSATIONS AND IMAGES 19

that images are related to objects in the world in a different way. Ryle says that imagining is not to be likened to the purely sensuous

element in perceiving but rather to such processes as the following of a tune; running through a tune in one's head is like hearing the tune in that both are utilizations of one's knowledge. He presumably means that, in both cases, one knows what note is coming next, at least if one has heard the tune before. Ryle distinguishes this from cases where we sense an object without knowing what it is. He points out that one may see a watercourse without recognizing it and without being able to say how it would look close up, etc. But he then says that it is im­possible to imagine something without knowing what it is.! This last does not seem to be true. If I can see a watercourse without knowing what it is, I can just as well shut my eyes, imagine what I have just seen, and still not know what I have imagined since I have no new information about it. Similarly, we can run through notes in our head without always knowing what notes are coming next; we sometimes have to stop and decide consciously what the next note is to be. Thus we must conclude that this "knowing what comes next" is not essential either to imagination or to sensation and does not explain away our belief that they have a great deal in common.

There is another aspect of the phenomenology of sensation and ima­gination which, though already touched upon, must now be considered in more detail. R. Firth in his article, "Sense-data and the Percept Theory," states and supports the thesis of certain psychologists in this matter.2 His first main point is that in ordinary perception we are conscious of more qualities than epistemologists generally have held in the past, and that we are conscious of these additional qualities exactly in the way that we are conscious of the old qualities. It generally has been held that when we see a tomato we sense only a red, round, flat patch. We also know that it has depth, insides, etc., but this knowledge belongs not to the sensation itself but rather to the interpretation we make of it. Firth, on the other hand, holds that we are conscious of its depth and certain other properties, such as taste and smell, in exactly the same way that we are conscious of its redness and roundness. Similarly, when from a warm room we see a distant mountain which is white with snow, we are conscious of its coldness in the same way as we are conscious of its whiteness. That is, our sensations will be like

1 Ibid., p. 265. a Roderick Firth, "Sense-Data and the Percept Theory" in two parts: Mind, Oct. '49

and Mind, Jan. 1950.

Page 30: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

20 FOUNDATIONS OF SOLIPSISM

those of coldness in certain respects, but the whiteness will probably be more intense. In fact, when we are perceiving an object by means of one of the senses our perception will often contain elements that usually come via other senses. Secondly, Firth admits that we can take up an attitude of abstraction or doubt that will cause us to perceive the bare minimum of qualities. That is, if one sets one's mind to it, one can perceive the same tomato and be conscious only of its redness and roundness. We can, as he puts it, effect a phenomenological reduction. But he also claims that the sort of perception which we have as a result of this reduction has no priority over any other kind of perception. We can affect our perception by taking up a great many different atti­tudes, but all equally are kinds of perception, and there is no reason to pick out some core of qualities and say that these are the only ones which are really perceived.

Phenomenologically, this seems quite correct, and it is also in the spirit of what we have said earlier, that we perceive whole objects and not intermediaries called sense-data.

It further follows from Firth's view that sensations cannot be ulti­mately divided up into the traditional five categories. For one thing, most sensations will not be limited to just one sense. Visual sensations will almost always have a tactual element and vice versa; similarly, olfactory experiences will generally have a gustatory element and vice versa. Even auditory sensations will often have a visual or tactual element; we usually do not just hear noises but rather the noises made by various objects. We will still, of course, be able to have sensations that are strictly limited to one sense by means of the process of reduc­tion, but these will be the exception rather than the rule. Further, we will usually not be able to carefully distinguish in a sensation its various elements according to the senses; that is, there will be no sharp boun­dary within the sensation between what is tactual, visual, etc. For instance, some of the most common properties, such as size, shape, position, speed, etc., belong both to the visual and the tactual. Thus when I perceive a large object with sharp corners coming towards me rapidly, it will be completely impossible to say to just what extent my sensation is visual and to what extent it is tactual; I will be having just a mixed visual-tactual sensation. Thus, there will be a great many sensations that do not fit in anyone of the traditional categories. In our ordinary speech we will still be able to classify them artificially according to their stimuli. If the large sharp object is out of my reach I will call my experience a visual one, and if it is within my reach and

Page 31: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

SENSATIONS AND IMAGES 21

my eyes are shut I will call it tactual. But this is artificial, because these facts may not actually make any difference to the experience itself; even if my eyes are shut and I feel the object only, the visual aspect of my sensation may be just as vivid as if my eyes were open. Insofar as we do classify sensations according to the five senses, we do it chiefly according to the relation of the object sensed to our bodies; we then decide which sense organ or organs are capable of detecting it.

If all this is true of sensations it will be equally true of images. When we imagine a mountain-top we will be at least as likely to imagine it as being cold as we would sense it as being cold when it is visually per­ceived. In fact, it is even harder to classify images into the five cate­gories because the relation between the imagined perceiver, oneself, and the thing imagined is often indeterminate. If I imagine a large sharp object moving toward me, it generally is not clear whether I imagine myself near enough to be touching it or not; thus we cannot use the criterion that we use in the case of sensations to say whether it is a visual or tactual image. But not all qualities are open to more than one sense in the way that size and shape are; thus we usually classify images by taking the most conspicuous qualities of the image and assign­ing the image to the sense which sensations having like qualities are ordinarily associated with. However, even in the cases where we can do this, the images generally are not limited to that sense but have other qualities which are characteristic of the other senses. We must conclude first that very few images are confined to one sense, and secondly, where an image combines elements from more than one sense, that it is impossible to draw a sharp line between these elements. In this respect images are complex rather than compound, as are sensations.

The next problem is that of the extent of imaginability. We have already seen that it is theoretically possible to have an image which corresponds to any given sensation; that is, there is nothing about images which prevents us from producing one exactly similar to any given sensation. But the next question is whether in fact we can always produce such images. First let us consider the possibility of a man who never has any images. However unlikely this may seem it is a theoreti­cal possibility, and if there were such a man it still would not follow that he is unable to produce any images; it might simply be that he has never tried to do so. This situation, where a person is completely un­able to produce any images, seems to be a psychological impossibility. There are ordinary situations that are very favorable to imagination.

Page 32: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

22 FOUNDATIONS OF SOLIPSISM

Suppose, for instance, we show an orange to a man and then ask him to close his eyes and imagine it; under such circumstances it seems clear that anyone would be able to produce at least a rough image of the orange. Further, it seems safe to say in general that any human being can produce an image which has something in common with the sensa­tion he has just experienced. At least, if there are people who cannot do even this, one never hears of them. How close the image is to the sensa­tion will, of course, vary with the individual. It also follows that under favorable conditions there will be no large gaps in the powers of ima­gination of any given person. That is, it is not the case that certain people will be completely unable to produce any olfactory images or any gustatory images, etc. Still, it is possible and probable that certain people are not very good at producing various sorts of images, in the sense that they can produce these images only with great effort and with very little precision. Thus, it does not follow from the fact that it is psychologically possible for any given person to have olfactory images that he can imagine the smell of a peach precisely enough to differen­tiate it from his images of the smells of other fruits.

There are, nevertheless, several factors which cause people to under­estimate their powers of imagination in certain respects. For instance, H. H. Price thinks he is deficient in psychosomatic tactual imagination in that he finds it impossible to imagine the sensations involved in the correct golf swing. We must admit first that it is not always possible to produce images where we have experienced no corresponding sensa­tions, although Hume certainly goes too far when he says that this is never possible. But it is entirely possible that Price and others in his situation have too high a standard and discount the images that they do have. One who plays golf knows that the sensations involved in the golf swing are themselves quite indistinct and blurred. During the swing there is not time for each muscle to report in with a distinct tac­tual sensation; it all happens too fast for that, and one has only an in­distinct sensation not particularly unlike those which accompany other bodily movements. Thus, one who has never had the appropriate sensa­tions might happen to imagine the sensations which he would in fact have if he swung properly without realizing it. Another sort of case might be that of someone who, having smelled garbage, is convinced that he cannot imagine how it smells, but only how it looks. In this sort of case it might well be that he can imagine the smell of garbage, but his images are only partly olfactory and also have a strong visual aspect; when it comes to describing complex images the person might overlook

Page 33: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

SENSATIONS AND IMAGES 23

the olfactory element and describe them as purely visual, possibly because he is under the illusion that they cannot be both.

However, when people talk about their imaginative deficiencies they usually do not mean that they can produce no images of such phenom­ena as the smells offruits; they mean rather that the images which they produce are so vague and indeterminate that they do not distinguish the different smells belonging to the various fruits. As we will see, both sensations and images can be, and often are, indeterminate in the sense that there are some qualities that they cannot be said either to have or lack. It also seems to be true that the typical image is somewhat less determinate than the typical sensation, even though this is not always the case.

We will now state two psychological hypotheses which seem to be true of images and sensations. First, there is a positive correlation be­tween the degree of indeterminacy of images and the indeterminacy of corresponding sensations. For instance, our images of smells are generally less determinate than our predominantly visual images, and our olfactory sensations themselves also tend to be less determinate than our visual sensations. Thus we must remember that where our imagination is deficient in providing us with determinate images, our sensations are also likely to be deficient in the same way. Secondly, there is a high correlation between inability to imagine a certain object precisely and the indeterminacy of images which are qualitatively similar to that object. This is to say that if I cannot imagine precisely the smell of a pear, the imagined smells offruits that I do have will be indeterminate. It will not be the case, provided that I have actually tasted all the common fruits, that I will be able to imagine very closely all except one, and that one not at all. IfI am unable to imagine a fruit that I am really familiar with, the imagined smells of like-smelling fruits will be indeterminate enough so as not to exclude it outright. We can conclude that, in a field where our imagination is weak, there generally are not qualities which are completely unimaginable; it is rather that the images we do have are indeterminate enough so that they might represent any of a whole range of qualities, but none to the exclusion of the others.

Page 34: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 2

LANGUAGE AND SENSATION

It will be remembered that the solipsist's main thesis is that every­thing which can ordinarily be said could, in theory, be said in a lan­guage which referred only to one's own sensations. In the last chapter we, among other things, established that sensations and images are not reducible to behavior. We can now see that the solipsist needs to have a language in which one talks only of phenomena that are non-behav­ioral, in the sense suggested by our investigation. In this chapter we will consider how such a language could be set up and see what difficulties may arise.

One of the most effective methods of teaching and learning language uses verbal definition. In pure instances a verbal definition says that two terms can be substituted for one another in any context without effecting any changes in meaning; at least, this will be true where the expressions concerned are used rather than mentioned in that context. The main purpose of this technique is to give someone a shorter expres­sion which he can use in place of a longer one.

It seems, however, that there are two things which this technique will not accomplish. On the one hand, it will not widen our range of discourse and allow us to talk about anything we could not talk about before. Secondly, it will not give our language greater precision than could have been achieved without verbal definitions. A verbal defini­tion just gives us an equivalence between two sets of terms. If the definition is to be of any use we must already understand one side of the equivalence, and if the two terms or sets of terms really are equiv­alent in meaning we have only acquired a new way of saying the same things and talking about the same entities.

The second point follows from this and is actually a special case of it. It might seem that a definition such as "Dachshund: small dog with short legs originally used for hunting moles" allows us to speak more precisely; we can now use a special term denoting a particular type of dog. As long as we construe the definition as an equivalence, however, we could have used the definiendum with just as much precision, if

Page 35: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

LANGUAGE AND SENSATION 25

with some inconvenience. This will be true however complex the defini­tion and however long our string of definitions. Of course, in actual practice a definition like the one mentioned above might not be taken as final. We might be willing to discard this definition in favor of a more involved and a more precise one, even though this latter defini­tion might be unintelligible to us without further training. For exam­ple, we might give priority to an expert's technical definition over the one originally accepted, but as long as this definition is unintelligible to us, we could not say that our use of the term "dachshund" has in­creased the precision of our own language. Someone else's finer skill in speaking of dogs will not sharpen the precision of our own language about dogs, even if we do use some of his words.

Because of this, the traditional view oflanguage concludes that there must be some other technique by which we can teach someone who does not yet have any language. More generally, the view requires a way to increase the range of things that can be talked about. This view assumes that this other process must be ostension or something that reduces to it.! The classic (and perhaps unrealistic) case of ostension would be where we point out an object to a child who is learning to speak, while we utter the name of the object, repeating it until the child learns the word and uses it himself to designate the object. In practice, it seems unlikely that anyone learns language in just this way; the actual process will not be as simple as this. It remains to be seen, however, whether this picture of the structure of language is consistent with a more plausible psychological account of the learning of language.

The traditional theorist might say, for example, that we do not begin by explicitly giving the child ostensive definitions. Instead we tend to use the names of objects which the child is confronted with in our conversation. Sooner or later this leads the child to associate the name with the object, even though he may at first make the wrong association as often as the right one. This could happen if, in the child's presence, an adult kicks a refrigerator and shouts, "This damned refrigerator is broken again." The child may then associate "refrigerator" with either the refrigerator or the foot kicking it, depending on how he directs his attention.

The child might get a start at language in this way, and, from a philosophical point of view, this is ostension of an implicit kind. We still

1 For an example of the traditional theory see C. I. Lewis, An Analysis of Knowledge and Valuation, pp. 96-168.

Page 36: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

26 FOUNDATIONS OF SOLIPSISM

designate the object, with our glance or some gesture if not with our finger, and we still utter the name of the object, even though it may be part of a sentence. The name is spoken with more emphasis or repeated more often than the accompanying words, and the child eventually will be struck by it. Thus, the traditional theorist may qualify his view psychologically and yet retain the picture: that we learn to talk about new things only because of ostension, that verbal definition only intro­duces abbreviations, and that no other important factors are involved.

But there is one very important point about ostension, in all of its variants, that holds true whatever its role may turn out to be. If we succeed in teaching someone the name of an object by means of osten­sion, it is because we have led him to associate in some way two events, the sensing of the object and the uttering of the word. Ostension could not be successful unless the learner senses the object concerned or is told, by reference to what he has sensed in the past, what it would be like to sense the object. The psychological association which is set up is not a simple one. A child who has been taught the name of an object by ostension does not automatically utter the word which has been taught whenever he sees the same or a similar object. Nevertheless, there are certain contexts in which the word will not be used unless a certain kind of sensation has occurred. It is a psychological fact that if the term "Gila monster" is defined ostensively to someone, he will not then sincerely say that he sees a Gila monster unless he has some of the same sensations he had when the ostensive teaching was taking place.

Even more important, the way the person uses the term will depend, not on the nature of the object which was involved in the ostensive teaching, but rather on the kind of sensation the person had. Ifwe want to teach someone the meaning of the term "magenta" we can do it ostensively by pointing out magenta objects. But we can do it equally well by showing him objects of a different color after arranging the light in such a way that he has the same visual sensations he would have had ifhe had seen a magenta object in normal light. Again, if we show him a magenta object in normal light, but the subject has a tem­porary visual abnormality, the upshot of the ostensive teaching will be to teach him that magenta is the name of some other color, depending on the kind of illusion he has had.

The traditional view of the structure of language thus has the conse­quence that the relationship between objects and their names, and more generally, between the world and language, always depends for its existence on prior associations between sensations and word types.

Page 37: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

LANGUAGE AND SENSATION 27

One of the most interesting criticisms of the traditional view of osten­sion is to be found in Wittgenstein's Philosophical Investigations. In com­mon with the traditional view, Wittgenstein is willing to say that osten­sion is an important part of the learning of language and that in some sense it establishes an association between the word and the thing. However, he is not willing to rest the matter there and asks what this association consists in. In particular, he argues that ostension can achieve its effect only if it is combined with a certain prior training. If the ostension had been combined with a different training a different effect would have been produced.

I think that a general illustration of this point lies in the contrast between simple descriptive sentences and simple commands. In our language we usually begin our ostensive teaching with single words and then graduate to descriptive sentences; thus, I might begin by pointing to a table, say "table" and then expand my utterance to "That is a table." Later on, I might introduce the command "Bring me the table," but the word ostensively taught would not in itself have this meaning. Its utterance alone would not have the effect ofa command, but would at most direct the child's attention to the object. It would be possible, on the other hand, to give the child an entirely different training before using ostension to teach him a language. We could begin without any language at all, teaching the child to bring to us any object to which we might point; this could presumably be accomplished eventually by punishment and reward. Mter this habit was established, we could utter characteristic sounds corresponding to different kinds of objects when we point; gradually we would put more emphasis on uttering the sound and less on the gesture of pointing, until the latter dropped out altogether.

A child trained in this way would interpret an ostensive definition in a different way from a child trained in the ordinary way. The former would always interpret the utterance of a single word in connection with ostension as a command to bring an object of a certain sort. Ordinarily, as Wittgenstein points out, ostensive definition is compar­able to putting a piece on a game-board as opposed to making a move in the game, but in this case it would already constitute a move because of the different prior training.!

For reasons such as this, Wittgenstein argues that ostensive defini­tions can always be variously interpreted. This being the case, it will then be always possible for an ostensive definition to be misunderstood.

1 L. Wittgenstein, Philosophical Investigations, pp. 13-17.

Page 38: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

28 FOUNDATIONS OF SOLIPSISM

As we will see, Wittgenstein goes even further, asking if it is possible for someone who has no language to understand such a definition in the way that we ordinarily do. In particular, he thinks that there are difficulties involved in such a person's successfully guessing one inter­pretation as opposed to another and even in his consciously focusing his attention on one aspect of the situation as opposed to another.1

Wittgenstein then seems to conclude that ostension alone, when not coupled with previous training, is psychologically an insufficient means of teaching language. It may even be argued that he draws the stronger conclusion that ostension is not basic to language in that one must al­ready have at least a rudimentary language in order to learn by it.

I think that the traditionalist might grant that the usual examples of ostension do not really depict the way in which one might learn language from scratch. However, we must then ask whether ostension is an essential part of the more complex process which constitutes this kind oflearning; this raises the further question of whether there is any other way of learning language from scratch, a way not involving ostension. In particular, we must look at the view, possibly suggested by Wittgenstein, that learning from scratch via os tension is self­defeating because it involves a process such as guessing or focusing the attention, which is possible only when we already know a language.

If we grant that an ostensive definition always might be misunder­stood, it nevertheless may be an effective means of teaching someone language from scratch if there is some chance that it will be under­stood, particularly if there is some way of correcting or inhibiting mis­understandings. Thus, we are tempted to say that the child hits on one interpretation rather than another, perhaps more or less randomly, and that there is a good chance of his being correct. If his behavior indi­cates that he has hit on a wrong interpretation, we can indicate dis­approval without making use of language, and we can repeat our ostensive definition. It seems likely that he will then hit on a different interpretation (as shown by the sorts of occasions on which he subse­quently uses the word). In this may we might be able to eliminate false interpretations, until the ostensive definition has the desired effect. Wittgenstein's objection, then, is that the pre-linguistic child is not in a position to distinguish different alternatives and choose one. He thinks that this is what is involved in "guessing" the correct interpretation. Undoubtedly he is correct in saying that the child cannot describe to himself the various alternatives and choose one among them; he could

1 Ibid., p. 16.

Page 39: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

LANGUAGE AND SENSATION 29

not, for instance, ask himself "Shall I take this word 'red' to refer to the color or the shape of this object?" Nevertheless, the success of the ostensive definition does not depend on the child's making a conscious choice between the various alternatives or even being aware that the ostension is having one effect on him when it might be having a differ­ent one. All that is necessary is that it should establish an association between a word and one aspect of the situation pointed at or otherwise indicated. We are then led to say not that the child guesses the right alternative, since this implies an element of conscious choice, but that he may hit on the right alternative in that his attention may be focused on the right aspect of the situation when the ostensive definition is given. At this point Wittgenstein implicitly asks the traditional theorist for an account of pointing to one aspect rather than to another, and more important, for an account of what it is for the learner to focus his atten­tion on one aspect of the situation rather than on another. He then argues, quite plausibly, that this amounts to different things in different circumstances, that there is no single thing that always happens in either of these two processes. Even so, we do not have to suppose that there is a single process which we might call "focusing the attention," in order to hold that ostension is basic to the learning oflanguage. We must suppose only that the child is capable, in any given situation, of focusing his attention on one aspect in some way or other. That is, it must be possible for him to be interested in one aspect much more than in any other aspect. We must then make the psychological assumption that if the word being uttered at this time is associated with any aspect of the situation, it is most likely to be associated with the aspect in which the child is most interested.

In ostensive teaching, we often do not really try to direct the child's attention to a certain object or aspect of an object; instead, we wait until the child is already interested in a certain object or aspect of an object, and then we give our ostensive definition. It is usually not very hard to tell what is occupying a child's attention, and the teacher is in a position to make guesses on this subject. Thus it is often the teacher rather than the learner who must guess right in order for the ostensive teaching to be effective. Suppose, for instance, that a child is sitting on the floor when a cat walks by. He immediately catches the cat by the tail, makes noises at it and examines it. If we choose this moment to point at the cat and utter the words, "There is the cat," our ostensive definition is reasonably likely to succeed; these words may be associ­ated in the future with a cat's being present. All this depends, of course,

Page 40: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

30 FOUNDATIONS OF SOLIPSISM

on the child's being of a suitable age. Even then, the ostension might have to be repeated in similar circumstances before it would be effec­tive. It is conceivable that the child might take the word "cat" or the sentence "there is a cat" to refer to the floor under the cat; or, more plausibly, to some part of the cat. In a situation like this, however, we can be fairly sure that the child's attention is on the cat. If he is un­familiar with cats and is not gazing fixedly at some particular part of the cat, it is quite likely that he will associate the word with the expe­riences of seeing, handling, or perhaps biting the cat. Thus, the osten­sive situation does not demand that the learner go through some special process of focusing the attention. Whatever that may amount to in particular cases, it is clear that pre-linguistic children do pay attention to various objects even though they may not know their names. Fur­ther, it is fairly easy to guess what they are attending to even though there may always be a possibility of mistake. Again, from the point ofview of the teacher, we do not need to postulate any special process of pointing to an object or to an aspect of an object. There are ways of attracting the learner's attention to various objects, but we do not need to make use of them to use ostension effectively. Instead we can just wait until the learner's attention is naturally attracted to the object in question.

The traditional theorist should give some account of the difference between ostensively defining an object-word like "cat" and a property­word like "red." He can agree with Wittgenstein that concentrating one's attention on an object may be quite different from concentrating one's attention on a property, and he need insist only that the pre­linguistic child is in a position to go through the latter process as well as the former. The alternative would be to say that property words cannot be defined ostensively from scratch but have to be introduced later in some other way. When we direct our attention toward an ob­ject such as a cat we actually may be conscious of the object to the ex­clusion of other things. That is, our visual sensations may be almost exclusively of the cat, and similarly for our other sensations. At least, over a certain period of time the object in question may be prominent in our sense fields, and perhaps this is a common element in most of our sensations during this period of time. However, when I am concen­trating my attention on the color of an apple, I need not be conscious of the color to the exclusion of the shape. In this case, attending to the color may consist in comparing it mentally with other objects having the same color, or it may amount merely to picking up the apple upon being asked to find a red object.

Page 41: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

LANGUAGE AND SENSATION 31

As Wittgenstein points out, concentrating the attention can amount to different processes in different contexts. This is especially true of paying attention to properties rather than to objects. Nevertheless, the point remains that it is possible for someone who does not have a lan­guage (typically, a child) to so behave that we can presume him to be concentrating his attention on a property, even though his power to distinguish properties may not be as great as it would be in an adult. In fact, in some sense this ability is already present in someone who is able to pick out such an object as a blanket when it is folded in different ways or not folded at all. The child who can pick up his blanket in these various circumstances has learned how to pay attention to certain properties of an object, and he may be able to do this long before he has learned any language at all. This being the case, there is a good chance for ostensive definition to succeed even when we are trying to teach the name of a property rather than the name of an object.

The question now is whether one can have a language which refers only to sensations. Since up to this point we have found no philosoph­ical grounds for rejecting the view that ostension is basic to the learning of language, that view will be assumed in what follows. Further, we have so far established that there are sensations to talk about and osten­sion and related devices make it theoretically possible to learn such a language. However, it is often believed that such a language would be a so called "private language" and that there are difficulties which are peculiar to such languages.

In discussion of the private language question, two different issues are often confused. One of the questions deals with the possibility of what I will call a personal-sensation language. By this I mean simply a language invented by someone in order to talk about his sense ex­periences, provided that this language is not derived from any other language. It must be independent of other languages in that all of its terms must be meaningful to its originator without his having to define them in terms of another language. The syntax of the language may be similar to the syntax of one of the standard languages; it may have in it words that have the functions that 'and,' 'if,' 'but,' etc. do in our language, again as long as they can be understood without going be­yond the personal language. It should be noted that if such a language as this is possible, a somewhat stronger sense of 'personal-sensation language' also will be theoretically possible. This would be a similar language which was evolved by someone who had never been in con­tact with any other language. But it is always possible that someone

Page 42: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

32 FOUNDATIONS OF SOLIPSISM

might be ingenious enough to develop the mechanics of language all by himself; at least there is no contradiction in the concept of his doing it. Therefore, if a personal-sensation language is possible in the weak sense, it will be in the strong sense as well. This language, in the weak sense, is basically the sort of language Wittgenstein had in mind; it is also that which most writers have described as a private language.

However, the possibility of such a language is a question separate from whether it would be logically possible for anyone else to under­stand it if it did exist. A language is usually called "private" if and only if the second condition is not fulfilled. It is true that the objects being talked about are private in the sense that a sensation belonging to one man cannot be identified with a sensation belonging to some other person. I will take it that two entities are logically distinct if there are any differences between them, even relational ones. In this case my sensation is distinct from anyone else's on the grounds that I come to be aware of it in a different way: by having it as opposed to observing it through behavior. Further, my failure to have a sensation would have had different causal effects. In this sense, the personal language is limited to talking about private entities. But this is not yet enough reason for saying that it is logically impossible for anyone else to under­stand the language.

A number of comparisons between personal sensation languages and ordinary natural languages will help us at this point. Suppose that someone originates a private language and keeps a diary of sensations which have occurred. Perhaps he includes predictions of other sensa­tions that he expects to occur from time to time. This diary might be "translated" by someone who came across it and who was closely ac­quainted with the life of the diarist. He might realize that the manuscript was a diary and think he could correlate sentences in it with events which he knew to have taken place in the life of the diarist. He would also, of course, misinterpret every sentence in the diary, and construe it as a statement in a different, public language. He would naturally suppose that the terms refer to objects or kinds of object, whereas in fact they have been introduced as the names of sensations or kinds of sensation. Hence, one might say that the diarist and the reader are talking about systematically different sets of entities. We might com­pare this situation with the ordinary one where two people are using the same public language to converse about the same objects; there are certain similarities between the two cases. Even in the public case a man would be willing to make such statements as "There is a chair over

Page 43: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

LANGUAGE AND SENSATION 33

there" only after having had certain sensations which are not identical with any sensations that another man may have; similarly, he will alter or take back his statement if he has certain other sensations. Further, the way in which he describes the objects that he sees will depend not on what sorts of objects are there, but rather on the kinds of experiences he has. Ifhe is subject to a sensory illusion the description may be false, or, if he is having a hallucination, there may not be anything there at all. Statements made by two men talking about the same objects in a public language will therefore be prompted by, and corrigible with respect to, events which are private each to each. This is explicitly true for the diarist and his personal language, since he claims to be doing nothing more than reporting and predicting his sensation.

There are also differences between the personal and the public lan­guages. Most important, the reader of the diary would say that the statements he is reading are about objects in the world whereas the diarist would say that they are about sensations. At this point we might suggest that within his language the reader makes the distinction be­tween talking about sensations and talking about objects, but that the diarist does not. It would be possible, however, for the diarist to make a distinction of this sort, although it might not be the same one. He might, for instance, incorporate within his language logical construc­tions and the other paraphernalia of phenomenalists; he could do this without ever leaving his language and without in any way introducing a term with reference to material objects. He could have within his language ways of saying 'and' and 'if ... then,' and he could then build up sets of statements which amounted to the phenomenalist's logical constructions, which are supposed to be equivalent to statements about physical objects.! He could then introduce other terms which would serve as abbreviations for the logical constructions and would thus be defined into the personal language and in no way go beyond it. The diarist would now be in a position to make a distinction between sensa­tion statements and object statements within his language; the latter would simply be more complex patterns of sensation statements.

The problem that now remains is to state the nature of any difference there may be between the language of the diarist and that of his reader. According to the diarist, the difference between talking about sensa­tions and objects will be that in the latter case our statements will be more complex in various ways. If the distinction is the same for the reader there will be no important difference in the two languages; if

1 The details of such a construction will be discussed in chapters 9 and 12.

Page 44: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

34 FOUNDATIONS OF SOLIPSISM

the distinction is different for the reader and he means something more by "talking about objects" than just making more complex statements about sensations, then there will be an important difference in the two languages, which might be grounds for saying that the reader really does misunderstand the diarist systematically and that the latter has a private language. But the important issue here is the analysis of what it means to talk about material objects as opposed to sensations. It is precisely the issue of the truth or falsity of phenomenalism. The ques­tion, in other words, is not about the personal sensation language, but of the proper analysis of material object statements in the public language. Assuming that it is possible for personal sensation languages to arise at all, their nature seems to be fairly clear. It is the status of part of the public language which is controversial.

The diarist might even have in his language logical constructions similar to the phenomenalist behaviorist's analysis of statements about other minds. Again, in asserting the logical construction, or the sen­tence which abbreviates it, the diarist will only be making predictions about his own future sensations and will be entirely within his personal sensation language. The reader might construe the diarist's sentence abbreviating a logical construction as a statement about another mind; again the controversial question is not the analysis of what the diarist means by the sentence, but the analysis of what the reader means by it. If we say that the reader misunderstands the diarist, or that anyreaderother than himself must misunderstand him, it will be because we reject this sort of behaviorism as an analysis of statements in the public language.

For the phenomenalist, then, there will be no important differences between personal-sensation languages and public languages, since he does not suppose that we ordinarily talk "about" objects in any way that the diarist cannot. That statements in the ordinary language are usually intended to cause their hearers to act in certain ways does not distinguish them from the statements of the personal diarist in any basic way. It is possible, of course, to keep a secret diary that is not supposed to affect anyone in the ordinary public language. On the other hand it is possible for the diarist in the personal-sensation lan­guage to write with the intention of influencing readers; all that is re­quired is that the terms of his language be introduced by associating them with sensations, or by defining them within the language so begun without reference to any other language. He might still be aware of the possibility that someone could "translate" his language even though they might misinterpret it.

Page 45: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

LANGUAGE AND SENSATION 35

If, on the other hand, we reject this sort of analysis of statements about material objects and other minds, there will then be a basic difference between the statements of the diarist in the sensation lan­guage and corresponding statements of the ordinary language. Just what this difference is will depend on what analysis of these statements, if any, one does accept.

Is this difference sufficient, under any analysis, to justify us in saying that it is logically impossible that anyone else should understand the statements of our diarist? It is clear that the reader does not fail to understand this writer in the way that an Englishman who does not understand German fails to understand a German. It would be more a case of misunderstanding than a case of lack of understanding, but it would also differ from the usual sort of misunderstanding. The usual misunderstanding always can have practical consequences; if a man misunderstands a command he may perform the wrong action, or if he misunderstands an indicative statement, he may be led to expect some state of affairs which the speaker did not intend. In the diarist's case, however, the misunderstanding would not have consequences of this sort; the statements of the diarist, if they are believed, would always lead the reader to have the same sorts of expectation that the diarist himself had when he wrote them. If the diarist asserted that he kept a dog, meaning thereby a long conjunction of statements about sensa­tions, the reader might interpret the statement as meaning something slightly different, but this difference would not come to light in ordinary practice. Whether it could make an empirical difference, even in theory, depends on our analysis of statements of the ordinary language. Thus, the misunderstanding would be at most very slight in any par­ticular case, but it would also be very general, as it would occur in the case of every such statement made by the diarist.

We might then suppose that the reader was informed in some way that the diary was written in a personal-sensation language (this might happen if the diarist could speak the public language as well as his own language). The misunderstanding would then disappear; the reader would realize that the diarist is talking about his own sensations rather than about objects. He could still draw information from the diary, because of the correlation between terms in the diary describing kinds of sensations and certain kinds of objects which the diarist had been known to have encountered. Our earlier assumption that the diarist was in the main talking about events in his life in chronological order, then turns out to have great explanatory power, for we can match

Page 46: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

36 FOUNDATIONS OF SOLIPSISM

sensation words with objects which caused the sensations to occur. The reader in effect would be treating the statements of the diarist as signs accompanying events.

Although there would be no misunderstanding left, it could be ar­gued that the reader understands the diarist only in the sense that he understands black clouds to be the signs of a storm, and not in the way in which he usually understands what someone is saying. The argument here would presumably be that the reader has no idea what the sensa­tions of the diarist are like, hence does not know what the diarist is talking about. For all the reader knows, the diarist may be conscious of a red apple in the way the reader is conscious of a green apple. This is the sort of possibility that arises all the time for the phenomenalist, but he has a weaker criterion for mutual understanding. He admits that the same object may consistently cause different people to have different sorts of sensation, but holds that understanding is assured by using the same words, not to describe the same sorts of sensations, but to describe sensations which occur in similar circumstances; and this is what people are taught to do in childhood. Even if one rejects phenomenalism one could still adopt this criterion of understanding, entailing in this case that the reader understands the diarist even though he is not acquainted with his sensations. Thus, whether it is possible for anyone else to understand the diarist depends on the sort of criterion we adopt for understanding.

Because this is a very extraordinary kind of case and there is no or­dinary way of talking about it, it seems clear that this question cannot be settled by looking at the ordinary use of the term 'understanding.' Whether we say that it is or is not logically possible for someone else to understand the statements of the diarist, we are adding to the concept of understanding; there is, of course, no reason why we should not do this, but if we do so we should be aware of it. The important point here is to notice the similarities and differences between cases of this sort and cases of the usual sort; having done this, it is not particularly im­portant whether we call the personal sensation language a private lan­guage on account of the differences, or a public language on account of the similarities.

Let us now ask directly whether there could be a personal sensation language. For convenience, we will follow standard practice and call such a language a "private language" whether or not it really is one in the sense delineated above. Many of the objections to private lan­guages are not logically dependent on the classification of personal

Page 47: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

LANGUAGE AND SENSATION 37

languages as private languages, but are based on the nature of a per­sonal-sensation language itself.

There are two main types of objection to such a language. First, there are those which seek to establish that a private language can never succeed in attaching meaning to terms which refer only to a person's own sensations independently of a public language. Secondly, there are those who argue that even if a private language could get started, there would be no criteria for the correct use of the language, hence no rules for its use; the conclusion is then that it would not be a language at all.

As an example of the first type of argument we will take some of the objections of Professor Malcolm in his review ofWittgenstein'sPhilosoph­ical Investigations. l Some of these arguments come from Wittgenstein directly, while others represent Malcolm's interpretation of the former's views. Since Wittgenstein and Malcolm seem to agree on this question, 1 will not attempt to distinguish their arguments.

First, Malcolm considers the case where a user of a private language resolves to use the same words for the same kinds of sensation. How­ever, we cannot suppose that the word 'same' means what it does in the public language; the word could be applied in various ways to sets of things which would not usually be called the same. "The point to be made here is that when one has given oneself the private rule 'I will call this same thing "pain" whenever it occurs,' one is then free to do any­thing or nothing. That 'rule' does not point in any direction. On the private-language hypothesis, no one can teach me what the correct use of 'same' is. 1 shall be the sole arbiter of whether this is the same as that. What 1 choose to call the 'same' will be the same. No restriction what­ever will be imposed upon my application of the word. But a sound that 1 can use as I please is not a word."2

Part of Malcolm's argument seems to be that there is no way of discovering whether the rule has been obeyed. We can deal with that later. His main point here is that since the usual meaning of 'same' cannot be taught to someone who knows only a private language, his resolution to call the same sensations by the same name is an empty one. The supposition here seems to be that one must be able to say the rule to oneself in the language before one can follow the rule. This would, or course, make it impossible to learn any language; if one had to be

1 N. Malcolm, "Wittgenstein's Philosophical Investigations," Phil. Review, Oct., 1954, pp. 530-559.

• Ibid., p. 536.

Page 48: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

38 FOUNDATIONS OF SOLIPSISM

able to state the rules for the use of 'Buch' correctly in German before one could learn the word and use it correctly, one could never learn the word. Complex notions such as "same," which are used in stating such rules, can never be the first parts of a language to be learned. The user of the private language need not say to himself "I will always use this word to refer to the same kind of sensation" in order in fact to use the word for the same sensations. We might say oj a language in some other language that if it is used consistently the same words will be used for the same sorts of entities; this admits of some ambiguity (how similar things must be to have the same name, etc.), but in practice we have no difficulty in understanding such statements. In general it is not necessary that people using the language follow this rule consciously or even that they should ever have thought about it. The case usually is that they just develop habits of speech which this rule describes; thus a man could in fact use the same words for the same things even if he did not know the meaning of 'same' or any similar word.

Thus when we say that if the private diarist speaks consistently he uses the same words for the same sorts of sensations, we are making a statement about his language in the public language; if there is a problem about the word 'same,' it is a problem for the user of the public language, but not yet a problem for the private diarist. It is true that the private diarist must have the ability to follow a habit once it has been initiated; he must in fact use the same words for the same sensations even though he will not be able to say in his language, at least in the beginning, that he does this. This ability need not depend on knowing another language, although we do not require that the private diarist be totally ignorant of public language; neither is this entailed by Malcolm's definition of "private language." A cave man who had no language whatever might discover by trial and error what is the best size of stone for throwing, and establish a habit of picking out a certain size of stone for this purpose. He would then presumably be able to abide by this habit, and this would involve picking out the right size of stone from among the many possibilities. This ability seems to be very similar to the ability to pick out a certain word in accordance with habit. The criteria for whether this has been done correctly are another matter, which we will consider later.

Malcolm also argues that if we learn a term like 'pain' privately as the name of a sensation, it will be contradictory to suppose that some­one else is in pain when the user of the private language is not; part of

Page 49: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

LANGUAGE AND SENSATION 39

its meaning is that it exists only when it is felt. l This conclusion seems to be drawn from the fact that in order to associate a word with a sensation, it is necessary that the sensation be present. In the public language, on the other hand, we can often explain to someone the meaning of a material object word when no object of the sort referred to is present. However, this happens only when we can define the new word in terms of other words which are already understood. In the case of a child or someone who knows very few other words, we usually have to introduce the names of things when the objects named are present. In this way a child may learn the word 'table' from hearing it talked about, and a table may always be present; he does not on this account conclude that there is a contradiction in the notion of a table's existing when it is not being perceived. In Malcolm's argument it is implicit that in an ostensive definition all the circumstances which obtain at the time it is given become involved in the meaning of the term and have to be re-enacted in order for the term to be correctly applied. It is conceivable that someone might interpret such a definition in this way, but it is not necessary that he do so. It is a psychological fact that when people learn a noun they usually do not even interpret it as a singular term, much less as being properly applicable to something which must have a particular location in space and time, or which must be in a certain relationship to the speaker, such as being present or being perceived. In the private language, the situation is somewhat different, since at least the primitive terms of the language are names of kinds of entities which are not supposed to exist when they are not present to consciousness. The user of the private language might teach himself these terms by making up sounds, or words, and associating them with certain sensations which he is having, or perhaps with corresponding images. This process would then be the counterpart of ostensive definition in the public language; however the nature of this process does not entail that part of the meaning of the term introduced is that the sensation referred to must be present to the speaker in order to exist. There is nothing about the private language which prevents the user from supposing, along with Russell, that his sensations continue to exist when they are not present to him, or from supposing that other people sometimes have the same sensations.

Of course, the private language was deliberately limited in certain ways when it was introduced. The private diarist cannot ascribe sensa­tions to other people within his language using the same words that he

1 Ihid., p. 538.

Page 50: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

40 FOUNDATIONS OF SOLIPSISM

uses to report his own sensations. This is a direct consequence of his practice of always using the basic terms of his language to refer only to his own sensations, and it cannot be objected against the private language that it does not accomplish more than it was intended to accomplish. But it does not follow from this that the private diarist believes that only he has sensations, or that no one else can be in pain. If he were to learn a richer language he could state his beliefs to the effect that other people suffer pain, etc. Further, to the extent that we can have beliefs without being able to express them linguistically, he might have had these beliefs all along. Nothing, therefore, about ostensive definition or its counterpart in the private language prevents us from associating a word with certain features of the context in which it is introduced, and dissociating it from other features. Whatever the user of the private language, or anyone else, may mean by another's being in pain, his having learned the term with reference to his own pain need not prevent him from thinking about other people's pain without falling into contradiction.

Another sort of objection used by Malcolm and others centers on the kind of private ostensive definition used in the private language. Here one would consider a sensation, which is either occurring at the moment, is remembered, or is imagined, and associate a word with its qualities; we have already seen that if this association is to be useful it must not be between the word and all aspects of the sensation, such as its occurring, being remembered, or being imagined. The association must be strong enough so that we do not apply the word to totally different kinds of sensations, yet loose enough so that the word can be applied to sensations which are closely similar to the example, but perhaps not exactly similar. In fact, we do seem to be able to direct our attention to certain aspects of a situation or sensation when receiving an ostensive definition in the public language, and there is even more reason to think that we can do this when giving ourselves an ostensive definition (associating a word with certain aspects of our consciousness at a given time). In this case there is no room for the sort of misunder­standing that can take place in the public case. It is still presupposed that we have the basic ability to focus our attention on certain aspects of consciousness while ignoring others, and this assumption is certainly borne out by our ordinary experience in dealing with public languages.

The objection is that a word in the public language is not learned by hearing it uttered in the presence of the object, because the sound could be taken to refer to anything or nothing. In other words, the

Page 51: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

LANGUAGE AND SENSATION 41

intended ostensive definition might be misunderstood or not understood at all. Malcolm thinks, to the contrary, that to be learned it must play a part in such activities as calling, fetching, distinguishing the object from other objects, distinguishing pictures, etc. l The point would then be that these cannot be duplicated in the private case. There are two advantages that such a process as telling a child to bring an x and pointing to it might have over just pointing to the object and uttering 'x.' First, one can tell by his behavior whether he has understood, and praise or blame him accordingly. Secondly, he may bring the wrong thing, in which case one can cause him to dissociate the word 'x' from things which are not x by expressing displeasure.

There is no difference in kind, however, between these activities and ostensive definition, but only a difference in degree. The possibility of misunderstanding or not understanding at all is still present. When the child is asked to bring an apple he might not understand that he is being asked to bring something at all, and perhaps thinking that his father is angry, bring him a nearby apple to appease him, without having any idea that 'apple' is the name of that kind of object. The father would then assume that the child understood and let the matter drop for the time being. Of course, the more this sort of thing is repeat­ed the less chance there is of misunderstanding, but, as we saw before, this is also true of repeated ostensive definitions where the non-essential parts of the environment are varied. In any case there is no possibility of this sort of misunderstanding in the private case anyway, so it is no objection that in a private language we cannot ask ourselves to fetch things in the same way.

On the second point, it can be granted that it is as important to dis­sociate a word from the things that it does not refer to, as it is to associate it with the things that it does refer to. Pedagogically there may be some advantage in using the processes Malcolm mentions, but again it is a difference of degree rather than kind. This sort of dissociation occurs whenever an ostensive definition has any effect, even when it is mis­understood. And, as we have seen, it is reasonable to suppose that an association can be set up between the word and that part of the situa­tion on which we focus our attention, and not with the rest of the situation. Further, as we have also seen, if such processes as calling for things and telling someone to fetch things were essential to the learning of language, there would always have to be someone skilled in language for anyone else to learn language; from this it would follow

1 Ibid., p. 553.

Page 52: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

42 FOUNDATIONS OF SOLIPSISM

that language had no beginning. While we do not wish to speculate about the origins oflanguage, we do not wish to assume that language either had no beginning on our planet or that it was brought here by a teacher from another planet, and so on, ad infinitum.

Throughout his arguments Malcolm seems to have a more restricted idea of a private language than ours. We have seen, however, that there are excellent grounds for calling the personal-sensation language a private language. Any other kind of private language would be im­possible, not for the reasons he has given, but because there would be no subject matter for the private language to talk about.

Assuming that it would be possible for the originator of a private language to attach names to sensations, we come now to question whether there is any criterion for the correct use of these words. Wittgenstein here argues in the following way. Suppose that someone is keeping a private diary about the recurrence of a certain sensation. For this purpose he associates the sign 'S'with the sensation,l perhaps by concentrating his attention on the sensation and repeating the sign. If this process establishes the meaning ofthe sign it must bring it about that the connection is correctly remembered in the future. There is no criterion except memory, however, for whether'S' has been used correctly. Thus, whatever seems right will be right, and this means that we cannot talk about "correct" here. Thus when the private dia­rist says that a sensation is an S, his sole grounds for saying this will be his recognition of the sensation, and there is seemingly no independent check for this.

As against Wittgenstein, A. J. Ayer tries to draw a parallel between the private language and the public language on this point. 2 First, he argues that it is sometimes possible to check one memory against an­other, so that the private diarist is not always totally without any criterion for detecting false recognitions of a sensation. More important, he points out that even in verifying a public statement about a material object, one's ultimate appeal must always be to one's own sensations. For instance, if we want to verify the fact that there is a table in the next room, our last appeal must always be to the fact that we see it there when we look, etc. It is always conceivable in such a case that we fail to describe and identify our own sensations correctly. When I look into the next room I might have the sensations I usu<;!-lly have when

1 Wittgenstein, op. cit., p. 92. • A. J. Ayer, Symposium on Private Languages, Aristotelian Society Supptement, No. 28, pp.

63-76.

Page 53: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

LANGUAGE AND SENSATIONS 43

observing a chair, but mistakenly classify them as table-sensations. Of course, other people then point out my error, but in order for me to understand and take account of what they say, I must recognize and classify correctly the auditory sensations of their words. Hence Ayer's conclusion is that even where the recognition of a public material ob­ject is involved we must always depend on being able to recognize sensations correctly, and there is no independent check for this, beyond memory, any more than in the case of the private language.

Despite this, there seems to me to remain an important difference between the criteria for the correct use of the two kinds of language. In the private language, even in the few cases where we can check one memory against another, our attempt to determine whether we are using a word consistently for the same kind of sensation must come to an end very quickly. Having called up all the memories one can, and having seen whether or not they are consistent, there seems to be nothing else to do, and our investigation comes to a dead end. The situation is different in the case of the public object. Here we can keep on investigating the question whether the object was correctly re­cognized as long as we like without ever going over the same ground twice. While we must always recognize sensations in the end, we can go about verifying the truth of a statement about a material object in an indefinite number of ways, and keep on adding to the probability of the statement. The sensations which have to be recognized in each in­vestigation will be different ones, and we may never reach a point where investigation is senseless. As the matter now stands we would be justified in saying that the criteria for the truth of statements in the public language are very much stronger than for statements in the private language, and we might make this the basis for saying that a private language is not a language in the same sense that a public language is.

We will now see if it is possible to conceive of a situation where a private diarist could have evidence independent of memory as to whether or not he is using a word consistently for the same kinds of sensation. Let us suppose that the subject is marooned alone on an island, and that he has developed a private language with a primitive set of words, which are letters of the alphabet. While he has been on the island he has kept a private diary of the sensations he has had during certain periods of the day. This diary does not mention images that he has, but only sensations. In looking over his diary he finds the word S entered fifteen years before. At the moment he would put down'S'

Page 54: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

44 FOUNDATIONS OF SOLIPSISM

only ifhe has a sensation characteristic of smelling an elephant. In fact he remembers giving that name to that sort of sensation (perhaps with the help of a memory image) since he had always lived among ele­phants before being marooned, and this was a familiar sort of sensation. But now on reading his diary he thinks that he very likely made some kind of mistake in putting down'S' in the past, or at least that its meaning has changed. He is quite certain that there are no elephants on his island, and he knows of nothing else that would produce the same peculiar odor. He could, of course, have been suffering from a hallucination fifteen years ago, but this is very unlikely, since he has never been subject to that sort of phenomenon that he can recall. Thus, he has good inductive evidence for thinking that the mistake he has made was to violate a rule of his language: that he did not at that time use S to record the sensation which S stands for today.

There can also be further evidence in either direction, just as in the case of statements in the public language. Suppose that he goes for a walk that afternoon and happens upon a cave that he has not gone into for fifteen years. He decides to go in and, immediately upon entering, becomes aware of a strong smell S (the elephant smell) ; he then traces it to a certain moss which does not grow elsewhere on the island. Now everything is explained; he concludes that when he went into that cave fifteen years before he had experienced the same smell and noted the fact in his diary. In the meantime he had forgotten this rather un­important incident and naturally supposed that there was nothing on the island capable of producing that smell. Hence he now concludes that he has used'S' correctly all along; the interplay of evidence is the same as it is in the public language. Hence there are at least some cases where there are independent criteria for discovering whether the rules of a private language have been obeyed.

Wittgenstein's example of a manometer1 suggests that he would here argue that the private diarist is no longer talking about his sensations, but is now using a public language to talk about the existence of a certain smell in the cave, and that the sensation is no longer important. We have already argued that the occurrence of sensations is crucial for the recognition of objects, but, apart from this point, there seems to be no reason to say that the private diarist all along must have meant to make a statement about a public object when he supposed that he was making a statement about a sensation. When he made the statement oS' he may have had no idea that there would be any way of verifying

1 Wittgenstein, op. cit., p. 95.

Page 55: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

LANGUAGE AND SENSATION 45

it in the future, as we usually suppose that smells in the air are transi­tory and not permanent. When he reconsidered the truth of the state­ment fifteen years later the later evidence indicated that he still attach­ed the same meaning to it, but he would never have guessed the man­ner in which it was to be at least partially verified. Hence there is no reason to think that by'S' he ever meant to talk about the existence of a smell in the sense of a material object as opposed to a sensation.

In fact, there is a general method which any private diarist could use for checking the consistency of his terminology. If he did not already know a public language he could learn one in the usual way and then compile a notebook in which, opposite each sentence in the private language, there would be its correlate in the public language. Thus opposite some such statement in the private language as "Chair-like sensations occurred when I had sensations characteristic of looking in that direction" might be "There is a chair" in the public language. Whenever the private diarist wanted to check a statement in the private language he could then utter the corresponding statement in the public language and verify the latter in the usual way by asking other people, etc. As we have seen, the two statements may have differ­ent meanings even if the statement in the private language were much more complex, but there will remain a strong inductive correlation between the truth of the two statements. In the great majority of cases one has these sorts of sensations if and only if there is a chair nearby; the private diarist would be able to infer with a high degree of proba­bility that the statement in the private language is true if the statement in the public language has been confirmed. Since there is no possibility of his being mistaken about the sort of sensation he is having, as opposed to the proper description ofit, he infers that he is using the terms of his private language in the same way that he did when he compiled his notebook, and that he is following the rules of the private language. He has, in this sense, checked his consistency.

Our conclusion, then, is that a private language, in the sense of a personal-sensation language, is possible if any kind of language is possible. Since our solipsist has never doubted the existence of lan­guages and will always presuppose the possibility of some kind of language, there can be no theoretical objection to a private language and to his using such a language as his ultimate language of analysis.

Page 56: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 3

IMAGINING INSTANCES

This chapter presents a general theory of meaning as a basis for the analyses to come. The solipsist has a number of choices here, but I think that a particular kind of image theory will provide the best foundation for his philosophy.

The reason for presenting an image theory is not because Berkeley and Hume held such theories; in fact, the present theory will be quite different from their theories and the reader is asked not to automatical­ly attribute the associated difficulties to the modern solipsist's position. This kind of theory can be made as plausible as any other which at­tempts as much, and it is also highly conducive to the economy which the solipsist seeks so intently.

One of the most important concepts of this theory is that of an in­stance. Starting with the concept as it already exists in our ordinary thinking, we will extend it in various ways for the sake of convenience. It is an important feature of language that we talk about instances in connection with certain words - in particular, predicate words like "red" and predicate phrases like "red barn." We will subsequently argue that we can talk about instances of certain larger linguistic units, but, for the time being, we will talk only about the instances of predi­cate expressions such as these. There are also different ways in which we can talk about instances. Concentrating on one side of the relation­ship, we can talk about instances of words, or, alternatively, instances of concepts or properties. Someone might insist that an object is not an instance of a word or larger linguistic unit, but is instead an instance of a concept or property for which the words stand. I find this an in­convenient way to proceed, because it already presupposes a theory: concepts are non-linguistic entities for which linguistic entities stand. Even if someone insists in talking in this way and so restricts the notion of "instance of," there will still be an indirect but definite relationship between the words and the entity which is an instance of the correspond­ing concept. This being the case, we could use some other term to denote that relationship; however, it will be more convenient to use

Page 57: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

IMAGINING INSTANCES 47

"instance of" to denote the relationship between the piece of language and the instance. We can then leave open questions as to the existence and nature of any intervening concepts.

Another feature of the instances of predicate expressions is that they constitute relatively small parts of the physical world, as opposed to comprising whole actual or possible worlds. For example, if we want to say that there are three instances of "green book" scattered through­out the room, we will be talking only about books and not about whole worlds. This, of course, does not prevent our talking about instances drawn from possible worlds instead of from the actual one, but here again, the instances would be found in relatively small parts of these possible worlds.

There are a number of other expressions which are more or less synonymous with "being an instance of." If x is an instance of y, we can usually also say that x is an example of y, or that x is a case of y. We do not often talk in any ofthese ways and such sentences sometimes seem rather awkward. It seems much more awkward to talk about the class of instances of a particular predicate, or to say such things as "There are sixteen instances of the property of being a book in this room." The awkwardness of this last sentence probably sterns from the fact that there are such simple ways of saying the same thing, but it is surely legitimate to talk in this way if we can speak of an instance here, and an instance there, and so on.

It should also be noticed that a predicate can be made indefinitely complex. Thus, in contrast to the predicate "red," we can form the predicate "is a red book the binding of which will decompose if left in salt water for 36 hours under such-and-such conditions."

On the other side of the relationship of instance hood, we can talk about physical objects as instances. Furthermore, in certain cases, we can talk about sensations, or even images, as instances. As it turns out, there is no reason to restrict instances to any class of entities. The most prominent instances of the word "red" are, of course, red things, but sensations can also be red in some sense, or, at the very least, can be of red things, and there is no harm in also speaking of sensations as in­stances of the term. Further, it makes sense to speak of imagining in­stances, and the images so produced can also be taken as instances. We have already seen that in the learning process the sensation is more intimately related to the word than is the object. Since the relation of "instance of" reflects the learning of words, we will, as it turns out, be more interested in sensation-instances and the closely related image-

Page 58: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

48 FOUNDATIONS OF SOLIPSISM

instances than in instances which are material objects. The two proces­ses in which we are most interested at the moment, then, are those of recognizing a sensation as an instance of an expression, and imagining an instance of an expression.

There is a close parallel between the two processes, since they both involve recognizing and accepting a mental state as an instance. In the one case we have to find an appropriate object and sense it, whereas, in the other case we imagine such an object. We will therefore talk primarily about imagining instances, and much of what we say will be directly applicable to the recognition of sensation-instances. It is the recognitional aspect of the process in which we are interested; we can consider it apart from the sub-process of finding a suitable object in the world.

The first question is whether there actually is such a process as ima­gining an instance of something, if by "imagining" the actual having of an image is meant. There are some cases where I might say some­thing like "I can imagine what would happen if Jones were to get on a horse" without intending to say anything about images at all. In the context this might just be a way of saying that Jones would be very unsuccessful if he were to try to ride. Similarly, the phrase "I imagine that. 0 0" is often just another way of saying "I believe that .. 0," and need not have anything to do with images. However, these cases in­volving the phrases "I imagine that.. 0" or "I can imagine what .. 0" are very special ones. At the other extreme there are cases which clearly do involve images; if someone says "I am trying to imagine a golden retriever in as great detail as possible," it is quite likely that he is ac­tually having images. Cases which involve the phrase "I can imagine" as in "I can imagine Jones in purple shorts" or "I can imagine an air raid on New York" are somewhat less clear. The speaker is only claim­ing a disposition, and whether it is a disposition to actually produce images generally depends on the context. It cannot be doubted that we sometimes have images, and it seems equally clear that in some cases we can imagine instances of various terms such as "golden retriever," "houseboat," etc. Even though the term "imagine" has some rather extended usages, we will use it in the narrower sense of "producing images" in order to avoid confusion.

Having assumed that there is such a process as imagining instances, we must ask whether it is reducible to some other process, or is ir­reducible. As we have seen, imagining an instance of a fiying squirrel might just amount to saying what a flying squirrel would look like,

Page 59: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

IMAGINING INSTANCES 49

but if we take "imagine" in the strict sense outlined above, there is no obvious way of reducing the process of imagining an instance to some other processes. In fact, we have argued that the process of having images is itself a mental process irreducible to anything else. The process of imagining an instance obviously goes beyond imagining pure and simple, since it involves imagining in a particular kind of context. We will try to specify in as great detail as possible exactly what that context is, but even if we are successful in so isolating the process, we will not claim to have reduced it to some other set of processes. We do hope to give a unique description of this context, and hence of the process of imagining instances, but in doing so we will only be trying to indicate the particular context and process which we have in mind, and to distinguish it from other similar processes. Our conclusion will then be that this process has to be accepted as a basic psychological mechanism or ability. Its status would then be similar to that of such processes as remembering (in its most primitive form) or of associating similar events. We undoubtedly have these abilities, but we can say very little about them except to isolate them and distinguish them from one another. The only possible explanation of the existence of processes and abilities such as these would be on the neurophysiological level, and we do not plan to attempt anything of the sort here. We are saying, then, that once we have isolated the process of imagining instances, our analysis will stop. Even ifI am incorrect in supposing that this is where the analysis has to stop, the theory would not necessarily be invalidated simply by someone else's carrying the analysis further.

There are different kinds of cases where we can be said to imagine instances. One of the most interesting distinctions centers on the ability to direct our imaginations. If, for example, we are trying to imagine an instance of "flying squirrel," we can, prior to imagining, give ourselves certain directions as to what to imagine. Whatever we imagine must have wings of some sort and it must be like a squirrel in certain respects. We may not be conscious of these rules, but, if they are not followed, the ordinary person will not be able to imagine anything he would accept as an instance.

There are other cases where no such rules seem to be available. If someone had difficulty in imagining a patch of grass, it is unlikely that he could give himself or be given any directions that would help. Definitions usually function as directions in this respect, but if a person did not know English well enough to find an instance of such a simple expression, he would probably find the terms used to define "grass"

Page 60: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

50 FOUNDATIONS OF SOLIPSISM

even more obscure. In the case of the flying squirrel we can imagine instances of the definiens more easily than instances of the definiendum, but the reverse is likely to be the case with the patch of grass. It is not being asserted, of course, that the expression "patch of grass" is a basic expression which has to be learned before other words can be learned, but just that for any person there are some words and expressions which are more familiar to him than any expressions in terms of which they might be defined. Unless these entities are hard to imagine for some reason (like chiliagons), these will be cases where directions will not help; a person either will be able to imagine instances right off, or he will not be able to imagine instances at all.

The situation here is similar to that oflearning a language by means of verbal definition. Verbal definitions ultimately will be circular. There must come a point where it is useless to give such a definition, and where we must resort to some other method. It is likely that directions for imagining will be useless in just those cases where we have learned the expression by ostension, although we need not assume this. At any rate the process of imagining an instance most interests us in these cases where we have no rules to follow, and we will call such instances "basic" instances.

Even though there are no rules to follow in imagining a basic in­stance, the process still differs from random imagining. Sometimes we get into an imaginative reverie and all sorts of images run through our heads. This random imagining is like the imagining of a basic instance in that there are no rules which we are committed to follow here either; the only principle of succession is a loose association of images. How­ever, the difference between the two processes is that in random ima­gination there is no question of success and failure, because there is nothing that we are trying to accomplish; we just let our imaginations go and take what comes. However, in the case of imagining a basic in­stance, the fact that we are not proceeding on the basis of rules which we can state does not mean that we will accept any image as an in­stance. In some cases we might have to try several different images before finding one that satisfies us as an instance: thus we may not be able to define "aquamarine blue" in any useful way, but we may still be very selective as to what we will accept as an instance of it. One can say that our habits of imagining basic instances are orderly and that there will be strong resemblances between different images accepted as instances. Hence this process is quite different from random imagina­tlOn where nothing matters and where the images which please us may

Page 61: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

IMAGINING INSTANCES 51

have nothing in common with one another. When imagining basic instances corresponding to some expression, the images which satisfy us will be closely inter-related.

Another important question which we must answer before proceed­ing further concerns the kinds of expression for which we can imagine instances. So far we have talked about imagining instances of single words and more complex predicate expressions, but the basic units of language are usually conceded to be sentences, and they cannot have instances. Consequently, we will now contrast predicate expressions with sentences expressing statements.

The most obvious difference is that sentences do not have instances, but are true or false (or at least the statements they express are true and false). Nevertheless, there is one common feature which predicate ex­pressions and sentences have. A predicate expression can either be left in the warehouse of unused language or someone can actually apply it to some object or state of affairs. Similarly, a sentence can just be left as something we might someday want to use, or it can actually be asserted. That is, a sentence is like a predicate expression in that it does not somehow apply or assert itself, but has to wait for some person to use it. This is, of course, a platitude which is true of all parts of lan­guage, but it nevertheless has important consequences which are some­times overlooked.

The point is that in the case of a sentence we must make a distinction between the act of asserting that a given state of affairs is the case and that state of affairs itself. In other words, there is a legitimate distinc­tion between what is asserted and the act of asserting it. We can start with a given state of affairs brought into consideration by a sentence or other locution, and do a number of things with that given state. We can either assert it to be the case, question whether it is the case, suggest that it might be the case without asserting it, urge someone that it ought to be the case, wish that it were the case, and so on. To take John Searle's example, there is a common element (or common speech act) in each of the following five sentences despite the differences:1

(1) Will John leave the room? (2) John will leave the room. (3) John, leave the room! (4) Would that John left the room. (5) If John will leave the room, I will leave also. The issue here is a rather confusing one, because the ordinary sen-

1 John Searle, "What is a Speech Act?" in Philosophy in America, Max Black, ed., p. 225.

Page 62: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

52 FOUNDATIONS OF SOLIPSISM

tence which merely brings into consideration the state of affairs that we are interested in also asserts it and makes it look as if there were no difference between just talking about a state of affairs and asserting it to be the case. But we can find locutions which we can add to a sen­tence, or special contexts into which we can put a sentence, thereby removing its assertive force and allowing us to remain neutral as be­tween asserting, denying, wishing for the possibility to be realized, and so on. Thus, we prefix the sentence, "The cat is eating the mouse," with "It's being the case that ... " to get the phrase, "it's being the case that the cat is eating the mouse." Or we can speak alternatively of "the possibility that the cat is eating the mouse." These constructions give us names for states of affairs, without our asserting anything about those states.

Better yet, we can invent a special context where it will be under­stood that we are not using sentences to assert statements, but are merely considering statements that might be asserted. We now need some name for this component of a statement that can be asserted, considered, wished for, etc. Let us therefore call this component a "preliminary statement," which can be expressed in the ways men­tioned above. These preliminary statements do not state anything themselves but are preliminary to stating, doubting, wishing, and so on. For short we will call them simply "preliminaries."

Actually, discussion of these issues is generally limited to the sort of case where we have two persons talking with one another. For our purposes, however, the case where a person is simply thinking to him­self in words is much more important. Most of the time, when we are thinking in words, we probably use only fragments of sentences, and the context therefore becomes even more important than it is in inter­personal discourse. We can leave a great deal out because we construct the context ourselves and are intimately familiar with it; many of the possibilities for misunderstanding which are present in interpersonal discourse are thus eliminated, and we need not be nearly so careful about what we say. Consequently, it is highly unlikely that in one's own thinking one would find it necessary to express statements in one way and preliminaries in another. If I am simply thinking of the cat's being on the mat without knowing or caring whether it is really there, it is unlikely that I would say to myself anything like "it's being the case that the cat is on the mat" or "the possibility that the cat is on the mat." More likely all that would go through my head would be the words "cat ... mat," perhaps followed by an "is it?", and perhaps

Page 63: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

IMAGINING INSTANCES 53

followed eventually by my going to see whether it is on the mat. This would be distinguished only by the general context of my thinking, from an assertion that the cat is on the mat, which could be expressed in much the same way. Hence, this difference in context is decisive and crucial.

Conversely, even predicate expressions, either in interpersonal dis­course or in one's own thinking, can be used to make assertions. When driving down a road I might merely say "big truck" to indicate to my companion that it is a large truck that is coming towards us, or I may think the same thing to myself and again be making the same assertion.

The difference between considering a state of affairs and asserting it to be the case is not a grammatical difference, and it is not the differ­ence between sentences and predicate expressions, or between sen­tences on one side and noun phrases and descriptions on the other. In thinking, it is a difference in attitude on the part of the thinker that makes the difference, and we will have more to say about this later.

In interpersonal discourse there are also a large number of behavioral concomitants which go with asserting as opposed to considering, and we will also have more to say about this in a later chapter Although an explicit preliminary, stated in one of the ways we have indicated, would occur only rarely in one's own thinking, we will find it convenient to count also as preliminaries those sentences that occur in thought­contexts which neutralize their assertive power.

Now that we have removed the assertive force of sentences by switch­ing to preliminaries, we must ask how strongly these preliminaries resemble the predicate expressions mentioned earlier. In the case of a preliminary like, "it's being the case that the cat ate the mouse," which involves proper names or definite descriptions, we can derive predic­ates such as "is the cat which ate the mouse," or "is the mouse which was eaten by the cat." Although there is a close relationship between the preliminary and these two predicates, they are not in any sense equivalent. The difference is that when we choose to talk about a possibility in explicit subject-predicate form, we choose some part of the situation as the subject and give it a privileged position; we say that we are talking about that subject even though we may convey informa­tion about other objects or other parts of the total situation. In talking about the mere possibility as such (by using a preliminary), we need not choose any particular part of the situation as the subject around which to focus what we want to say. If one likes, one can say that the difference is between talking about the possibility that something is

Page 64: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

54 FOUNDATIONS OF SOLIPSISM

such-and-such and simply talking about the possibility that such-and­such. That is, it is the difference between talking about the possibility of something's being in a certain condition and simply talking about the possibility without identifying the important features of the situation as modifying some part of that situation.

It seems clear that the similarities between preliminaries and predi­cate expressions are more important than the differences. The main similarity consists in the fact that in both cases we have a description of a state of affairs, however it may be couched, and in both cases we leave open the question of if, where, and how that description may be applicable. The main difference consists in the fact that a description represented by a predicate expression may be realized over and over again at different times and at different places. This possibility is left open no matter how complex and exclusive the predicate may be, as long as it is stated in general terms and is not a definite description. On the other hand, we are tempted to say that a preliminary either applies to the world and is the case, or that it does not apply and is not the case. Thus, we can consider the sentence, "No roads are perfectly straight," in such a way as to make it a preliminary, and it then seems that the question we ought to ask is whether it would be true if asserted, and not where and when it may apply. It would seem odd, for exam­ple, to say of that sentence, even when considered as a preliminary, that it applies to North America and Europe, but not to certain islands in the Pacific where the only roads are disused landing strips. If some­one were to say that, we would probably get his meaning, but we would think that he had made some sort of category mistake. If the sentence is like "The cat is on the mat" in that it contains definite descriptions, we might be more tempted to say that it applies, or is true, under certain sets of circumstances and not under others. But here we could argue plausibly that the sentence has a different meaning when different cats are being referred to; there would then be no single state­ment that is sometimes true and sometimes false. Nevertheless, we clearly can say that other possible worlds exemplify a given preliminary while the actual world does not, or vice versa. It then looks as if the difference is that a predicate expression can have instances drawn from parts of the world while the "instances" of a preliminary are whole worlds, either actual or possible. In the latter case we would not ordi­narily speak of these worlds as being instances of the preliminary, but rather we would speak of worlds which exemplify the preliminary, or of worlds in which the corresponding statement would be true. How-

Page 65: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

IMAGINING INSTANCES 55

ever, the concept of a preliminary is not part of our ordinary conceptual equipment, so there is really no ordinary way of speaking about its "instances." Further, it appears that the difference between instances of predicate expressions and the "instances" of preliminaries is mainly one of size, so there is nothing to prevent our extension of the concept of an instance to cover the latter case and our concomitant removal of the quotes from "instance" when so used.

Some additional explanation is required when we so extend the concept. As mentioned above, physical object instances of prelimina­ries will be much larger than physical object instances of predicate ex­pressions. Similarly, image and sensation instances will tend to be larger in the former case in the sense that they will be of relatively lar­ger chunks of the physical world. Nevertheless, in the next chapter we will introduce the concept of a partial instance. Speaking very roughly, a partial instance which is an image will turn out to be an image of a part of a state of affairs which would be counted as a complete instance. Hence, the partial image instances of preliminaries need not "take in so much ground," and will be the sorts of images that we ordinarily have.

Even when we are dealing with physical object instances it turns out that instances of preliminaries will not be quite so all-encompassing as it first appeared. We do speak quite naturally ofa sentence's being true given a particular universe of discourse, and of its being false relative to some other universe of discourse. Suppose, for instance, that aJudo instructor is explaining to a student that if he gets his opponent in a certain position, the latter will quickly fall heavily to the mat. The generalization that the instructor is asserting would not be true of Judo as practised by future spacemen whiling away their time during weightlessness. However, the instructor would not be deterred by such counter-examples and would insist that he had never been talking about extra-terrestrial Judo.

Consequently, it seems that the instances of preliminaries are not often whole worlds, either actual or possible, but rather, parts of actual and possible worlds determined by the universe of discourse which is operating in the context. There will remain a great deal of difference between the instances of predicates like "red barn" and instances of such preliminaries as "No roads are perfectly straight," but the differ­ence is no longer that between a part of the world and the world as a whole. Thus, there is even more reason to extend the concept of an instance so that we can talk about instances of these preliminaries.

An interesting feature of preliminaries is that for every sentence ex-

Page 66: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

56 FOUNDATIONS OF SOLIPSISM

pressing a statement there will be a corresponding preliminary which we get when we remove the assertive force from the statement. Thus, while we cannot talk about instances of sentences asserting statements, we have extended the concept of an instance so that we can talk about instances of the preliminaries which correspond to those statements. It also follows that what we said before about imagining instances of predicate expressions will, in its general outlines, hold true for the imagining of instances of preliminaries. Of course, the instances here will be much larger and it will take much more imagining to produce them: nevertheless, it is theoretically possible to imagine on a scale larger than the usual one. When a preliminary is determined by a relatively large universe of discourse, however, it is extremely unlikely that it would have basic instances; there would almost certainly be rules which would guide us in the imagining of these instances. For example, any statement which is either universally or existentially quantified would provide us with rules which we would have to follow in imagining instances of the associated preliminary. Hence, since the difference between predicate expressions and preliminaries, with respect to imagined instances, is relatively unimportant, we will, for convenience, sometimes speak of the imagining of instances of prelim in­aries when we actually have in mind the imagining of instances of either predicates or preliminaries.

In fact, there are many cases where a predicate, such as "red," abbreviates a preliminary such as "the possibility that this is red." If we were to utter the former to someone else, we would either be saying that some nearby object is red, thus using "red" to abbreviate an assertion, or we would be urging someone to consider the property, probably in order to say something further about it. Thus someone might say, "Liquor? Well I think that it is the source of all evil." But since we all know that redness is a property that things have and most of us do not suppose it to exist independently, we are really urging the listener to consider the possibility of something's being red, only we do not say what. In many cases the context would determine this for us: suppose, for instance, that the context is a discussion about the proper color for the new living room carpet. Hence, the preliminary (however it might be expressed) might serve the same purpose, except that it would be much more specific in that it would have us thinking about "its being the case that this is red" instead of just thinking of red. If someone instructs us to think in the former way as opposed to the latter, he is giving us more detailed instructions concerning what he wants us

Page 67: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

IMAGINING INSTANCES 57

to think about. Simple predicate expressions call to mind possibilities in much the way that preliminaries do, and in the sort of context we just mentioned it is impossible to say whether a word like "red" is just operating as a predicate or is abbreviating a preliminary. Indeed, the utterance of such predicate expressions is very likely the usual way of introducing preliminaries into our thinking and into our conversation. It is only in very special contexts that locutions such as this serve as assertions or ostensive definitions. We can generally treat them as ex­pressing preliminary statements in a very vague way.

We have seen that there is a one-to-one correspondence between preliminaries and statements. Further, every cognitively meaningful proposition is true of some possible state of affairs, and the consequence is that every meaningful preliminary has at least one possible instance. This is not true of all single words. The word "if" is clearly meaningful in some sense, but it has no instances, and thus differs from the word "red." The difference is that the single utterance "red" can easily be construed as being an abbreviation for either a statement or a preli­minary. In the latter case we let it take just those instances which the preliminary that it abbreviates has. The word "if" gives no hint as to what a statement or preliminary containing it might be about; hence it will not serve as an abbreviation and we are at a loss to say what instances it might have.

We are now left with a number of different types of linguistic ex­pressions: those that can have instances, those which can be true of some possible states of affairs, those which are meaningful but cannot either have instances or be true, and nonsensical expressions, the clear­est cases of which are simply such words as "sysx." The first three types of expression are meaningful, but the question remains whether they are all meaningful in the same way.

The relationship between the first two is fairly simple. Both kinds of expression stand on their own feet, and it is very easy to convert one into the other. Thus a statement and its corresponding preliminary statement mean different things, but they have a great deal in common, and we can say that they have the same kind of meaning. On the other hand a word like "if," which can neither have instances nor be true, has its meaning (or use) because it can be an essential part of prelim ina­ries or statements. To say what such a word means is to say how its presence affects the meaning of larger units of discourse. We can con­clude, then, that the basic units of meaning are predicates, statements and preliminary statements, all of which have independent meaning.

Page 68: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

58 FOUNDATIONS OF SOLIPSISM

The single words which are their constituents have meaning only because they can be used in statements or preliminaries or because they abbreviate one or the other. We can then say more generally that any expression which does not express a statement or preliminary can at most have derivative meaning. While such expressions are often single words like "if," they need not be; the question really is not ofgramma­tical form, but depends on whether the expression has instances or can be true.

The main object is to say in general what determines the meaning of an expression that has independent meaning. As a preliminary to trying to answer this question, we can say that the meaning of such an expression depends on the kind of instances it has or might have or, if it is a statement rather than a preliminary, on the kinds of actual and possible situations it would truly describe. This does not, of course, imply a denotation theory of meaning, since we have said that meaning depends not on actual instances, but on actual and possible instances; further, we have not identified the meaning of an expression with any entity whatever. In fact, saying this does not really get us very far, because we have not analyzed the notion of an instance. However, we are approaching this problem by describing the processes of imagining instances and recognizing instances. Once these psychological processes have been isolated, we may then be able to analyze the relationship of "instance of" in terms of them.

So far we have merely distinguished between imagining an instance and random imagining. There are other processes, however, which are also distinct from random imagining and which we have not yet distin­guished from imagining instances. One such process involves the rela­tionship of similarity. We can set out to imagine something which is similar to a given object, sensation, or image. As in the case of imagin­ing an instance, this process is surely different from random imagining since we would accept some images under these circumstances and reject others. Again, this attempt can be attended by success or failure, as opposed to random imagining. It does happen, however, that there is a general difference between imagining an instance and imagining a resembling object. Since the relationship of similarity is symmetrical the latter process will be reversible. If I have a certain sensation, x, and set out to imagine a similar one, I may accept an image, y, as being similar; then in start with y, try to produce a similar image, and have an image exactly like x, I will accept it as being similar. This is not true, of course, of the process of imagining an instance. If x is an instance of y,

Page 69: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

IMAGINING INSTANCES 59

then y is not an instance of x; in general, the kinds of entities which can be instances are not the kinds of entities which can have instances.

Nevertheless, we have only distinguished the process of imagining an instance from one other imaginative process, even though it is an important one. There are other non-random processes which are also irreversible. These are also "looking for a ... " or "trying to imagine a ... " processes where the relationship between the given object and the end product does not happen to be symmetrical. An example of such a process might be that of trying to imagine something darker in color than a given sensation or image. Here, there will again be the question of success and failure, and the process will also be irreversible. Hence, nothing we have said so far would distinguish this process from imagining an instance. There certainly is a difference, however, and we will try to say what it is. This difference need not consist in some­thing that happens while the process is going on, but the different processes will at least involve different dispositions.

The relationship of "instance of" is said by someone to obtain be­tween an expression and a state of affairs ultimately because of the existence of a particular psychological disposition in him involving the two terms of the relationship. This disposition is the result of his ex­periences when learning the language and may have originally been set up by ostension or by some other means. In fact, finding an in­stance is the inverse process corresponding to ostension. In ostension we start by attending to an object or set of objects, and suitable word tokens are then produced for us. In finding an instance we start with the predicate or preliminary and are able to produce the objects our­selves because of our previous training; in imagining an instance we produce an image instead of a real object. Of course, associations of some sort also are involved when we try to imagine something darker than a given sensation, and these reflect a different aspect of our train­ing. There are certain logical differences in the two cases. Before one could train a child to always come back with an object darker than a sample, one would have to train him to recognize the sample itself. Similarly, one could not ostensively define a relational term like "darker than" before one had taught the meanings of terms that could stand in this relation. This in turn implies that one must already understand the notion of an instance (understand some phrase which fulfills the function of "This is a ... ") before one can understand the concept of being darker than a sample. If, in teaching someone ostensively, we showed him objects darker than the objects actually denoted, while

Page 70: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

60 FOUNDATIONS OF SOLIPSISM

saying "This is blue," "This is green," etc. we would not succeed in teaching him the meaning of "darker than." We would instead be teaching him to treat these objects as instances of the terms or preli­minaries concerned, but his set of instances would be systematically darker than anyone else's. If, instead of saying "This is blue," etc., we told him "This is darker than blue," we would still not be teaching him the meaning of "darker than" unless he already knew what it is to be an instance. He might quite correctly gather that the sample is correct­ly described by the sentence, "This is darker than blue," but he would take "darker than blue" to be the name of a color in the same simple way that "red" is the name of a color. He would not be in a position to grasp the fact that "darker than" represents a relationship between colors and that the same relationship holds between colors other than blue.

Let us now consider a case where someone is confronted with a particular word token or tokens which either constitute or abbreviate a preliminary, and let us further suppose that this preliminary (or predicate expression) is basic in his language. We can then compare two processes which he might go through. On the one hand he might try to imagine a basic instance, and, on the other, he could try to ima­gine a word token which is written more compactly and is shorter and fatter, and which might not even belong to the same type. In both cases, the process is irreversible and non-random, but there is a difference in the kind of psychological disposition involved in the two cases. In the case of imagining the instance, the disposition is directly from the linguistic expression to the right kind of image; the only stimulus that is needed is the awareness of the expression, provided that one is in a mood to imagine at all. In the other case, however, one would never imagine a word shorter and fatter than a given one unless one had some special reason to do so, such as being asked to do it. This amounts to saying that the one kind of association is much more dominant than the other and much more readily aroused.

There can remain extremely strong associations involving images which have nothing to do with the process of imagining instances. Suppose, for example, that in an experimental situation a child always hears a loud noise whenever he hears the word "dog"; this experience, repeated many times, might give rise to a disposition for the child to imagine the noise even when he hears the word unaccompanied by the noise. Thus, in a sense, there might be a stronger association between the word "dog" and images unconnected with its meaning than there

Page 71: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

IMAGINING INSTANCES 61

is between the word and images of its instances. However, the child might still understand the term in the usual way, and this association which has been built up in him might have no effect on what the term means for him. This case also differs from random imagining, since the occurrence of the image of a noise is presumably involuntary; in random imagination the production of images is voluntary in some sense even though it does not matter which ones are produced. Thus an image may occur as a conditioned reflex, but this is not the way images occur when we imagine instances. In this last process the occur­rence of the images is voluntary and we have to want to imagine an instance before we have one. A conditioned reflex, on the other hand, takes place whether we want it or not and is often painful. It is perhaps incorrect to say that we have to try to produce an image which is a basic instance before one will occur, for the same reason that we do not say that I have to try to lift my finger before I can do it. The process is so easily performed that no conscious effort is needed. However, we can always arrange to make any bodily movement difficult enough so that it will require conscious effort, and there are similar cases of imagining basic instances. A child just learning language may have difficulty in imagining an orange, and someone with bad auditory imagery may have difficulty in imagining "wind in trees" noises, and these may also be basic instances. The kind of association that goes with conditioned reflexes can therefore be distinguished from that involved in the rela­tion of "instance of" on the grounds that the latter is voluntary and the former involuntary. This also brings in the factor of success and failure; association which has been established by this sort of conditioning is like random association in that there can be no question of success or failure in producing the associated image.

Another important distinction between imagining an instance and other imaginative processes would again hinge on the fact that, in imagining a basic instance in a certain sense no directions are of any use. This is not true of trying to imagine a word that is shorter and fatter, or of other similar processes which can be compared with ima­gining an instance. In these cases we can stipulate criteria which the image we accept will have to fulfill. We can say that the desired image must be dissimilar to the given one (unless we are trying to imagine an identical one), and we can say in what respect it must be dissimilar. Usually there are respects in which it must be similar as well.

We must now make some qualifications. The most important of these hinges on a distinction between "natural" meaning and "con-

Page 72: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

62 FOUNDATIONS OF SOLIPSISM

trived" meaning, and the different kinds of instances which correspond to these two kinds of meaning. In most cases a term or preliminary has a meaning for a person prior to his thinking about it. Usually this meaning depends upon a person's past habits and training, although he need not be consciously aware of that when he uses an expression meaningfully. At a particular time we have dispositions to accept certain kinds of cases as instances of a given preliminary and that settles its meaning. There are also some cases where we set out to let a term or preliminary have a certain meaning. This happens particularly when we introduce a technical term. At the beginning it has no meaning for ourselves or anyone else, but we decide to give it a certain meaning and then use it accordingly. Hence I will say that such a term used in such a way has contrived rather than natural meaning. This phenomenon need not be confined to technical or scientific language; in exactly the same way one might decide to give a term like "magenta" a particular meaning where it previously had a vaguer meaning, or even where the new meaning obliterates the old.

It is even conceivable that basic terms and preliminaries could some­times have contrived meaning. Suppose, for example, that we realize that the meaning of one of our own expressions is changing, and that we no longer use it in quite the way that we used it before. Or we may just be afraid that this will happen. It is, after all, possible for the mean­ing of some terms to shift gradually over time. If, for some reason, we want to prevent this, we may set aside the natural meaning that a term has at a particular time and determine to mean by it exactly what had been meant on a previous occasion. We might, for instance, re­member that color images of a certain kind had once been accepted as instances of a given color-predicate and determine to accept them in the future. However, we might still not have rules telling us what sorts of images to accept as instances, or which would be at all helpful in guiding us in imagining an instance. That is, one might be determined to accept certain images as instances and still not be able to form general rules for what to accept. Thus, the distinction between basic and non­basic instances may not quite correspond to the distinction between natural and contrived meaning.

It should not be taken, however, that the sort of phenomenon we have just mentioned is very common. It would occur only when we have some sort of special reason to use the same term with a very high degree of consistency over time, and we are not usually that concerned about preserving consistency of usage. We do, of course, suppose that

Page 73: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

IMAGINING INSTANCES 63

our "natural meanings" remain fairly constant, for we find it unneces­sary to compare regularly the object to which we predicate a quality to other objects which have that quality. We can say that something is red without implicitly comparing it to some sample. Even when I say that something is three feet long, I would not ordinarily be asserting that it bears a measured relationship to the standard meter, or any similar object. We establish and exercise our linguistic habits in an un­conscious way. I apply a predicate to an object without trying to remember how I last applied the predicate.

There will be a corresponding difference between natural and con­trived instances. When a term has contrived meaning one cannot try to imagine instances of it in the same way, since one has already decided to give it a certain sort of instances. We will soon come to analyze the relationship of instance hood, and what we say will be intended to apply to the concept only as it relates to natural meaning.

We can now discuss another process which is similar to trying to imagine instances: trying to form an image of some object or event which was previously seen or otherwise experienced. This process will be a particular kind of remembering. For example, I might try to form an image of the automobile accident I saw two weeks ago. One is tempted to say that I would be trying to imagine an instance of the phrase "the automobile accident I saw two weeks ago." However, this brings in definite descriptions which can have only one instance, and we have up to this point been considering the imagining of instances for preliminaries, which can have any number of instances.

There are many difficult problems which surround definite de­scriptions, and it would be unwise at this point to assume that the process of imagining instances of them is the same as the imagining of instances of general terms. Rather, it is more plausible to say that the process of imagining or producing a memory image of this accident has two elements; first, one must produce an image of an automobile accident, and, secondly, this image should match the original sensation or sensations and is subject to alteration according to certain criteria. Everything that has been said about imagining instances generally will hold true of producing images of automobile accidents, but if we are trying to imagine a particular accident there are additional criteria.

If I witnessed the accident, I would remember certain facts about it m the form of sentences rather than images, and these facts may lead me to tone down an over-dramatic image, or modify it in other respects. It may also be modified by things I learned from the newspaper. If I

Page 74: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

64 FOUNDATIONS OF SOLIPSISM

have confidence in someone else's account and think it likely that I was subject to a hallucination or illusion and saw things that did not happen, I may be inclined to alter my image in order to make it more like the sensations I think that I must have had. In this kind of case, the relationship between imagining an instance and imagining a past event is fairly clear. The former is involved in the latter, but the latter involves something else as well.

There is also another kind of case. Suppose that one is to look fixedly at an object such as a doorknob, close one's eyes, imagine the doorknob, look back at the doorknob, imagine it again, and so on. Here it seems that we are only trying to match a sensation with an image, and many of the factors which arose in the case of remembering the accident are not relevant. However, according to the argument in the last chapter for the case of private languages, there is always some evidence which bears on the question of whether any two images or sensations are similar in content. In the present case, it is quite possible that there would be physiological evidence; if the person involved suddenly had an image completely unlike the other images and sensations in the chain, this might be correlated with a brain event. Even if such evid­ence were lacking, there would always be the possibility of empirical evidence which would be relevant.

The second case differs from the first in that the second one is en­tirely non-verbal. Someone could attempt to match an image with a sensation without saying anything about either. In the first case, how­ever, we are committed to imagining an instance of an automobile accident and matching a present image with a past sensation. In this case I would not be remembering the accident if I did match the image with the sensation but the sensation was not of the automobile accident. Hence the process has a verbal element. This case of remembering is intermediate between the cases of "pure matching" and the more sophisticated kinds of remembering which do not involve images at all, but only the ability to make certain kinds of statements, to respond in certain ways, etc.

Our conclusion, then, is that if images are involved in remembering at all, then matching also is involved, and there are then criteria which are relevant for the correctness of the matching. The process of match­ing is distinguished from the process of trying to imagine basic instances because in the latter case we have seen that no criteria are relevant. Again this is only true of the imagining of instances of basic prelimina­ries which have natural meaning. In cases of contrived meaning, the

Page 75: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

IMAGINING INSTANCES 65

process of trying to imagine instances may reduce to a kind of re­membering.

The important point remains that criterIa are never relevant when trying to imagine an instance of a basic preliminary which has natural meaning, and that criteria are always relevant in any kind of re­membering. We have also remarked that contrived meaning is a more complex phenomenon than natural meaning, and is unlikely to arise until the learner of a language has acquired a certain sophistication. Thus, it is no cause for concern if imagining instances of preliminaries with contrived meaning does sometimes collapse into a kind of re­membering. We can still distinguish the imagining of instances from remembering in the most important cases.

When we are trying to imagine something which bears a given rela­tionship to a sample other than being an instance of it, we are always applying general rules to a particular case, and there will always be criteria which must be satisfied. Thus, if we are trying to imagine something darker than blue, we can say in advance that it must stand to blue in the same way that black stands to brown, and similar things. Further, the object to be imagined must be predominantly of a single color and cannot be multi-colored or transparent. It is true that when imagining an instance we can also say that the object to be imagined must stand to the word tokens as other instances stand to other word tokens. However, there is an important difference between the two cases. If someone is shown that certain colors are darker than other colors he may grasp the principle involved and be able to say what co­lors would be darker than certain new samples which had not been used in teaching him the meaning of "darker than." However, I may be familiar with any number of linguistic expressions and know what their instances are, but this does not help me in any way to say what would be an instance of some other basic preliminary. Further, I can change my meaning of "red" and let it have the same instances that "black" has without affecting the general relationship of "instance of"; all other terms and preliminaries can continue to have the same in­stances. However, if I decide that I will count brown as being darker than black, then all other instances of the relationship will be affected. We can then conclude that we class cases of "darker than" and other similar relationships together under the same concept because we can actually see similarities between the cases. However, the various in­stances of the relationship "instance of" are not classed together for this reason; we do not say that an object, sensation, or image is an instance

Page 76: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

66 FOUNDATIONS OF SOLIPSISM

of a basic preliminary because of any empirically recognizable general relationship which holds between instances and bits of language. We count certain kinds of things as instances because we have been taught to do so; however, we could have been taught different habits of counting objects as instances, and it does not really matter what we mean by expressions of the language as long as all the people in the community are uniform in their habits. It is for these reasons that we cannot say in advance what criteria a basic instance will have to fulfill; we can only state criteria for a particular basic instance by first looking at our own habits and describing the sorts of things we already do count as instances. In the case of other relationships that we know, we begin with the rule and apply it to the particular case, but the situation with "instance of" is reversed, at least when we are concerned with basic preliminary statements.

In this chapter the process of imagining instances has not been re­duced to anything else, but we can claim to have isolated and described a very important mental process which is basic to our theory of mean­ing. We can now refer to it freely in analyzing the concept of under­standing.

Page 77: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 4

COMPREHENSION

In this chapter I will begin to explain the relationship between the processes of imagining instances and understanding units of discourse. In order to do this, we must consider another important process, an extension of imagining an instance, which is imagining a range of instances.

While it is easy to imagine a single instance of a basic preliminary it may be more difficult to imagine a whole range of instances, par­ticularly as we approach the extremes of the range. We have already seen that we may be either satisfied or dissatisfied with an image we produce. In simple cases, of course, it may be so easy to produce an image of the required type that we feel very little satisfaction at having done it, and we pass on to other, more rewarding, activities. Even here, however, we may sometimes not be able to imagine immediately just the sort of situation we want, and this is likely to be irritating. If, on the other hand, we are trying to imagine all the different things which can be called "red," we will come to various attitudes towards the images produced, just as we would if we were dealing with various samples of red objects. Some will be clearly accepted as red, others will be felt to be red but "not as red" as others; others will be rejected as orange rather than red, and borderline images and samples will leave us undecided.

We need not comIIlit ourselves as to whether the attitude of accept­ance is just a feeling or a set of dispositions as well. Dispositions might include an avoidance of looking further, making certain responses to questions, etc. In any case the rejection of an instance need not entail an attitude altogether different from one associated with acceptance; it will be merely the absence of the feeling of satisfaction and perhaps the absence of the relevant dispositions.

It might be that in rejecting an image there is a feeling of dissatisfac­tion, and that merely not having this feeling constitutes acceptance. However, this seems implausible to me, since the feeling of disgust which one might have with a bad sample seems only to occur in special

Page 78: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

68 FOUNDATIONS OF SOLIPSISM

cases and to be incidental to the process. The role of the attitude of in­difference also bears on this point. One can be indifferent as well as being undecided as to whether a given sample or image is an instance. If one is indifferent one feels nothing at all. If acceptance amounts to just not feeling dissatisfaction, then we would accept as instances all samples toward which we are indifferent; this cannot be correct, since we ordinarily feel that there is a great difference between acceptance of something and indifference towards it.

In fact, an attitude of indifference is not likely to arise as an initial reaction at all, when we try to find or imagine instances. In general, this attitude does not usually accompany a project which we initiate and carry out ourselves; it is a much more typical reaction to a ques­tion raised by someone else which we may be asked to answer. Thus if someone asked us to find instances of Wittgenstein's term "bububu," we would probably reply that we could not or would not, and not even attempt to find any. However, if we were encouraged or compelled to try to do this, we might then be indifferent toward any sample we might produce. Ifwe were interested enough to set out to do this in the first place and came across a borderline case, our reaction would be that of indecision rather than indifference. Thus it seems that an attitude of indifference would arise initially only in farfetched cases.

Questions as to the nature of acceptance and rejection in this con­text are basically questions of introspection. For our purposes it does not greatly matter what the answers are. All we need to suppose is that there is such a thing as acceptance and its absence, or perhaps rejection and its absence. It also appears that there will be no sharp break be­tween acceptance and rejection. Ifwe are dealing with a feeling, it will have varying degrees of intensity. Even if we are dealing with disposi­tions, they would have to be described in a rather vague way, and it will be possible for them to hold in varying degrees. If we can specify the dispositions precisely, they will probably amount to dispositions for feelings of satisfaction and/or dissatisfaction to occur under certain circumstances. Again, these feelings will admit of degrees and there will be no clear minimum degree of satisfaction.

Up to this point we have been talking about the imagining and finding of instances of predicate expressions rather than of prelimina­ries. Of course, we saw in the last chapter that the difference between a predicate expression and a preliminary is not crucial for our pur­poses, but we will nevertheless say something here about imagining instances of preliminaries.

Page 79: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

COMPREHENSION 69

When we have a predicate expression in mind, and no special con­text is operating, our having of an image together with a feeling of satisfaction amounts to the acceptance of that image as an instance simply because there is nothing else we could be doing with the pre­dicate expression except looking for an instance of it. Similarly, one might start with an image in mind, again without any particular context being present, and a predicate expression might then occur to one together with a feeling of satisfaction. In either case, we can talk about the application of a predicate expression to a case which then turns out to be an instance. However, there is then no specific action which goes with this application of a predicate expression. Rather, there are just the two images (one a verbal image and one a non-verbal image) and a feeling of satisfaction.

When, by contrast, we are looking for an instance of a preliminary, again outside any special context, we are prepared to assert the pre­liminary in our own thinking when we find a suitable instance, either an image or a sensation. There still will be the feeling of satisfaction, but this time it is closely bound up with the act of asserting.

The fact that asserting is a kind of action can be brought out more clearly by considering a peculiar kind of language. Suppose that two persons have a system of communication whereby one throws stones of different sizes and shapes at the other to indicate events that are taking place. In this system of communication, stones of varying sizes and shapes would correspond to the preliminaries in that they are en­dowed with different kinds of cognitive content. Still, nothing is assert­ed by picking up one of these stones, just as nothing is asserted by having a preliminary in mind. In order to assert anything, one must throw the stone.1

In our language the act of asserting is not marked off quite so clearly, but in interpersonal conversation we nevertheless can isolate acts of uttering or writing locutions where the uses of particular grammatical forms count as acts of asserting. In the next chapter we will discuss ex­actly what is involved in understanding that an utterance is an asser­tion. For present purposes, however, we must concentrate our atten­tion on what it is to assert a preliminary in our own thinking. This is because the process of accepting an imagined instance of a preliminary is just the process of moving from a preliminary to an assertion where the universe of discourse is simply the imagined state of affairs.

1 This very useful example was suggested to me by H.-N. Castaneda. He would, how­ever, disagree with the conclusion arrived at here.

Page 80: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

70 FOUNDATIONS OF SOLIPSISM

On one hand it is clear that we can make assertions in our own thinking because certainly we can have beliefs all by ourselves, and the way we have beliefs is by asserting things to ourselves. On the other hand, it is clear that asserting something in one's own thinking need not involve any overt physical action, such as uttering or writing words. It should also be seen that, although believing presupposes asserting, there is a difference between believing and the kind of assert­ing involved in accepting an imagined instance. Most beli ~fs are pre­dictive, can be verified, and are corrigible; and, of course, predictions are a kind of assertion. When we are imagining an instance of a pre­liminary, however, we are asserting the preliminary only for the uni­verse of discourse fully constituted by our imagined state of affairs. There is nothing predictive about this assertion since there is nothing in the imagined state of affairs which is hidden from us, and which could turn up on the occasion of some further thinking. What, then, is comparable to the throwing of the stone?

The distinction between asserting and merely considering in one's own thinking, then, will be a more subtle one, as one would expect from our earlier comparison of thinking and inter-personal conversa­tion. This difference seems to be comparable to the utilization of cer­tain abilities and is reminiscent of some of the things that Ryle said about imagining. We rejected his view that to imagine is simply to exercise certain abilities, but, once having a preliminary in mind (which would amount, at least in part, to having certain visual images of written words or auditory images of spoken words), the asserting of it might amount to going through the imagined string of words without any sort of hesitation and, at any point in the string, being confident of what comes next. That is, it might be comparable to the difference between making up a song in one's imagination and going through a song that one already knows in the imagination. Even if one is making up the song, one may never have to stop until the song is finished, but one would still not have that "being confident of knowing how to go on" feeling at all points. In the latter case one is utilizing an ability, the ability to hum to oneself or imagine the song, while in the former case one is acquiring an ability rather than utilizing it.

When we first consider a preliminary before we have found or imagined an instance of it, we may utter it to ourselves (or have a few odd snatches of words that serve the same purpose go through our heads), but we do it with a feeling of uncertainty and with a willingness to "take back" these words. When we find our instance, the feeling of

Page 81: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

COMPREHENSION 71

uncertainty disappears, and the attitude of uneasiness is replaced by one of confidence. There are similar cases which are drawn from inter­personal discourse, but which will illustrate what is meant. On the baseball field I may hold up two fingers to inform the other players that there are two outs. On the other hand, if I am not certain that two are out, I may look at someone and again hold up two fingers in the expectation that he will correct me if there is only one out. In the latter case, I will probably have a quizzical look on my face, but I may be too far away for this to be seen and, in any case, it is not essential to the process of questioning. In these two situations one would have very different feelings or attitudes while giving the same sign. This reflects the fact that the sign is being used for very different purposes. Again, in the latter case, I would be willing to "take back" my sign (assertion) by quickly lowering my fingers without any hesitation or argument, while in the former case the sign is made confidently and I will not re­move it without some sort of further persuasion. It is this same differ­ence of attitude which marks the difference between asserting and considering in our own thinking, and which is consequently integral to the process of imagining and finding instances of preliminaries. As in the case of the instances of predicate expressions, it is a difference of feeling or attitude that is crucial; however, when dealing with pre­liminaries, there are certain additional nuances of feeling which we have felt worth notin?;.

We will now see that the process of imagining a range of instances is more closely related to understanding than is the process ofimagining a single instance. I will argue shortly that the notion of understanding is more basic than that of meaning, and it turns out that there is an­other state which I will call "comprehension" which is more basic than understanding. This turns on the difference between understanding a sentence or expression, misunderstanding it, and not understanding it at all.

There clearly is an important difference between the latter two cases. If we misunderstand someone who is trying to teach us an expression, we nevertheless attach some meaning to it and are prepared to use it in some way, even though this use may be a deviant one. Thus, the logically prior question is whether someone attaches meaning to an expression. If he does, we can then ask what sort of meaning he attaches to it, or whether he means the same thing by it that other people do, and so forth.

It is this basic notion that we want to deal with, but there seems to

Page 82: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

72 FOUNDATIONS OF SOLIPSISM

be no term in English which covers both understanding and ITIIS­

understanding and excludes only not understanding at all. Thus I will introduce the term "comprehend" with the meaning "to attach some meaning or other to." We can later analyze understanding, and ulti­mately meaning, in terms of this notion.

It may appear to the reader that we have already made a mistake in making this distinction and bringing in the term "comprehension." It might be said that it is impossible to attach meaning to an expression in isolation, as is suggested by the notion of comprehension thus defined. However, even someone who takes an institutional view of meaning must admit that it is possible to misunderstand a teacher and to use a particular expression in the way that other people use some other ex­pression. Again, no one could deny that this is different from failing to learn to use the expression in any way at all.

In examining the relationship between comprehension and the process of recognizing instances, whether images or actual samples, we must notice the part played by partial instances. These cases, in which we are only partly satisfied when looking for instances, resemble the cases where we accept instances without reservation to a large ex­tent, but differ from them in two main ways. Taking first the situation where we are imagining a range of instances for redness, we might be doubtful about admitting a red-orange image and only be partly satis­fied with it. At other times, however, we might be doubtful about an instance not because it is a borderline case in this way, but because it is only part of what would constitute a full-fledged instance. For example, we might try to imagine an instance of a batter hitting a home run and produce an image of a batter swinging at and hitting a ball, but leave out the flight of the ball. This image would be satisfactory as far as it goes, but, of course, hitting a home run involves more than just hitting the ball, so we would not accept it as a complete instance. In this particular example we can easily appeal to the definition of a home run, so it is unlikely that we are here dealing with basic preliminaries or instances. However, it seems possible for the same thing to happen in the case of basic preliminaries. It need not necessarily be the case that events which are instances of our basic preliminaries are very limited in spatial and temporal extension. If they have any very con­siderable extension in either respect it may then be easy to imagine parts of the events, but difficult to imagine them aU at once, and we will again have the occurrence of partial instances. For example, in a certain society in which fishing was a very important activity, a single

Page 83: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

COMPREHENSION 73

word might denote the process of catching a fish, and this word might be one of the first and most basic ones in a child's vocabulary. Thus, in the child's vocabulary, the term or terms which would be the equiv­alent of "catching a fish" might be prior even to the word which func­tions as the equivalent of "fish"; this last term might be defined later in terms of the word which denotes the whole process. If such a child were trying to imagine an instance of the term in his language which has the meaning of "catching a fish," then an image of a fish would be only a partial instance. Thus we can get partial instances when trying to imagine instances of preliminaries which belong to very different levels of language.

A similar situation again holds when we switch from imagining in­stances to finding instances in the world. If we were trying to find a range of samples of cloth all of which are red, we would react in the same way to the red-orange sample that we did in the case of the red­orange image, and they will both be partial instances. In the other kind of situation, however, we have more complexity. If we are trying to find an instance of "catching a fish" in the world, we may still come upon a fish, but instead of classifying it as a partial instance of "catch­ing a fish," we will simply look to see if the fish is in the process of being caught; we will then either accept it as an instance without reservation or reject it completely.

This indicates that there are significant differences between thing­instances and image-instances of a preliminary. With the important exception of the borderline cases mentioned above, it seems that an actual state of affairs must always be either an instance of a preliminary or not. There is a class of partial image-instances for which there is no corresponding class of partial thing-instances. If we are in doubt as to whether a thing is an instance, we can, at least in theory, investigate further and settle the question; further, it makes sense to speak of the probability of its really being an instance. This is not the case with possible image-instances. If we cannot initially classify an image as either an instance or a non-instance, there is no possibility of being able to do so later. This is partly because there is nothing further to know about the image. In one sense it is impossible to know the quali­ties of an image more determinately than they are given in the first place. We cannot examine an image more closely to see what its quali­ties are because any increase in its determinacy will only mean that it has changed. Second and more important, it is impossible to explore the physical context of an image. If I have an image of a house, it is

Page 84: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

74 FOUNDATIONS OF SOLIPSISM

impossible for me to find out what is behind the house in the way that one can find out what is behind the actual house.

The situation is again changed when we think of instances that are sensations rather than physical objects. Ordinarily we would say we are looking for instances in the physical world; however, we have seen that our actual judgments as to whether or not a certain entity is an instance will depend on our sensations of the object rather than on the object itself. This is true especially if we think of the recognition of instances as an immediate and instantaneous process. There will then be no time to find out whether the sensation is illusory or not, and our reaction of satisfaction or the lack of it will depend entirely on the sen­sation itself. The empirical process of finding out whether a certain object really is what it appears to be is a familiar one, but it is distinct from and goes beyond the process which I have in mind in speaking of the recognition of instances.

Thus, there are really two different processes involved, the recogni­tion of things and the recognition of sensations; these are often con­fused with one another, but it is the latter process which is more im­portant for our purposes. Sensations are like images in that a sensation may be indeterminate and neither have nor lack certain qualities. They are also like images in that we cannot find out anything more about them by exploring their background or conducting further in­vestigations. Sensations, of course, occur in a context. We ordinarily know what other experiences preceded them or were preceded by them; we could even try to discover what other sensations might have occurred at the same place and time but did not. Nevertheless, the point is that we treat this information as being irrelevant to the nature of the sensation; by a sensation we mean a conscious event which took place between certain time limits and, as long as we remember it clearly, there is nothing more about it to be known. Hence, unlike the process of recognition of things, we will get the same kind of partial instances in the case of recognition of sensations that we did with the recognition of images.

Thus far we have treated recognition of instances and imagining of instances as parallel processes, but we must now look into the relation­ship more closely. In the former process it has been noted that the stimulus which is recognized will always be a sensation of an object. That is, the recognition or the lack of it will depend not on the actual qualities of the object, but on the sensations of it. Thus, in both these processes the entities which are accepted or rejected are mental

Page 85: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

COMPREHENSION 75

phenomena, and we have already seen the close correspondence be­tween images and sensations. Further, both of these processes can be broken down, at least logically, into two parts. On the one hand there is the production of candidates for instancehood, and, on the other, there is the question of whether the candidates are in fact instances. It might be argued that this is an artificial distinction. In practice we do not first produce an image or find an object and then sit back and judge its merits as an instance; rather, we produce an image or an object which is an instance and the very fact that a given image or object is produced in a certain context testifies to its instancehood. On the one hand we can admit that the two processes are often teles­coped into one in practice, but the logical distinction still remains. It is possible to try to imagine an instance, produce an image, and then reject it as an instance, particularly if we are trying to imagine extreme cases of a concept. In the same way, when dealing with color samples and trying to find instances ofred, I might pick up an orange object to this end, consider it, and then decide that my notion of redness does not extend that far.

This distinction is particularly important for the analysis of compre­hension. The important question will always be not whether an in­stance can be found or imagined, but rather, given an image or a sam­ple, whether it would be accepted. In the processes of recognition of instances and imagining of instances the aspect of acceptance or rejec­tion appears to be the same. We have already seen that images and sensations all can have the same qualities, and whether or not a mental phenomenon is accepted seems to depend solely on its qualities and our prior habits and training; the difference between an image and a sensation does not itself seem to affect the issue. It must be in the other aspect of the two processes that the difference lies.

Whether we produce images or look for samples, we are in both cases trying to find or become conscious of a mental phenomenon which will satisfy us; however there are two different ways in which we can go about this, and the means to the end differ considerably. In fact, some­one might simply want an instance of a certain preliminary and not care whether it is a sensation or just an image; to this end he might look at objects nearby and, not finding a suitable one, he might then begin to imagine other cases, or he might even alternately imagine cases and look for cases. It could be argued that really there is only one process which we might call finding instances, and that imagining instances and finding them in the world are just variations on the basic process

Page 86: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

76 FOUNDATIONS OF SOLIPSISM

in the same way that we can use different kinds of shovels to shovel dirt, and so forth. We will nevertheless find it more convenient to continue to distinguish between the two processes since each has its advantages and disadvantages as a criterion of understanding; the important point is that in both cases the samples found or imagined can be rejected or accepted as instances, and the criteria or mechanics of acceptance or rejection are the same in both cases.

Having thus distinguished these two processes, we must now ask how they are related to comprehension and which relationship is more fundamental. We can say generally that comprehension is the disposi­tion to recognize instances, and that the way in which we comprehend a preliminary is determined by the sorts of instances we would accept for it. This is a direct consequence of the view, expressed earlier, that the meaning of a preliminary is determined by its range of denotation over actual and possible states of affairs. If a possible state of affairs is included in the range, then, if we imagined it when looking for in­stances of the preliminary, we would accept it. Similarly, if the possible situation became actual and we encountered it when looking for in­stances, we would also recognize it. Further, any possible state of affairs will be such that there is no logical contradiction in supposing it to be imagined or realized.

However, this leaves a number of questions open. In particular we have not said which of the two processes, the recognition of images or the recognition of sensations, is more important. Further, there is the question as to whether comprehension is a matter of degree on the one hand, or if there is a sharp break between comprehension and the lack of it on the other.

Taking the second question first, we can begin with a case that is obviously one of complete comprehension. Suppose that when trying to imagine instances of a preliminary, someone can produce a range of images varying continuously as to quality, that all these images are accepted without reservation, and that no other images would be accepted even as partial instances. Further, let us suppose that he can immediately classify all possible sensations in the same way, and that the acceptable sensations have the same qualities as the acceptable images. In this case we could say that the range of meaning of the preliminary is exceptionally clearly marked out, and that there are no fuzzy edges; consequently, there will be no partial instances. Let us also suppose that he is able actually to imagine on the spot all the different kinds of instances it might have. Thus, we could say that this is an

Page 87: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

COMPREHENSION 77

example of perfect and complete comprehension, but it seems unlikely that it is ever realized.

In the first place, we have not taken account of indeterminate images and sensations. Even if we have a concept of a chair which allows for no vagueness and is sharply distinguished from a stool, we might still have an image or sensation which is blurred and indeterminate and is not clearly either a chair or a stool. In this case it will be a partial instance, in spite of what we have said above. Further, it seems unlikely that any of our concepts are so clearly defined that they admit of no borderline cases as far as images and sensations are concerned. We can use numerical methods in order to settle the range of application of a concept with regard to physical objects, but since these methods do not apply to sensations and images, even if they are determinate, there still will be borderline cases and partial images. Lastly, the ideal of complete comprehension also demands that we be able to produce a representative range of images, and in some cases our powers of ima­gination may not be up to this.

Thus the maximum, though probably unattainable, degree of com­prehension eliminates vagueness altogether and we are completely clear on what is an instance and what is not. Comprehension is not to be identified with lack of vagueness, since we may also be perfectly clear that an expression to which we attach no meaning has no in­stances; hence, we can fail to comprehend without there being any vagueness involved. Nevertheless, complete comprehension does seem to imply the lack of vagueness, and when vagueness increases, compre­hension diminishes. Thus, if an expression is so vague for us that there are only borderline cases and no clear instances at all, we do not have a very clear comprehension of it; our concept is at best confused if we cannot find any possible clear instance of it. Thus the continuum of comprehension is related to a corresponding continuum of vagueness.

We can now indicate some of the main stages of this continuum. A degree of comprehension which is probably fairly easily attained would be where there are some partial instances both among images and sen­sations; we would also be able to recognize as complete instances a wide range of images and sensations which vary continuously in quality between the various borderlines. As we progress down the scale of comprehension we will find that fewer and fewer complete instances can be imagined, and an increasingly larger proportion of the instances accepted will be only partial instances. We will eventually reach the point where any sensation or image which is encountered will at most

Page 88: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

78 FOUNDATIONS OF SOLIPSISM

be accepted as a partial instance. It could happen that even if all possible states of affairs were paraded before us or imagined, we would not accept any as a complete instance of a preliminary and yet would have some comprehension of it. Thus I might have some vague idea of what a squeeze play in baseball is and admit that some sequences of events come reasonably close to being squeeze plays, but, on the other hand, I might be selective enough so that I would not unquali­fiedly admit anything that could happen as a case of a squeeze play. This would happen if I am determined to mean by "squeeze play" not just anything, but want my meaning to be in accordance with estab­lished tradition. However, I also realize that it has a complex meaning, and that I do not know exactly what would be a squeeze play. On the other hand, I know what it is to run from a base and know that this is involved in some way, but not in what way. Thus, if I try to find or imagine instances, I will find partial instances but not complete in­stances. Someone might object here, in accordance with the distinction made above, that I comprehend the expression but do not understand it. This would be the case if! settled on some other meaning of "squeeze play" for which I could find complete instances. However, in this case I am not doing that because of my determination to mean by the ex­pression whan everyone else means. Consequently, since I have only a partial notion of what they mean, I have only a partial or confused comprehension of the expression myself. In this case I would be aware of this fact, but there could be other cases in which I might be unaware of the fact that I could think of no complete instance of a certain preliminary. Even this does not mark the minimal degree of com­prehension, and there is no such degree. Someone who recognized even fewer partial instances of the same preliminary and was even more hesitant about them would comprehend it even less. Thus, the degrees of comprehension form a continuum stretching between the ideal mentioned above and total non-comprehension.

Going back now to the question as to whether the ability to recognize images or sensations is more basic to comprehension, we will see that both have their advantages. An important related question concerns the ways in which we can or cannot be mistaken and be in doubt when we think that we understand and comprehend preliminaries and their associated sentences which make assertions.

When someone says that he is not sure whether he understands a piece of discourse or not, he generally means that he attaches some meaning to it, but is not sure whether this is the intended meaning.

Page 89: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

COMPREHENSION 79

Let us now consider the more interesting case of whether it makes any sense to be in doubt as to whether one attaches any meaning to a given piece of discourse, or comprehends it. In one sense we might be in doubt as to whether what we comprehend by it is consistent; we could be convinced that we attach meanings to different parts of the discourse, but we might wonder whether we attach any meaning to it as a whole. This doubt would be especially likely to occur if the discourse is very complex logically; we would then have to work out the relationships slowly and methodically before it could be determined whether the discourse as a whole is comprehended in any way. It seems clear that if meaning is attached to all the parts, (including whatever connectives are used to string the parts together), and it is consistent, then some meaning will be attached to it as a whole. This need not imply that it will be comprehended as a unit, since certain parts may be totally unrelated. But the discourse will be comprehended either as a conjunc­tion of assertions, a disjunction, etc., or as a unit. Here, of course, one would not really be puzzled as to whether one attached meaning to the discourse as a whole; at first the answer would simply be negative and the question rather will be whether one can come to such a com­prehension. This will occur when one realizes that the various parts are consistent with one another.

There are other cases where we might not comprehend the discourse at the present moment, but might suspect that we are coming to com­prehend it for a different reason. In such a case it might be that we are unfamiliar with the terms used in the expression, and with the help of a dictionary or other device we are translating it into a different and more familiar terminology. We would then not comprehend until the translation had been completed. When dealing with complex subject matters, we often have to translate sentences uttered by others into our own language before we can decide whether we attach meaning to them. Again, we do not comprehend the sentence initially and know this, but we also know a method by which we may come to compre­hend it.

In any of these cases we are able to answer the question, "Do you attach any meaning to this sentence?" either immediately or after go­ing through certain mental processes. These processes, however, never involve any appeal to outside information or experience. Thus it is just a case of the application of rules, not experimentation. In very simple cases, there seems to be no possibility of doubt, and any mistakes can be immediately corrected. The discourse is not long or complex

Page 90: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

80 FOUNDATIONS OF SOLIPSISM

enough to hide a contradiction, and if it needs to be translated into more familiar terms, this can be done instantly if at all. It will, of course, happen that a simple sentence which was not at first compre­hended will be comprehended later when we are told that it is syno­nymous with some more familiar sentence; but this does not involve ever being in doubt as to comprehension or the lack of it at any par­ticular time. Thus, the question of whether any meaning is attached to a simple sentence is somewhat like the question whether one feels pain. There is comprehension-behavior which is correlated with com­prehension in the same way that pain-hebavior is correlated with pain. However, it is always logically possible for someone to be in pain with­out displaying any particular behavior, and it is always possible to comprehend a piece of discourse but to show no effects of it. At least the question about comprehension resembles the question about pain more closely than questions about the nature of physical objects. These questions about comprehension even differ from questions about bodily abilities such as "Could I throw a silver dollar across the creek?" I can be prevented from settling this last question by physical circumstances, but no matter what the world is like, I can find out whether I attach mean­ing to a sentence which expresses either a preliminary or a statement.

In the more complex cases it is always possible to apply the rules of logic, grammar, and arithmetic incorrectly, and, as a result, to give a false answer to the original question. But the same possibility of mistake arises in arithmetic without detracting from the fact that arithmetical statements are certain if they are true at all. In all cases where doubt could arise, this doubt can also be settled conclusively without having to wait to observe our behavior, or any other external phenomena. In the more complex cases we generally do not take the trouble to make the conclusive verification; as will be described later, we make certain short-cuts which are usually reliable, but which can lead to error.

All this is consistent with our earlier conclusion that comprehension is a matter of degree, and that there is no sharp line between com­prehension and non-comprehension. We outlined a scale of compre­hension above, and our present conclusion is that we can settle any doubts we may have as to approximately where our comprehension of a given preliminary belongs on the scale, and that we can do this without empirical experimentation. Having determined this, the question of whether to call it "comprehension" or "non-comprehen­sion" is an entirely separate one; it is also a purely verbal issue and is not very important.

Page 91: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

COMPREHENSION 81

The upshot of the foregoing discussion would be that there is an advantage in stating our analysis so that the recognition of images is most important. In that case any doubts about comprehension could be settled on the spot, as we would expect. If, on the other hand, we stated it in terms of sensations, we would have to hunt around for ob­jects which would make us feel the required sensations before we could settle our doubts. Of course, the situation would be reversed in the case of someone who had a very poor imagination and immediate access to all sorts of objects; however, this latter case represents the exception rather than the rule.

In concluding we must mention the rather strange possibility that someone's dispositions to accept images as instances of a preliminary might be quite different from his dispositions to accept sensations as instances of the same preliminary. Thus, if we are looking for instances of something's being a red barn, it is conceivable that someone might accept a certain image as an instance and then reject an exactly similar sensation, or vice versa. This might even happen in a systematic way and reflect a prejudice against either sensations or images. In connec­tion with this it is interesting to inquire what will happen when we intentionally look for one class of instances to the exclusion of the other. Thus I might comprehend "red barn sensation" and "red barn image" in the usual way and try to find instances of them. If I take the latter expression, I will then be less satisfied with a sensation than with an image, and in the former case I would be less satisfied with an image than with a sensation. There is a certain imbalance here, however; if I am looking for an image, I would be unlikely to have a sensation of the required sort at all, but in the opposite case there would be a ten­dency, if I could not find an object to produce the required sensation, to imagine an instance instead and to be partly satisfied with it. Never­theless, the parallel is still close between the two processes, and because of this kind of case, it would be a mistake to define comprehension in terms of one kind of recognition to the exclusion of the other.

If we are not explicitly looking for one kind of instance, it is very improbable psychologically that we would systematically favor one kind of instance over the other, but it is still an interesting logical possibility. If someone had a systematic prejudice against either sensa­tions or images, but could still recognize instances of some sort, I think we would have to admit that he still comprehends the preliminary con-

Page 92: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

82 FOUNDATIONS OF SOLIPSISM

cerned to some degree. If he will only recognize images as instances, he is simply refusing to talk about sensations or physical objects, and pre­liminaries for him might have the same meaning that they would have for us if we added the phrase "image of" to them. He might still have a language, but it would tend to be a private language. In the opposite case something rather similar would happen. We would not find it as surprising if someone said that his powers of imagination were very poor, and that he could not imagine instances. But here we are dealing with someone who has normal powers of imagination, but is unwilling to accept any image as an instance just because it is an image, even though it may have all the required qualities. It is implied that he cannot talk about images as such, and, again, we might make his preliminaries synonymous with those of other people by inserting "sensations of" into them.

In either case we would be dealing with someone who comprehends perfectly well, but who, for some reason, refuses to talk about a certain class of entities at all. Thus, even though there is some reason for saying that the ability to count images as instances is more basic, our criteria for comprehension of a preliminary will be a disposition to recognize either images or sensations as instances in a systematic way. We cannot make the criteria more exact because we have already seen that com­prehension is a matter of degree, and that there is no minimum degree which we can specify.

Page 93: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 5

UNDERSTANDING AND SYNONYMY

Having discussed the concept of comprehension in the last chapter, I will now try to relate it to the understanding of sentences. The term "understanding" is used in two quite different ways. We can speak of understanding what someone says, and we can also speak of under­standing a sentence apart from any particular person who may utter it. The first usage takes account of the possibility that while the person who utters the sentence may not attach the conventional meaning to it, still we may speak of understanding him. In the second case we speak of understanding the sentence out of context and apart from any par­ticular user; this will then amount to knowing its conventional mean­ing, which may vary from context to context and from user to user.

With respect to the first sense of understanding, we can start out by saying that the listener understands the speaker completely if and only if he comprehends the speaker's utterance in exactly the way the speaker comprehends it himself. Later a few qualifications will be need­ed. According to the above analysis, speaker and listener both will be disposed to accept as instances of the associated preliminary exactly the same images and sensations, and each person would have to be disposed to accept them to the same degree. This implicit comparison of images and sensations belonging to different people will, of course, raise the traditional problems of other minds, which will be discussed in a later chapter; the analysis given there will try to make sense out of such a comparison, and, if this is accepted, we are provided with an easy transition from the concept of comprehension to the more impor­tant sense of "understanding."

It will follow from this, however, that we very seldom understand someone completely. Our understanding usually falls short of this upper limit in one of two ways. First, the listener may comprehend the sentence to the same degree as the speaker, but attach a somewhat differ­ent range of meaning to it (recognize a different range of instances for the associated preliminary), and, secondly, his comprehension of it may be of a lower degree than that of the speaker. There will then be

Page 94: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

84 FOUNDATIONS OF SOLIPSISM

degrees of understanding just as there are degrees of comprehension, the main difference being that there are cases where we can under­stand an utterance completely without comprehending it completely. If the speaker has only a vague idea as to what he means by his own utterance, the listener can understand him completely if he attaches the same vague meaning to it; even if the listener comprehends the utter­ance more completely than does the speaker, the former does not under­stand the latter any better, since he already understood him com­pletely.

In the second sense of understanding, we are said to understand a sentence if and only if we comprehend it in the way that most people in the language area or the special field involved comprehend it. Here again there will be degrees of understanding, but in technical contexts it is usually assumed that the concepts are precise, and in these cases complete understanding would imply a high degree of comprehension. In non-technical contexts, on the other hand, a quite low degree of comprehension might suffice for complete understanding.

It is clear that we also speak of degrees of understanding in ordinary speech. For instance, it may be appropriate to say that Tom under­stands what Dick is saying better than does Harry, or that Tom under­stands psychiatric terminology better than does Dick. In ordinary lan­guage, however, we usually talk this way only when we are dealing with subject matter that is supposed to be hard to understand. Thus, I would not say that I understand the word "red" better than Psmith unless we both happen to be foreigners or children learning the lan­guage, or some other special context. Still, we can conceive of cases where even simple predicates and sentences would not be compre­hended very well. Hence, there could be degrees of understanding even in such cases.

Further, when we distinguish the first sense of understanding from the second, the possibility of frequent relatively slight misunderstand­ing, on almost any analysis of meaning, becomes much greater. Since our ordinary language does not take into account all these possibilities, the word "understanding" is not always used in such a way that it ad­mits of degrees, and the logic of the term is somewhat confused. In order to clarify it, we will always treat this concept as one involving degrees.

It is now necessary to analyze the comprehension, and hence the understanding, of sentences expressing statements in terms of the com­prehension and understanding of preliminaries. We must first distin-

Page 95: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

UNDERSTANDING AND SYNONYMY 85

guish the understanding of a sentence which makes an assertion from the understanding of a sentence or phrase (usually a noun clause) which expresses a preliminary. It will be recalled that the difference between the two is that while they both may be about the same state of affairs (if they correspond as explained above), the one simply considers it while the other asserts it to be the case.

In the last chapter we considered the difference between a prelimi­nary and the corresponding assertion in the context of one's own think­ing. We will now be concerned with what it is to understand that someone else is making an assertion as opposed to mentioning a possi­bility, raising a question, etc. Surely the difference in our understand­ing of locutions doing these various jobs will not always be traceable to a difference in comprehension, but will sometimes have to do with the intentions with which the locution is uttered. Since a locution ex­pressing a preliminary is not really a complete move in most language games, it can be filled out in various ways: as a request to consider a possibility, as a question, as a statement that someone is considering a possibility, etc. In fact, if someone utters a phrase expressing a pre­liminary or even says something such as "It might or might not be the case that .... ," we might feel that the question of understanding has not yet arisen, and will do so only when he fills out his utterance in such a way as to make an assertion, ask a question, or give a command. Then, in order to understand, the listener must not only comprehend the preliminary in the same way, but also know in which of these various ways the preliminary is being used. We must now inquire what it is to know that a preliminary is being used in a certain way, that is, to make an assertion as opposed to asking a question, etc.

The function of assertions as opposed to other kinds of utterances is closely connected, clearly, with the phenomenon of belief. Nevertheless, it is not logically necessary that one believe all the propositions one asserts, because it is possible to lie and still make the same assertion one would have been making in uttering the sentence sincerely. In order to say what is involved in asserting, we also have to refer to the intentions of the speaker. It might seem logically necessary that when someone makes an assertion he at least intends his listeners to believe him. Even here there are exceptions. If two people are competing with one another to see who can tell the biggest lie, and one says that he caught a fish the size of a whale, he does not expect or intend that he be believed. However, in this case we can plausibly say that he is not really making an assertion in a straight-forward manner, but is just

Page 96: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

86 FOUNDATIONS OF SOLIPSISM

testing his own powers ofimagination. Taking the context into account, we might say that if he is asserting anything, it is that this would be a very improbable situation, or that this is the most improbable situation he can think of at the moment, and he could expect to be believed about that.

In such a context the preliminary normally expressed by the sen­tence would not be the one which is asserted to be the case, and all concerned would be aware of this fact. However, there are other cases which are not so clear. Suppose that a diplomat in a certain situation thinks that he will be disbelieved in a certain question, no matter what he says. He would nevertheless be careful to choose a falsehood, or even a true statement, which he thinks will have the right effect. In this case he certainly does not intend to be believed, but we are not so ready to say that he is not asserting the sentence he utters. In particular, we do not want to say that he is misunderstood by some naive person who takes him to be asserting that the state of affairs his sentence describes is the case. Our puzzlement in this case arises because, while there is some connection between beliefs, intentions, and asserting, we feel that it should not be necessary to inquire into someone's secret motives in order to know whether he is asserting anything; rather, we tend to think that an utterance is an assertion if it has a certain kind ofform.

On the other hand, the difference between an assertion and other utterances is not purely a matter ofform, since we could easily imagine a convention by which we could use the form of questions to make assertions. Thus, while it is not a matter of any particular form, we can say that if there is to be a community of people who consistently under­stand one another, there must be some difference of form in their language between the various kinds of utterances that exist in the language. We can then say that assertions constitute that class of utterances which people ordinarily use when they intend to influence the beliefs of their listeners in factual matters. The consequence is that we do not have to know the individual speaker's motives in order to know if he is asserting anything, but we do have to know the uses of certain grammatical forms within the language community, and this will amount in the end to knowing the intentions of the majority of the community. Our devious diplomat is able to operate effectively only because he utters sentences which are used more directly and sincerely by most people.

In conclusion, understanding that an assertion is an assertion is a very different matter from understanding the content of the assertion.

Page 97: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

UNDERSTANDING AND SYNONYMY 87

Assertions, questions, and preliminaries do not ordinarily say of them­selves that they are assertions, questions, etc. When we identify a sen­tence correctly as an assertion it is perhaps better to say that we know how it would ordinarily be intended as opposed to saying that we thereby partially understand the sentence itself. All this, of course, presupposes an analysis of belief which does not presuppose the notion of assertion­hood, and I will try to provide this analysis in a later chapter.

Individual words or groups of words not comprising a sentence or noun clause must be comprehended in terms of the parts they play in forming these entities. It must be the case that such a word or group of words comprehended or understood in itself is being used as an abbre­viation for a sentence or phrase expressing a preliminary. We can also speak ofa derivative kind of understanding and comprehension attach­ed to words we know how to use in larger expressions. Here again, the range of comprehension will be a matter of degree, depending on how widely one can make use of the word. Some words, such as com­mon nouns, might be taken to be abbreviations for preliminaries and sentences, while others, such as prepositions, conjunctions, etc., cannot be converted so easily, but can serve various purposes in larger units of discourse.

Having talked about comprehension and understanding, we now come to meaning. Rather than attempting to pick out any entity which is supposed to be the meaning of a preliminary or statement, we will try to indicate how questions about meaning can be answered, except, of course, the question "What is meaning?" In fact, our version of the image theory is completely independent of the view that images, or any other entities, constitute the meaning of sentences.

The first important question is "Does x have any meaning?" where x is any statement or preliminary. The answer will in general be "if and only if there is anyone who comprehends x." We must recognize, how­ever, that in most contexts this question amounts to "Does x have a generally recognized meaning?" The answer will then be "If and only if most people in the group concerned comprehend x and comprehend it in roughly the same way."

The question "Do x and y mean the same thing to the same person?" is to be answered in a different way. The answer will be unqualifiedly affirmative only where the same images and sensations would be counted as the same kinds of instances of each; otherwise, there will be less than complete synonymity, or perhaps none at all. Again, two sentence types as used by two different people will have exactly the

Page 98: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

88 FOUNDATIONS OF SOLIPSISM

same meaning if and only if the two people would count the same sorts of images and sensations as the same kinds of instances of the two sentence types. (In my discussion of the problem of other minds I will give an analysis of what it means for two people to count the same sorts of images and sensations as instances.)

Thus, we are now furnished with a criterion for synonymity, which is something any general theory of meaning must provide. If we must define "meaning," we might do it in the way that the integers are de­fined; thus a particular meaning would be the class of all sentence tokens that have the same meaning.

The last and most important question, at least in practice, is simply "What does x mean?" This amounts to a request to be taught the meaning of the sentence, and it can be satisfied in two main ways. The simplest method is to give the questioner a sentence which is synonymous with the given sentence, and which he is known to under­stand. A more usual variant of this is to define one or more key words in the sentence; the questioner is then able to construct a sentence which he understands, and which is synonymous with x.

The other method would consist in either giving an ostensive defini­tion, or in letting the questioner see how the sentence is used within the language community. This creates in the learner a tendency to count images which correspond to sensations characteristic of that object or situation as instances of the preliminary corresponding to the sentence. Since verbal definition must end somewhere, the latter sort of teaching is more basic. Its acknowledged purpose is to bring the learner to recognize phenomenal situations to which the linguistic ex­pression is applicable. Since sensations and images have so much in common, this teaching will determine his habits for images as well as sensations.

Until now we have been dealing with relatively simple preliminaries. It is important, therefore, to indicate in what way the analysis stated in the last section will be applied to more complex preliminaries. To this end we must distinguish three main uses of analytic statements.

First, we have those statements which are analytically true by arbi­trary convention. I may define "bwack" to mean "three-legged bear"; this having been done, the statement "a bwack is a three-legged bear" is analytically true, in the same context at least. Statements of this sort are not backed by anything except a rather arbitrary convention which anyone may lay down at any time, and they do not occur very often outside formal systems.

Page 99: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

UNDERSTANDING AND SYNONYMY 89

An example of the second use of analytic statements would be "a sibling is a person who is either a brother or a sister." This too rests on a convention, but it is rather different from the previous statement. For one thing, we immediately recognize the latter statement to be analytically true whereas we do not know what to make of the former until supplied with a definition. The convention on which the second depends is ingrained and well-established in a way that the other con­vention is not, and for this reason the second statement does not strike us as being arbitrary in the same way.

There is still a third sort of case which can be expressed by the same sorts of sentences. It is conceivable that a person might not have learned "sibling" as a defined term; he might instead have learned it in exactly the same way that he learned "brother" and "sister," per­haps by means of ostensive definitions. He might then discover that he meant exactly the same thing by "being a sibling" and "being a brother or sister." Under circumstances such as these, "A sibling is a person who is either a sister or a brother" might be thought to be an empirical statement by which this person would mean "everything that I call a sister or brother always turns out to be what I also call a sibling." But it would probably mean more than this. The person concerned might realize that it is not accidentally true, but that it is a consequence of his teaching. He might see that any material object that is an intance of brotherhood or sisterhood must also be an instance of siblinghood, and not because the one has been defined in terms of the other, but rather because he has learned that the two preliminaries independently mean the same thing.

In this latter case we would want to say that statements such as this, which represent discoveries about one's own meaning are true because any image which is an instance of the one preliminary is the same sort of instance of the other. Such statements are usually regarded as analytic; but on the basis of the image theory of meaning, it can be argued that they are empirical statements about the kinds of images which are in fact associated with given preliminaries under the proper conditions. This is to say that according to the above account it is practically (though not theoretically) impossible to pin down suffi­ciently the meaning of terms which have come to have their meanings independently of one another in order to know whether a supposedly analytic statement of the third type is even true, much less analytically true. That is, we would not be able to say whether one of the terms could be defined in terms of the other without making more specific

Page 100: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

90 FOUNDATIONS OF SOLIPSISM

the meaning of the terms. Thus it may be more profitable in the minor­ity of cases where the key terms were learned independently of one an­other to regard analytic statements in ordinary language as empirical hypotheses which, if true, represent psychological laws.

Previously we have laid down a general criterion for synonymity of meaning in terms of images. In the case of any genuine analytic state­ment, furthermore, we can find one or more synonymities on which the truth of that statement depends. This does not mean that we can always distinguish a subject and a predicate which are synonymous, but simply that any genuine analytic statement will ultimately rest on verbal de­finitions as opposed to other sorts of teaching. In fact, we decided that the third kind of supposed analytic statements, of which this is not true, are not really analytic at all.

From these two facts it might then be inferred that the truth or falsity of any analytic statement can be determined solely by imagining in­stances. But there is a difficulty: it is only in the third case that we can have any hope of deciding whether a statement is true solely by this kind of introspection, and without making use of any verbal definitions. In the first case it is likely that no images would be initially accepted as instances of "bwack." But if we accept the definition and suffer the word to enter our vocabulary, we are then initiating a habit of count­ing as instances of it whatever images we had previously counted as instances of the definiendum. That is, upon encountering "bwack," I will tell myself that it means the same as "three-legged bear," and refer back to the meaning of the latter; as a result I then accept as an in­stance of "bwack" any image or sensation I had previously accepted as an instance of "three-legged bear." I do not discover the fact that the two terms have the same instances, and hence the same meaning. Having already decided to give them the same meaning, I then count the same images and sensations as instances of both. Thus this is a kind of contrived meaning and our criterion for synonymity still holds.

Similarly, much the same difficulty arises with the second kind of analytic statements, since we do not usually have two sets of images to compare in this case either. In the second and most common use of analytic statements we have acquired habits of going from one ex­pression to another. These habits do not always involve single words, but often whole sentences, although the two habits will be closely related. When this happens it is generally the case that one sentence is on a lower level oflanguage than the other. That is, one sentence has been learned primarily by verbal definition while the other has not

Page 101: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

UNDERSTANDING AND SYNONYMY 91

been learned in this way; at least, if both have been learned by verba definition, one usually involves more complex definitions, and the more complex expression is often partially defined in terms of the less complex expression. An example might be the sentences "An elephant is walking" and "A non-extinct pachyderm is perambulating"; in this case we go back and forth between the sentence which might have been learned by ostension and the sentence whose terms were probably learned by verbal definition, and we consequently view them as synonymous. We have in effect, agreed to accept the verbal definition and we consequently give the second sentence the same meaning that the first one already has; in short, the meaning of the second sentence is contrived. Thus, this habit of going from one sentence to another in fact and imagination without any hesitation gives rise to the analytic state­ment, "Whenever an elephant walks a pachyderm perambulates." We have in these cases acquired the habit of identifying the meaning of the two sentences, and the analytic statement itself reinforces this habit. It also sets forth a rule which the speaker can be presumed to follow, and which he would ordinarily be recommending to his lis­teners.

All this could be true on any account of meaning. On the theory stated here it will mean simply that we have acquired the habit of counting the same images and sensations for the same sorts of instances of the preliminary associated with the second sentence as we already counted for instances of the preliminary associated with the first. If anyone questions an analytic statement of this sort, we appeal to certain definitions, which amount to other simpler analytic statements. If the questioner admits the truth of these latter analytic statements, he is admitting that he already abides by linguistic rules which will insure the truth of the first statement under all possible circumstances and hence will make it analytic.

We must remember in all this that synonymy between sentences is a matter of fact: whether the sentences are synonymous depends upon the person's psychological dispositions to recognize instances of them. Thus it follows that any so-called "analytic" sentence is a factual one at bottom. But if an analytic statement of the usual sort purports to describe our linguistic behavior and we always make our behavior conform to it by design, then it has a kind of universality which other factual statements do not have and we are justified in distinguishing it from the others. Nevertheless, the solipsist can claim at this point to have done away with "analytically true" as an irreducible category

Page 102: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

92 FOUNDATIONS OF SOLIPSISM

and made it a matter of psychological fact. The same treatment could easily be given the other logical modalities.

Despite this, there still remain genuine differences between the first and second classes of "analytic" statements on one side and the third class on the other hand. Similarly, there is an important difference between this third class of statements and ordinary factual statements even though all these statements are ultimately factual ones. Although a statement of the first or second classes refers to a psychological habit that we have, it is always a habit that we are aware of having, and one which we are reinforcing by the very utterance of the sentence. This is different from the third usage where we simply discover that we have a habit which we have not set out to have. In order to see how the statements of all three classes differ from ordinary factual statements, let us recall what the traditional analytic-synthetic distinction is supposed to be. An analytic statement is one which does not assert any­thing about the world and whose truth or falsity is not affected by anything that might happen while a synthetic statement is not in­dependent in this way. On the traditional distinction all these supposed analytic statements are synthetic because any such statement which is true would be false if we had different habits of counting images and sensations as instances. On the other hand, all these statements are un­affected by anything which might happen in the outside world and which did not involve a very narrow range of psychological dispositions. In other words, the truth of these supposed analytic statements is un­affected by anything which happens outside our own heads. In prac­tice, this makes even analytic statements of the third class very different from ordinary factual statements. Thus, the traditional analytic­synthetic distinction does not stand, but we can make a corresponding distinction which would divide statements up into much the same two camps. This turns out to be a very useful distinction and we will there­fore continue to talk of " analytic statements." But it must be remember­ed that these statements are factual in a very special way.

There are certain obvious similarities in the above account to Quine's views on the analytic-synthetic distinction.1 With respect to the third class of "analytic" statements I would agree with most of Quine'S arguments and follow him in his questioning of the traditional analytic­synthetic distinction. Quine does except explicitly one class of state­ments which he calls logical statements from his general critique of

1 W. v. Quine, "Two Dogmas of Empiricism," reprinted in From a Logical Point of View, pp.20-46.

Page 103: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

UNDERSTANDING AND SYNONYMY 93

analytic statements; these are statements, such as truth table tautolo­gies, which remain true on any interpretation of the non-logical terms. He also excepts another class of statements implicitly. He remarks that there are some cases of genuine synonymy which rest directly on abbre­viations, particularly in formal systems. Quine seems to be thinking of such abbreviations as "p == q" for "(p::J q). (q::J p)" where we merely introduce a new notation as a sort of shorthand. I take it that Quine would be willing to make the same exception for such cases when they occur outside oflogic and mathematics. An example would be my abbreviation of "bwack" for "three-legged bear." Since these genuine synonymies will give rise to analytic statements which differ in kind rather than degree from synthetic statements, (although they are still factual) it seems that Quine does not intend his doctrine to apply to what I have called the first class of analytic statements. We can look, then, to the second class, where I diverge somewhat.

In general Quine does not think that definition other than notational abbreviation provides an adequate foundation for synonymy. In support of this claim he points out that we cannot really settle the question as to whether two terms are synonymous by going to the dictionary. The lexicographer is simply one who examines past usage of the terms in an empirical manner and tries to capture as much of it as possible in his definitions. He does not make two terms synonymous by defining one by means of the other; on the contrary, his definition, if correct, will only be a consequence of a previously existing synonymy. Quine and White1 further argue, I think correctly, that if we found the analytic-synthetic distinction on any distinction between different types oflinguistic behavior, the distinction will be one of degree rather than kind.

I think, however, that Quine and White have overlooked an impor­tant point in all this. In the case of a term like "sibling," most of us cannot say what it means to identify siblings without at least being inclined to say to ourselves something such as "It means the same as ------, so this must be one." We might say that it does not have meaning in its own right, but has meaning only because it goes proxy for some other usage. In this respect it is like a notational abbreviation, but it really is not one as it does not depend on any particular person's definition and, by this time, it has a fairly secure place in our language. Further, we cannot always interchange the ex-

1 Quine, op. cit., pp. 24-25 and M. White, "The Analytic and the Synthetic: An Unten­able Dualism," reprinted in Linsky, Semantics and the Philosophy of Language, pp. 272-286.

Page 104: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

94 FOUNDATIONS OF SOLIPSISM

pressions "male sibling" and "brother" without affecting the truth value of sentences in which they occur, even in an extensional language ; this is because of what Quine calls the referential opacity of quotation, belief, etc. Thus, it would be very difficult to find a general definition of "sibling" which would satisfy Quine's rigorous criteria in the way that "p == q ~ (p :> q) . (q :> p)" does in a particular syste ffi. It is nevertheless important to notice that even the latter definition holds only within a particular system or systems, and that the "three bars" may be defined in other ways in other systems. Similarly, we might say that "sibling" may be genuinely and completely synonymous with some other expression in a particular context in that, in that context, we may be willing to mentally substitute for it an expression which we take to be equivalent and which does not mention the word "sibling." The trouble occurs when we try to find another expression which can be substituted for the given one in arry context. Quine's as­sumption seems to be that a genuine synonymy is one that always holds independently of the context, and he is then forced to the conclusion that there are no such synonymies aside from the special cases men­tioned above. On the view of meaning I have taken, however, we can look for instances either in a particular context or in the abstract, and we can also speak of synonymy within a context. Thus it seems that there can be genuine synonymy, at least within certain contextual limitations, and we are then led to conclude that analytic statements (on our redefinition of the term) of the second class differ in kind from ordinary synthetic statements.

We must now examine the relation of contexts to meaning in more detail, both in order to make our treatment of analytic statements clearer and in order to deal with more complex sentences than the ones we have considered up to now.

A good example to begin with is the following. Suppose that two men, Psmith and Mjones, are at a political rally. One of the candidates says that he believes Ohio does not need any additional taxes and that, if elected, he \\ ill oppose such taxes. At this point Psmith says to Mjones "Politicians always lie about taxes and things like that." It is conceiv­able that the listener might deduce "This candidate is lying" from Psmith's utterance together with the suppressed premise "This candi­date is a politician." However, this is in fact unlikely to happen. The listener does not go through any deduction, but simply construes Psmith as saying that this politician is lying. Psmith is also saying that the practice is widespread, and he is perhaps also hinting that the politician

Page 105: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

UNTERSTANDING AND SYNONYMY 95

in question should be partly excused for his deliberate falsehood. Psmith probably does not mean to assert that the practice is absolutely uni­versal, and he would not ordinarily be construed as asserting this in the context. In this case, however, the deduction could have been made if we stick to the form of the original utterance strictly and add an obvious premise.

Let us now suppose that Psmith had said instead "Politicians some­times have to lie about taxes, or they would never be elected." In the context this sentence also says that the politician is lying, but it goes somewhat further in pardoning him and giving extenuating circum­stances. But in this case the corresponding enthymatic deduction could not have been made because of the inclusion of the word "sometimes." Suppose now that the candidate is elected and it turns out that he really will go to any lengths to prevent new taxes from being levied. Mjones might now remark to Psmith, "You were wrong about that politician; he wasn't lying after all." Psmith might now reply, "I never said that he was lying; I just said that politicians sometimes lie about that sort of thing." Mjones would at this point justifiably consider that Psmith is just trying to back out of his original statement; if Psmith really meant what he now says that he meant, he should have made some qualification. In this particular context the general statement and the particular statement mean much the same thing, and the statement intended could not be strictly deduced from either in itself. In fact, cases without number can be found where the literal meaning of our words have implications far beyond or short of what we really mean to assert. Moreover, it is very difficult and inconvenient to use lan­guage which literally never implies more or less than we mean to assert; but we are seldom misunderstood because of this, since in inter­preting us people always take into account the context in which our sentences are uttered.

One can trace the literal meaning of sentences as we have done above, but the so-called literal meaning has no priority over any other possible meanings the sentence might have had. Actually, the literal meaning is seldom the intended meaning, since we generally let the context speak for us to some extent, and are correspondingly careless about the exact choice of our words.

The upshot of all this is that some sentences tend to reduce to others when uttered or written in a given context in two general ways. First, if the sentence contains such defined terms in our language as "soluble," "intelligent," "bachelor," etc., it will be analytically equivalent to

Page 106: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

96 FOUNDATIONS OF SOLIPSISM

sentences not containing these words. These latter sentences will often be longer and will generally contain truth-functional connectives, and, in some cases, non-truth-functional connectives. They may be simpler, however, from the point of view of analysis. Second, and even more important, some sentences are often merely interpreted as others according to the context, without our going through any definitions. Genuine analytic sentences can be taken to express rules which allow us to go from certain sentences to certain other sentences in a wide variety of contexts, or in the case of analytic sentences of the first class, in a precisely specified range of contexts.

There are many more linguistic rules of the same kind which do not apply to as many contexts, but only to a relatively small number. These usually apply when we interpret a statement in such a way as to take account of only a few of its literal implications. Even in such a simple case as "there is a cat" it does not occur to one that the speaker is asserting the presence of certain feline anatomical niceties, and in ordinary contexts this is not being asserted.

There is a similarity between this account of meaning which takes into account the context in which any sentence may be uttered and Nowell-Smith's doctrine of contextual implication. Instead of saying that sentences change their meaning from one context to another, he says that a sentence retains its meaning but implies different things in different contexts. He defines this contextual implications as follows: "I shall say that a statement, p, contextually implies a statement, q, if anyone who knew the normal conventions of the language would be entitled to infer q from p in the context in which they occur. Logical im­plications are a sub-class of contextual implication, since if p logically implies q, we are entitled to infer q from p in any context whatever."l Thus logical implication would express a general rule of the language, while contextual implication would express a special rule.

The above passage, in fact, seems to suggest that this is the only im­portant difference between the two: that the contexts to which the one applies are more extended than those to which the other applies. How­ever, in the case of logical implication we can say in exactly what contexts q may be inferred from p; namely, in all contexts within the system or in all substitution instances of the rule. Thus we are entitled to make the inference in a particular context because we know that the rule holds for all contexts including that one. In the case of contextual implication, on the other hand, it is not clear that we can do this.

P. H. Nowell-Smith, Ethics, pp. 80--81.

Page 107: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

UNDERSTANDING AND SYNONYMY 97

Nowell-Smith is able to give examples only of particular inferences by contextual implication, and he is not able to say in exactly what con­texts a given case of contextual implication would hold. Thus we can­not form a hypothetical analytic statement of the first or second class from the contextual implication by specifying in the antecedent the range of contexts for which it holds. Logical implication, on the other hand, always gives rise to a hypothetical analytic statement. This is not to say that these inferences by contextual implication may not be justified in particular cases, but rather that there is a basic difference between logical implication and contextual implication; thus a logical implication cannot be regarded as a conjunction of contextual im­plications.

There is a basic difference between Nowell-Smith's contextual im­plication, which is a connective between two statements that keep the same basic meaning in all contexts, and the account I have given, which involves logical implication within a context. In the former, the con­stant meaning of the statement p is such that certain other statements can be inferred from it in certain contexts. In the latter case the sen­tence p changes its meaning from context to context and, given a cer­tain meaning for the sentence at some one time, other statements are deducible from it in the usual way according to logical rather than contextual implication.

So far we have pointed out what conditions have to be fulfilled in order to make various kinds of analytic and semi-analytic statements true, but this does not tell us what it is to understand an analytic state­ment as analytic. There is a very important difference between the understanding of an analytic statement, even in a particular context in which it holds unqualifiedly, and the understanding of an ordinary synthetic statement. In order to understand the former it is necessary that we know whether it is true or false. Thus, if someone says that he understands "All husbands have wives" in a context where "husband" operates as an abbreviation for "man having a wife," and it later turns out that he thinks the statement is false, we then conclude that he did not understand the statement at all. With synthetic statements, on the other hand, it is always possible to understand a statement and still be mistaken about its truth value.

With comprehension there is a different situation; one may com­prehend "All husbands have wives" in the same context and still think that it is false; one may comprehend it in the way that most people comprehend "Some husbands do not have wives," or one may com-

Page 108: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

98 FOUNDATIONS OF SOLIPSISM

prehend it in the way that most people comprehend certain synthetic statements. One problem, then, is the distinction between compre­hending a statement analytically and comprehending it synthetically.

The main distinction between analytic and synthetic comprehension is that in the former case we recognize the tautologous or contradictory nature of the sentence. More accurately, in order to comprehend a statement as analytic we must, when looking for or imagining instances with the context in mind, either realize that the negation of the state­ment can be true of none if it is analytically true, or that the statement itself is true of none if it is analytically false. In the case of an analytic statement of the first or second classes, we would not even try to ima­gine cases of which its negation would be true, since we would realize that our own conventions prevent there being such cases. The analytic statement is still factual because, in this case, it says of oneself that one abides by certain conventions as a matter of fact although one could imagine abiding by a different set of conventions for the recognition of instances of the associated preliminary. In the case of an analytic statement of the third class, we would begin to try to imagine such in­stances and would simply not find any. We would not comprehend the sentence as analytic until we realized the pointlessness of looking for such instances.

The comprehending of a statement as analytic seems to be quite a sophisticated process which often does not come to us when the sen­tence is first encountered. The child probably treats the statement "All husbands have wives" as being roughly equivalent in meaning to "Anyone who is called a husband has a wife." The latter is clearly a synthetic statement. At this stage the question of whether it is logically possible for there to be a husband without a wife never arises, and if it did arise the child would probably not be able to answer it. Until our habits of accepting instances become more determinate, and we come to realize that certain preliminaries have no instances, even imagined ones, no statements are comprehended as being analytic.

Of course, there is more to comprehending an analytic statement than recognizing its analyticity. When we comprehend an analytic statement, and comprehend it as analytic, we must be able to say why it is analytic; this amounts to knowing that we comprehend certain words which appear in the sentence according to certain conventions by which we abide. We have analyzed the understanding and com­prehension of individual words in terms of the understanding and comprehension of statements in which the words appear. Thus, the

Page 109: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

UNDERSTANDING AND SYNONYMY 99

comprehension of an analytic statement as analytic will entail the comprehension of statements which truly describe certain linguistic conventions by which the person abides in that context at that time. Since comprehending a statement or preliminary as analytic involves knowing why it is analytic in terms of our own linguistic conventions, it is here that analytic statements differ from one another in meaning. Insofar as they refer to conventions to which the speaker commits him­self they are factual and, since the conventions they refer to differ, they are not all synonymous. Since comprehending statements as analytic implies knowing why they are analytic as part of the process of com­prehension, most pairs of analytic statements will differ in meaning.

We must now look at the ways in which logical concepts might arise in a language. Many terms will be ostensively defined or taught which are implicitly conjunctive. An example would be the phrase, "red barn," since something must be both red and a barn to qualify. Such a term might nevertheless be taught not by defining it in terms of "red," "barn" and "and" but directly in the way that the predicates "red" and "barn" are taught. Someone who had so learned these words would now be familiar with an example in terms of which one could now begin to teach him the word "and." That is, one could encourage him to use the sentences, "That is a red barn" and "That is red and it is a barn" interchangeably. It is presumably by multiplying examples of this kind that we learn our truth functional concepts. This teaching would have the effect of encouraging someone to take as an instance ofa preliminary of the form "p and q" only things which are instances of both p and q. Once these habits have been established we uni­versalize them by committing ourselves to the appropriate analytic statements. Once we have "and" and "not" or "not" and "or" we can use analytic statements to define all the other truth functions.

It should also be remembered that universal quantification can be reduced to conjunction, and that existential quantification can be reduced to disjunction, without any problems as long as the universe of discourse is finite. We will later adopt a finitist form of phenomen­alism and our analysis of other minds will also be finitist. Hence, any statement about the world or about minds will involve quantification over limited domains and the range of the variables will always be finite. The kinds of quantification that will be needed, therefore, will also be truth functional and can be introduced by definition once we have "and" or "or" and "not." We are not saying that we can reduce all statements to truth functional statements. But at least there will

Page 110: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

100 FPUNDATIONS OF SOLIPSISM

be no difficulty in introducing "all" and "some" in addition to the overtly truth functional connectives.

We have now seen how logically complex statements can be built up definitionally out of logically simple statements. Since nothing really "new" has been introduced we can apply the same analysis of comprehension and understanding to these logically complex state­ments. However, as far as synonymy is concerned it will be more convenient to limit the account given above to statements which are not truth functionally complex. We can then say that logically com­plex statements are synonymous if they have isomorphic logical structures and if there is a one-to-one correspondence between ulti­mately simple components so that they are synonymous each to each. Further, we will allow logically complex statements to be equivalent if they can be put into that relationship with the use of truth functional equivalences.

Page 111: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 6

VERIFIABILITY

The image theory having heen stated we will now compare it to the verifiability theory of meaning as stated by Ayer in Language, Truth, and Logic. l In the first place there is a basic ambiguity in Ayer's state­ment of the theory. It can be viewed as a general method for deciding what a statement means, if anything, or it can be viewed as a criterion for deciding whether a statement has any meaning on the assumption that its intended meaning is perfectly clear. In this book Ayer in fact proceeds in the second way. When dealing with metaphysical and theological statements he assumes that their intended meaning is such as to exclude any possibility of their being verified.2 According to the verifiability theory it then follows that if the intended meaning is the only meaning they have, then they have no cognitive meaning at all. However, the people who assert these statements often do take certain facts about the world to be evidence for their assertions; thus on the verifiability theory these metaphysical and theological statements will have a meaning even though it is quite different from the intended meaning in that those who make these assertions are in fact only making empirical assertions about the physical world. Thus to be plausible the verifiability theory must be construed not just as a crite­rion of meaning but as a general method of determining the meaning of any utterance. If it is treated solely as a criterion, theological state­ments which can be verified in part empirically will be meaningful and will not be reduced to empirical statements. It should be noted that the image theory stated above is both these things; it first gives general rules for determining the meaning of any expression and the expression has meaning if and only if a meaning is found in this way.

Another difference between our version of the image theory and some versions of the verifiability theory is that these theories analyze the notion of being meaningful in general whereas we have considered chiefly the notion of being meaningful to an individual, that is, an

1 A. J. Ayer, Language, Truth and Logic, particularly pp. 33-45. • Ibid., see particularly p. 41.

Page 112: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

102 FOUNDATIONS OF SOLIPSISM

individual's understanding a sentence. One cannot very well speak of an expression's being meaningful in general unless one supposes that the expression is intended by everyone to mean the same thing, and this is an assumption which we prefer not to make; thus we always speak of what an expression means to those who comprehend it in­stead. Nevertheless it is easy to imagine what analysis of comprehension would be given by the verifiability theory. To comprehend a sentence would be simply to know how to verify it, and in case it is comprehend­ed, its meaning would be determined by the way in which it would be verified. Then Ayer and others would be in a position to say substan­tially what they want to say about meaning by asserting that if meta­physical and theological statements are to be comprehended at all, they must be comprehended in such a way as to make them synonym­ous with certain empirical statements.

According to the verifiability theory then, comprehension would amount to knowing how to verify statements. Let us now examine the logical relationships between this kind of comprehension and com­prehension as viewed by the image theory. In the first place it is clear that if one is able to imagine a partial instance of the preliminary corresponding to a given sentence and recognize complete instances of it, whether they be images or sense experiences, one would be able to pick out these instances in any process of verification. That is, if one who comprehended a sentence in the image sense were to witness what Ayer would admit as a verification of the sentence, this person could accept some entity involved in the verification as an instance, or more accurately, he would not have a feeling of rejection in connection with some of his sense experiences of the process of verification, provided that he had adopted the attitude of looking for an instance when he set out to observe this process. If the verification is conclusive then a complete instance would be found; if it turns out to conclusively refute the original statement, no instance will be found; in other cases a partial instance will be found. This point can be made more clearly simply by noting that in the case of a preliminary of any complexity imagining an instance of that preliminary will amount to imagining what the process of verification would be for the corresponding sentence. Sup­pose that I want to imagine an instance of "its being the case that there is a chair in the next room" ; in order to do this I will have to imagine going into the next room and seeing and touching a chair. This is what Ayer would count as a verification of "there is a chair in the next room." If I accept this compound image as a complete instance of the pre-

Page 113: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

VERIFIABILITY 103

liminary then the actual process of verification as opposed to the imagined process will be what Ayer calls a "strong" (conclusive) veri­fication of the corresponding statement. If the image is only accepted as a partial instance, then the verification will be a weak one. Thus we can, if we wish, reformulate the image theory stated previously in terms of imagining verifications instead of imagining instances without changing it in any way. Thus the requirement of minimal compre­hension would be, roughly, the ability to imagine weak verifications plus the ability to recognize strong verifications.

We must now clarify one point which Ayer complicates unnecessa­rily. He distinguishes on the one hand strong and weak verifications.1

This distinction is quite clear; a verification is strong if having been completed it is then either logically necessary on the basis of the veri­fication that the original statement is false or logically necessary in the same way that it is true, and if neither of these is the case, the verifica­tion is weak. The other distinction is between verifiability in principle and verifiability in practice2 ; this distinction is also quite clear and needs no explanation. However, Ayer says that a statement is cogni­tively significant if and only if it is verifiable weakly in principle, and he would presumably say in our language that a person comprehends a given statement ifand only ifhe knows how to (or perhaps, is able to) verify it weakly in principle. It is obvious that the second qualification is needed in the statement of the verifiability criterion; if all meaning­ful statements had to be verifiable in practice, much of science would turn out to be nonsense. This qualification is built into the image theory and does not have to be made separately because we can easily imagine verifications which could never in practice be made. But Ayer also thinks that the other qualification of weak verification is needed because of universal statements and statements about the past. 3 With respect to the former, he argues, contrary to the view we will take, that statements such as "arsenic is poisonous" are universal and have an infinite application. This being the case, they can never be conclusively verified in practice, and Ayer then seems to assume that they can never be verified conclusively even in principle. Here he seems to be assuming a very narrow sense of "in principle" because there is certainly no contradiction in the notion of an infinitely long verification, and such a process is hence logically possible. If Ayer does not mean "logically

1 Ibid., p. 37. 2 Ibid., p. 36. 3 Ibid., p. 37

Page 114: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

104 FOUNDATIONS OF SOLIPSISM

possible" by the phrase "in principle," it is hard to think what he could mean by it. The next question is whether statements about the past are conclusively verifiable in principle. Ayer thinks that they are not, presumably because it would involve going back in time, which is impossible. At this point we must make a distinction between going back in time and being in the past.1 I would admit that the former does not literally make sense. If one tried to imagine going back in time, one would imagine instead a world which kept going forward in time, but in which the buildings, fashions, and everything else gradually changed and became like the objects of the past. This would be simply to imagine a reactionary world without imagining a reversal of time. On the other hand we can imagine having been born and living in a past time without having to imagine any reversal of the time series. It is an empirical fact that one is not able to get up to the Rubicon in time to see Caesar cross it just as it is an empirical fact that one does not have time to get to the bank before it closes. Thus there seems to be no need to incorporate the qualification of weak verification in the statement of the verifiability criterion. In our statement of the image theory we require that a person who comprehends a statement at least recognize complete instances (conclusive verifications) whether ima­gined or real: more accurately, the meaning of a statement which he comprehends is determined by the conclusive verifications which he would recognize and cannot go beyond them. The making of this sort of qualification seems to weaken a theory of meaning because it suggests that there is something more to the meaning of a statement, which cannot be conclusively verified even in principle, that is not taken account of by the theory.

Thus on any analysis of this sort, comprehension and understanding as well as the meaning of statements will be ultimately reducible to the notion of ability to verify or knowing how to verify. The difficulty is that these notions are themselves left unanalyzed. There is, of course, a point where any analysis has to stop and leave us with unexplicated primitives, but we can at least require of an analysis that the unanalyz­ed notions it leaves us with be already reasonably clear. But the notion of knowing how to verify a statement is subject to the very same problems as was understanding in the first place. We can raise the same questions about the former that we discussed previously in connection with the latter: "Can 'knowing how to verify' be reduced to the actual

1 For a discussion of these qualifications see C. G. Hempel, "Problems and Changes in the Empiricist Criterion of Meaning" in Linsky, op. cit., pp. 163-185.

Page 115: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

VERIFIABILITY 105

and possible behavior of the subject?"; "Does it amount to the ability to recognize verifications when encountered?"; etc. It is true that it eliminates a few possibilities. We could not hold that understanding a statement involves an intuitive grasp of a non-natural fact and the verifiability theory at the same time. But we must nevertheless require that a theory of understanding go much further than the verifiability theory.

There is an interesting parallel between the verifiability theory and the newer dictum that understanding an expression consists in knowing the rules for its proper use. Knowing rules in this sense is ambiguous in that same way that knowing how to verify is in the verifiability theory. It should be pointed out that the former is sometimes coupled with a behaviorism which clarifies it, but more often it is simply left unanalyzed. Thus the image theory stated in the last chapter can be looked upon as an attempt to clarify either or both "knowing the rules for the use of an expression" or "knowing how to verify an expression." It is apparent that some philosophers in their attempt to avoid the psychological side of Bume have in so doing left unanalyzed some key notions which are at least dealt with by Bume, and which ought to be dealt with in some way or other. Whatever one may want to say about our analytic solipsist, one cannot accuse him of avoiding these problems.

Let us now ask just what a criterion of meaning is supposed to do and how it can be justified. The purpose of such a criterion is often supposed to be the separation of sense from nonsense. We must then ask what it means to say that an expression is nonsensical. This might be answered by giving some examples of sense and nonsense: "red," "weiss," and "verde" would be examples of sense while "abqj" would be an example of nonsense. One distinction here is that some people claim to comprehend the first three while no one claims to comprehend "abqj." But a criterion of meaning is meant to do more than make this distinction, and we must take a more complex example. "This is red" would be an example of sense according to any criterion except, per­haps, that of Par men ides; on the other hand "God loves his creatures," understood in the traditional way, would be an example of nonsense on some criteria. On the verifiability criterion, for instance, anyone who made this statement and then consistently rejected all verifiable interpretations of it would be declared to be talking nonsense. It may seem somewhat paradoxical to say of a statement, as the positivist some­times does, "If that is what you mean, then you are talking nonsense," since the antecedent suggests that something is meant and the conclu-

Page 116: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

106 FOUNDATIONS OF SOLIPSISM

sion denies it. However, there is no contradiction here; in a case such as this the "meaning" is not made clear by producing it, but rather by rejecting all possible candidates, and the process is thus one of deciding that the statement has no meaning because none can be found. But, be this as it may, there is a difference in the two pairs of examples since in the latter pair the criterion of meaning is in the position of saying that someone who claims to comprehend a statement cannot really comprehend it. The person who makes the claim of comprehension will then be justified in asking for ajustification of the criterion, and we will now see what can be done here.

There have been three main types of justification for the verifiability criterion, and much the same sorts of justification would be equally appropriate to the other criteria. The first would be simply that it is intuitively obvious, in this case that we cannot talk sense about things of which we can have no knowledge. Though this still has some force, it has been weakened by the elaborate qualifications introduced into the criterion (the more complex it becomes, the less intuitively obvious) and particularly by the problem of other minds where we do seem to be talking sensibly about things we cannot know first-hand at least. Another sort of justification is really a normative one. This would be that our heads will be cleared and our language clearer if we limit our­selves to the making of statements which are verifiable. But it could be objected here that while this may be good policy for practical purposes, there are some things which are interesting and important which can be talked about only by going beyond this limitation; that is, the pro­gram can only be carried out in certain fields, and there is no reason to extend it to other fields.

Thus the only sort of justification which can reasonably be given of a criterion of meaning is an analytic one. We can justify any given crite­rion only by showing that it follows from a general analysis of under­standing and meaning. As we will see, the image theory may itself be looked upon as such a justification. But at this point it will be objected that our analysis has not been entirely true to ordinary usage of such terms as "understand" and "meaning"; instead, we have not attempted to analyze all uses of these terms, but only certain ones. We have in fact singled out uses of "understand" such that if a person understands a sentence, that sentence will be either true or false and in the former case some information will be conveyed. The image theory stated in the previous chapters is meant to apply only to this sense of "under­stand" and the corresponding sense of "meaning." But there are at

Page 117: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

VERIFIABILITY 107

least two other sorts of understanding with which we have not dealt. The various kinds of understanding can be treated as species belonging to a genus; some things can be said about understanding in general, but in the previous analysis we have gone beyond this and described one of the species. One is the understanding of imperatives, and another would be the understanding of statements which are not supposed to be about anything and which convey information only incidentally, but are rather part of a game. As an example of this latter kind, let us think of a game played by two contestants with the following rules. One person starts by uttering any sentence in English, whereupon the other must reply by uttering any other English sentence with one more word within two seconds; the first person must then produce another sen­tence with one more word, and so forth until someone gets stuck. There is an extended sense of understanding according to which one contestant might be said to understand a sentence uttered by the other if he knew the number of words in it, even if he did not understand it in the usual sense. Thus sentences can function in the same way as football plays, although this sort of meaning is not usually isolated in this way, but is mixed in with other sorts of meaning. The objection to the criterion of meaning will now be: "Why give one sort of under­standing and meaning a privileged status?; to call one sort of meaning sense and the others nonsense is just to make an ethical judgment of their relative importance since we generally use "understanding" and "meaning" to cover all three cases. This objection must be granted since it is the job of the analyst to distinguish and analyze the different sorts of meaning without evaluating them. However, there is something to be said on the other side. Statements which have the sort of meaning we have been primarily concerned with have several interesting and very useful features. In the first place they are true or false in an ob­jective way; if a statement is true for one person, it is true for all, and ifit is true at one time, it is true for all time (at least if it contains a time reference); secondly statements of this sort contain information which can be practical in that there is always the possibility that, if believed, the statement will, in conjunction with out desires, determine our voluntary action. Imperatives may also, of course, affect our action, but in a different way; if successful they either change our attitudes or cause us to act contrary to them while the other sorts of statements redirect our attitudes without changing them. Often when the terms "understand," "meaning," and even "true" are extended beyond their purely cognitive use the intent is to associate these features of cognitive

Page 118: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

108 FOUNDATIONS OF SOLIPSISM

meaning with some other sort of meaning. But the same sort of truth and falsehood and the same imparting of information just do not ac­company statements insofar as they are meaningful in the other two ways. Thus the distinction between sense and nonsense is useful insofar as it tells us not to mix up the different sorts of meanings. But having noted this we will not label any piece of discourse nonsensical unless no one claims to understand it in any sense, as is the case with "abqx." We can still set up criteria for any of the different kinds of meaning, and we will now investigate the exact relationship between the veri­fiability criterion of cognitive meaning and the image theory of cogni­tive meaning.

The criterion of cognitive meaning which is derived from the image theory will be that a statement has cognitive meaning to a given person if and only if he is able to produce an image which he will not reject as an instance of the preliminary corresponding to the statement. This is logically very closely related to the verifiability criterion for cognitive comprehension by a given person which we constructed earlier. This is so because it is always possible to have sense experiences which corre­spond qualitatively to any given set of images. It is not as easy as the inverse process, but it is at least theoretically possible to construct something in the physical world which will produce in the observer sense experiences corresponding to any given images. Thus the set of all possible physical worlds is just as rich qualitatively as the world of imagination, even though some of these possible worlds will be much more than hammer and nail constructions. We can even arrange to have sense experiences which have the same degree of determinacy as the images. One can produce more determinate sense experiences by constructing the object to be observed in more detail and one can generally arrange to have a less determinate sense experience by ob­serving the object at a greater distance, or by interfering with the sensory organs in some other way.

From all this it follows that if any statement has cognitive meaning for any person according to the image criterion, it will also have cogni­tive meaning for that person according to the verifiability criterion.

Page 119: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 7

OBJECTIONS

One of the chief objections to the image theory has always sprung from the fact that the world seems to transcend what we can imagine in various ways. Descartes pointed out that there are some things in the world such as chiliagons which cannot be imagined in a way that dis­tinguishes them from similar objects.1 Others have pointed out that our imaginations are less active in some directions than in others; it is easier to have visual, auditory, and tactual images than olfactory and gustatory ones. Thus it would seem to follow on an image theory that statements about smells and tastes cannot be understood as well as statements about objects which are encountered in other ways. Lastly, the real world seems to some to be much more complex than anything we could imagine. The objection then is that if we could only talk about the things we can imagine, there would be many objects in the world about which we would not be able to talk meaningfully, but that we in fact do talk meaningfully about these objects.

Let us now ask why the world seems to be more complex than any­thing we can imagine. This may be compared with the following ques­tion: "Does the sensation which we might have at any given time seem to be more complex (as opposed to being more determinate, vivid, etc.) than anything we can imagine at the present?" It is interesting to note that we tend to give different answers to these questions. With respect to the former, one is tempted to say that the insides of a rocket are completely beyond imagination at the present moment, but with respect to the latter question, one has the feeling that given proper directions one could imagine any sort of sensation with which one is reasonably familiar, at least if it is not a smell or taste. Certainly if one were to have a sense experience, one could then imagine it. And pre­sumably in the case of examining the insides of a rocket one would not come across any radically new sorts of sense experiences. Thus the insides of a rocket defy with their complexity any single sense expe­rience we might have just as they do any single image we might have.

1 R. Descartes, A Discourse on MetJwd and Selected Writings, p. 143.

Page 120: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

110 FOUNDATIONS OF SOLIPSISM

The difference is that we are used to having continuous sense experien­ces while we are at the same time used to having only momentary images. In order to get an idea of the insides of a rocket it would no doubt take days of poking around inside it, feeling wiring, pulling levers, etc. On the other hand most people seem never to go through any sort of sustained process in their imaginations. The obvious ex­planation is that it would take just as long to imagine poking around the rocket as it would to really poke through it, and no one has that much time to spend on something that will not give him any informa­tion. Thus we come to think of imagination as a process which is in­herently limited in time and lose sight of the fact that anyone, if he first arranged to be fed intravenously, could lie on his back for the next ten years and imagine them instead ofliving them. Thus, if it be grant­ed that we are capable of imagining any single sense experience we might have, the assertion that the real world is beyond imagination because of its complexity implies that it is also beyond sensation in some way.

But the objection can be put in a different way. It might be ad­mitted that it is theoretically possible to imagine any event which might take place, but this does not in itself imply that anyone who is capable of understanding statements about complex events actually knows how to imagine them. Even if I could imagine the insides of a rocket upon being given proper directions, I may not be given direc­tions and may come to a point where I do not know what to imagine next. Ordinarily when this is the situation we are said simply not to understand the statement concerned. If someone says "That steam locomotive is articulated" and I am unable to imagine an instance of an articulated locomotive, not because there is anything unimaginable about it, but because I do not know what to go about imagining, I do not understand the statement and may not even comprehend it. The question then is whether there are any complex cases in which a given person does not know how to imagine a certain object, but can still be said to understand statements about it. First let us note that not all statements about a complex object require that we be able to imagine it in the same degree of detail according to the image theory. The state­ment "The insides of a rocket are white" requires only that we be able to imagine looking inside a rocket and seeing white substances in order that we understand it; the statement "The insides of a rocket contain gyro-stabilizers" obviously requires much more to be understood, at least according to the image theory. However, it may be argued that

Page 121: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

OBJECTIONS III

there are some complex statements about complex objects which are understood by people who do not know how to imagine the complex­ities involved. An example might be "A hydrogen bomb of a given weight is more powerful than an atom bomb of the same weight." When we are taught the meaning of "more powerful" we are taught to construe such statements as equivalent to the statement "If an atom bomb and a hydrogen bomb of the same weight are both exploded, the hydrogen bomb will cause a greater area of destruction." This is clearly the sort of statement which nearly everyone is supposed to understand, but it is also true that almost no one in our society outside of the scientists who deal with these objects is able to imagine a hydro­gen bomb in such a way as to distinguish it from an atom bomb. The average person does not know what chemicals go into them, and even if he did know, he would probably not know how to distinguish them; consequently he cannot imagine the distinction. It also seems clear that in order to understand the statement at all, one must make a distinction between an atom bomb and a hydrogen bomb, but since the average man does not on the image theory make this distinction, it would seem to follow that he does not really understand the state­ment. This conclusion seems partly false, but not entirely so. We cer­tainly want to say on any theory that the average man understands such a statement as made by another average man, but it also seems likely that the average man misunderstands the scientist when he makes the same statement. By a hydrogen bomb a scientist means a bomb which contains certain components which display certain testable properties, and by an atom bomb he means a bomb which contains any of a variety of other components. But for the average man these terms do not have anything like these meanings; a hydrogen bomb may be simply a bomb which is called a hydrogen bomb by the scien­tists, and similarly for an atom bomb. On a somewhat higher level a hydrogen bomb may be a bomb which is called that by the scientists, and one which contains substances which are in some way derived from hydrogen. It is impossible for the layman to mean more than this by such terms because he doesn't know the chemistry which would allow their meaning to be enriched; and until we know enough about a hydrogen bomb to be able to distinguish it from similar objects, we will have to build the appeal to authority into the very meaning of the term as understood by non-scientific persons.

The conclusion then is that the average man can imagine the distinc­tion between a hydrogen bomb and an atom bomb as he understands

Page 122: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

112 FOUNDATIONS OF SOLIPSISM

the terms. More generally, we can be sure when people are said to understand complex sentences without being able to imagine instances of key notions which are essential to the understanding of the sentences, that they mean something much simpler than is being supposed when it is thought that the key concepts are beyond their imagination.

The other aspect of this problem involves certain deficiencies of imagination. Some people, for instance, have great difficulty in ima­gining smells. This does not mean that they do not have any olfactory images, and it does not ordinarily happen that there are gaps which cannot be imagined at all; rather it means that the only images which can be produced are indeterminate enough not to represent any qual­ities very well. In fields where our imaginations are weak the images which we do produce will be less determinate and "cover" more qual­ities. In the case of visual images there seems to be a minimum degree of determinacy. It is, as Berkeley pointed out, impossible to imagine a triangle which is neither equilateral, isoceles, nor scalene, 1 but this is also a case where our imagination is relatively strong, and one can generally produce quickly and easily a fairly determinate image of any sort of triangle whatever. Conversely, in those cases where our ima­ginations are weak, as in the imagining of the smells of fruits, there seems to be almost no limit to the degree of indeterminacy which an image can have. One can produce an olfactory image such that we can say of it only that it is sweet and nothing more, and in this way three or four images might cover the whole olfactory field.

Thus the case of such simple sentences as "I smelled a peach" will present no problem for the image theory since we will always be able to produce an image which is an instance of the relevant preliminary. It is true that the image will be an indeterminate one and will probably be accepted as only a partial instance, but this is all that we require for a minimal comprehension of the sentence. Someone who can produce a more determinate image will have a better comprehension of it. But let us now consider a more complex type of statement involving smells; a good example is "peaches smell sweeter than pears." It might be objected here that since many people cannot produce an image of the smell of a peach which is distinct from the image which they would produce as an instance ofthe smell of a pear. Thus such a person could not produce images of both one of which is sweeter than the other. Thus it might seem that the image theory forces us to say that we cannot understand such statements as "peaches are sweeter smelling than

1 G. Berkeley, Principles of Human Krwwledge, pp. 14.

Page 123: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

OBJECTIONS 113

pears" because we cannot imagine an instance of the preliminary "peaches smelling sweeter than pears." However, we must note here that on any analysis the understanding of such sentences does not imply that we know the distinction between the two smells or even that we know what either fruit smells like. In this respect the sentence will be like ''Jackson is taller than Jones" where we have no idea who either Jackson or Jones is. We can understand the sentence despite this. Therefore, the meaning of such statements as "peaches smell sweeter than pears" cannot be such that understanding them logically im­plies that statements such as "there is a peach" are also understood. It is clear that understanding the latter sort of statement on any theory presupposes some knowledge of what peaches are like. The meaning of "peaches smell sweeter than pears," at least as comprehended by someone who cannot imagine smells distinctly enough to differentiate between them, will be equivalent to "If I sniff something that looks like a peach (or is called a peach in case I don't know what one looks like) and I sniff something that looks like a pear, the former will smell sweeter than the latter." Thus the original statement is reducible to a hypothetical statement of which the antecedent is an ordinary matter of fact statement involving the sort of preliminary of which it is easy to imagine instances: it is easy to imagine how peaches and pears look and it is easy to imagine sniffing at them as long as we do not have to imagine the smells sniffed. The consequent, however, is the sort of statement with which we have not explicitly dealt since it involves partial variables represented by the words "former" and "latter." These are here referring to sense experiences which have not yet oc­curred, and they are only described in the most general way possible: nothing is said as to what qualities they will have. Thus the correspond­ing preliminary will be simply "one sense experience's being sweeter than another." And even someone who has a minimal smell imagination will be able to produce an image of a sweet smell and one which is not sweet provided that he does understand the notion of "being sweeter than." It is only necessary, then, in order to understand "peaches smell sweeter than pears" to be able to imagine a sweet smell. It is possible that we might understand sentences asserting such a delicate relation­ship between smells but that we could not produce the relevant images. The normal person could probably imagine instances of "being much sweeter than" and "being a little sweeter than," but might not be able to imagine instances of more minute differences. Even though we would have to invent new words to talk about these minute differences

Page 124: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

114 FOUNDATIONS OF SOLIPSISM

in smell, we might still understand the resultant sentences. Here we would imagine not the differences in smell themselves, but rather the reactions which experts (possibly ourselves) would have on actually encountering these differences. Thus the image theory is not committed to saying that it is impossible to understand sentences about smells, tastes, etc. where our imaginations are weak.

A good deal of attention has been given lately to Wittgenstein's remarks concerning the image theory. His central thesis is that it is not psychologically necessary that images precede certain events, such as the recognition of objects. If one is told to pick out a red bolt of cloth from a number of bolts of different colors, it is not necessary to form an image as an intermediary; one may simply pick out the red bolt without the aid of an image, and if one does, this need not be accidental.1 Similarly, even if images were necessary in order to enable us to pick out the right object, we would then have to have other images in order to pick out the right images.2 However the view that a word which is not accompanied by an image is meaningless is not implied by anything in the present statement of the image theory. It may have been held at various times that if one is to understand the words one is using, one must have images in connection with them on the grounds that we would otherwise be unable to pick out the objects named by the words. Wittgenstein then is certainly right in saying that the picking out of an object does not always presuppose the previous picking out of an image. And it would be equally incorrect to say that the picking out of an image presupposes the previous picking out of an object. Although there are strong psychological connections between these two processes, they are logically independent and one can take place without the other. The more interesting question, which Wittgenstein does not ask, is whether the ability to non-accidentally pick out a red bolt of cloth presupposes the ability to produce an image of a red bolt, and vice versa. Our thesis of a previous chapter was that the former presupposes the latter with certain qualifications. And this seems perfectly consistent with Wittgenstein's assertion, with which we agree, that the occurrence of one process does not presuppose the actual occurrence of the other process.

Another part of the image theory which is sometimes a source of objections is the relationship of the image to the object which it re­presents. This problem may be seen more clearly by going back to

1 L. Wittgenstein, The Blue and Brown Books, pp. 85-87. 2 Ibid., p. 12.

Page 125: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

OBJECTIONS 115

Berkeley for a moment. He held, at least in some places, that one idea (image) was associated with a word such as "red"; having associated the word and the idea, we are then in a position to pick out red objects. l

As we have seen, it is mistaken to think that our picking out objects psychologically presupposes our having had images; but even if we suppose that Berkeley took the view that the ability to have a certain idea is the criterion of whether we understand a given word, a problem still arises. Our highly determinate image of redness will be exactly similar to only one shade of redness. If we accept only exact similarity as our criterion of representation, "red" will be the name of only one particular shade of redness. We could adopt a more liberal criterion of representation, but this will involve something more than the rela­tion of similarity, and, as we will see, this leads to additional problems. As against Berkeley we might admit indeterminate images, which we would want to do in any case, but we still have no assurance that our image of redness will be indeterminate enough to cover all shades of redness. The obvious solution to this problem is as follows. We must not have just one standard image for each word, but rather a whole range, all of which are associated with each other and with the word. Our version would be as follows: a whole range of images would be accepted as actual instances of a preliminary such as "that object's being red." Thus for each particular shade of redness it will be possible to imagine a complete instance which is exactly similar to it. In this way we account for the full range of such a word as "red" without going beyond the notion of exact similarity.

A variant of this problem is raised by Chisholm in his book, Per­ceiving - A Philosophical Stuqy.2 We will discuss his views more fully later in connection with belief, but what he says seems also to be relevant to understanding. His central thesis, which he attributes to Brentano,3 is that the statements describing certain mental phenomena are irre­ducibly intentional. A statement in general is intentional if it leaves open the existence or non-existence of an object mentioned by a key term in the statement, as with "I want ice cream." Unfortunately Chisholm does not explicitly discuss understanding, and it is not clear from the criteria of intentionality which he gives whether statements of the form "A understands 'the grass is green'" are intentional or not. They come closest to the second criterion: here it is said that a sentence

1 This view occurs sporadically in Berkeley's Principles of Human Knowledge. 2 R. Chisholm, Perceiving: A Philosophical Study, pp. 168-185. 8 Ibid., p. 168.

Page 126: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

116 FOUNDATIONS OF SOLIPSISM

containing a propositional clause is intentional if the sentence does not imply either the truth or falsity of the clause.1 Assuming that a state­ment in quotation marks counts as a propositional clause, statements about understanding will then be intentional since neither the state­ment as a whole nor its negation implies either the truth or falsity of the statement in quotation marks. But since we are analyzing statements about undersi.anding in terms of the ability to have images of certain sorts, we will, contrary to Chisholm and Brentano, be analyzing inten­tional statements in terms of non-intentional statements.

When confronted with an analysis of this sort, Chisholm seems always to show that in analyzing away the intentional term, the analysis pre­supposes some other notion which is itself intentional. 2 The only notion in this analysis which could conceivably be intentional would be that of "instance of." Thus when the analysis says that in order to under­stand a preliminary one must be able, among other things, to imagine a partial instance, this might involve the belief that an image is an instance. However, this notion of being an instance has been defined in purely psychological terms, the having of a certain emotion given a certain attitude, and it is not obvious that any belief is involved. This is why considerable time was spent in previous chapters in isolating the required attitudes. Further, belief, at least in Chisholm's intentional sense, always involves the possibility of being mistaken. But there is no possibility of mistake here; if an image is accepted under the specified circumstances, then it is an instance at that time, and nothing which happens later could possibly refute this.

There is another objection which he might make, which is closer to our original problem. This would be that the image theory has the consequence that statements concerning understanding presuppose additional statements. In every case the additional statement would be that the image instances which might be imagined "represent" the actual object instances which may exist. We have previously argued that this is not the case, but, for the sake of argument, let us suppose that there is such a presupposition. Chisholm might then say that this representation involves belief. This belief would presumably be the belief that the image represents the object about which we are talking; but according to the image theory, the sorts of images we have deter­mine the sorts of objects about which we are talking. Thus such a pre­supposed belief would have to be a psychological belief about oneself;

1 Ibid., p. 171. I Ibid., p. 183.

Page 127: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

OBJECTIONS 117

namely that one would talk about sense-experiences in the same way that one talks about images. But this is not the sort of belief about which we can be mistaken as we can make it true whenever we want.

Lastly, we will consider some objections to the image theory made by H. H. Price in Thinking and Experience. 1 In the first place he states what he takes to be the image theory in a different way from that which we have given: " ... the thesis of the Imagist philosophers may now be formulated as follows: Mental images are the primary symbols, and all other symbols are secondary and derivative. In particular, words are secondary symbols. Of course, the Imagist does not deny that words have meaning, but he holds that they have it only indirectly, as sub­stitutes for images. These substitutes are needed, because words can be manipulated much more quickly and easily than images can. Hence they are normally used in an uncashed manner; indeed that is precisely their function."2 It should be obvious how different this is from the theory we have asserted. For one thing Price talks of words as substitutes for images and we have nowhere said anything like that; on the con­trary the relation between words and images where there is any has always been that of the latter being instances of the former. In Price's language, we have treated images as objects for which words are symbols in themselves. At this point one may wonder whether the two versions of the image theory are even consistent with one another, and it seems that they are because they are intended as explanations of different things. The image theory we have stated is intended as an analysis of understanding and an explanation of how it is that verbal expressions mean what they do. Price's version, which he goes on to criticize, is intended, however, as an analysis of what it is to think. There is, of course, some relation between the two theories, but it is clear that one could hold a dispositional image theory of meaning and understanding and some other sort of theory of thought. It might be recalled that the former does not actually entail that we ever have images; we might at the same time hold that thinking not only does not involve any images, but that words are not just substitutes for images. We must now explore more closely the relationship between the two sorts of theories and decide what might be meant by a word's being a substitute for an image.

Thinking may be regarded as a genus of which the two most im­portant species are word thinking (either public or private) and image

1 Price, op. cit., chapters VIII and IX. 2 Ibid., p. 239.

Page 128: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

118 FOUNDATIONS OF SOLIPSISM

thinking. The relation between the analysis of understanding and the analysis of thinking hinges on two questions. The first is whether all thinking is intelligible to the person who is doing the thinking; the second is whether it is literally possible to think, but think about nothing at all. One might suppose that the two questions amount to the same thing since thinking which is unintelligible to the thinker would also be about nothing. The distinction would be that unintelligible thinking would occur only when the thinker sets out to think about something and does not succeed, while in the latter case the thinker would never have intended to think about anything. The answers to these questions obviously depend on whether we are considering word-thinking or image-thinking.

Let us first take the case of image-thinking. It is clear that the process of having images in itself is never unintelligible to the person involved. More precisely, there is no difference in the imagining process itself between understanding the images which one produces and not under­standing them, since it is impossible for any aspect of an image to be hidden from the imaginer. Thus the question does not even arise and the words "intelligible" and "unintelligible" are not properly appli­cable to images. There is, however, another possibility: image-thinking may involve more than just having images and be intelligible or un­intelligible in some other way. We will defer this possibility for the moment. At this point we must ask whether it is possible to have images without thinking about anything. Of course thinking about an object is a form of being conscious of an object and the rules which apply to the latter also apply to the former. Nlost important, we need not sup­pose that the objects of thought really exist in every case. Thus we cannot conclude that we sometimes think about nothing on the grounds that we do sometimes think and have images of objects which do not exist. Then too, we do often say that we have been thinking of nothing in particular or just about nothing. But the question here is whether this is just a way of saying that we are thinking about something un­important, or perhaps that we are not thinking at all.

On the other hand, in the great majority of cases where we have images it is usual to say that we are thinking about the sort of object that corresponds to the image. In a few cases we might prefer to say that we are thinking about a sense experience rather than an object: having an image of a statue of Siva, one might say that he is thinking about what it would be like for him to see a statue of Siva. But in thi~ case one would still be thinking about something removed from his

Page 129: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

OBJECTIONS 119

present state of mind. There is still another possibility. An example would be a Buddhist monk who produces an image, perhaps of a lotus flower, simply as a preliminary to the attainment of a certain psycho­logical state. And when he has the image he devotes his attention ex­clusively to it and experiments with it in certain ways. In other words he purposely makes the image itself the sole object of his attention. In such a case we might be tempted to say that the monk is not thinking about anything except the image. Since the monk has banished all words from consciousness he is image-thinking insofar as he is thinking at all, and he is therefore thinking only about the process of thinking itself. Two questions then arise: the first is whether thinking about the process of thinking itself is an example of thinking about something; the second is whether we are willing to call the process which the monk goes through thinking at all. We might instead call it only having an image and entering a trance since it involves no multiplicity or manipu­lation of any kind. I do not think that any definite answer can be given to the last two questions, and it is not important to do so. We need conclude only that most imagining involves thinking about something and that if there are any contrary cases, they do not involve any cases of thinking which are unintelligible to the thinker since the question of intelligibility cannot then be raised.

It is still possible that image thinking might involve more than just the having of images as such; the images might have a symbolic func­tion, and the conclusions we have just reached suggest that this is often the case. This would mean that the images we have in image-thinking would function in the same way as the words involved in word-thinking. Thus, instead of saying the word "barn" to oneself, one might produce an image of a barn. A naive image theory of thought would hold that all word-thinking could be replaced by image-thinking in this way; but, of course, we might admit that some word-thinking, perhaps very little, could be replaced by image-thinking without being committed to this thesis. In any case, insofar as we do this sort of image-thinking we are intentionally using images to think about something other than themselves. The possibility of intelligibility or unintelligibility might then seem to arise for this reason; we know that word-thinking can be unintelligible and if image-thinking can have a parallel role, and can in some cases replace word-thinking, then it may be possible for it to be unintelligible as well. However, we must examine this sort of image­thinking more closely. A given image occurring in this sort of thinking might be the symbol only of a certain sense experience which is expect-

Page 130: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

120 FOUNDATIONS OF SOLIPSISM

ed to occur, and which is very similar to the image. But in fact when images function symbolically at all they usually symbolize much more than a sense experience; rather they are the symbols of complete material objects. Thus a barn-like image would usually symbolize a barn rather than the experience of seeing a barn. However, there is only one way in which symbolic image thinking can fail to be under­stood by the thinker; this would be the case where it is not possible for an entity which would be thought about to exist. That is to say that if symbolic thinking is unintelligible to the thinker it must be because it is self-contradictory. If it is simply the case that the image is not in­tended to symbolize anything, then we are not dealing with symbolic thought at all. But it is also obvious that symbolic image thinking can never be self-contradictory. A round square cannot exist in our imagin­ation any more than it can exist in the world. Thus we conclude that symbolic image thinking cannot be unintelligible any more than non­symbolic image thinking, provided that we choose to call the latter thinking at all. This in itself is a sufficient ground for rejecting the naive image theory of thought. Word-thinking can sometimes be self­contradictory and since image-thinking cannot, we cannot image-think everything we can word-think.

In fact word-thinking can not only be self-contradictory but can also be nonsensical in other ways. This is clear because word-thinking operates with the same words and sentences as inter-personal discourse, and since the latter contains nonsensical and contradictory statements, so must the former. Word-thinking often consists just in uttering to oneself (whether silently or aloud) sentences which might be used in discourse with others. Thus the general analysis we have given of the understanding of statements and the meaning of statements will apply also to the statements which we make to ourselves.

It is now clear that there is an important distinction between a theory of meaning and a theory of thought. Since not all thinking is intelligible even to the thinker, a theory of thought must take in more than a theory of meaning, as the latter sort of theory can never give any account of unintelligible thinking. In fact a given theory of mean­ing commits us only to making the distinction between intelligible thinking and unintelligible thinking in a certain way; it does not entail any analysis of thinking in general.

We will now distinguish the image theory of meaning we have stated in previous chapters from certain image theories of thou~ht. In the first place there is the naive image theory of thought which holds that all

Page 131: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

OBJECTIONS 121

word-thinking can be replaced without loss by image-thinking; there is also the even more naive theory which holds that all our thinking is in fact image-thinking, and we will call this the non-dispositional image theory of thought. Then there is the image theory which Price con­siders and eventually rejects; as we have seen, the essence of it is that images are the primary symbols in thinking and that words are sub­stitutes for images. It is not clear what is meant by these phrases and the naive theory first suggests itself. However, Price immediately qualifies the theory he is considering so that it is not necessary that all word-thinking be replaceable with image-thinking. He also explicitly rejects the non-dispositional theory outright and seems to have in mind some more plausible image theory of thought. Although he never states it precisely it seems likely that the theory he has in mind would amount to the following: all meaningful word-thinking may be replaced by image-thinking, and the image-thinking will always be at least as clear and precise as the word-thinking. The theory might also involve a second and stronger assertion: image-thinking could be carried on even if there were no words and word-thinking, but there can be no intelligible word-thinking unless it has at one time or another been preceded by the corresponding kind of image-thinking. Let us compare this with the thesis of the image theory of meaning. This would be that if we are able to word-think intelligibly concerning a certain subject­matter, we will also be able to imagine that subject-matter (with the qualifications stated above). In other words, if we are able to word­think about something, we will also be able to image-think about the same thing. But this does not imply that the word-thinking can be re­placed by the image-thinking; on the contrary we will soon argue that word-thinking does certain jobs which image-thinking can never do. Thus our interpretation of Price's version of the image theory is much stronger than the image theory of meaning in that the former asserts something about the nature of thought which the latter does not. Price's theory also presupposes a theory of meaning since it makes an assertion only about intelligible thinking and thus presupposes a dis­tinction between intelligible and unintelligible thinking. For the same reason it is less than a complete theory of thought in that it does not say anything about unintelligible thinking.

The image theory of meaning is also distinct from the second or optional thesis of Price's theory which says that word-thinking must have its origin in image-thinking. The image theory of meaning says only that intelligible word-thinking implies the ability to image-think

Page 132: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

122 FOUNDATIONS OF SOLIPSISM

about the same things; it does not imply that the image-thinking always comes first. In fact we generally word-think about things before we have occasion to image-think about them.

Most of Price's objections to the image theory center on its role as a theory of thought. The two which seem the most important are the following. First he points out that there is no reason to give priority to private image-thinking over word-thinking which is sometimes at any rate public.1 He thinks that the motive behind this might be the view that thinking is a kind of inspection, and in thinking an image must always be present or readily available to be inspected. 2 Again, there is no reason to make thinking a kind of inspection. Secondly, an image theory of thought will have difficulty with such logical notions as "and," "or," "if ... then," etc.3 These objections do constitute diffi­culties for an image theory of thought, but there seems to be an even more basic objection. If the image theory of thought is to make any distinction between merely considering a state of affairs and actually thinking it to exist, this distinction must be made in the way that Hume made it, in terms of vivid images. This is not generally accepted, for fairly obvious reasons, as an adequate account of belief, and hence of thought. Save for pointing out these difficulties in the image theory of thought, we will not attempt to evaluate it, but merely point out its logical independence from the image theory of meaning. We have al­ready indicated that the latter does not attempt to reduce thinking to the having of images in any way. In fact, if the image theory of thought were true and word-thinking could be dispensed with in favor of image­thinking, the problem which the image theory of meaning is intended to solve would have less importance, if it did not disappear altogether. This is because the analysis of the meaning of verbal expressions is interesting only if we assume that thinking in terms of words involves something that could not be done merely by the having of images.

A good deal of space has been devoted to the theory of meaning even though the views stated have not been peculiarly solipsistic and even though they do not directly entail the "solipsism proper" which is to be found in the chapters on the material world and other minds. Thus while the solipsist still has some choices left as to what sorts of theories to adopt, the stage has been set and anyone who has followed him this far will find it difficult to object to the rest of his Weltanschauung. As

1 Ibid., p. 241. • Ibid., p. 259. 3 Ibid., p. 297

Page 133: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

OBJECTIONS 123

with other sorts of metaphysical viewpoints, the crucial part of solips­ism is the theory of meaning on which it is based, and this has been our main concern up to this point.

In the remainder of the book we will state solipsism proper and con­sider the smaller objections which arise in connection with particular analyses. We have already forestalled the more general and more basic objections.

Page 134: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PART II

CAUSALITY

Page 135: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 8

CAUSES AND COUNTERFACTUALS

We must now analyze the concept of a causal law, both for its own sake and because the analyses to follow will involve the concept. As usual, the solipsist is on the lookout for the most economical analysis, and theories of the Humean sort are unquestionably the most econom­ical.

In this chapter we will therefore examine Humean extensional theo­ries of causation. Particular attention will be paid to objections to these theories based on the notion of an accident, since it is objections of this sort that are most crucial. The generic theory of this type is the one which equates a causal connection between two events, a and b, with a constant conjunction between events of the type A and events of the type B. Correspondingly, a general statement asserting a causal connec­tion between two types of events will actually describe a causal law if and only if there are no exceptions to this generalization anywhere in time or space. Many qualifications will be necessary if this theory is to be at all plausible and such phrases as "of the type A" would have to be analyzed further; however, this is the basic theory upon which ex­tensional analyses of causation have drawn, and this statement of it is sufficient to bring in the important objections.

The most important of these is that on this analysis we cannot distin­guish between a causal connection and an accidental connection which happens to be unbroken. l Thus it happens that everyone on a certain park bench is Chinese, but this does not establish a causal connection between sitting on the bench and being Chinese. Of course, no ex­tensional analyst holds that a constant conjunction which has only held for the past and will not hold for the future constitutes a causal connec­tion; his initial reply can then be that if the connection is merely acci­dental it will not hold for the future and will thus be distinguished from a causal connection which holds for both the past and the future. How-

1 On this point see R. Chisholm, "The Contrary-to-Fact Conditional," Mind, 1946, N. Goodman, "The Problem of Counterfactual Conditionals," Journal of Philosophy, 1947. pp. 122-123, and A. Burks, "The Logic of Causal Propositions," Mind, 1951.

Page 136: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

128 CAUSALITY

ever, this answer is not likely to satisfy the objector who can reply that an accidental connection can hold for both the past and the future, but is still different from a causal connection. It is this disagreement which I will try to elucidate in the course of this chapter. However, I would like to note at this point that the Humean can make some distinction between a connection which is accidental and one that is causal; namely, that the latter is one which holds for the future as well as the past. The question is whether this is the correct distinction, and, in case it is not adequate, whether some less naive distinction can be made from the point of view of this theory.

Our next job is to look at some of the relationships which we suppose to hold between the notions of cause and accident. One obvious differ­ence between the two concepts is that we can talk about ing Ie events as accidents whereas we have to talk about causal connections between sets of events, or at least about the cause of some event which is specified by the context. We never talk about a cause outside of any such context. The result of this is that the word "accident" is a relative term in that we always have to know what it is an accident relative to before we understand the particular assertion that is being made. Thus an auto­mobile accident is an accident relative to the intentions of the driver, but it is not an accident relative to the physical laws of moving bodies. In general it seems better to talk about accidental connections between events rather than just talking about accidents. However, in order to transform our language in this respect we must first find the implicit relative term and re-state the original assertion so that it says, not that A is an accident, but that the connection between A and B is an acci­dental one. In practice, however, it is not usually very easy to pick out the relative term that was intended, as there are always many possi­bilities.

Thus when the anti-Humean points to the fact that everyone sitting on a certain bench is Chinese and says that this constitutes an accident, it is not yet clear exactly what he means. One possibility in such a case is that the event in question is an accident relative to anything and everything that has happened in the past. This would be tantamount to saying that there is no way in which the event could have been pre­dicted and that it is a chance event for which there is no explanation. It is often the position of philosophical indeterminists that there are such events, even though the indeterminist may not claim to know just which events are inexplicable in this way. In any case this is a philoso­phical position which does not seem to be involved in our everyday

Page 137: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CAUSES AND COUNTERFACTUALS 129

assertions concerning the existence of accidents. One does not have to be an indeterminist in order to say that it is an accident that the men sitting on a bench are Chinese. Further, we are not necessarily saying that there is no empirical explanation of this fact. We might be given an explanation of how it happens that they are all Chinese and still call the phenomenon an accident. Even in a case of supposedly "pure chance," such as the outcome of a tossed coin, we do not want to deny that there is an explanation of the outcome. On the contrary, we would normally suppose that if we knew the exact characteristics of the coin, its initial position, the exact way in which it is tossed, the resiliency of the surface on which it will land, etc., it would then be possible to predict the outcome. Again, even after we are informed of the way in which the outcome could be predicted we continue to think of the outcomes as chance events or accidents. Hence we must look for some other meaning of "accident" which will permit an event for which there is an explanation to be accidental.

It may be that we have picked out the wrong relative term. Thus, instead of saying that the coin's coming up heads when tossed is acci­dental relative to anything that has ever happened, we might say that its coming up heads is accidental relative to the fact that the coin is normal and is tossed in any normal way. This is in effect to say that there is no direct and complete causal connection between the tossing of the coin and its coming up heads, and this might be what it means to say that a connection is accidental. However, at the same time it leaves open the possibility that there is a causal connection between the tossing, a great many other unstated conditions, and the outcome of heads. Thus we can now say that the outcome is an accident without saying that there is no explanation for it. While this analysis has this desired consequence it has some other disadvantages. For one thing, it seems that even when we are confronted by invariant constant con­junctions between sets of properties we almost never experience directly all the properties involved in the conjunction. In most such cases we do not even know what some of these properties are; this often takes the form of asserting that two sets of properties are always conjoined under "normal conditions" which are unspecified, but are believed to obtain most of the time. It may also be possible to state completely laws which involve ideal conditions such as perfect vacuums or the complete lack of friction; however, when such laws are restated so that they apply to particular situations it is very difficult to state a complete list of conditions which wiU make the law hold without any exceptions at

Page 138: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

130 CAUSALITY

all. Thus on the analysis of "accidental connection" most, or perhaps all, of the connections which we take to be causal will turn out to be accidental instead, so this cannot be the right distinction.

In the light of this we can now clarify the ordinary use of the terms "causal" and "accidental." When we speak of a causal connection between properties we are usually prepared to admit that the causal connection is not really between the properties which we mention, but involves many other properties as well. Similarly, when we are faced with a high correlation we suppose that if it is indicative of a causal relationship then there will be invariant connections which underlie it. Thus in speaking of causal relationships we have a certain ideal in mind which we only approach in our statements of causal laws.

However, the situation is somewhat different in the case of accidental connections. When I say that properties A and B are only accidentally connected I do not usually mean that these properties do not in them­selves form an invariant connection. In this sense almost all properties which we suppose to be causally connected would also be accidentally connected because our statements of causal laws are almost always in­complete. On the other hand we would not want to assert that there are no causal laws in the strict sense connecting the two properties among others; we would normally suppose on the contrary that almost any two properties are ultimately connected causally, however indirect the connection may be. Rather, when we say that a connection is acciden­tal we usually admit that there are such connections but assert that these connections are relatively indirect. Similarly, when we say that a generalization is accidentally true we are saying that it is a relatively incomplete statement of one or more causal laws. Thus even the notion of "accidental connection" is a relative one and the standard of com­parison is more or less implicit. While there may be borderline cases, we still have no difficulty in classifying the connection between tossing a coin and getting heads as an accidental one and the connection be­tween pouring water on a fire and the fire's extinction as a non­accidental one. It may thus be that the distinction is one of degree rather than kind, and we will explore this question later on. For the moment, however, we will examine further the objection to the Hume­an theory based on the notion of an accident.

It appears that none of the connections we describe as being either causal or accidental are invariant except over a very limited range of space and time. When we talk about a causal connection we suppose that there is a strict invariance, but that it is not identical with the one

Page 139: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CAUSES AND COUNTERFACTUALS 131

we are confronted with or the one we are stating completely. In the ordinary use of the term "accidental" we assume that the connection is not invariant in the strict sense, as we would even if we were asserting it to be a causal connection, and assert that it is a connection which is relatively easily broken or one that is broken often. It must also be re­membered that we are ordinarily talking about repeatable properties. In his example of an accidental constant conjunction Goodman speaks of the property of being a coin in his pocket on V.]. day, and this pro­perty is not repeatable since it cannot conceivably apply to anything after that date. 1 It may be possible to find a set of properties which do not involve indices, but which apply only to those coins on that date. Nevertheless, if we want to find an invariant conjunction which comes close to being accidental in the usual sense we should make use of those properties which can in principle have later instances even if they do not in fact. We should also try to find a better example than the Chinese on the park bench, since it is not plausible to suppose that only Chinese have sat on this bench and that the connection is acci­dental, as it might not be if the park were in China.

Let us suppose instead that a certain penny is minted, but is very lop-sided and has other defects which in fact distinguish it from any other coin which ever has or will exist. Due to these deformities it is quickly melted down again, but before this is done one of the minters playfully tosses it once and it comes up heads. Thus we have no ex­ceptions whatever to the rule that a penny of this sort when tossed comes up heads; all the properties could in theory be repeated, but it is plausible to think that they in fact do not have other instances. On the other hand we believe that while the penny is deformed the deform­ities are not such as to make it always turn up heads, and we would suppose that ifit had been tossed many times it would have sometimes turned up tails. This connection seems to be accidental, but we must now compare it with more ordinary accidental connections and see what the relevance of this kind of case is for a causal analysis in terms of constant conjunction.

This case differs from what we would ordinarily call an accidental connection even though the properties involved are repeatable and the conjunction is believed to be one that could easily be broken. We are usually just not faced with rules involving only non-ideal properties which have no exceptions; we do not expect that much even of our causal connections. Thus it is something of an extension of the common-

1 Goodman, op. cit., p. 122.

Page 140: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

132 CAUSALITY

sensical notion to say that such a connection is accidental, but it appears to be justified in view of the fact that such a connection is believed to be one that could very easily be broken even if it is not broken in fact. However, this in itself has its ramifications. Before, we said that an accidental connection was one that was easily broken, but we could just as well have said that it was often broken, and perhaps this is all that we meant. In this case, however, we can say that the connection is easily broken, but we cannot say that it is often broken, since, in point of fact, it is never broken. Thus it becomes clear that the man who uses this case as an objection to the Humean theory is himself making a rather complex counterfactual assertion which need not be involved when in ordinary language we say that a connection is accidental. The objector will be saying that while two sets of properties, A and B, are in fact constantly conjoined over all time and space, there are other circumstances C which have never and never will be realized such that if C had been realized the properties A and B would be separated, and these circumstances C represent a causal possibility in some sense. This will, of course, involve all the usual problems of counterfactual con­ditionals, but, apart from these, it is important to note that the objec­tion, when fully expanded, is not quite so simple and obvious as it seemed at first. On the other hand, it does seem that the minter could have tossed the coin many times instead of once, and that under these circumstances heads would not always have resulted. Thus the problem is still with us and we must look at the various ways in which it could be met.

There are various alternatives left open. The most obvious one is just to give up the Humean theory of causation. However, another alternative is to keep the basic theory, but to try to get around the difficult case by placing limitations on the properties that must be in­volved in a causal relationship. Other possible approaches would be to deny that this case really is unique, or to deny that it is accidental.

First we will examine the position that such a case does not represent an accidental connection at all. Here it would be argued that the special coin could not have come up tails. It might be that this special coin can be minted only under very special circumstances and that these special circumstances are causally connected with the outcome of the toss and the behavior of the minter who tosses the coin. It does seem that there are some combinations of events which seem to be causal possibilities at first, but turn out not to be when we think about them more carefully. Thus we might suppose it to be possible

Page 141: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CAUSES AND COUNTERFACTUALS 133

for a dinosaur to catch and eat a monkey (particularly if we are used to seeing such beasts hob-nobbing in science fiction movies) until we realize that the physiology of the two animals is such that they could not have lived at the same time. However, it is a great jump to the position that all combinations of events which do not occur are causal impossibilities, and some independent argument would be needed even to make plausible that position. The real trouble is that the present analysis implies that there is a causal connection between any two properties which are constantly conjoined. Even if we found some obscure causal connection between the tossing of the queer penny and its outcome, other properties would be involved, but these would not be counted as part of the connection on this analysis. Thus the position that the case is unique but not accidental in this simple way is antece­dently very implausible, and we have yet found nothing to counteract this implausibility.

A similar but more plausible approach can be derived from the view, mentioned above, that the difference between a causal connection and an accidental connection is one of degree rather than of kind. The force of the objection is grounded on the commonsensical view that there is a sharp difference between the two kinds of connections, that this clearly represents an accidental connection, and that the theory wrongly makes it into a causal connection. If we can show that the view of common sense on this matter is confused and that even on the commonsensical view of causation the difference is only a matter of degree, the force of the objection will then be considerably weaker. The theory would not then confuse two different things, but would include more things as causal connections because of certain common features, and it might be possible to make some adjustments of theory to practice.

We already argued above that when we look closely at our ordinary usage of the terms "causal" and "accidental" the difference is only a matter of degree, even though we may not be aware of this fact about our usage; in fact, despite our use of the terms, we tend to think of the difference as an absolute one. On the Humean theory the difference really turns out to be also one of degree, but for quite different reasons. Here we have a continuum of true generalizations each admitting of no exceptions, but with varying degrees of exemplification. This de­pends on the assumption, which seems plausible, that we can describe anything that happens uniquely without using indices or spatio­temporal references directly, but by mentioning peculiar features of the event itself or its surroundings. This means that we can take any

Page 142: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

134 CAUSALITY

accidental connection between a set of properties and form a true generalization describing that connection by putting into the antece­dent conditions which allow the rule to apply to exactly those cases where the connection happens to hold. That is, if we know where a certain rule holds and where it does not, we can, after the fact, doctor the antecedent clause so as to make the rule universally true over all time and space even though it incorporates so many ad hoc hypotheses that no one would take it seriously as a causal principle. Thus from the Humean point of view there are some generalizations which are vacu­ously true, some which have one instance, some which have two in­stances, etc.; as we will see, there may be some which have an infinite number of instances. Hence it may be that the connections we call accidental are those which fall under generalizations of relatively small extension while those which we call causal come under generalizations oflarger scope. It could then be argued that it does not make very much difference where we draw the line. The important common feature of all these connections is that they are invariant, although some general­izations are much more useful for prediction because of their greater range of instances. Thus the consequence would be that an accidental connection is a particular kind of causal connection which is relatively useless and is distinguished from the others by degree only. If one is willing to accept this the objection loses its force because it presupposes that an accidental connection cannot also be a causal one.

The chief objection against this approach is likely to be that causal laws are reduced simply to descriptions of conjunctions of events and that they have no legislative power. On the other hand, we cannot deny that they have predictive power. All the laws hold for all times so that if we believe a certain law to hold and we are located at a particular time we can then deduce predictions from our belief. How­ever, the law that all A's are B's on this analysis will not have legislative power in the sense that while we can deduce "If something is to have the property A it will have the property B," we cannot deduce "If something had had the property A it would have had the property B." We cannot deduce the counterfactual statement because of our analysis of a law as a generalization which holds for all time and space. Counter­factual situations are not included in the range of situations the general­ization must cover in order to be a law, and there is no reason to believe that it does extend to these situations. Ordinarily one of the differences between generalizations which we believe to describe causal connec­tions as opposed to those which are believed to describe only accidental

Page 143: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CAUSES AND COUNTERFACTUALS 135

connections is that we deduce counterfactuals from the former, but not from the latter. Thus the objection would be that this analysis reduces all kinds of connections to the level of accidental connections. The Humean who takes this approach would reply that we can in fact never deduce a counterfactual statement from a corresponding generaliza­tion; this jump from the generalization to the counterfactual always represents a further inductive argument, and this argument may be good or bad in varying degrees. In general, the more instances the generalization has the better the argument will be, but in no case will it be a strict deduction. Thus the Humean can relate the causal general­ization to the counterfactual, but he cannot give us quite the deductive relationship which seems to ordinarily be presupposed.

One's final attitude toward this approach is likely in the end to depend on one's view of philosophical analysis and what it is supposed to do. The "continuum-of-causes-and-accidents analysis" does involve a certain shift from our ordinary notion of causation and we would have to talk somewhat differently about causation, particularly when we are inferring counterfactuals, and probably in other respects as well. On the other hand this analysis preserves most of the usual causal relation­ships and is not destructive in the way that the suggestion that we should stop talking about causes and instead talk about correlations is. Nevertheless, the solipsist is arguing that his view is an adequate inter­pretation of our ordinary beliifs and he should try to make sense out of the ordinary notion of causation and see if some better distinction between causal and accidental connections can be made in a Humean way.

A remaining possibility is to admit that the case cited above re­presents an accidental connection, but to question its uniqueness. We might say that since we have an accidental connection there is reason to believe that there have been or will be similar pennies which are tossed and sometimes come up tails. It seems that the only way in which we are likely to be able to support this assertion would be to establish the general principle, touched on above, that any combination of events which constitutes a causal possibility is actually realized at some time or other. I think that there is some evidence for this principle, and that it should not be dismissed offhand as an ad hoc support to a theory, even though it will have to be qualified. In particular, let us consider the following argument. If anything is a causal possibility, it will be a logical possibility, and the probability of any logical possibility is greater than 0, at least if the set of relevant logical possibilities is finite.

Page 144: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

136 CAUSALITY

However, we also know that the relative frequency of the occurrence of an event will, in the long run, approach its probability; we can then be assured that the event in question has occurred or will occur at some time.

The basic notion in this argument is obviously that of a causal possi­bility. There are various ways in which this concept can be treated, but we will first define a narrow sense of "causal possibility": any com­bination of sets of properties constitutes such a possibility if it does not in itself violate any causal laws. This just means that the state of affairs does not violate such laws in its internal composition; it does not say that it is causally possible that a state of affairs which is itself a causal possibility could occur in the world, or in any world which we might specify. If we want to know whether a state of affairs X which is causally possible in itself could occur at time t in our world we must add to X all the states of affairs which have taken place in the world up to t and ask whether the augmented X still represents a causal possibility. This latter sense is important for us since the argument is that any causal possibility would occur in our world sooner or later. The question then is whether or not it is causally possible that a queer coin similar to the one described could be tossed and come up tails at some time. First, with respect to the future, if the possibility in connection with the history of the world up to the present violates a causal law, it is not logically possible that it should occur in the future. It is part of the minimum meaning of "causal law" that if anything is a causal law it holds for all times and we cannot consistently suppose both that the law holds and that there are exceptions to it. Determinism, of course, would imply that the coin could not have come up tails when it in fact came up heads, but even on that assumption it might be causally possible that a future toss would come up tails, since the circumstances would be d.ifferent and different causal laws would apply. Further, even if some events which have occurred make an outcome of tails on such a pennyforeverimpossible in the future, it might still be causally possible that it could have occurred in the past prior to the occurrence of the events which eliminated its possibility. Thus it is not so easy to know whether the combination is causally possible in the broad sense even though we might feel sure that it is causally possible in the narrow sense. We feel that such a penny could turn up tails without this com­bination in itself violating a law, but, given the world as it is, can there be any other such pennies, will they be tossed, and how will they come up? The last questions are not so easily answered. Since we do not

Page 145: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CAUSES AND COUNTERFACTUALS 137

know all the causal laws we cannot say exactly what will be causally compatible with a certain specified world, and when dealing with the remote past and future, we do not even know what sort of a world to specify. However, there seems to be very little reason to believe that a state of affairs which is internally compatible and is compatible with the relatively permanent features of the world should be incompatible with all the future and past states of the world even if it should be in­compatible with the present state. The objector may be saying that this case is very likely unique, but he would not be best advised to commit himself to this much more extreme position in the case of this or any similar example.

Let us now see where the argument takes us. Having assumed that the event is causally possible there can then be no good inductive reason for thinking that this event will never occur. However, for any given time, there will be other causal possibilities having the same status such that only one of these possibilities can actually occur; the antecedent probability of the event in question will depend on the number of competing possibilities. But if the number is always infinite and the probability of tails on the queer coin approaches 0 whenever it is a causal possibility we cannot be assured that tails will ever occur. Thus some restriction must be made on the properties which the Hu­mean can talk about as being causally connected. This brings us back to the very first problem in the Humean analysis. We there said that a connection between two properties was causal if all properties similar to the first are invariably accompanied by properties similar to the second. The question then was how close the similarity must be; this can be restated as the question of how general or specific the properties must be, and the Humean must answer this question in some way. We have also seen that he should restrict these properties in such a way as to avoid having an infinite number of causal possibilities at anyone time, and this will be avoided if there are a finite number of relevant properties. Thus the properties which can enter into causal relations must be general enough so that there are only a finite number of them. This is still only a partial answer to the question and we must try to clarify the necessary restrictions further.

The basic idea behind the analysis is that if certain properties are connected only accidentally, then sooner or later they will be separated in other circumstances. In cases such as the one mentioned above some of the properties occur so seldom in the history of the world that they never have an opportunity to be separated. We must then distinguish

Page 146: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

138 CAUSALITY

what I will call "durable" properties from others. Durable properties in general will be those whose occurrence is not limited in the way that the special properties belonging to the above-mentioned penny are. We must now see how the notion might be defined. The strictest defini­tion of durability would be that a property is durable if, given any finite period of time, the property has occurrences outside of that time. Thus a durable property would be one that has occurrences in the infinite past and in the infinite future. Given this definition we can plausibly argue that no invariant connections between durable properties are acci­dental. Someone who asserted that such a connection is accidental would again be asserting that there are circumstances under which the properties would be separated. However, the plausibility of this objec­tion before came from the fact that these circumstances, C, were known not to be operating when the malformed penny was tossed. If the pro­perty of being a similarly malformed penny which is tossed were durable then those very circumstances would be realized, and presum­ably heads would not always turn up. In general if the antecedent properties are durable we will never have reason to think that the set of circumstances, C, which would be likely to break up the constant conjunction, will never occur along with the antecedent conditions. The objection turns on having the knowledge that the conditions which might have broken up the conjunction never in fact have a chance to operate. If the antecedent event is durable, however, we will not have this knowledge and it is plausible to suppose that if the connection is only accidental and occurs an infinite number of times, then sooner or later the connection will be broken. While it seems unlikely that there would be an infinite number of durable properties, we can make the added restriction that they be general enough to be finite in number. It then appears that causal laws involving only durable properties can be analyzed in terms of constant conjunction, but it turns out we must make further assumptions of an empirical nature.

In order to make any sense out of the ordinary concept of causation we have to make at least one empirical assumption. Putting it loosely, we have to assume that there are conjunctions which hold for all times, and do not hold only vacuously. If we do not make this assumption we will have to junk the notion of cause and fall back on that of a corre­lation. However, in order to preserve the sharp distinction that com­mon sense makes between a causal and an accidental connection we either have to go beyond the Humean analysis in some way or we must rely on the notion of a durable property as defined above. That defini-

Page 147: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CAUSES AND COUNTERFACTUALS 139

tion results in the desired sharp break between durable and non­durable properties, and it seems to be the only definition of durability which has this consequence. The disadvantage in using this concept in the analysis of causation is that we have to make further empirical assumptions. The most obvious of these assumptions is that there are durable properties; it is logically possible that the occurrence of any property is restricted to a finite range of time. Secondly, we must assume that the properties we need in order to state basic causal laws are durable. On this analysis all genuine causal laws will involve such properties and all other connections will either be accidental or deriva­tive. The required picture of the world will then be that there are a number of very basic permanent properties in terms of which anything can be explained if there is any explanation for it all. If we refuse to make this assumption or adopt this ideal we will either have to go back to the "continuum of causes and accidents" analysis or give up the Humean approach altogether.

We made it clear at the outset that the sort of modern solipsist we are considering is not averse to making inductive assumptions or at least to admitting that they are embedded in our ordinary beliefs. However, he does not want to make any additional assumptions if it can be avoided and he also wants to insist that our ordinary beliefs can be translated into a language which does not mention physical objects or the sensations of others. For this reason it would be unfortunate for him to suggest that we dispense with a concept such as cause and re­place it with another concept such as correlation; it might then appear that the solipsist is not translating the statements of ordinary English into his solipsistic language but just constructing a different language in which one might not be able to say all the same things.

Hence, the solipsist's best strategy at this point is to concede that a causal statement may say more than any truth functional statement which can be formulated, and, in particular, that a causal statement may always entail a counterfactual statement. We can then go on to analyze counterfactual statements, their comprehension, and under­standing. Even if we have to admit a non-extensional element in the end, we may not have the maximum degree of economy of analysis, but the solipsist will still be able to carry out his program of analyzing the statements of our ordinary language in terms of statements about his own sensations and images.

We have now taken the position that to the extent that "Any occur­rence of an A causes a B" goes beyond "All A's are followed by B's" it

Page 148: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

140 CAUSALITY

entails "If an A had occurred at any time a B would have occurred." This last hypothetical statement is counterfactual whenever A is not occurring and the antecedent is false. When we assert "If A then B," knowing that A did not occur or will not occur, certain conditions have to be satisfied for the sentence to be plausible enough to be interesting. Specifically, it must be supposed that the speaker is not just making a guess, and that he has some sort of evidence for his assertion. This goes with the assumption that the occurrence or non-occurrence of A and B is not a matter of chance, but that there is a causal law connecting them. When we consider the counterfactual possibility of A's occurring, we usually assume that there are causal laws which, together with the occurrence of numerous past events, in fact brought about A's non­occurrence. Thus, in order to suppose that A occurred, we have also to make other counterfactual suppositions concerning either the causal laws or the prior chain of events. We have to suppose that a whole range of events reaching into the past to an undetermined extent did not occur when we know that it did, or we have to suppose that certain causal laws which we ordinarily take to be true are in fact false, or we have to combine both suppositions in some way. The hypothetical counterfactual then tells us that if the universe had been different enough to determine A's occurrence rather than non-occurrence, B would then have occurred. This conclusion holds even if we take causal laws as being only universally true generalizations.

We now come to the question of determinism and its relation to our ordinary language. I do not think that there are any very decisive arguments to support either determinism or indeterminism, but we can at least try to see whether either position is presupposed by the beliefs of common sense. A basic point seems to be that even if we be­lieve that some events cannot be explained, we are not in a position to say which particular events. If we are indeterminists it is not because we think that there are some unexplainable events, but because we have doubts about all events being simultaneously completely explain­able. This is reflected in the fact that it is never improper to ask for an explanation of anything that happens. In fact, as Kant pointed out, empirical investigation of any subject matter presupposes that there is an explanation to be found. Since there is no facet of the world which we commonly suppose to be uninvestigable, we collectively suppose at one time or another that any event which is open to ordinary observa­tion does have an explanation. This is not quite a deterministic assump­tion because common sense never seems called on to make such an

Page 149: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CAUSES AND COUNTERFACTUALS 141

assumption about every subject matter at once; rather, we only sup­pose that the event we are talking about has an explanation. When we assert a counterfactual we are engaged in a kind of explanation, and we would be defeating ourselves if we supposed the events surrounding it to be undetermined and unexplainable.

The problem which now arises is that in supposing A's occurrence we have set aside either some of the causal laws which seem to govern the world, or some of the past history of the world, or both. We then assert that on the basis of the altered past and/or the altered causal laws B will occur; this is what is meant by "If A then B" when it is explicitly counterfactual. However, in practice we never know to what extent we must change the actual world in order to let A occur, and it may be that in doing this we have set aside the very causal laws and/or past history which might bring about B's occurrence, or for that matter, B's non-occurrence. As an example, suppose that someone asserted the following: "If that ball had been hit three feet higher Jones would not have caught it." If asked, he might defend this statement by pointing out that Jones cannot jump very well. This seems to be a relatively simple empirical statement which would be false if, for instance, it turns out that Jones can jump much higher than had been supposed; let us assume, nevertheless that there is good evidence for the statement and that we would be justified in believing it according to the usual stan­dards for counterfactuals. However, the situation becomes more com­plex on examination. In particular, there must be ways in which the ball could go higher than it did without affecting greatly the evidence for thinking that Jones could not catch the ball.

Let us begin with a very adverse case. One way to allow for the ball's having gone higher would be to suppose a difference in the laws of gravitation so that there would be less attraction between the ball and the earth, which would result in a higher trajectory for the ball. How­ever, it is obvious that this counterfactual assumption would alter the whole pattern of our solar system and might well result in an earth which was not only unsuitable to baseball, but to all other forms oflife as well. This is clearly a bad assumption to make. On the other hand, we cannot just suppose that the ball went higher and that nothing else was altered; we would then be supposing that in a particular instance laws were violated which we believe to hold generally. Since neither of these courses is attractive we would have to make some other kinds of changes in the hypothetical world. We might want to say that in the counterfactual situation the ball would go higher and farther because

Page 150: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

142 CAUSALITY

it is more resilient than the actual ball. Thus we are making a relatively local change this time which does not turn the world upside-down. However, even here things could turn out badly for the wishful thinker. We know that all balls which are used are carefully tested for uni­formity, and if this ball is to be more resilient, then it must be that all such balls are more resilient. In that case the hypothetical Jones would presumably be used to this fact and might play deeper. Thus he could have caught the ball even if it had been higher, and for an unsuspected reason. Of course, this is only one way in which the altered resiliency might affect the situation; we would also ask for the circumstances under which the baseball authorities might have decided on a livelier ball, and the answer to this question might affect the situation still further. It is easy to see how the diverse effects will ordinarily in­volve further alterations, and so on. Of course, it might be that the original situation would be unaffected by these changes, but we will never know this. There will always be the logical possibility that the situation would be affected in unforeseen ways, and any at­tempt to decide the question inductively will result in inconclusive arguments which become increasingly obscure the further they are pushed.

The situation is further confused by the fact that there seem to be alternative sets of counterfactual assumptions that we can make. Since we have supposed that there is a class of counterfactuals which we are justified in asserting on the basis of evidence at hand, our problem is to find a set of counterfactual assumptions which would not affect the relevance of this evidence. Having found such a set of assumptions we should have reasons for thinking that the changes they would involve in the world do not affect the evidence. Even if we grant the assump­tions which justify ordinary prediction (and these are not supposed to be counterfactual assumptions), we are not thereby provided with these reasons. We need not be concerned with the possibility of there being more than one adequate set of counterfactual assumptions. These assumptions are not entailed by the counterfactual and we could switch from one set to another without altering the meaning of the counterfactual. It is just that our reasons for thinking the counter­factual to be true evaporate unless this evidence is shown to be un­affected by at least one set of counterfactual assumptions. Even then, if the hypothetical world corresponding to these assumptions is too different from the actual one, the sustained counterfactual may no lon­ger be interesting or relevant. All this need not imply that counter-

Page 151: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CAUSES AND COUNTERFACTUALS 143

factual induction is mistaken or unreliable, but there is a gap in our inductive assumptions which has not been filled.1

It is particularly important to notice the role of generalizations in this process. Whenever we change an event we usually violate at least one general principle which holds true of the actual world. In order to save it we have to change other events, but they are instances of other generalizations which are supposed to hold, so the process is repeated over and over again, apparently with expanding consequen­ces. The more we try to allow for the original alteration the deeper we get. These problems do not arise explicitly in common sense reasoning, but they are presupposed by it in that our ordinary language is filled with counterfactuals of all kinds, asserted without any qualifications whatever. Insofar as we do consider this problem at all we probably try to minimize it in the following way. We tend to make alterations in the hypothetical world until we come to a case of human agency, and there we stop. In the above case we would probably suppose that Smith swung earlier or harder for reasons of his own, and there our inquiry ends. It is not really that we suppose that his actions are un­determined at this point. On the contrary, the process of a batter swinging at a ball is a fairly mechanical one which seems much easier to explain than a phenomenon such as a complex ethical decision. Rather, it is the fact that the causal laws which operate here are less well known than the laws of gravity and resiliency, and much harder to state precisely. Psychologically it is much easier to set aside an un­familiar law which we suppose to exist than it is to abrogate a perfectly well known law. It is also more attractive to many than the alternative process of changing an indefinite number of events reaching perhaps infinitely far back in time in order to keep the generalizations the same. A scientifically oriented person might be willing to keep the generaliza­tions the same at any cost, but the generalizations have no logical priority over the particular facts, and there is no reason why this has to be done. The more natural tendency seems to be to abrogate some largely unknown and seemingly innocuous generalizations. As we have seen, from the point of view of counterfactuals neither process is particularly innocuous. Itis logically possible that the consequences of changing one event in the hypothetical world would upset the apple­cart.

We can now summarize the problem apart from any particular

1 For a discussion of the inductive assumptions see A. Burks, "On the Presuppositions of Induction," Review of Metaphysics, June, 1955.

Page 152: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

144 CAUSALITY

examples. When we allow for the possibility of the antecedent's being true in the case of a counterfactual, we are hypothetically substituting a different world for the actual one. It has to be supposed that this hypothetical world is as much like the actual one as possible so that we will have grounds for saying that the consequent would be realized in such a world. There are two extreme forms that this assumption might take. On the one hand we might suppose that exactly the same causal laws hold for the hypothetical world. Since we are altering the event represented by the antecedent of the conditional, we will then have to alter an indefinite number of other events according to these causal laws. This chain will presumably reach back indefinitely far in time and spread out a great deal in space with the sorts of problems that we have noted above. The opposite assumption would be that we need alter only the event represented by the antecedent; we would then try to find distinct but similar causal laws which would allow the antece­dent to be true. As we have also seen, the likelihood of being able to do this is small.

Rather, the most plausible assumption which we can make in any particular case seems to be a mixture of the two kinds mentioned above. We alter a series of events until we get to a "natural" breaking off place, and it is there that we alter the minimum number of causal laws so as to limit the number of changes we have to make in the history of the world. Such a leaving off place is one that provides as much continuity as possible. In the aforementioned example we would not keep the trajectory of the ball exactly the same and then make it "jump" three feet because this would break up the continuity of a single event. It is more plausible to alter the causal relation between different kinds of events than it is to violate the simple laws of motion or make objects appear and disappear mysteriously. Thus the view of reflective com­mon sense is not too far off the mark since it does seem to be a combina­tion of the two kinds of assumption that gives us the best hope for an orderly hypothetical universe which will be as similar as possible to the actual one. It is just that there is no reason to pick out a case of human agency as the breaking off point. The hypothetical world will thus have most of the same laws and much the same history as the actual world, but the differences which do exist between the two worlds might still involve very different futures.

These assumptions are affected by the fact that one can take various views of the actual causal structure of the world. On the one hand we might suppose that there are a very few causal laws which are very

Page 153: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CAUSES AND COUNTERFACTUALS 145

basic; this would, of course, imply that there are only a very few kinds of uniformity in nature and that everything, or almost everything, that happens is an instance of one of those uniformities. In this case the alteration of one event would either be very destructive of these uni­formities, or it would force us to change a great part of the history of the world. We would then have to assume in asserting a counterfactual that there are other uniformities, we know not what, which would still produce the expected result, or that the world had a different history, we know not what, which would produce the result. In either case a counterfactual for which we had good reasons in the ordinary sense becomes groundless. On another view of the world, however, causal laws would be less basic and more numerous; the uniformities would be of the form, "All events of the type A which occur under conditions B are characterized by C," where the relevant properties might not be sub-atomic. In this case the alteration of a single event or small num­ber of events need not upset the whole causal structure of the world, but only one pattern of uniformity. The latter seems to be the view closer to common sense, and if true, would make it easier to justify the counter­factuals we assert by appealing to other uniformities which we suppose to be unaffected. It is interesting to note that this assumption is very different from the one mentioned earlier involving durable properties; in fact, they represent opposite pictures of the world. Fortunately, there is no need to make both assumptions at once. If we decide to go with the Humean analysis, we must assume durable properties and we can then translate "If a had occurred then b would have occurred" as "All As are Bs and a did not occur." Ifwe decide that causal laws entail counterfactuals and counterfactuals come to play too important a role to be analyzed in this way, we must then make the present assumptions.

This whole issue seems to be logically independent of the issue of freedom and determinism. Even if we suppose that there is a multi­plicity of limited causal laws each describing different domains, it might still be that the causal laws all operate without exception within that domain. Determinism does not demand that the alteration of one event affect everything else in the universe; rather it demands only that the substitution of one event for another affect all the events within a certain class.

It will now follow that the usual inductive assumptions do not en­tirely support counterfactual induction, and, in order to see why this is so, let us take for granted a very strong inductive principle that would completely support induction by simple enumeration. This

Page 154: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

146 CAUSALITY

would be that all constant conjunctions of events which have held in the past will also hold for the future. This is such a strong principle that it is patently false and would never allow a prediction on the grounds of simple enumeration to be false, no matter how bad the sample; however, it will do all the better as an example. The counter­factual is not supported by this principle for several reasons. First, it is not a prediction at all, and cannot be directly verified in the future in the way that statements about white swans can be. Of course, there are many non-counterfactual predictions which are not based on simple enumeration either, at least in any obvious way. In particular, we make many hypothetical predictions which are not meant to be counter­factual, and the basis for these is usually the supposed operation of certain causal laws. If the evidence for these laws is indirect then the inductive principle which supports enumerative induction may not support these hypothetical predictions. However, we do suppose that the law, whatever it may be, has operated in the past, and we can now support the hypothetical prediction by adding the second assumption that the same laws which have held true of the past will hold for the future.1 This second assumption does not provide a basis for the counterfactual because we cannot be sure that the changes in the hypothetical world involved in the counterfactual antecedent will not conflict with and undermine the evidence for thinking the counter­factual to be true. The relevant law brings about its effect only if the antecedent is realized and another list of conditions, only part of which can be specified, is fulfilled. In the first place, there might be a conflict between the antecedent and one of the unspecified conditions according to the laws which hold for both the actual and hypothetical worlds; thus we could be assured of the universal truth of the law, but still not know that it applies to the counterfactual situation. Secondly, it might turn out that it is convenient to set this law aside in the hypothetical world; in this case the law might always be true of the actual world, but not of the hypothetical world. The counterfactual asserts more than that the law holds for the actual world. It asserts that the same or simi­lar law holds for the hypothetical world, and that it would operate to produce a certain effect in that world. It is in this way that it goes beyond the ordinary hypothetical prediction.

Our conclusion then is that counterfactual reasoning, although a kind of induction, is so different from other kinds of induction that the

1 For this purpose one can think of the laws either as universally true generalizations or as Burks construes them in "The Logic of Causal Propositions," op. cit.

Page 155: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CAUSES AND COUNTERFACTUALS 147

usual inductive assumptions are not even relevant to it. On the other hand, it is not easy to state precisely the assumptions that are required. It seems possible only to lay down certain necessary conditions for the plausibility of a counterfactual. These, as stated above, will be that the respects in which the hypothetical world differs from the actual one, apart from the counterfactual antecedent, are not related causally to the consequent to any great extent, and further, to the extent that there is such a relationship, these laws are independent of other laws, and can be set aside without severe consequences. Thus it seems that, even if the solipsist tries to avoid making the assumptions involved in re­ducing causal laws to universal generalizations and takes to counter­factuals instead, he must still make factual assumptions beyond the usual inductive ones.

Given the fact that we can legitimately use counterfactuals, we might note the difference between the comprehension of them and the comprehension of the corresponding material conditionals. Since p :::J q is an abbreviation of '" p v q, any imagined or sensed state of affairs which is a complete instance of either '" p or q will be a complete in­stance of p :::J q. However, in the case of "If p then q" when it is inter­preted as having counterfactual force and implying "If p were the case, q would be the case," a complete instance of '" p will only be a partial instance of "If p then q." Similarly a complete instance of q will only be a partial instance of "If p then q" in this sense. In fact the rules of instance hood for p :::J q simply do not apply at all to "Ifp then q." As has been suggested earlier, a complete instance of a counterfactual involves a whole hypothetical world in which the counterfactual ante­cedent is consistent with the laws that obtain in it. In fact, a counter­factual of the form, "If p then q" has been analysed as follows: There is a hypothetical world such that p is true of it and this world contains a causal law such that q will follow on p and other facts about this world; further, this hypothetical world is similar to the actual one in most important respects. Since we have to refer to "important respects" a counterfactual assertion turns out to contain a valuation among other things, and this will sometimes make it impossible for different people to agree as to its truth. Nevertheless, it will be possible to get a greater measure of agreement on what is an important respect than is possible in most ethical contexts.

We cannot in fact sense or imagine such a hypothetical world in its entirety, but it can be claimed that we know how to go about it and that there is no single aspect of such a hypothetical world which is

Page 156: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

148 CAUSALITY

unimaginable; it is just that life is too short for such academic exercises. If causal laws are analyzed extensionally, we can imagine states of affairs in which they are literally satisfied. However, even if they are felt to have some "extra strength" we can still imagine states of affairs which do not violate these laws and for which they would hold. This latter is all that is required for the comprehension of counterfactuals. As with the comprehension of other kinds of statements we must be able to imagine a range of partial instances, and it must be possible with sufficient effort to imagine a complete instance.

We must now see what implications the above discussions of counter­factuals and causal statements have for the theories being put forward here. One conclusion was that counterfactuals are an indispensible part of our everyday language. If additional assumptions are required to support counterfactual induction, then these assumptions simply must be made and they represent the presuppositions of our language. We have already made use of counterfactuals and dispositions in the analysis of meaning and will make use of more in subsequent analyses. In all these analyses the aim is to formulate sentences which have the same meaning as certain sentences of ordinary English. Since counter­factuals are an essential part of everyday speech, there is nothing ob­jectionable to their presence in the language of analyis. I am not trying to support the extensionality thesis and my language of analysis is not restricted to truth-functional connectives. As will be seen, the language of analysis has a certain kind of economy in that there are names in it for only a very few kinds of entities, but the logic of this language is no more economical than that of the ordinary language with respect to the causal modalities. Of course, I am not necessarily denying the truth of the extensionality thesis and the possibility of an exclusively truth-functional language of analysis. In our discussion of causation, for instance, we came to no definite conclusion as to whether causal connections are reducible to constant conjunctions. This being the case, I will abstract from this question and leave it open. Someone who feels that counterfactual and causal statements can be reduced to truth functional statements could very easily carry my analyses a step further than I intend to take them.

In other words, one can go the whole way with solipsism and concur in the extensionality thesis as well as the other doctrines which are associated with reductionism or one can stop short at this point with a language of analysis which is not as economical as it might be. My concern will be to show that even without the assumption of extension-

Page 157: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CAUSES AND COUNTERFACTUALS 149

ality one can be a solipsist about other minds and physical objects and still retain our everyday speech and the beliefs that go with it. Hence we will forsake logical economy for metaphysical economy in the following chapters.

Page 158: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PART III

SOLIPSISM PROPER

Page 159: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 9

A QUASI-AXIOMATIC SOLIPSISTIC SYSTEM

In developing our solipsistic system we will use a quasi-axiomatic approach. For reasons that will shortly be appreciated we cannot use a fully axiomatic method, but it will be helpful to go as far in this direc­tion as possible. The axiomatic method gives us the considerable ad­vantages of increased clarity, rigor, and elegance, but the main reason for using the method here is that it allows us to see more clearly what it is that we are doing while we are doing it. The most prominent use of the axiomatic method in philosophy was that of Spinoza, but Spino­za's system lacks rigor and many of the theorems do not really follow from the axioms by methods of deduction. Further, the kind of method he had in mind was that of Euclid and we here want to use the modern axiomatic method. We will therefore begin with a brief discussion of changes in the axiomatic method which have taken place in modern times.

Euclid's conception of the method was to define the key terms of the subject-matter, in his case geometry, as completely as possible so that they would be intelligible to anyone who studied the definitions. Hence he defines a point as "that which has no parts" and a line as "breadth­less length." Next the axioms are statements which involve these key terms and which will be intuitively obvious to anyone who has followed the necessary definitions. Hence we get such principles as "That all right angles are equal to one another" (Post. 4) and construction axioms such as "To describe a circle with any centre and distance" (Post. 3). The theorems are then to follow deductively from the axioms and if the axioms are certain, as they are supposed to be, then so are the theorems. Finally, the hope was that all the truths of geometry would appear in the system either as axioms or as theorems. Histori­cally, of course, the concern over the parallel postulate and the sub­sequent invention of the non-Euclidean geometries had much to do with the abandoning of this conception of the axiomatic method. How­ever, the factor that was ultimately responsible for this rejection was the realization that geometrical statements, particularly the axioms of

Page 160: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

154 SOLIPSISM PROPER

a geometry, are not just true or false but that they either apply to a subject matter or do not apply to it. Consequently the axioms of differ­ent systems could be interpreted as describing different subject matters. Thus one can see that the axioms of no geometry can be intuitively self­evident nor can all the principles which could be correctly applied to some subject matter or other be consistently combined in one system. Further it was realized that if one alters the definitions of certain key terms one can gradually apply a geometry, or some other kind of mathematical system, to more and more subject matters. These various realizations were most permanently combined for mathematics general­ly by Hilbert in his formalist program. The axiomatic conception here is that certain terms are chosen as primitive and that these terms have no meaning in themselves; they may be regarded just as marks on a piece of paper. The primitive terms can be given a meaning if we wish to apply the system to some subject matter, and this is called "inter­preting" the system. A system may be capable of being interpreted in many ways, but the primitives remain meaningless in themselves and do not in any way "absorb" these interpretations. The system itself can be constructed without interpreting the primitives at all. The axioms then become, not statements which are self-evidently true, but state­ments which are in themselves meaningless. When the primitives are interpreted the axioms take on meaning and become true or false state­ments about a certain subject matter. If there is any interpretation in which all the axioms are simultaneously satisfied, the system is then said to be consistent. If one of the axioms were self-contradictory it could never be satisfied at all and if two or more axioms were mutually inconsistent they could never all be satisfied at once; hence the finding of such an interpretation has special significance and it is called finding a "model" for the system. The system can no longer be complete in Euclid's sense of containing all the truths of the subject matter since it is no longer supposed that the axioms and theorems are truths at all, at least in themselves. Hence the new conception of completeness is that a system is complete if any statement that can be formulated in the language of the system at all is such that either it or its negation is provable as a theorem. The method of deduction constitutes the respect in which the axiomatic method has changed the least, and this is where we can see the historical continuity between Euclid and Hilbert. In both cases the method of deducing theorems from axioms is just the method of deductive logic. Even in modern mathematics it is not usual to specify the particular logical rules which must be followed in deriving

Page 161: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 155

theorems from axioms. There are some disputes about the legitimacy of certain techniques of proof, but for the most part, mathematicians use the principles of modern deductive logic without question. It is only in systems of logic that particular rules of inference are likely to be specified.

As an example of a modern axiom system we will take R. L. Wilder's formalization of a fragment of geometry. 1

Primitive terms: Point; line

Axiom 1. Every line is a collection of points. Axiom 2. There exist at least two points. Axiom 3. If p and q are points, then there exists one and only one

line containing p and q. Axiom 4. If L is a line, then there exists a point not on L. Axiom 5. If L is a line, and p is a point not on L, then there exists

one and only one line containing p that is parallel to L.

Definition 1. If a point p is an element of the collection of points which constitutes a line (cf. Axiom 1), then we say, variously, that L contains p, p is on L, or L is a line con­taining p.

Definition 2. Two lines Ll and L2 are called parallel if there is no point which is on both Ll and L2•

Wilder then goes on to prove such theorems as "Every point is on at least two distinct lines" and "Every line contains at least one point." The proofs are stated without giving any interpretation of the primi­tives. When we come to finding a model of the system and proving its consistency (in one of the many senses which this term ultimately turns out to have), we do have to give the primitives an appropriate inter­pretation. We will therefore interpret "point" to mean "circular blob of ink printed below." There will then turn out to be the four points arranged and named below.

p. .q r .

. s Axiom 3 then tells us that we must have lines between every pair of points and axiom 1 tells us that these lines will be sets of points. Hence there will be the following lines: [p, q], [p, r], [p, s], [q, r], [q, s], [r, s]. It is important to notice that we have used set terminology here so that "[p, q]" should be read "the set which contains only p and q"

1 R. L. Wilder, Introduction to the Foundations of Mathtmatzcs, p. 10.

Page 162: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

156 SOLIPSISM PROPER

and not just as "the line pq." In fact if we were to draw with a pencil a line from p to q that line would have nothing to do with "the line pq" under this interpretation. Axiom 1 forces us to interpret "line" so that this particular line is a conceptual entity, a set, and not something one can see such as a line on a piece of paper. In fact, axiom 1 operates as a definition of "line" and Wilder could have taken just one primi­tive and treated it as a definition. The difference between this sort of system and a traditional one could be seen even more clearly if we arranged our four points as follows:

p. q.

r. s.

We would still have the same lines and the diagonals would not only not intersect but they would be parallel. The lines in question, [p, s] and [r, q] are sets which have no common members; thus they are parallel according to definition 2. By applying the axioms to the situa­tion of the four points and six lines one can easily see that all the axioms will be true, and we consequently have here a model of the system. Wilder points out, however, that we can get a model by taking an interpretation which has no connection at all with geometry. Thus we can interpret a point to be a person or a book and a line would be a two-person club or a two-book library. As long as we take four persons and six clubs or four books and six libraries and they are related in the same way as the points and lines above, we will still get models and the models will be isomorphic. That is, all the same statements that can be formulated within the language of the system will be equally true of our three models.

Another example of an uninterpreted system, this time from "work­ing" mathematics, would be the axioms for group theory. Here we can take the primitive terms as G (the set of objects which constitutes the group), "0" (read "operation"), "e" (read "identity element"), and '"'' affixed to any object (x' is read "x inverse"). We could get along with fewer primitives by using definitions but it will be clearer this way. The axioms are:

AI. (x)(y)((x e G. y e G) :::> (x 0 y) e G) For any x and y, if they belong to the group then x 0 y also belongs to it.

A2. (x) (y) (z) ((x 0 (y 0 z) = (x 0 y) 0 z) A principle of association. A3. (x) (x 0 e = x) For any x, x 0 e is equal to x. Hence e is called

the identity element.

Page 163: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 157

A4. (x) (x 0 x' = e) For any x, x operation its inverse is equal to the identity element.

Let us start by interpreting "G" as the set of integers (both positive and negative), "e" as 0, the inverse sign as the minus sign, and "0" as addition. We can then see that the axioms are all satisfied. The sum of any two integers is an integer, and the principle of association holds for addition. Any number plus zero equals itself and any number plus the corresponding negative (or positive) number equals zero. Hence this system is consistent provided that the system of arithmetic on the integers is consistent.

A different model for the same system can be obtained by interpret­ing G as the set of rational numbers (fractions constructed out of integers) greater than 0, "e" as 1, the operation as multiplication, and the inverse as the reciprocal of a number (formed by inverting the fraction). The product of any two such numbers must also belong to the group and multiplication is also associative. Again, the product of any member of G and 1 will be equal to the member itself and the product of any such number and its reciprocal will be 1. Usually systems have an intended or favored interpretation which has to do with the motives for the construction of the system, but in this case one would have to speak of both these interpretations as intended ones. We are now in a position to compare the sorts of philosophical systems we have in mind with mathematical systems.

Like a mathematical system our philosophical system will have primitive terms which in themselves have no meaning, and also like a mathematical system our system will have an intended interpretation. Of course, an interpretation, intended or not, always involves an appli­cation of the system, but it often happens in mathematics that we find our model for one system in another system. This is what happened when we gave arithmetical interpretations to the primitives of group theory. However, our interpretation of Wilder's system was quite different in that the model produced there consisted of a set of physical objects in the world - a set of blobs of ink on a piece of paper. When we find a model in the world in this sense we cannot actually prove that the axioms on that interpretation hold for the situation which consti­tutes the model in the way that, for instance, the associative law for addition could have been proven. On the contrary, establishing the truth of the axioms, as interpreted, is in part an empirical matter even though, as in the present case, we may be able to obtain practical certainty. This is to say that the axioms of the system, as interpreted,

Page 164: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

158 SOLIPSISM PROPER

accord with empirical beliefs about the world which were already held with great tenacity. And, of course, if we believe the axioms we must also believe all the theorems which can be derived from them. Thus when one applies a mathematical system outside of mathematics to empirical subject matter one can speak of systematizing our beliefs about that subject matter. In the case of a philosophical system the set of beliefs we are trying to systematize consists simply of all our beliefs about the world and any part of it. In fact the relevant set of beliefs is so vast that we cannot really construct the system itself but can only sketch out directions for its construction.

As in the case of mathematics there are various different ways of axiomatizing this system of beliefs, and it is possible to choose different sets of primitive terms and different sets of axioms. Nevertheless, in whatever way one proceeds, roughly the same set of theorems should be derivable. These theorems should include, on the intended inter­pretation, not only our scientific beliefs but our common sense beliefs about the world and other persons. In our system the primitive terms will be chosen in such a way that, again on the intended interpretation, they will refer to sensations, and the axioms will then be statements that ultimately talk about sensations.

We said before that the function of philosophy is just to give direc­tions for axiomatizing our beliefs about the world. The axioms will themselves represent some of those beliefs but the actual statement of the axioms lies outside the scope of philosophy, at least as traditionally conceived. Rather we merely characterize them and say what sorts of principles would be axioms. In fact, since different persons are in different situations and have differing beliefs about the world, there would have to be a somewhat different set of axioms for each person. Further, since we accumulate more information as time goes on and lose other information, the axiom system might have to be JOT a partic­ular person at a particular time. However, if our directions were follow­ed all such axioms would ultimately talk about a person's own sensa­tions. Hence our system will be solipsistic in that the axioms refer to such a limited scope of phenomena.

As it turns out, the crux of our system will be in the definitions and the defined terms. It is typical of a mathematical system such as group theory to start with concepts which are very simple on their intended interpretation and to gradually add by definition very complex con­cepts. Theorems are then proven which reflect these definitions and their relations to the axioms. Similarly, in our philosophical system we

Page 165: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

QUSI-AXIOMATIC SOLIPSISTIC SYSTEEM 159

start with primitive terms which are, as interpreted, very simple, and we ultimately develop complex definitions which allow us to translate statements about the material world, persons, and beliefs into our primitive language. In an axiomatic system as such the definitions are a mere convenience in the sense that they abbreviate and allow us to say in a short way what would otherwise consume many lines of sym­bolism. This is also true of our system and we can look upon the defined terms merely as convenient abbreviations. However, there is a great difference involving the use of definitions between a mathematical system and a philosophical one. In mathematics it does not matter what definitions we bring in; some turn out to be more useful and interesting than others but the system can be developed in any number of different directions and, as long as one can prove a sufficient body of theorems to make it interesting, it can be a successful system no matter how developed. In a philosophical system, on the other hand, we do not have this freedom. The difference is that a mathematical system does not have to have any particular application in order to be a perfectly good system in itself. But a philosophical system, being only a sketch of a system, cannot really be developed in the abstract and is only of interest if it has a certain sort of application. In particular, it is only of interest if it can be used to systematize the beliefs we already have. It is as if one were to develop a mathematical system with an interpretation in mind before choosing the primitives and axioms and to count the system a success only if it turned out to have the desired application. Hence, the philosophical system which we will sketch must be such that we can derive as theorems in it our ordinary beliefs about persons in the next room, the sun's rising tomorrow, and so on. The system must also provide the foundation for however much science we want and it must not be too powerful and provide us with theorems which we actively disbelieve. Further, the theorems that are derived must actually say, as interpreted, the same things as the statements of our ordinary language which are used to express these beliefs. Thus it is not just that we have to derive the needed theorems, but the defini­tions we provide, while they are just notational abbreviations, must nevertheless be abbreviations for statements within the system which are synonymous with the statements of ordinary language. In order to say that they are synonymous we need a criterion of synonymy and that is the reason for our prior development of a theory of meaning which provides it.

We will now compare a philosophical system with an axiomatized

Page 166: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

160 SOLIPSISM PROPER

scientific theory. It is characteristic of a scientific theory that at least some of the important inferences are inductive rather than deductive. Since all theorems in an axiomatic system must be deduced from the axioms the structure of the scientific theory has to be re-arranged when it is axiomatized. The key inductive inferences will be stated as axioms of hypothetical form and applied to other axioms and theorems so that the rules of inference will be entirely deductive. Again, the primitives are uninterpreted in themselves but may be given interpretations as desired; the intended interpretation will probably be more important than it would be in a mathematical system and alternative interpreta­tions will be rare. On the intended interpretation the axioms will, of course, be statements about the world and they must be consistent with observations that have been made. It must also be possible to deduce predictions about the future from the theory and the accuracy of the predictions is a test of the adequacy of the theory. Strictly speak­ing, if even one of the predictions turns out to be false, the premises from which it is deduced must also be false and at least one of the axioms must be false. At this point it may be possible to construct a new system which is very much like the old one, but which differs enough so that the theorem in question can no longer be deduced. On the other hand, it may be that too many of the axioms were involved in the ori­ginal deduction for such a reconstruction to be possible.

Here then we have the notion of explanation and comparative ex­planation. An axiomatized theory explains whatever observations and predictions are deducible from it, and the verification of these predic­tions constitutes confirmation of the theory. Two theories may both explain a set of phenomena which has been observed to occur, but one may explain it more economically than the other in that the axioms of one may be fewer and/or simpler than those of the other. There is here a somewhat misleading parallel with a philosophical system. There, too, there seems to be a kind of explanation. We know that in our philosophical system we must in the end be able to derive the beliefs of common sense and in an axiomatized scientific system we have similarly to be able to derive predictions which turn out to be true. One difference is that in the case of our philosophical system we are ultimately trying to produce statements that we believe without ques­tion while in science we are trying to produce statements that will turn out to be true under scrutiny. This is reflected by the fact that in a philosophical system we know in advance what statements will have to be deducible as theorems. In the scientific theory one does not know

Page 167: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 161

in advance what predictions one will have to make; one just has to make some set or other of predictions which can and will be verified.

The most important difference concerns the meaning of the state­ments which will have to be proven as theorems. A scientific statement which says, for instance, that a particular planet will. be at a particular position at a particular time, says what it means in a clear-cut way and the meaning of the statement does not usually have to be settled within the system itself. There are exceptional cases where a phenomenon can be observed in more than one way with different results and the theory itself has to decide between various possible operational definitions; however, the necessary operational definitions are usually embedded in scientific practice prior to the formulation of any particular theory. Now it might seem that there is even less room for dispute about the meaning of such a statement as "There is a grocery store down the street." But the fact is that the scientist can abstain from a great many disputes into which the philosopher must plunge headlong. In fact, the dispute between philosophical systems is ultimately going to rest on the interpretation of just such statements as this one about the grocery store. It will be provable in all the competing systems and perhaps with various degrees of economy, but that is not the whole story as it would be in science. One philosopher may assert that we cannot do justice to our ordinary assertions about grocery stores with­out mentioning some sort of substance or something which underlies what we see and hear. Another philosopher may analyze such state­ments in such a way that they refer only to the sensations we would have under certain circumstances. This difference will be reflected in the two systems which are subsequently constructed, and the latter will probably be more economical than the former. If the systems were scientific ones, and we assumed them consistent, we would choose the more economical system, but in this case the substance philosopher can always say that his opponent is able to derive the statement about the grocery store economically only because he leaves out or analyzes away part of the meaning of the statement. The opponent will undoubtedly deny this, but it can be seen that there is an additional question here beyond economy of explanation. The situation is clarified if one thinks of a philosophical system as having two main jobs. On one hand, it must produce as theorems statements which correspond to our common sense beliefs and, on the other hand, it must within the system produce as defined terms concepts which are equivalent to our ordinary con­cepts. In our example the substance philosopher argues against the

Page 168: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

162 SOLIPSISM PROPER

phenomenalist to the effect that the latter may have accomplished the former task but that he has neglected the latter. There seems to be no corresponding demand in the case of a scientific system. All that is needed there is that predictions be produced and that they be con­firmable. The theory itself may specify how the predictions are to be verified, or the means of verification may be specified outside the theory. In either case what is demanded is that the predictions and concepts have an empirical meaning, but it is not demanded that they have some already specified meaning - that they have the same meaning that certain concepts and statements belonging to our everyday conversa­tion have. Hence one can speak of explanation in a philosophical system and compare it with explanation in a scientific system as long as one remembers this basic difference.

There is one other respect in which a philosophical system resembles a scientific system. In both cases it is not necessary that the axioms be believed prior to the construction of the system or even that they be particularly plausible. An axiomatized system of physics may contain axioms which postulate the existence of entities which are completely unknown otherwise. As long as the system explains the observations that have been made and which can be made, and is reasonably economical, it is nothing against the axioms that they have no intuitive plausibility. Similarly, if a philosophical system yields as theorems the beliefs of common sense and these theorems really are logically equiv­alent, as interpreted, to the corresponding statements of ordinary lan­guage, the axioms of the system need not be particularly plausible. We will see later that there is a decided advantage in choosing for our system axioms which are close to our common sense beliefs, but this is not a requirement of a philosophical system as such.

Of course, there is still the consideration of economy which allows us to prefer one system over another, and in a philosophical system economy has a great deal to do with the number of different kinds of primitive terms admitted, although this is not the only factor involved. But economy in this sense has nothing to do with whether the axioms talk about familiar sorts of objects and say things that we already be­lieve or are inclined to believe. Even if the axioms assert the existence of such things as unsensed sensations and a satisfactory interpretation of that concept can be provided in the metalanguage (that is, a theory of meaning is adopted that makes the concept make sense), it is not a defect in the system that the axioms postulate the existence of such entities. In a system constructed according to our directions many of

Page 169: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

QUASI-AXIOMATIC SOLIPSISTIC SYSTEM 163

the axioms will be very simple statements about sensations and will either be certain or close to it. However, other axioms, while they still talk about sensations, will assert complex relationships to hold among them and will be considerably removed from intuition. Even if all the axioms were very intuitive we would not claim it as an advantage for the system - the important thing is not that the axioms be common­sensical but that they provide the basis for common sense belief. Of course, the axioms cannot be directly contrary to common sense belief or the system could not then give us what we want and still be consis­tent. Since anything can be deduced as a theorem in an inconsistent system any sort of axiomatic system, including a philosophical one, must be consistent to be at all interesting.

All this relates to the way in which ontological commitments arise. It seems that each of us has a large body of common sense beliefs which we are unwilling to give up. I should not, perhaps, claim that the state­ment, "I am now sitting in a wooden chair," is certain, as Moore might have done, but, certain or not, it is not one that I am going to give up very easily. There are a very large number of other beliefs of similar kinds: "I am now on the second floor of my house," "I recently painted my house," "There is a country called England," "England contains a queen," and so on. We are not absolutely committed to all these be­liefs and under very extraordinary circumstances we might occasionally give one up. We would have to do so if, for instance, we found out that two or more are inconsistent. The point is that even if one were forced occasionally to give up a few of these beliefs, they would be replaced by others of similar kind. It is a psychological fact about us that we do not suspend judgment wherever it is logically possible to do so; in fact we do not suspendjudgment in many cases where it would be practical­ly feasible to do so. It is not necessary for me to believe that there is an England in order to go about my everyday business, but I am never­theless very strongly committed to that belief and to many others which are not really needed.

Our biological nature is such that we cannot just act but have to have beliefs and articulate them in order to survive in a world to which we are not very well adapted physically. Further, we have many more strongly held beliefs than we need to have to get along in our immediate environment. These beliefs vary from person to person and from time to time as was pointed out before; whenever a person goes from one place to another he can hardly avoid acquiring additional beliefs of this sort, and, as time passes, we almost automatically acquire addi-

Page 170: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

164 SOLIPSISM PROPER

tional such beliefs about events of the day. The philosophical sketch must tell us how to construct an axiom system in which these beliefs appear either as axioms or as theorems. It is also from these beliefs that ontological commitment is to be derived. While we do not want to say that these beliefs themselves constitute ontological commitments, it is obvious that such commitments will ultimately have to come from them in some way. If our commitments were not ultimately founded in these most basic beliefs, it is hard to see where they would come from.

We have also just seen that the question of what these statements we believe so firmly mean is always one of philosophical dispute. Hence, our ontological commitments will depend on how these statements are analyzed. In our previous example we pointed out that these beliefs can be analyzed in such a way that they assert the existence of sub­stances, or they can be analyzed in such a way that they assert the existence of irreducible physical objects, or so that they imply the existence only of sensations, or so that they imply the existence of un­sensed sensations. All these analyses will go with different sorts of systems. If we want to get the common sense beliefs, as analyzed in the various ways, as theorems in a given system, we will then have to have axioms which assert the existence of the required entities. We will also have to have either primitive terms which refer to those entities or a primitive language which is rich enough to permit the definition of terms which do refer to the entities in question. In order to determine ontological commitment one has to look not only to the body of com­mon sense beliefs or to the sorts of entities that we quantify over in ordinary language but to some system or other which has been con­structed according to a philosophical sketch. One can then see, by looking at the primitive terms and the axioms of the system, what it is that we are ontologically committed to. The most significant category is generally that of primitive terms; the axioms will ordinarily assert the existence of the entities to which the primitives, on the intended interpretation, refer, and it is impossible for the axioms to assert the existence of any entities which are not logical constructions out of these basic entities. Assuming that there are no classes of primitive terms which are not actually used in the system, we can then largely deter­mine the kinds of ontological commitment which the system has us make with our ordinary beliefs by seeing what sorts of primitives the system contains and how they are interpreted.

It is, of course, traditional in philosophy to claim that one's system holds true not only for oneself but for everyone else, including those

Page 171: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 165

philosophers who have different systems. We can imagine a philosophi­cal world in which everyone would just claim to analyze his own beliefs and in which two philosophers might compare and contrast their common-sensical beliefs which could differ considerably if they were analyzed in two different ways. One might suppose this to be a more rational philosophical world, but it turns out that there are some criteria according to which we might conclude that someone has failed to analyze correctly even his own beliefs on a certain subject. For in­stance, some philosophers analyze beliefs of categorical form about material objects ultimately in terms of hypothetical statements. It has been argued by Isaiah Berlin that categorical statements perform certain functions for all of us and that these functions can never be taken over by hypothetical statements'! In particular he thinks that categorical statements allow us to implicitly "point" to objects in a way that hypothetical statements do not. This is a legitimate sort of objec­tion; the objector points out that common sense statements ofa certain kind have a function which the analyst has not noticed but which he himself makes use of along with everyone else. If he has analyzed away that function, and we think that the statements in question betray such a function by their ordinary usage, we then have grounds for criticizing the philosopher's analysis of his own beliefs. After having set up our system we will try to make it plausible as an analysis of anyone's beliefs by considering objections of this sort and showing how they can be met. One could, for instance, meet an objection like this one by adding a categorical element to the analysis of such statements, by showing how we can use hypothetical statements to accomplish the same thing, or by arguing that our common sense categorical statements have no such function and no such use in the first place. Hence it does not seem that we are forced to analyze only our own beliefs on the grounds that there are no criteria for a correct analysis of any other person's beliefs. We can at least analyze our own beliefs, suggest the analysis to others, and in some cases intelligibly wonder whether others have correctly analyz­ed their own beliefs.

In developing our system we will first set forth the various classes of primitive terms. Since the system has an intended interpretation, which amounts to an explanation of the way in which the primitive terms are expected to be used, we will give the interpretation for each type of primitive term along with it instead of waiting until the end. For the sake of convenience we will use as much English as possible in

1 1. Berlin, "Empirical Propositions and Hypothetical Statements," Mind, July, 1950.

Page 172: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

166 SOLIPSISM PROPER

constructing the language of the system and where necessary combine it with logical symbolism and other symbolism of our own devising. It should not, of course, be assumed that the English used is in all cases ordinary English or that it is being used in an ordinary way. In every case the explanation will tell how the terminology is to be used on the intended interpretation. The explanations themselves will not be given in the language of the system but rather in the metalanguage, a lan­guage that is used for talking about the system; the metalanguage will consist simply of ordinary English with necessary technical terms intro­duced into it by definition.

The first class of primitive terms consists of expressions which will be interpreted as descriptions of particular kinds of sensations. These will consist in the word "sensation" preceded by an expression which will pick out a quality. When we earlier discussed independent and dependent sensations, it was pointed out that for any independent sensation there corresponds a dependent sensation and vice versa. The difference was that in dependent sensation the object sensed was be­lieved to be identical with the conscious state itself while in independent sensation the object sensed was supposed to be logically distinct from the conscious state itself. We will therefore interpret "sensation" in such a way that it is neutral between the two - that is, the term will be taken to denote what is common between an independent sensation and a corresponding dependent sensation. It will not include the element of belief or disbelief which distinguishes the one from the other. Further, it follows that the term denotes an exclusively mental or conscious process and will not denote any phenomena of the brain or nervous system or any behavioral pattern. There is, of course, a dispute over the possibility of talking about exclusively mental phenomena in this way. Before it was argued that not all mental phenomena are reducible to physical phenomena and in a later chapter the legitimacy of this use of language will be examined in much greater detail. Thus the term "sensation" has been introduced into the language as a primitive term.

More complex primitive terms could be constructed by preceding the word "sensation" with a quality word such as "blue." However, there are not really enough such words, which we can apply directly to sensa­tions, to allow us to completely describe our sensations and discrimi­nate them from each other. There are a few quality words such as "painful" which are applied primarily to sensations as opposed to physical objects, others such as "blue" which can be applied naturally to either sensations or physical objects, and many others such as

Page 173: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 167

"saturated" which are applied exclusively to physical objects. In many cases sensations are described in terms of the physical objects which ordinarily produce those sensations in observers. Thus there is a visual sensation whose content is made up oflittle multi-colored squares and is one that we ordinarily have when we look at a person in traditional Scottish dress and look somewhere between the waist and the knees. We do not have a name for this kind of sensation and do not often have occasion to talk about the sensation as opposed to the object. When we do have to talk about it, we describe it as a sensation "characteristic of looking at a kilt" or as "a kilt-like sensation." At this point many people wonder if sensations can be described at all without also de­scribing physical objects. Of course, it is fairly obvious that some sensa­tions, such as pains, are described without reference to physical objects. But ultimately, in order to say how we can and cannot talk about sensations, we have to look to a theory of meaning. On our theory quality words have meaning only because we can eventually recognize sensations and images as instances of them; on this theory images and sensations are more basic to meaning than are physical objects. We can find instances of words and propositional functions among physical objects only derivatively as a result of being able to find instances among images and sensations. Further, according to this theory, the basic mechanism of learning language revolves around ostension.1

The idea is that we learn quality words in the first place by associating them with sensations, which leads to their association with qualitatively similar images, and it is only later, and as a result of this association, that we can apply the terminology so learned to the sorts of physical situations that produce the relevant sensations. Hence it follows imme­diately from this theory of meaning that we can apply quality words directly to sensations without having to go via physical objects. We can therefore invent as many new such quality words as we need in order to fill out the sensation language. In theory there is nothing to prevent our using the same ostensive technique to mark out any distinction at all between kinds of sensations. These differences will always be recog­nizable, since there could not be any unrecognizable differences be­tween sensations, and the only practical limitations on developing such a vocabulary would be the difficulty of remembering slight differences between sensations and associating so many different kinds of sensations with so many different quality words. However, the practical difficul-

1 For a related discussion see W. Todd, "The Theory of Meaning and the Learning of Language," Inquiry. Vol. 8, Winter, 1965.

Page 174: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

168 SOLIPSISM PROPER

ties are not important since we are not suggesting that anyone actually use such a language. The theoretical difficulties have been dealt with in our theory of meaning and in our private language argument.

We do not have to be able to use a language as interpreted in a particular way even in theory in order to base an axiom system on it, but when we are dealing with a philosophical system we want the language, as we interpret it, to be one we could use without supposing ourselves to have any special abilities beyond the ability to use osten­sion, the ability to remember, to imagine, and so on. Thus while the language is not intended to be usable in practice we want it to be one that we could use without postulating anything more than a much better memory. But even if we would need better memories in order to use such a language we would still remember in the same way that we do now.

All this being the case, the convention will now be adopted of placing in front of "sensation" not a standard quality word but a series of dots as in " ....... sensation." Any particular number of dots will be as­sociated with a sense quality and the expression could be learned in the way that we suppose "red" to have been learned. We will never have need for more than a finite number of dots. We must first note the fact that all our senses have thresholds and we can never take note of differences in stimulations which fall below a certain minimum level. Further, all sense fields have maxima and minima of some sort: there is a maximum intensity oflight that makes a difference to our visual sensations (when we are blinded) and a minimum intensity (when it is too dark to see at all); as regards hue, there is no way of escaping from the color spectrum, and so on. Thus there are only a finite number of different kinds of sensations that we can have and a finite number of dots will always serve. We can now give a very simple recursive defini­tion for the membership of any locution in this class of primitive terms.

Definition for Class I of primitives: (1) "sensation" belongs to I (2) If a belongs to I then .a belongs to I

This definition allows us to generate "sensation" preceded by any number of dots as a well-formed formula belonging to the first class of primitives.

So far we have a way of talking about sensations, but we have no way of asserting anything about them in the language of the system. Thus we will introduce the term "occurs" which can be preceded by

Page 175: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 169

any well-formed formula (wff) of class I. On the intended interpreta­tion the complex so formed will be completely timeless and will say only that a sensation of the required type has occurred or will occur at some time or other. Hence, the use of the present tense here is mis­leading, but cannot be helped.

However, it will be necessary in the axioms to refer to sensations which are occurring at the present time. We will therefore also some­times say not only that a particular kind of sensation occurs, but say that it "occurs now." The term "now" differs from the others so far introduced in that it will be a defined term rather than a primitive term. The definition will have to wait until a later chapter when we deal with time and tenses specifically, but we mention the phrase here since it will appear in the axioms. There is, of course, no difficulty in having defined terms in the axioms (they appear in Wilder's axioms for example), but it does tend to interfere with the order of exposition since we must ask the reader to wait for the rather complex definition that will be needed. We will consequently here try to give some idea of what will ultimately be meant by the phrase "occurs now."

When we say that a sensation "occurs now," we will not necessarily imply that it is occurring at the present instant, but rather that it is either now occurring (in the literal sense) or that it has occurred in the very recent past. It is difficult to say exactly how much of the past is to be included under the heading, "the very recent past." On the one hand, the kind of solipsism we are constructing is not a solipsism of the present moment, and we allow ourselves to remember something of the past and refer directly to it. Hence, the axioms of our system which assert the occurrence of sensations will not stick only to those which are presently occurring but will include some references to the past. Since we will want to use "occurs now" in those axioms we therefore allow it some temporal latitude. On the other hand, we remember physical events much better than we remember sensations. One can easily remember events which occurred several years previously, but it is very hard to remember just what sorts of sensations one had at the time. One can, of course, infer what sensations one must have had, but that is a different matter. Thus our direct memory of sensation is very limited and perhaps our "specious present" should be limited to something like two minutes. The exact length of time is not important and it is obvious that we could specify it exactly only in an arbitrary way. What is important is that it be short enough so that we can remember easily and accurately the sensations that took place within it and long enough

Page 176: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

170 SOLIPSISM PROPER

SO that we can conduct simple observations of material objects and be able to talk directly about the earlier stages of the observation while engaged in the later stages. In our language there is no need to distin­guish sensations occurring at the present moment from those which occurred a minute or so previously, so we will use "occurs" to cover them all.

We will still have to have some primitive temporal predicates. One of these will be "is simultaneous with" placed between wffs (well­formed formulas) of class I. This will be interpreted to mean that two sensations are simultaneous if and only if they have parts which are simultaneous - that is, if and only if they overlap temporally. This also includes the case where one sensation is entirely included in the time­span of another. It should be noticed that the two simultaneous sensa­tions may belong to the same sense. We have not defined "sensation" in such a way that a description of a sensation has to be a description of one's whole sense field at a time. What we call a visual sensation may then be just a part of one's visual field; one could then have another visual sensation at the same time and the description of it would turn out to be a description of another part of the visual field. Again, if there is any temporal overlap we say that they are simultane­ous. This predicate of simultaneity is like the ones introduced earlier except that it is relational. Its introduction raises no theoretical prob­lems and it will be learned in connection with pairs of sensations rather than single sensations. The same thing will be true for the other rela­tional predicates to be introduced.

We also need a concept of temporal priority and we will therefore add the primitive relational predicate "precedes" placed between two descriptions of sensations. This relationship applies only where two sensations are not simultaneous and where the two sensations do not overlap temporally at all. If we want to assert that part of one sensation precedes the other but that they also overlap we then have to talk about parts of the sensations as sensations in themselves. This suggests the need for still another predicate "is tern. part of" placed between two descriptions of sensations. If one sensation is a temporal part of another it need not constitute a spatial part of the other but must be entirely included in it temporally. Of course, there are also spatial relations within sensations and we will introduce the primitive relational predi­cate "is sp. part of" and interpret it to denote this relationship.

Before proceeding to the logical primitives we will give a recursive definition for Class II of wffs:

Page 177: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

QUASI-AXIOMATIC SOLIPSISTIC SYSTEM 171

(1) If a belongs to I then a belongs to II. (2) If a belongs to I then "a occurs" belongs to II. (3) If a and {J belong to I then "a is simultaneous with {J" belongs

to II. (4) If a and {J belong to I then "a precedes (J" belongs to II. (5) If a and {J belong to I then "a is tern. part of {J" belongs to II. (6) If a and {J belong to I then "a is sp. part of {J" belongs to II. Let us now take stock of what we have so far. We can talk about any

of the kinds of sensations that can occur, we can draw the distinction between kinds of sensations as finely as we wish, and we can assert certain simple relationships to hold between these sensations. We can even make a description so explicit that it may in fact apply to only one sensation. In this language, however, we have as yet no indicator words such as "this" and we have no way of naming a particular sensa­tion as opposed to a kind of sensation. In this respect our artificial language is not yet as rich as English in that we can talk about partic­ular sensations in English. This can be done by using phrases such as "this pain," or we could give proper names to sensations. This, in fact, will be done in a later chapter.

A simplifying feature of our language is a direct result of its solip­sistic character. Since it is a private language it is used only for referring to the sensations of the speaker himself; hence it is never necessary to identify a sensation as belonging to any particular person. In our ordinary language we have to use descriptions such as "the sensations Roseboro felt when Marichal hit him" ifwe do not know exactly what kind of a sensation the former had. But in our solipsistic system there are on the primitive level no terms referring to persons, not even the word "I," and we do not need definite descriptions or similar devices as applied to sensations in order to sort out sensations belonging to different persons. More will be said on this subject when we come to discuss Strawson's views.

In order to complete our language we must now add some primitive logical notation. We will first introduce the stroke function and give it the usual interpretation - given any two statements, p and q, p/q is true in all cases except when p is true and q is true. It is then possible to define all other truth-functional connectives, such as negation, con­junction, alternation, and material implication, in terms of the stroke. We will next introduce quantification and set membership. Both these notions involve the concept of existence and we will have to say how we will interpret this concept. We will take the assertion of the existence

Page 178: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

172 SOLIPSISM PROPER

of a sensation to mean just that it has occurred, is occurring, or will occur. Actually we need not introduce the concept of existence as such into our language with a special predicate, and we will instead bring it in with quantification. In order to use our quantifiers we must have variables over sensations and the letters, x, y, and z with subscripts as needed will do for this purpose. The simplest sort of quantified state­ment that could be made would be one such as "(Ex)( .... x)." This would say that a sensation of the sort denoted by four dots has or will occur at some time. One could also say, "(Ex)( ...... x occurs)," and here the quantification is redundant and could be replaced by " ..... . sensation occurs." On the other hand, "(x)( .... x)" would simply say that no sensations of any sort except the four dot kind ever occur and that some sensations of this sort have occurred or will occur. The fact that we do not yet have any meaningful individual constants to sub­stitute for the variables does not mean that we cannot use the latter in quantification; it just means that we will have to introduce one of the quantifiers as primitive instead of defining it in terms of an infinite conjunction or disjunction.

Since we are constructing a language which will be adequate for stating our common sense beliefs about the world and also the state­ments of science, we want it to be powerful enough to contain a mathematical language. We will therefore introduce the concept of set membership. When this concept is combined with quantification we will be able to say things like "(Ex)(x e ... )"; this statement will be interpreted as saying that there is a sensation which belongs to the class all of whose members have the property ... , and that this sensation has occurred, is occurring, or will occur. This involves a new usage of the dots but there is no possibility of ambiguity since the context will always make clear the proper reading. We will also be able to make such statements as (x) (x e .. :::> x e ... ) which can be read, "any sensa­tion which belongs to class .. also belongs to class ... ". Hence this state­ment asserts class inclusion.

Let us now formalize what we have and define class III of wffs : (I) If a belongs to II then a belongs to III. (2) If a and P belong to III then alP belongs to III. (3) If a belongs to III and contains the word "sensation" then the

result of replacing "sensation" with a variable belongs to III. (4) Where a is a variable and P is any specific number of dots, a e P

belongs to III.

Page 179: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 173

(5) If a belongs to III and f3 is a variable free in a then ('3: f3)(a) belongs to III.

Universal quantification can now be introduced by definition as fol­lows: (x)(a) is an abbreviation for ~ (Ex) ~ (a). We now have the primitive terms of Quine's Mathematical Logic. He there defines such concepts as zero, number, successor, and the other concepts of arith­metic; the resulting language is sufficient for a very large part of mathematics. Hence all this is implicit in our system since we could introduce all of Quine's definitions. We are still not quite finished with the development of our primitive language, however, and there is one more condition that must be added to III. This involves the counter­factual conditional which we have discussed at length. We will want to be able to make hypothetical statements which have counterfactual force when the antecedent is false and thus transcend material implica-tion. 1 We will therefore introduce the locution "If ..... then ...... " into our language to denote this conditional as distinct from the material conditional for which we use the horseshoe. In general we will inter­pret this conditional according to our hypothetical worlds analysis of the previous chapter. Of course, if the antecedent happens to be true the relevant hypothetical world turns out to be the actual world and the task of evaluating the truth or falsity of the conditional is greatly simplified. We will, of course, combine the use of the conditional with quantification and a typical such statement would be "(x) [If ((x e ... ) & (Ey) (y e .. & y is sim with x)) then ((Ez)(ze .... & z is sim with x) )]." This says "Given any sensation, if it is of type 3 and if a sensation of type 2 occurs simultaneously with it, then a sensation of type 4 will also occur simultaneously with it." We now complete our definition of III

(6) If a and f3 belong to III then "If a then f3" belongs to III. Since class III includes the other two classes it comprises the primitive language of the system and anything else that is needed can be intro­duced by definition.

Besides the language of the system we will frequently use the meta­language, which is ordinary English. The first part of the book was spent in analyzing such concepts as synonymy and understanding; given these concepts it is fairly easy to define such other concepts as entailment, logical possibility, etc. These will all be treated as technical terms which we have added to our natural metalanguage, and we will often appeal to these concepts in justifying our analyses. Two chapters

1 For a fuller discussion see W. Todd, "Causal Laws and Accidents" Theoria, Vol. XXXI, 2: 1965.

Page 180: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

174 SOLIPSISM PROPER

hence, after we have finished our sketch of the stsyem, we will give detailed consideration to the role of the metalanguage in a philosophi­cal system and its precise relationship to the object language.

We next come to the axioms and rules of inference. The principle of division between the two is largely arbitrary since a principle eliminat­ed as an axiom can generally re-appear as a rule of inference and vice versa. We will follow the usual mathematical procedure and take as our rules of inference simply the whole set of generally accepted rules of deductive logic. This will include techniques for manipulating truth­functional connectives, quantifiers, proof by contradiction, and mathe­matical induction (actually a kind of deduction). We could be much more economical about our rules of inference if we wished, but since we are not primarily interested in constructing a logistic system, there is no point in sacrificing convenience for economy in this respect. It is another kind of economy in which we are interested - metaphysical economy.

Since the axioms vary from person to person and from time to time, we will divide them into various classes some of which show more variation than others. The first class will be the one that varies the most from person to person. Each axiom of this class says that a sensa­tion of some specified sort occurs now. That is, all the axioms of this class will be of the form "a sensation occurs now" where a represents some specific number of dots. Thus we simply take each of the sensa­tions which is presently occurring or which is easily remembered, assign it to a class, and say that it occurs now. This set of axioms then entirely reflects the sense experience of the person involved. It is as­sumed that talk about different axioms for different persons in connec­tion with solipsistic systems will not bother the reader. Many persons can be solipsists and have their own solipsistic languages; in that case all other persons, and the person himself, will be logical constructions within a particular such language. We can still give directions for constructing a language and system of this sort to anyone who is interested since, even if they follow our directions, they will not end up denying our existence. This first set of axioms constitutes a starting point and it is from here that we have a choice as to our further proce­dure. We could add as an axiom a very strong principle of inductive inference and try to deduce everything we need just from that principle and this first set of axioms. However, such a principle of inference would have to come close to saying that we can, in Goodman's terms, project every observed regularity in our sense experience and, while

Page 181: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

QUASI-AXIOMATIC SOLIPSISTIC SYSTEM 175

we would have observed the regularity only over a very short period of time, we would be projecting it over a very long period of time. The difficulty is that there is no reason to believe that the statements so deducible would be anything like our ordinary beliefs about the world. In fact, there is not even any reason to think that this procedure would give us a world stable enough so that we could identify material objects within it. The point is that, if we want to come out with theorems which represent our beliefs about the world, we must start with the sorts of axioms we ordinarily argue from as premises and transpose them into our artificial language. Then we can take principles of in­ductive inference which are close to the ones that we ordinarily use and we will not need the kind of super-strong principles of induction which would be likely to get us into trouble of one sort or another. While the primitive concepts of our system do not have to be those of ordinary language, the system must be so constructed that the inferential patterns are the same.

We mentioned before that for every person there is a body of beliefs about himself, the world, and others which is held so tenaciously that he will not even try to justify these beliefs and will regard any question­ing of them as absurd. Then, of course, there is a much larger set of beliefs which he infers inductively and deductively from the first set of beliefs. Many of the beliefs which form this starting point may be memories, but they are likely to be memories which go beyond the first class of axioms. Suppose that I remember having gone from Chicago to Milwaukee on the old Chicago, North Shore, and Milwaukee Elec­tric Railway. At a distance of several years I may still be able to re­member vaguely certain sensations such as those of looking out over the yards at Milwaukee, the jiggling feeling that the train produced at high speed, and so on. But I also claim to remember perfectly clearly episodes of the journey for which I cannot remember corresponding sensations at all, and I claim to remember the whole thing much more clearly than the remaining memory traces of sensations would warrant. In a sense what is remembered is not so much a set of sensations but a set of statements: I remember that I got on the train at about 10 AM, that we went by a large naval base, that there were drunks on the train, and so on. These statements may reflect remarks that were made at the time to companions or to oneself, or statements that were made reason­ably soon afterwards, and are consequently produced very easily. It may even be a mystery how we can produce these statements as memo­ries, but it is certainly something that we can do. None of these state-

Page 182: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

176 SOLIPSISM PROPER

ments can be justified exclusively on the basis of the sensations re­membered, but they are believed just as tenaciously as statements about recent sensations. Thus it can be seen how much our starting point would be broadened if we included non-sensous memories of this sort. Not all our most tenaciously held beliefs are memories of this kind; some are predictions, and some may even come by hearsay if we trust our conversant sufficiently. Further, we may have as much un­questioned faith in certain generalizations such as "All metallic objects, when released in air, fall to the ground," as we have in statements of any kind.

It seems appropriate to create a second class of axioms representing those beliefs which we do not ordinarily attempt to justify and con­cerning which we will not take seriously any doubts raised by others. Some of these axioms will be dependent in the sense that they will be deducible from the first set of axioms and the principles of inductive inference which we will shortly add. Most of the dependent axioms will be those concerned with the objects immediately surrounding us, but since we cannot specify exactly either set of axioms, it is impossible to say just which ones will be dependent. However, the discovery of a dependent axiom causes no difficulty in any sort of axiomatic system; one simply erases it as an axiom and adds it as a theorem. The axioms of the second class are all stated in terms of material objects instead of sensations and are therefore not in the primitive language of the system. Thus we must analyze the concepts involved in material object state­ments and introduce them into the system as definitions before we can get the second class of axioms into primitive notation. This practice is not unusual in axiomatic systems; even in mathematical systems one often has to add defined terms before one can state all the axioms, and there is no difficulty in doing this. This was done, as noted before, in Wilder's system where the defined terms "parallel" and "p is on L" appear in the axioms. Even though these axioms, when stated in abbreviated form, are in ordinary language, they do not take us beyond solipsism. Our analysis of statements about material objects and other persons will be such that they will be transformed into statements which refer only to one's own sensations. The taking of these statements as the axioms of class II just means that we are taking as axioms a set of state­ments which are rather more complex than those belonging to the first class.

We must now talk briefly here, and more extensively later, about the definitions which are the culmination of the system. In order to state

Page 183: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 177

the axioms of second class as well as the necessary theorems three definitions, or really classes of definitions, are most important. First, we must find a way of reducing sentences about material objects to sentences about sensations without loss of meaning. Secondly, we must reduce statements about minds, particularly other minds, to statements about sensations (or, as we would say in the metalanguage, about our own sensations), and thirdly, statements about belief will have to be reduced to statements about sensations. In these cases our units of translation will be larger than they have been in the past - that is, one will not be able to take a single word, or single symbolic expression, and state a string of words or symbolism which will be adequate to replace the first word or expression in all contexts. Rather, it will be necessary to take a whole sentence about a material object or objects and replace it with the conjunction of a set of sentences which refer only to sensations. The same thing will be true for the analysis of other minds and belief.

Here too, we will only be giving a sketch of the system which would ultimately be required. There are so many different kinds of statements about material objects that we cannot possibly give a definition for each one. It is not theoretically impossible to do this as one can see when one realizes that there are only a finite number of statements of this sort, or of any sort, in English as it is actually used. A natural language contains only a finite number of letters, characters, or other building blocks, and we can arbitrarily state a maximum length which we never need or want to exceed in a single sentence. Therefore there can only be a finite number of sentences. Even if we allow the language to be arbitrarily extended by inventing new words, we can only invent them at a finite rate of speed and the language will at any given time still only contain a finite number of sentences.

Traditional phenomenalists, such as C. I. Lewis and A. J. Ayer, have said that a statement about a material object can only be replaced by an infinite number of statements about sensations. On the contrary, our analysis of any statement about material objects will make it equivalent to only a finite number of statements about sensations. Hence it can be seen that the impossibility of actually giving a translation is in our case a purely practical one. Nevertheless, it is a real one and we will be able to give only directions for the making of such translations.

We will now give a brief outline of such a translation and three chapters hence we will give more detailed instructions for meeting some of the contingencies that can arise. In giving these directions it

Page 184: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

178 SOLIPSISM PROPER

will help if we distinguish in our metalanguage three different classes of sensations. One class constitutes those that go with our ordinary bodily movements and are called somatic sensations. When I move my arm up and down I do not have to look at it to know that it is moving, and there seem to be some feelings that go with that particular kind of muscular exertion, perhaps the feelings of some bones rubbing together and various mixed tingles; these sensations, or at least the sequence of them taken as a whole, differ from those which accompany other bodily movements such as shaking the head. One might even say that one does not know that one is moving one's arm at all - one just moves it. But the only point we wish to make is that moving it is not an uncon­scious process which follows a conscious decision to move the arm. We do not find out about the moving in any indirect way, such as see­ing it move; moving one's arm is something one feels as one does it. It is the presence of somatic sensations that makes the difference be­tween this conscious process and other bodily processes which are un­conscious. The same thing is true for such other conscious processes as turning one's head and sniffing with one's nose.

A distinction is often made between perception and observation; looking at a naked woman who happens to be in front of one is a case of perception and climbing a tree to look in her bedroom window is a case of observation. We are implicitly making such a distinction here because the class of sensations we are trying to mark off as somatic are those characteristic of perceiving rather than observing. However, the distinction is obviously not a hard and fast one and there will be border­line cases. One might in certain circumstances see the woman straight on without having to do anything, but in other circumstances one might have to turn one's head a little, look through the corners of one's eyes, pretend to look in another direction, or be looking in another direction but be looking at a mirror. One could then invent more inter­mediate cases until one got to the point of climbing up the tree. Actual­ly, somatic sensations will be involved in observation as well as percep­tion since observation involves perception, but in the pure cases of observation an added class of sensations will be involved.

The next class of sensations includes those which we customarily have at a certain location. When I am sitting in a particular room and talking with someone, but not particularly looking at anything, I never­theless have any number of random sensations of walls, ceiling, book­cases, desks, and so on, which taken together are characteristic of that room and a certain position within it. It is unlikely that the same

Page 185: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 179

pattern is characteristic of any other location and, even if it is, there would be other sensations which we could have which would ultimately distinguish one of two different locations from the other. We have seen that the process of observation involves somatic sensations but it also involves sensations characteristic of being at a certain location - the observation point. Thus when our Peeping Tom has finished climbing his tree and is looking in the window, he is having somatic sensations characteristic of focusing his eyes on an object at fairly short range, craning his head to get the right angle on the object inside, and so on. But Tom is also hanging onto the tree and is having sensations of the bark pressing against his body and also those sensations characteristic of being thirty feet above the ground. Again, it might be said that Tom does not need to have these particular locating sensations in order to know where he is, but it is still true that if he never had sensations of this general sort he would never know where he is. This second class of sensations is essential to observation but not perception since observa­tion usually implies taking up a particular position from which to ob­serve; we will therefore call these sensations location sensations when we refer to them in our metalanguage.

The third class of sensations which we will distinguish are those characteristic of the object at which we are looking. For the observer in the tree the object sensations would be those visual sensations which are characteristic of seeing the woman from various angles and, per­haps, those auditory sensations which Tom would have when she dis­covers him in the tree. Again, it may sometimes be hard to distinguish members of this last class, which we will call obJect sensations from certain location sensations, but a rough-and-ready distinction is all that we need.

Another useful metalinguistic concept will be that of a sensation sequence. This will allow us to refer not just to single sensations but to whole sequences of sensations; the only restrictions are that all the sensations of a given sequence must belong to one or another of the classes defined above, and they must all belong to the same one. Fur­ther, any sensation belonging to the sequence must be simultaneous (in our special sense) with at least one other sensation belonging to it and the sequence as a whole must form a chain with no temporal gaps. When we mention a sequence we will ordinarily attribute shared quaIties to it - to all the members of the sequence - in much the way that we attribute them to individual sensations. These metalinguistic notions will make it much easier for us to sketch out our analyses, but

Page 186: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

180 SOLIPSISM PROPER

it is important to see that we could get along without them ifwe had to. They are all just devices for picking out relatively large classes of sensa­tions which it would be inconvenient to do in the object language since sets of dots are assumed to refer to relatively specific qualities and clas­ses. However, the fact that we have conjunction and disjunction in the object language allows us to form larger classes and more general qualities, and to introduce by definition new predicates which are as general as we wish or which apply only to a previously selected class of sensations. Thus we could define into the object language a predicate which would do the job that "somatic" does in the metalanguage and we could also reduce the other metalinguistic predicates to the object language in case of need. In seeking this kind of help from the meta­language we are in line with the standard procedure used in construct­ing logistic systems.

In general, a statement about a material object will be replaced by a finite conjunction of statements about sensations and the sensations most prominently referred to will be of the three classes mentioned above. Let us now take an example and sketch out the various stages in the analysis making use of both object language expressions and metalinguistic abbreviations. The material object statement to be analysed will be:

There is a gray cat under my car. The analysis will have the general form of a conjunction and a typical conjunct might be:

(x) If { (Ey) [(y is a sequence oflocation sensations characteristic of being in front of my house and includes visual sensations of a white car with a bluish dent in it) & (x is a sequence of somatic sensations characteristic of walking towards the car, squatting, turning one's head upside down, ducking it below the level of the car, and sweeping with the eyes the area underneath) & (x is simultaneous with y)] then ((Ez) (z is a sequence of object sensations characteristic ofa gray cat) & (z is simultaneous with x) }

This says, in effect, that ifl were to go to the white car which I identify as mine and look underneath it, I would see the cat. But this is only a rough first attempt and would have to be refined. First the phrase "somatic sensations characteristic of walking towards a white car with a bluish dent" contains hidden material. When we approach an object more closely this affects our sense fields, particularly our visual field, in that, when we look towards the object, the object-pattern comes to occupy an increasingly large proportion of our visual field. Among our

Page 187: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 181

primitive descriptive predicates there will be some that describe the sorts of sensations we have when we look at the particular car in ques­tion side-on from a distance of twenty feet. We also have in our primi­tive notation a way of talking about spatial parts of sensations and we can therefore have predicates which say how large a part of a sensation is relative to the total sensation. Hence we can describe in purely sensuous terms the sequence of getting ever closer to the white car until we can touch it. We can also describe in sensuous terms the process of walking in a straight line or turning towards something and then walking. Further, we have a purely sensuous way of talking about direction relative to ourselves. We have a primitive predicate which says that a part of a visual sensation is in the center of the visual field and we can also say, in somatic terms, that we are looking straight in front. By combining assertions of these two kinds we can, in effect, say that we are looking at something that is directly in front of us. Since there are somatic sensations characteristic of turning the head and the eyes, or the whole body, we can apply primitive predicates to them and thus talk about all the other directions relative to ourselves, or anything else. It is in this way that physical space is constructed in our object language. It can now be seen that we have the materials to re­place the phrase quoted above with a more detailed analysis, and it could be done along the lines indicated above. In fact, of course, there are many starting points from which one could verify the existence of the gray cat under the car and one would get a different hypothetical statement for each of these starting points - for each route taken from the starting point to the car and so on. Thus the conjunction will be very long and complex but still finite. Our starting points will be limit­ed to the general vicinity of the car and there are only a finite number of spatial positions within that area which we can discriminate from one another on the basis of our location sensations; similarly there are only a finite number of routes from starting point to car which we can discri­minate, and the product of two finite numbers is a finite number. In fact, we have not reached the end yet. Even after one gets to the car, many different sorts of visual sensations would count as seeing the cat. One might see its profile, or one might see its shining eyes, or one might even feel it with one's hand behind a tire. In this case the simplest course would be to expand the consequents of the whole class of hypo­theticals into disjunctions which would describe the various ways in which the cat can be noticed. Thus it would take a very considerable amount of energy to specify all of the sensations mentioned in even a

Page 188: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

182 SOLIPSISM PROPER

single hypothetical statement belonging to the analysis, and we would only then be in a position to replace all the metalinguistic terminology with primitive predicates.

A further difficulty is that even if there is a cat under the car it i8 terribly easy to look there and still not see it. In order to protect our statements we might narrow down our original statements so that they require us to look under the right hand side of the car; three other groups of similar statements would then be added which would have us looking under the left hand side of the car, the front and the rear. We would also have to have a provision which would require the ob­server to keep watch as he goes from one position to another so that the cat cannot move. We would then disjoin the four groups of statements so that verification of anyone would count as a verification of the original material object statement. Thus we are protecting ourselves against concluding that the cat is absent when he is really present. If the disjunction is falsified by all four components turning out to be false, then the chances are that the cat is really not there. There is still a possibility of mistake and the problems generated by this problem will be discussed later.

Basically different ways of verifying the cat's presence will add a further complicating factor. For instance, one might put food behind the car to attract the cat out from under it or drive the car away and have someone watching to see if the cat runs out or gets squashed. Complex hypothetical statements analogous to the one discussed above would then be needed to represent these alternatives and would be added to the hypotheticals we already have in order to get the final analysis. There is, nevertheless, one complicating factor which we will be able to avoid. In traditional phenomenalism, as mentioned before, there are an infinite number of statements in the analysis on the grounds that no matter how far afield one goes in either space of time from the original event one can still find evidence that bears on it. In our phenomenalist analysis we will set up criteria for limiting the relevant evidence, or at least limiting what we will put in the analysis, and thus maintain finitude.

Two sorts of intuitive objections to even our finite brand of phenom­enalism are likely to arise at this point. One is that we have not actually produced the required definitions and the second concerns the stagger­ing complexity of the whole undertaking.

We have made much of the point that the failure to produce these definitions is due to practical rather than theoretical difficulties. Let us

Page 189: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

QUASI-AXIOMATIC SOLIPSISTIC SYSTEM 183

make this claim more specific. We make the claim not just because everything involved is finite, although that is very important, but be­cause the complexities involved are of a special sort. It is easy enough to give the rules according to which the definition has to be constructed. For instance, we laid it down as a rule that one has to have a hypothet­ical in the analysis describing the process of verifying the presence of the cat under the car from any given starting point in the area. Similar­ly, it is easy enough to order the relevant starting points. One dis­covers first the least difference in spatial position that one can detect using our ordinary sensory cues, and one then divides a map of the neighborhood up into grid where each square represents such an area. Perhaps the squares would be some six inches on a side. We then order each square using numbers and, taking each square in turn, one dis­covers empirically the location sensations for being in that square and forms a set of statements, perhaps itself including hypothetical state­ments, which describes in phenomenal terms the situation of being in that square. Having done this one constructs the required hypothetical statement which will be almost exactly similar to the previous one. One can also deal in a piecemeal way with the other factors involved in the analysis. No one is particularly complex in itself - it is just that there are so many pieces! The problem in constructing the analysis is not in knowing how to proceed but in actually going through all this for each six inch square in the neighborhood. This is what is meant by saying that the problem in actually constructing the definitions is not one of theory but one of practice.

The fact that we are not able to actually produce the required defini­tions shows that we would have failed if we had sought to produce the system itself as opposed to a sketch of the desired system. The fact that we cannot provide the axioms either, but only say what sorts of axioms would be required already necessitated our having to be satisfied with a mere sketch of the system; being able only to characterize the sorts of definitions we need does not seem to take away anything else of im­portance. Further, an axiomatic system which systematizes our com­mon sensical and scientific beliefs about the world is not a system that we actually want to use. Even if we had such a set of axioms and defini­tions we would not take the trouble to deduce theorems. Some of the theorems would be ones which we are not prepared to question in any case, and have no need to justify, and others would not be sufficiently vital to us to be worth taking the trouble to deduce. The only point of setting up such a system is to see what sorts of ontological commitments

Page 190: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

184 SOLIPSISM PROPER

we will have to make in the end. Our point was that one can distin­guish that part of our language which is eliminable and does not com­mit us ontologically to anything from that part which is not eliminable and does so commit US, only by systematizing the whole language and seeing what can be eliminated in favor of what. But, in order to see that, we are interested not in particular axioms, particular primitives, and particular definitions, but in the sorts of primitive, defined terms, and axioms that we would have. For these purposes a sketch of the system does as well as would the system itself.

We have been using mathematical systems as our paradigms for axiomatic systems, and it should be noted that even there things are not always so tight as one might think. A common method of proof is mathematical induction and, as it is generally used by mathematicians, it constitutes not so much a proof in itself but a set of directions for giving a proof. Thus, if we want to prove that the product of any odd number and two is an even number, we first prove it for 1 X 2, assume the theorem true for some odd n, and prove it true for (n + 2) X 2. We nowhere prove that the product of five and two is even, but we show how to adapt the proof of the initial case to get a proof of that fact.

On the score of complexity of the system as such, the most important thing is to see that the complexity occurs in the right place - in the definitions. There are some extremely powerful and economical systems, such as group theory, where a vast body of theorems is prova­ble from four simple axioms. This is achieved primarily because the primitive language is enormously extended, at least in one sense, by the use of some extremely complex definitions. In fact, this is generally true of economical systems; if one wants economy of axioms and primi­tive language then, in order to get anything out, one must have com­plexity in definitions and proofs. If one wants to keep one's definitions simple and still get significant theorems one has to add to the axioms. But the point of a solipsistic system is to see how little we can get along with in the matter of primitives and axioms. However, we still have to be able to derive the beliefs of common sense, so the definitions in such a system will have to be extremely complex whether the system be a phenomenalist one, a physicalist one, or something else. This being the case, it is no objection to the system, or its intended uses, that the complexity which one would expect appears in due course.

A related objection has to do with the assertion that the primitive language is one that we actually could use. Of course, the "could" here

Page 191: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 185

might involve such things as having better memories and quicker minds, but we do not really need to appeal to that. Even in the actual human condition, the processes which a child must go through in order to successfully recognize the stable objects of his environment are very complex ones.1 In some way or other a very large number of sensory cues which appear at one time have to be so ordered that cues belong­ing to that set can be associated with different cues belonging to a different set which must be similarly organized. Further, the associa­tions which must be made for the child to learn anything are mixed in with all sorts of extraneous factors which have to be disregarded. Actually phenomenalism is a formalization in logical terminology of the relationships which the child has to take in unconsciously in order to be able to recognize objects and get around in the world. One could construe it as a theory of what has to be learned in order to recognize physical objects even though we are not committed to that hypothesis. We have now said enough about phenomenalism to indicate how the system is set up; later on we will discuss particular problems that arise for the phenomenalist and give related analyses for statements about other minds and statements about beliefs.

We saw before that for the axioms of group II we need statements about material objects which go beyond the sensations that we can remember but these statements still lie within the general area of com­mon sense belief about the world. Once stated, the definitions of ma­terial object statements can be used and the axioms can be put into primitive notation. The axioms belonging to this class will vary from person to person in the same way as those belonging to the first class. For instance, one has many very basic beliefs on the material object level about the city in which one lives, and these beliefs would not be shared by someone who lives in a distant place. There will be common elements, but in almost all cases there will be differences as well. It is an interesting feature of science that steps are taken to minimize differ­ences in basic presuppositions on the part of different scientists. Many of the observations on which a science like physics or chemistry depends are ones that can be made by anyone anywhere. Thus the behavior of the elements and the compounds into which they enter is not a feature ofa particular part of the world but is open to very general observation. Sometimes crucial events, say in astronomy, can be observed only from a particular place. Nevertheless, where science depends on observa­tions of this kind pains are taken to ensure the accuracy of the observa-

1 For more on this point see W. Quine, Wl11'd and Object, pp. 80-85.

Page 192: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

186 SOLIPSISM PROPER

tions and to publicize the results. Thus a large degree of common belief is obtained and, even when scientists advocate differing hypotheses, there is a high degree of agreement about the facts to be explained. This is not true of our common sensical beliefs about the world since there is so much relevant evidence that different persons are bound to be impressed by different selections of evidence and come out with differing beliefs. We could, if we wanted, incorporate scientific axiom systems into our systems of belief and the axioms of such systems would be added to group II. Thus these added axioms would represent those parts of our total belief about the world where there is a very high level of agreement, and the other non-scientific axioms would represent the divergencies. On the other hand, it should not really be necessary to have the axioms of scientific systems included among the axioms of group II. A successful scientific system is one that explains the phenom­ena, reflected by our common sense beliefs, and we will certainly want principles of inductive inference which will allow us to infer the truth of scientific theories which explain the phenomena more successfully than their competitors. That being the case, the axioms of such systems would appear as theorems in our philosophical system.

We could imagine a species of men who would be so scientifically oriented that they would be willing to set aside the sorts of common sensical beliefs we have mentioned if by so doing they could achieve greater economy or greater elegance in their scientific systems. Of course, there would still have to be observations for the scientific systems to explain but the class of observations regarded as relevant might be narrowed down to those made under very special conditions; the rest of day-to-day experience would simply be ignored. In con­structing a philosophical system which did justice to the beliefs of such people the axioms of group II would be comprised exclusively of the axioms of scientific systems, and the axioms of group I would describe only those sensations which are involved in making observations of the required sort. But it is obvious that we are the sort of people who cling more tenaciously to the beliefs of common sense than we do to any scientific beliefs; in fact, we check the latter against the former. One of the reasons that we do this is that we understand the statements of common sense so much better than we do the statements of science. However, there is another reason, and a better one. If we draw up scientific systems which conflict with the beliefs of common sense and prefer the former to the latter we have to delineate the class of observa­tions which are to be regarded as important and worth explaining and

Page 193: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 187

disregard all the others. But we could divide phenomena into the important and the unimportant only from the point of view of the scientific theory which we are testing. Of course, our procedure would then be circular and we would not really be testing the theory because we would be antecedently throwing out all possible adverse evidence. In fact, it does not seem possible to draw up any such criteria which would allow us to disallow any carefully made observations; on the contrary, we regard all observations, and the common sense beliefs that reflect them, as being potentially relevant to science. We hold up to science the task of explaining these phenomena and these beliefs. Hence, in our system, and in our behavior, we must start from these beliefs and have inductive principles which allow us to infer the theories of science.

One might next wonder whether the common sense beliefs of any single person would provide a broad enough base for science. Here we must remember that the most important beliefs from a scientific point of view tend to be those which are very widely shared. Thus Gallileo's experiments with falling objects, while surprising to common sense, could be duplicated by anyone with coins of different sizes, and many of the greatest advances in science have arisen from questions about such ever-present phenomena as fire, the sun and moon, and the more obvious features of the human body. The sort of knowledge which, like knowledge of the streets in one's neighborhood, is peculiar to an in­dividual is usually not relevant to science at all. We observed before that, partly as a result of the success of science, there is a high level of agreement about scientific statements, facts, and observations. We can now go further and say that there is an unusually high level of agree­ment concerning even the sorts of common sense beliefs from which science begins and agreement about the outward character of the sorts of phenomena concerning which the first scientific questions are asked. Then, too, one has to remember that the axioms of group II are con­cerned not only with particular facts and observations but include many generalizations, such as statements about released objects falling, which are a part of pre-science rather than science itself. We could not confidently base science on the axioms of group I, but, in view of these considerations, we feel justified in taking the axioms of group II as a base for inductive inference despite the fact that, in any given system, they only represent the beliefs of one person.

We come now to the third set of axioms which will allow us to make inductive inferences. It might seem more natural to state these prin-

Page 194: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

188 SOLIPSISM PROPER

ciples as rules of inference rather than as axioms at all. However, when we use these principles, the passage from premises to conclusion is one of inductive probability rather than deductive certainty. Since we want to preserve a traditional quality of axiom systems, that the theorems are deduced from the axioms, these principles will be stated as axioms. What is really needed here is what Goodman calls a theory of projec­tion. On the intended interpretation the axioms of groups I and II commit us to accepting a certain set of beliefs as facts about the world. We will then find that certain generalizations hold true of all these facts; the question is which of these generalizations should be projected to cover the future and all spatio-temporal regions not covered by the initial premises. Goodman thinks that there are great difficulties in doing this, and one of his examples is as follows: 1 Let us define the predicate "grue" so that it applies to anything which is green up to time t and blue afterwards. We then take time t as the present. Thus it is equally true that all emeralds are green and that all emeralds are grue. How then can we project the hypothesis that all emeralds are green rather than the hypothesis that all emeralds are grue, which would imply that they will be blue, when both hypotheses explain the observed phenomena equally well? Goodman proliferates this kind of example and then tries to develop a theory of projection which reflects our actual inductive practice. This theory seeks to show the difference between hypotheses which we project and those which we refuse to project. The examples of the grue and bleen sort have aroused a good deal of controversy and, although it seems to me that Goodman is basically right, even if we do not accept his examples there are obvious­ly going to be cases where we have to decide which of two competing hypotheses to project. Thus some sort of criteria will be needed at some point and these criteria will be reflected in the axioms of group III. Goodman's own account makes considerable use of the concept of an "entrenched predicate" which amounts, roughly, to one that has been favored by past inductive practice. Since, in developing an axiom system, one starts from scratch and has to make the first inductive inference somewhere, we would have to interpret an entrenched pre­dicate as being one that occurs frequently in axioms of groups I and II. Thus "green" might easily appear in those axioms but it is extremely unlikely that "grue" would choose such a locale to make its first appear­ance. However, there are other possibilities. Carnap, in his Logical Foundations oj Probability, has developed a way of assigning non-relative

1 N. Goodman, Fact, Fiction, and Forecast, pp. 74-75.

Page 195: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 189

probabilities to possible events on the assumption of a finite language. Our language also has a finite number of primitive predicates and, except for the presence of counterfactuals which have to be treated separately anyway, is the sort oflanguage to which Carnap's system is applicable. The primitive predicates he has in mind are not those which are applied to sensations but are applicable to material objects or spatio-temporal regions. However, since we can define material object terminology within our language, this constitutes no problem. Burks, in his review of Carnap, argues cogently that the latter's method of assigning probabilities presupposes a certain principle which is not too unlike certain traditional inductive principles. That is, we could conceivably assign an event a lower probability the more often that sort of event has occurred in the past, but since Carnap's procedure, and our ordinary procedure, does the opposite, an inductive assump­tion is involved. In his "On the Presuppositions of Induction"! Burks states the assumptions which he thinks are needed. If we were to make use of Carnap's system plus Burks' emendations and inductive assump­tions we would put the whole thing into the axioms of group III. We would then have axioms which would allow us to infer any statement belonging to the language of the system which has more than a certain specified probability. We would then have to decide what probability to specify. What at first looks like a difficulty then turns out to be an advantage. We could begin by taking an arbitrary probability, say .6, and then look to see what theorems could be deduced. If we are able to infer as theorems all the common sense beliefs which are em,bedded in ordinary language and then a great many other statements which we do not believe or which we disbelieve, then our probability is too low and we must raise it. If, on the other hand, a great many of our common sense beliefs about the world were excluded and were not deducible, then our specified probability would be too high and we would have to lower it. One of the attractions of this approach is that we could, by adjusting a number in one of our axioms, make the class of theorems approximate very closely the class of common sensical and scientific beliefs. Of course, for any particular individual, there would probably be some discrepancies since it is unlikely that we are com­pletely consistent in our ordinary inductive practice, and our standards of evidence probably vary from time to time. In fact, it is not possible for any axiom system to reproduce exactly the irregularities and in­consistencies of our actual practice, nor do we want it to. In this respect

1 A. Burk.~, "On the Presuppositions of Induction," op. cit.

Page 196: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

190 SOLIPSISM PROPER

the system, if ever fully developed, would be corrective of our ordinary practice and tell us that if we want a certain statement, p, we must take q too or reject them both.

In fact we have been somewhat unfair to our actual inductive prac­tice because there are times when it is reasonable to adopt different standards of evidence. Thus an "ordinary prudent citizen" would bet on a horse if he thinks the probability of its winning is .8, but he will not fiy in an airplane ifhe thinks the probability ofits not crashing is .8. In order to take account of this factor we could add a probability quali­fication to some of our theorems so that the statement that a particular proposition has a given probability will itself have a probability suffi­ciently high to allow its deduction as a theorem. We have a primitive language which is basically powerful enough to allow us to form proba­bility statements, but this would involve an analysis of the meaning of statements about probability, and that is beyond the scope of this essay. It should be noted merely that we have said nothing which would be inconsistent with the adoption of any of the usual analyses of proba­bility statements.

The question of what inductive axioms to choose is in itself a topic for a very large book and we will make no final decision in the matter. The main point to remember is that the language of the system contains as a part the material object language which we believe to be eliminable in favor of our primitives; since principles of induction are stated within the material object language any set of inductive prin­ciples, from Mill's methods on, could be added to the system without creating any contradictions. While we have indicated a slight prefer­ence for the Carnap-Burks system the reader is quite free to plug in his own favorite set of inductive principles.

In our earlier discussion of counterfactuals we argued that counter­factual induction makes presuppositions which go beyond those re­quired for ordinary induction. Since counterfactuals are primitive in our language and we are heavily committed to them at many places we will need to add to the axioms of group III a principle which sup­ports counterfactual induction. We saw that such a principle is hard to state precisely, but the essence of the matter is that the causal structure of the world be resolvable ultimately not into a very few very basic and very general causal laws which operate everywhere but rather into a large number of relatively specific causal laws which are in­dependent of one another. If this is the case, then, in our hypothetical world, we will be able to delete one or more of these laws without

Page 197: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

Q.UASI-AXIOMATIC SOLIPSISTIC SYSTEM 191

throwing the whole causal structure of that world into chaos. We therefore add this principle as an axiom. Our sketch of the axioms is now complete.

It should be noticed here that we have been trying to do the same general sort of thing that Carnap did in his Aujbau and that Goodman did in his Structure of Appearance. Still, our position differs considerably from that of Carnap. First, we have an altogether different theory of meaning and are more concerned with the theory of meaning as such as opposed to a criterion of meaningfulness. Secondly, our view of counterfactual conditionals is different from that of Car nap, and almost all the analyses to come will be built on counterfactuals. Thirdly, it will be seen in the chapters to come that the detailed working out of the system is entirely different.

It will also be noticed that Carnap, and particularly Goodman, have made a special effort to construct an economical language for talking about sensations. Our language is not particularly economical in this respect because we end up talking in our primitive language about particular sensations, an indefinite number of properties of sensations, and about classes of sensations. Both Goodman and Carnap would say that it is possible to say everything that can be said about sensations in a much more economical language, even though they propose different sorts of languages to this end. On this issue we will make no attempt to say whether either Carnap or Goodman is correct. However, the main point is that if either is right we can then simply eliminate some of the terminology from the primitive language of our quasi-system and introduce it by definition instead. In fact, practically everything in this book could be viewed as an extension of what Good­man says in the Structure of Appearance. We would then be trying to do for the middle and upper steps of Carnap's system what Goodman did for the lower steps.

Page 198: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 10

ALTERNATIVES TO SOLIPSISM

In this chapter we will consider some principle avenues which one might take in order to flee from solipsism, provided that one comes to philosophy with the same sorts of methodological assumptions. In the next chapter we will discuss the attitudes towards solipsism of those philosophers who have quite different methodological outlooks.

We will begin our discussion of the alternatives to solipsism by con­sidering a question which has been raised by Quine and which is rele­vant to what we have been doing. In the second chapter of Word and Object Quine argues for what he calls the indeterminacy of translation. He would admit that if one were considering translations of a French sentence into English and two translations came out very differently they could not both be correct. However, Quine is more interested in something that he calls radical translation. This is where the two lan­guages have no common roots and where the translator, who knows one language to begin with, learns everything he knows about the second language by observing the verbal and non-verbal behavior of the people who speak that language without any other aids. Thus he has to piece together their language by trying to correlate their re­cognizable utterances with their recognizable actions. In such a situa­tion Quine thinks that two translators could make very different trans­lations of the same sentence in the strange language, but that there might be no way of deciding which translation is correct. He then goes on to argue that a single sentence is too small a unit of translation and that the indeterminacy of translation arises when we choose such a small unit. If we choose a larger unit, such as a whole scientific theory, then the indeterminacy disappears. This is a view which is held more or less implicitly by many philosophers but which has been made ex­plicit by Quine. l

In the system which we sketched we claimed that it was possible, using our primitive language alone, to produce complex locutions which are logically equivalent to the material object sentences of our

1 Quine, op. cit., p. 78.

Page 199: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ALTERNATIVES TO SOLIPSISM 193

ordinary language. This being the case, we then introduced material object sentences into the language as defined terms, or, better, as de­fined sentences. We then get interesting consequences if we combine this view with Quine's indeterminacy principle. The conclusion would be that, while we have correctly translated the material object lan­guage into our solipsistic language, it might be possible to translate the same material object sentences into a quite different language which would give them a different meaning. Further, one might be able to interpret the solipsistic language in each of two different ways, in both cases doing justice to our material object sentences, but still have no way of deciding between interpretations. One might then conceivably draw the conclusion that translation of any sort would be impossible, but the Quinian sort of conclusion, and the more reasonable one, would simply be that there is a range of correct translations among which we have to choose arbitrarily if we choose at all. With each system and its interpretation goes the philosophical assertion that the system can do the job as far as our ordinary beliefs are concerned. But if it is possible to translate a statement of our ordinary language in two different ways at the same time, and do it correctly each way, then the consequence would be that two different philosophies, say phenomenalism and physicalism, could be correct at the same time. This is, in fact, an idea which has long attracted Quine.1 The situation would not then be just, that two different philosophies have roughly equal claims for "truth" but the way would be open to claim that two different philosophies are correct at the same time. All this is not directly incompatible with any­thing that we have asserted. We have claimed that the sort of system we have sketched is, on its intended interpretation, adequate, but we have not denied the possibility that some competing theory might also be correct. This possibility of holding two different philosophical posi­tions at the same time is certainly an interesting one and it deserves considerable attention.

This position has recently been attacked by Patrick Wilson who summarizes what he takes to be Quine's view as follows: 2 "Two persons could be completely bilingual in, say, Martian and English, their English speech habits could be exactly alike, and also their Martian speech habits, and yet the following situation might arise: one of them might make a statement in Martian, a statement which both of them took in exactly the same way, both of them understood, so far as all

1 Quine, "Two Dogmas of Empiricism," in From a Logical Point of View. S P. Wilson, "Quine on Translation," Inquiry, no. 2, Summer, 1965. pp. 206-207.

Page 200: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

194 SOLIPSISM PROPER

possible tests for understanding (conducted in Martian) could deter­mine: but if they both translated the statement into English, the results might be extremely different, might even be incompatible, although each understood the other's English translation perfectly, so far as all possible tests for understanding (conducted in English) could deter­mine."

In attacking this position one of Wilson's chief arguments rests on the possibility of a bilingual translator. Quine insists that such a trans­lator, even if he is bilingual, must at various points form what he calls analytical hypotheses about the synonymy of various pairs of words in the two languages. He also thinks that there may ultimately be no way of deciding which of two such hypotheses is correct; as Quine puts it, they might ultimately be tied for first place. Wilson, on the other hand, says that if the translator is bilingual, it is not then a hypothesis for him that two sentences or words belonging to the two languages are, or are not, synonymous. Both locutions belong to languages of which the translator is a master and one can know whether two words that con­stitute part of one's ordinary usage mean the same thing or not without forming any hypotheses. Wilson says, "It is, of course, true that all of us continually assume, what may very well be mistaken, that others attach the same meaning to words as we do; and it is also true that we may well be mistaken in reporting that we regularly attach the same meaning to two different words. But that on a particular occasion when we use two words as simple alternatives to each other, we mean the same by them, is a fact of which we can have perfect knowledge; like­wise, I should think, when a linguist sets himself to say 'the same thing' in two different languages, he cannot be mistaken in supposing that he means the same by the two utterances, however much he may be mis­taken in supposing that he is using words as others use them, or as he himself usually does."! It seems to me that Wilson is correct in this view, but that Quine could admit this point and still maintain his position about the indeterminacy of translation. The upshot is that a bilingual person can translate sentences and words from one of his languages into sentences and words in the other language without in­determinacy. But if the translator does not attach the same meaning to a Martian word that other Martians do, or the same meaning to the corresponding English word that other English speakers do, we could not then be assured that his translation holds for the words as used by others. For instance, the bilingual could well have unconsciously im-

1 Ibid., p. 205.

Page 201: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ALTERNATIVES TO SOLIPSISM 195

ported the peculiarities of his native language into his second language and always use terms of the latter to refer to "temporal segments" of rabbits instead of rabbits, or some such. Quine would argue that we can never know, apart from some special kinds of cases, whether the bilin­gual translator does attach the same meanings to the terms of either language that others do. Thus the indeterminacy will remain. The question really becomes whether it is possible to translate without in­determinacy from one person's language to another person's language. Whether or not they speak "the same language" in the conventional sense is not really relevant to the issue. So the main problem is not one of translating between different natural languages and arises even within the same language community as long as there are two or more persons. Both Quine and Wilson seem to recognize this in the end, but Quine apparently thinks that he can make a better case for his view when more than one natural language is involved, and he draws his examples accordingly.

At this point Wilson argues that we can get determinate translations from person to person by starting with what Quine calls "occasion sentences" and matching more complex sentences for synonymy by comparing their structure. Thus Wilson says, "If the metaphor of the 'network' be allowed, it must be allowable to speak of comparing two 'networks' point by point. Quine would admit that the 'bottom layer' of statements might be shared by two systems, and if the 'upper' ele­ments are introduced, in no matter how complex a fashion, in terms of 'lower' elements, it must be possible, and possible to notice, that in two systems elements are thus introduced in the same way, on the same basis. Relatively 'unobservational' linguistic elements being somehow indirectly related to experience, it must be meaningful to say, and possible in theory to discover, that items in two different structures are similarly related to experience."l

Quine thinks that there is a looser connection between these complex statements and stimulus meaning or, as Wilson puts it, experience. Quine thinks that the best illustrations of this come from science, but he does not give examples or explain in detail why he thinks there is such a looseness of connection. Since I think that there is some plausi­bility in Quine's position, I will try to set up a case where the sort of relationship Wilson wants is lacking.

Suppose that we have a theory (it can be in any scientific area) which contains hypothetical constructs which cannot be directly ob­

I Wilson, op. cit., p. 209.

Page 202: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

196 SOLIPSISM PROPER

served. We will call them "motons." The theory contains a large num­ber of statements describing the properties that these motons have and setting forth the relations which hold between them. A certain state­ment, p, can then be inferred inductively, but not deductively, from the theory. This statement says that the moton with the least mass (within a defined area) will be at position x at time t. While p is not deducible from the theory, it follows with a high degree of probability and the inference rests on very widely accepted assumptions. Hence the testing of p can be looked upon as a crucial experiment. Of course, p cannot be tested directly since it concerns a moton, but there is experi­mental evidence which bears on it. The relevant experiments are next performed and turn out very adversely for p and for the theory. Since p does not seem to be true there must be something radically wrong with the theory, and it will have to be either given up altogether or greatly modified. But since the theory posits the existence ofmotons in the first place and tells us all that we know about them, the concept of a moton would have to be rejected along with the theory. Since the statement, p, involves a definite description it presumably now takes its place with statements about the present king of France; it is actually in a worse position because its rejection may cause us to wonder what it is to be a moton. But when a statement is disconfirmed it is shown to be false, or at the least, its probability is lowered. However, in this case, the adverse evidence has not had this effect but may have rendered the statement meaningless so that it is neither true nor false and has no probability. Thus one can see immediately that this is not what is usual­ly meant by disconfirmation. Let us now suppose that later investiga­tors, instead of abandoning the theory, reconstruct it substituting the concept of a "soton" for that of a moton. Sotons differ from motons in many ways, particularly as regards the way their position is detected, but they are also similar to motons in many ways. Most important, they differ in such a way that, if they are substituted for motons in statement p, the evidence which was adverse to p is now favorable. The new theory also has all the advantages of the old one. We can just say that we have a new theory, that sotons are different from motons, and that the new theory is better. On the other hand, we could say that so tons are really the same as motons - not the same in all respects -but that they are still defined in the same way although the new theory says different things about them. If we take this line, the consequence would be that p never really was affected adversely by the evidence, but that some other parts of the theory were made implausible. In

Page 203: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ALTERNATIVES TO SOLIPSISM 197

particular it might be said that the old theory wrongly construed the observable signs of a moton's being at x at t. It is important to see here that the situation is not that of having a conjunction of p and the other statements of the theory, a conjunction which is disconfirmed as a whole so that we just have to decide which part of the conjunction to reject. Part of the theory tells us the conditions under which other parts can be confirmed or disconfirmed, and if we were to reject that part of the theory without substituting something in its place, the theory would cease to be empirical altogether. Similarly, when we substitute some­thing else for that part of the theory, evidence that was unfavorable to the rest of the theory may become favorable or vice versa. The situa­tion here is totally unlike the disconfirmation of a conjunction of such statements as "John's father told John he would get him a new switch­blade ifhe won at the races," "John's father won at the races," "John's father always tells the truth." In that case, if the conjunction is dis­confirmed, we can reject any of the three components without pre­judice to the other two.

In our example it is a question of whether or not the hypothetical construct changes when the ways of detecting it change; and, of course, this decision would be made purely on the basis of convenience. There is no logical compulsion either way. The trouble is that whether p was confirmed or disconfirmed (even if we allow these terms to be stretched sufficiently to cover this case) depends on a human decision. This is what is odd and this is what suggests that confirmation and disconfir­mation are not so simple in this case.

This seems to be the sort of thing that Quine has in mind when he talks about certain scientific statements, such as our statement p, being divorced from stimulus meaning. Statements describing the experi­mental processes would be directly connected with stimulus meaning but p would not be so connected. It would be a "standing sentence" in Quine's terms, and one which is very far removed from stimuli. Quine holds that such statements are not arbitrary and there is no indication that he holds them to be unverifiable in the traditional sense. Rather, when adverse evidence turns up, he says that there are alternative ways of proceeding. In our case we can either reject state­ment p and the theory and develop a new theory with new concepts, Jr we can keep the same concepts and adjust the theory in the back­ground thus retaining statement p. Since motons and so tons do not refer to directly observable entities, we can change them as much or as little as we like and keep the same names and some of the same proper-

Page 204: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

198 SOLIPSISM PROPER

ties. But there is still a non-arbitrary element here. The theory really does have to be changed or replaced so that the net output will square with the experimental evidence, but there is more than one way to do this. The difference between the various ways of doing it may be purely verbal in the sense that it does not make any difference to the totality of assertions made by the theory which course we choose, but it makes a real difference to the meaning and status of individual statements within the theory. It is not just that we always have a choice concerning which statements to keep and which ones to reject, but that no state­ment belonging to the theory is confirmed or disconfirmed in a simple and direct way. In the case of Quine's examples of occasion sentences, "Gavagai" and "Rabbit," one knows immediately whether the present stimuli are confirmatory, and one still knows whether certain evidence is favorable or unfavorable to a given theory as it stands, but one does not know automatically how the evidence will affect a particular part of the theory. The effect it will have depends on a complex decision which will have to be made and which will be concerned with maxim­izing convenience within the conceptual framework.

Many persons, perhaps Wilson among them, would say that em­pirical statements of any degree of complexity will have ultimately to be related to experience in the sense that they would be confirmed by some things that might happen and disconfirmed by others. In order to see what is confirmed by what, one might use the sorts of criteria provided by Hempel or Goodman.1 One would then say that two theo­retical statements are synonymous if they are confirmed and dis­confirmed by all the same things even if they belong to different theo­ries. Or, to be more sophisticated, one might say that they are syno­nymous if, when conjoined with the same statements, the resulting conjunctions are confirmed and disconfirmed by the same things. However, the standard criteria of confirmation are only applicable where we have either logical entailment from the confirmed statement to the confirming statement or when we can talk about the probability of one relative to the other. But in our present case we have to make choices before we know what is confirmed by what and we cannot talk about confirmation at all apart from these choices. Hence we cannot even apply such a criterion of synonymy until the choices have been made, and then the element of indeterminacy has already entered.

In the case of our motons and sotons we not only have to have a

1 For a defense of one of these criteria see W. Todd, "Probability and Confinnation," forthcoming in Mind.

Page 205: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ALTERNATIVES TO SOLIPSISM 199

theory in the background before we can talk about confirmation, or the relevance of stimulus meaning, but we also have some ideas about how to change the theory in case of emergency. Presumably for Quine two theoretical statements could be synonymous in a determinate way only if they formed part of the same theory and if, whenever two differ­ent persons used the statements, they would have identical ideas as to how to adapt the theory or conceptual framework to circumstances. Since some of these ideas are probably unconscious, it becomes almost impossible to determine whether such sentences could ever be com­pletely synonymous. And, of course, these are only necessary conditions and may not turn out to be sufficient conditions of synonymy. Before we could know that we had complete synonymy between theoretical statements, as used by different persons, there would have to be some statements belonging to the theory whose confirmation, unlike that of p, is not relative to other statements in the theory. It might then be possible to find other statements whose confirmation is relative only to those. Perhaps this is the typical situation in some sciences, but I think that the case we mentioned earlier is more typical of the more highly developed sciences. In that case there is no starting point where we can confirm some sentences in an absolute way.

Thus we can see where Quine gets his conclusion that one can only compare theories as a whole for likeness of meaning and that individual sentences belonging to them are not comparable. Part of the meaning of a sentence within a theory concerns its priority within that theory. It does not have a set independent meaning, but it has a role within the theory, and part of this role involves the price at which it will be given up in favor of something else. This role and priority enters into its confirmation and disconfirmation relations, hence into the deter­mination of its truth, and hence into its meaning.

Philosophers frequently say that the primary unit of meaning is the sentence, and that words only have meaning derivatively because of the role they play in sentences. The view here would be that when we are dealing with certain subject matters the primary unit of meaning is the theory, and that scientific sentences only have meaning because of the role they play in theories. Intuitively it is plausible to suppose that as the subject matter becomes more complex we need larger units of meaning.

However, Quine tends to liken scientific theorizing to our common sense beliefs concerning physical objects which can be viewed as hypo­theses intended to explain the phenomena. I think that there is a cer-

Page 206: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

200 SOLIPSISM PROPER

tain parallel here, but one cannot even assume that all scientific theo­ries are like our example of the motons and sotons. And even if we do compare common sense beliefs to theories there is no reason to think that they will be the same in this respect and that the confirmation of individual sentences about physical objects will be subject to the diffi­culties mentioned above. In fact, it is intuitively implausible to suppose that they are, and since Quine thinks that the indeterminacy of trans­lation applies to the statements of common sense as well as the state­ments of science, some additional argument should be supplied.

There is another factor, which was mentioned briefly above and which might even allow us to grant Quine's main point and yet deny its applicability to our translation of material object sentences into sentences in the solipsistic language. When Quine talks about the possi­bility of very different translations being made by two different bi­lingual translators, he has in mind the sort of situation where one translator takes "Gavagai" to refer to something like a brief temporal segment of a rabbit and the other takes it to refer to the whole rabbit. We could say at this point that both translators are translating correct­ly but that they are not really beginning with the same language. That is, if they understand the language in such different ways that one is systematically attaching object words to brief temporal segments of objects while the other attaches them to the whole objects, we might conclude that at least one of them has actually created his own lan­guage and that he is translating sentences from that language rather than Martian. It might even be that some Martians do use "gavagai" to refer to brief temporal segments of rabbits while others do not; there would then be a curious mixture of two languages among the Martians themselves. In that case indeterminacy of translation only arises when we have more than one person speaking the languages in question. If one man speaks both languages he automatically translates correctly from his version of one to his version of the other and there is no possi­bility of indeterminacy. We might then suppose that the philosopher who constructs the artificial solipsistic language is analyzing his own ordinary language; there would then be no indeterminacy of analysis in Quine's sense although he still might not be sure that his analysis is correct. Thus Quine'S thesis, even if correct, has limited applicability and, on the whole, we still have to be convinced that there could be two "contrary" correct philosophical theories.

Even so it will be of interest to look at some alternatives to solipsism. It is a feature of axiom systems in mathematics and elsewhere that the

Page 207: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ALTERNATIVES TO SOLIPSISM 201

system itself can be given various interpretations in which case the axioms turn out to be about different subject matters and make differ­ent assertions. Our solipsistic system can also be given various inter­pretations, and we can get a different philosophy simply by inter­preting the system we have in a different way as opposed to construct­ing an entirely new system. This time we will interpret the primitives in such a way that they will not refer to mental events but to events of a behavioral or physical nature. Thus the system itself will remain the same but under its new interpretation it will have exactly opposite philosophical implications. We will therefore define the notion of a stimulation! in such a way that a stimulation will consist of purely phys­ical events within the nervous system, the body, and the world at large. However, we still want to be able to analyze material object language in terms of the language of the system, or at least make a plausible claim for so doing. Thus, even though there is a great difference between sensations, as we have previously interpreted the term, and stimula­tions we will define the concept of a stimulation in such a way that there will be a one-to-one correspondence, or something very close to it, between sensations and stimulations. The analysis of a material object statement will then remain the same, as will the analyses of statements about other minds and beliefs, which have yet to be presented.

Stimulations can also be split up into classes of somatic stimulations, location stimulations, and object stimulations according to the same fairly rough metalinguistic procedure that was used in the case of sensa­tions. Having done this, the rules for constructing the analysis which we sketched out before will be just as applicable to the new inter­pretation as to the old.

Recent work by James Taylor in the psychology of perception allows us to give a more rigorous interpretation than would otherwise have been possible.2 His object is not to give a definition of a stimulation in our sense, but we can adapt his procedure. His first undertaking is to consider the situation where an infant is lying on his back and reaches out his hand to touch a visually perceived object. The perceptual variables here are listed by Taylor as follows:

l. The angle at which a ray oflight from X (the object) enters the left eye.

2. The corresponding angle for the right eye.

1 This concept is similar to one used by Quine in Word and Object, pp. 31ff. • James Taylor, The Behavioral Basis of Perception, chap. 2.

Page 208: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

202 SOLIPSISM PROPER

3. The horizontal deviation of the left eye from the median plane of the head.

4. The corresponding deviation of the right eye. 5. The vertical deviation of the eyes from a horizontal plane through

the head. 6. The rotation of the head about its vertical axis. 7. The backward or forward tilt of the head from the vertical axis

of the body. 8. The sideways tilt of the head from the vertical axis of the body. 9, 10, 11. The initial position of the hand referred to a convenient

coordinate system. It should be noticed that all but the last three variables are concerned with the focusing of the eyes on the object in such a way that its distance and direction can be gauged. If we then consider just the first eight variables, it can be seen that, given a normal brain and nervous system, they determine the somatic visual sensations which the observer has when he uses his eyes. Taylor uses the language of set theory to state his results and he claims that in the case of each of the factors the rele­vant variable is determined by a function in the strict mathematical sense. The most significant variable in each case is the amount of afferent energy directed through the nervous system to the brain. If the brain is then functioning normally it is assumed that the same signals sent to it will produce the same sensations on different occasions. It might be possible to measure this kind of afferent energy directly, but it would defeat the purpose of this kind of system to interpret the primitives in such a way that the proposed definition of them involves means of measurement which presuppose, for instance, the whole scientific theory of electricity. This system will be a physicalist one in that the primitives will refer to physical events, but the point is that the definition of the primitives should refer only to simple phenomena occurring in the immediate environment which can be easily measured without special equipment the relevance of which would presuppose great chunks of science. Then the system will tell us how to construct complex statements about material objects out of simpler statements about material objects.

The functions which Taylor sets forth allow us to give the required sorts of definitions. Let us now consider his treatment of the horizontal deviation of the left eye from the median plane of the head. He defines A 3 as the set of all horizontal angular deviations of the left eye from the median plane of the head and p3 as the set of all afferent impulses that

Page 209: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ALTERNATIVES TO SOLIPSISM 203

can arise from the lateral rectus muscles of the left eye. He then provisionally gives the function as (a;p) {a e A3, p e P3, ta = t p} where ta and tp are the times at which a and p occur. If this is actually a func­tion then the afferent impulses would be dependent solely on the angu­lar deviation of the eye. However, there is another relevant factor which prevents it from being a function. "It is known, however, that when a stimulus is applied continuously to a receptor organ the discharge of afferent fibers rises rapidly to a maximum and then diminishes more slowly. Consequently, the afferent discharge at any moment is a joint function of the energy of the stimulus and its duration".1 That is, it is not only a matter of the degree of angular deviation, but also of the duration of the angular deviation and of the temporal position of the relevant moment within that duration. For this reason Taylor suggests that we choose a set, T, of convenient temporal durations in order to standardize this factor. It seems that he should go further and require that we measure these durations from the instant that the deviation of the eye first occurs (this instant would mark the beginning of the opera­tion of the rectus muscles). He can then expand his ordered pairs into ordered triples. The function could then be defined as the set of ordered triples (a;b;p) such thataeA3, b. (the set of durations just discussed), p e p3 and ta = tp = tb' Then the third variable, P, is a function of the other two variables. This then allows us to define the afferent variable in terms of the other two and hence interpret the primitive terms of the system without ever referring to afferent energy in any non-eliminable way. The other variables, unlike the afferent variable, can be easily measured without much in the way of theoretical assumptions; in fact, all that we need is a protractor and a stopwatch. Taylor then goes on to state similar functions in the case of the other factors, and these could give rise to similar definitions. Having given simple operational defini­tions for each of the variables, we can then define a stimulation as a set of values for the variables at any given time. This stimulation will correspond to the somatic sensation oflooking in a particular direction in a particular way (turning the head or turning the eyes, etc.). Of course, we will have an infinite number of stimulations where we have only a finite number of sensations and it takes one further step to get an exact correspondence. We would have to discover the least differ­ences in stimulations that we can notice and then form sets of stimula­tions such that each set contains an infinite number of possible stimu­lations but such that the maximum difference between stimulations

Ibid., p. 16.

Page 210: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

204 SOLIPSISM PROPER

belonging to the same set is less than the least noticeable differ­ence.

Of course, we have only dealt with one particular kind of sensation, and a somatic one at that, but there is nothing to prevent the same method being used in the case of any kind of sensation; Taylor, in the course of his book, does define functions for many different kinds of sensations. All this does not allow for abnormalities in the state of the brain which would produce deviant sensations and behavior even in the presence of normal afferent impulses. However, there is nothing in the phenomenalist interpretation which prevents the occurrence of illusions and hallucinations, and the techniques which we will develop later for dealing with these phenomena are also open to the physicalist interpretation at least insofar as problems arise with the analysis of statements about material objects.

Now that we have re-interpreted "sensation" and our primitive pre­dicates in such a way as to apply them to stimulations instead, we have left out the original meaning of these terms and we must now introduce terms referring to mental events as defined terms. This is obviously go­ing to require a behaviorist assumption, but we are in a very good position, relative to most kinds of behaviorism, because there is such a close correspondence between sensations (in the original sense) and stimulations, or sets of stimulations. That is, trying to define sensations in terms of stimulations is likely to be a more rewarding enterprise than trying to define sensations in terms of behavior at horse races or a propensity to say "ouch." If we could be sure that the sense organs, nervous system and brain always functioned in the same way, we could add a principle stating that fact to the axioms and then get a derived equivalence between the occurrence of stimulations and the occurrence of sensations. However, these gadgets do not always function normally, and we are aware of this fact, so such a principle would not be an appropriate axiom. Instead we would have to formulate a description of physiological normality which would involve a whole series of tests designed to discover whether the linkage between the stimuli, as ap­plied to the receptors, and the end product in the brain is the normal one. This description must contain no reference to sensations and it would probably include such things as shining lights in the subject's eyes and watching them move, the application of rubber hammers to knees, etc. Since there is no reference to sensations the description can easily be put into the system. We will then define a statement such as " .... mental event (sensation on the original interpretation) occurs" as

Page 211: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ALTERNATIVES TO SOLIPSISM 205

" .... sensation (interpreted as a stimulation) occurs and physiological normality is satisfied."

It is interesting to note that the definition does not say who is physio­logically normal nor does it have to. The new interpretation of the system does not change the fact that the system is constructed to refer to only one person's sensations (stimulations) on the primitive level; even on the physicalist interpretation, all other persons are defined entities and statements about them are constructed in a logically different way. The only oddity is that on this interpretation one does not have to choose oneself as the favored subject, and it may be more natural not to do so. The most natural procedure might be to choose someone that is believed to be honest, co-operative, and a good physical specimen and to then verify his physiological normality. He would then be set to keeping a diary of his stimulations and one could send him around the world to observe phenomena. Hence the world could be constructed as seen through the eyes of an ideal observer! Of course, the obvious objection would be that one has to trust oneself at some point anyway (in choosing the observer and checking him), so why not trust oneself to observe the world oneself? This sort of objection would ordinarily be decisive. However, such a philosophy might seem more natural to someone who was shut up in a cave and was only per­mitted one visitor; one might then be tempted to "see" the world through his eyes rather than through one's own. Thus it seems that, under any kind of normal circumstances, the asymmetry in the treat­ment of persons which goes naturally with a phenomenalist account becomes a disadvantage when we reinterpret the system and give it a physicalist-behaviorist slant. On the other hand, while many of the objections to behaviorism remain, the jump from physical phenomena to mental phenomena is not as great as it might be.

One of the main difficulties with such an interpretation has to do with how much economy we are really going to achieve even apart from questions about minds. The interpretation of a stimulation which we gave is one of the more favorable ones and it gave hope that we could analyze statements about complex physical phenomena in terms of statements about the occurrence of simple physical phenomena. However, there are times when Taylor has to talk about the energy of reflected light. Given his purposes, this constitutes no difficulty, but it does for us. In the case of stimulations corresponding to object sensa­tions rather than somatic sensations, it should be remembered that we can determine shape and color, which are essential to our visual per-

Page 212: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

206 SOLIPSISM PROPER

ception, only because of rays of light impinging on various parts of the retina. Even in dealing with the first variable mentioned above (the angle at which a ray oflight enters the left eye) Taylor winds up talking about both the direction of rays of light and their energies. He thinks that wave length is irrelevant here and he does not think that it is nec­essary to talk about the absolute energy of reflected light but only about the ratio of the energy of the given light to that reflected by some stan­dard substance, such as matte black paper. It might be possible to measure this ratio in a fairly simple way but it can be seen that added complications will arise here. Thus it is very doubtful whether the primitive predicates which would ultimately be needed would be simpler and easier to apply than the predicates we ordinarily use in talking about physical objects. One could claim only that the sorts of predicates used in the more abstruse sciences are reducible to the kind of predicates we would need in order to describe perceptual situations. Thus it seems that if the phenomenalist interpretation is successful at all it promises to give us more than this particular kind of physicalism.

There is another interesting difference between the two interpreta­tions. On our physicalist interpretation it will be impossible to deduce a theorem asserting the existence of a material object without the favored subject's sensing the object or sensing something related to it. And this sensing means not just having a mental event but also involves the occurrence of a stimulation. It also seems to be true of our ordinary speech that we never do assert the existence of anything unless we have some sensations we can relate to it even if these are only sensations of other people telling us that they have heard other people telling them that other people have told them that .... and so on. Or if people do occasionally assert the existence of such things as gold mines on intui­tion without claiming any kind of evidence they at least imply that there is a physical path which, if followed long enough, would bring our bodies into physical contact with the gold. It is an essential part of the physicalist interpretation that it be possible for the favored sub­ject to bring his sense organs into physical contact with any object that exists or, in the case of vision, hearing, and smelling, into appropriate range under appropriate circumstances. Even in the phenomenalist interpretation the rules of analysis which we gave and would give, have the same consequence, since we are trying to "match" the statements of ordinary language. But it is not essential to phenomenalism as such to analyze particular statements of ordinary language in this way. Suppose, for instance, that we had bodies very different from the ones

Page 213: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ALTERNATIVES TO SOLIPSISM 207

we actually possess and that instead of sense organs we had something like radar. Let us further suppose that certain results within our organic radar systems produced the same sorts of sensations that our sense organs produce. Suppose further that our radar systems scanned automatically without having to be focused and that we spent our lives whizzing through space. We would then have relatively few sensa­tions and the ones we did have would come effortlessly so that there would be no difference between perception and observation. It would then be a general rule that we would never be close to the objects we perceive and we would never be conscious of having to be in any kind of special position in order to perceive them. We could still give a phenomenalist analysis of statements about material objects, and it would be changed relatively little. The material object statement, "Mars bears 40° to starboard and is 7 million miles distant" would still imply such hypothetical statements as "If such and such sensations of the earth and Venus, then such and such Mars-like sensations will immediately follow" ; it would just be a matter ofleaving out any referen­ce to somatic sensations which would no longer occur or be needed. With our radar sets we would determine direction by the order of occurrence of sensations instead of turning our heads and moving our eyes. In short the solipsistic interpretation ultimately makes reference only to mental events and it makes relatively little difference how our bodies bring about our conscious experience. Some of our concepts, particular­ly those relating to our own bodies, would be changed in such a world, but phenomenalist analyses could be given for the new concepts. On the other hand, there would be no such things as afferent energy and angular deviations of the eyes in such a world and the physicalist inter­pretation would have no application at all. Concepts from physiology would have to be replaced by electronic concepts, but this would amount to finding a new and altogether different interpretation. Thus there is a sense in which a philosophy based on the physicalist inter­pretation is less general than one based on the phenomenalist inter­pretation. The phenomenalist could not claim this as an advantage if the substitution of a world such as the one just mentioned involved the scrapping of our ordinary language and the substitution of some other sort of language for it. But in fact it seems that most of our ordinary concepts would remain; whether we have radar sets or brains and nerv­ous systems in our heads, we still want to identify and re-identify physical objects of many different kinds, and there is nothing in the strange world which would keep us from doing this. This being the

Page 214: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

208 SOLIPSISM PROPER

case, our analysis of the language of common sense should be as flexible as that language is itsel£ Even a kind of physicalism which did not in­volve the "favored subject" anomalies might have difficulty in pro­ducing a sufficiently flexible analysis. Nevertheless, it must be remem­bered that the kind of physicalism we have stated here is of a rather peculiar sort and most of the difficulties we encountered are not in­herent in physicalism as such. Part of the reason for considering this theory at some length was to indicate that there are more and stranger possible philosophies in heaven and earth than one might have supposed. A more important reason was to show that the quasi-axiom­atic method we have followed gives our system a great deal of power and flexibility. Once the system has been sketched out, even though it has not been fully stated, it is possible to create any number of new and different philosophies with, so to speak, a mere flick of the wrist.

Let us now pass to a much more respectable philosophy, that of P. F. Strawson. In Individuals he presents an interesting theory which must certainly be counted as competing with solipsism and which has collected many more adherents than any kind of solipsism is ever likely to attract. While Strawson does not present his views as a system in our sense, he comes much closer to it than most philosophers of his general outlook, and it is possible to see how his views might be systematized even though he probably would not want to do so. He starts from quite clearly defined premises and we will be able to agree that almost all his conclusions do follow from those premises. In fact, we will be able to accept some of his more important conclusions categorically quite apart from those premises. The only point at which we will disagree is at the very beginning, and that concerns the relationship which Strawson believes to hold between his view and other alternatives, such as our solipsism. The basic premise is that the language we speak is a public language which is not reducible to, or constructed out of, some con­catenation of private languages. Given this language he then asks how a public language is possible and most of what he says in the first half of the book can be looked upon as an answer to this question. A related premise is that we all make use of a single spatio-temporal framework which we use to locate bodies. In this connection he briefly discusses the sort of private language position we have taken as follows: A different, but not unrelated, error is made by those who, very well aware that here-aM-now provides a point of reference, yet suppose that "here" and "now" and "this" and all such utterance-centred words refer to something private and personal to each individual user of them. They see how for each person at any moment there is on this basis a single spatio temporal network; but see also that, on this basis,

Page 215: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ALTERNATIVES TO SOLIPSISM: 209

there are as many networks, as many worlds, as there are persons. Such philosophers deprive themselves of a public point of reference by making the point of reference private. They are unable to admit that we are in the system because they think that the system is within us; or, rather, that each has his own system within him. This is not to say that the schemes they construct may not help us to understand our own. But it is with our own that we are concerned. So we shall not give up the platitude that "here" and "now" and "this" and "1" and "you" are words of our common language, which each can use to indicate, or help to indicate, to another, who is with him, what he is talking about. l

In the first place, Strawson speaks of this position as an error rather than as a rival theory, and it is easy to see why he does this. In his own view he is doing descriptive metaphysics rather than revisionary meta­physics, and he thinks that this involves describing the conceptual framework that we have rather than constructing some other one that we might have instead. In general, Strawson's analysis attempts to conserve our ordinary usage as much as possible. While he does not outline an axiomatic system he does talk about primitive terms and if such a system were to be constructed using his account as a guideline, the primitive terms would refer to such entities as persons and bodies. Thus the axioms of such a system, unlike some of the axioms of our system, would, even in primitive notation, be recognizable and would be understood by the man in the street. For this reason Strawson tends to suppose that his analysis is descriptive while others, such as ours, are revisionary. However, on our view, any kind of metaphysics or any kind of systematic analysis of our ordinary concepts leads ultimately to the kind ofaxiomatization that we have discussed. Further, at various points Strawson asks what sorts of notions should be taken as primitive and this reveals some tendency on his part in this direction. But, given various different axiomatizations of our ordinary language, one should not then say that the correct system is the one whose axioms, in their primitive notation, look most like the statements of common sense. To see whether an analysis is adequate one has to look at the kinds of theorems that can be derived and use the definitions in translating these theorems so that they will be recognizable. If the theorems turn out to be the beliefs of common sense and we cannot put our finger on any­thing that is ordinarily asserted in English and which is left out of the system, then we have merely to worry about consistency, circularity, and economy. We do not have to worry about the presence or absence of any superficial resemblance between the axioms on one hand and the beliefs of common sense on the other.

If one wanted one could be even more descriptive than Strawson. 1 P. F. Strawson, Individuals, p. 30.

Page 216: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

210 SOLIPSISM PROPER

For instance, instead of taking the concept of a person as primitive, one could take the concepts of man, woman, and child as primitive and not attempt to define any in terms of the others. In fact we may very well learn the words which represent these concepts independently of one another rather than by definition; hence, a system which took all these and other concepts as primitive might more closely describe the conceptual framework as we learn it. But a system which attempted to preserve all such relations oflinguistic independence in the learning process would unnecessarily proliferate primitive terms. In fact it is obvious that Strawson is not completely impervious to considerations of economy, and the whole issue depends on how far one should go in narrowing down the class of primitive terms. One cannot stop at some particular point and say that it is just there that one has captured the meaning of the sentences one sets out to analyze, and this is what Strawson seems to do. Our basic disagreement with Strawson could, perhaps, best be put by saying that for us it is not obvious what the key terms of our ordinary language mean or what the conceptual framework embedded in ordinary language is like; this is something that we can only theorize about and, having constructed rival theories, we can compare them in the way indicated above. Strawson thinks, for instance, that a public spatio-temporal network is necessary for the kind of identification and re-identification of particulars that we or­dinarily engage in. On his theory this is, of course, true, and we can take it as one of the consequences of his theory. But it is not something that one should assume before one starts theorizing at all. Our con­clusion is that one cannot say which systems are descriptive and which revisionary unless, of course, one just picks out the one which is believed to be the best and says that only it is descriptive.

Still, we may have over-emphasized the points of disagreement. The passage quoted is the only one where he talks about private languages in any detail and, even there, he does not attempt to show the im­possibility of a private language. Hence it may be fairer to interpret Strawson simply as trying to construct his own system which can be compared with other systems without meaning to engage in any detail­ed discussion of the alternative systems. This is, of course, a legitimate enterprise.

We do not have the space to criticize the theory in detail, but there are two points about it which are of particular interest. First, one of Strawson's most important theses is one that we can accept ourselves and which is true on any system. Secondly, we will try to show that

Page 217: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ALTERNATIVES TO SOLIPSISM 211

Strawson's system is less economical than our own. The thesis with which we agree is: But how is it that one can ascribe them [states of consciousness] to others? Now one thing here is certain: that if the things one ascribes states of consciousness to, in ascribing them to others, are thought of as a set of Cartesian egos to which only private experiences can, in correct logical grammar, be ascribed, then this question is unanswerable and this problem insoluble. If, in identifying the things to which states of consciousness are to be ascribed, private experiences are to be all one has to to go on, then just for the very same reason as that for which there is, from one's own point of view, no question of telling that a private experience is one's own, there is also no question of telling that a private experience is another's. All private exper­iences, all states of consciousness will be mine, i.e. no one's. To put it briefly. One can ascribe states of consciousness to oneself only if one can ascribe them to others.1

Thus, at any level of language, we must either ascribe conscious states both to ourselves and others or not ascribe them to anyone. In our system we choose not to ascribe them to anyone at the primitive level, and this is reflected by the fact that words such as "I" and "my" do not appear at all in our primitive language. Strawson does criticize some­thing which he calls the "no-ownership" view, but this turns out not to be our position.2 The philosopher who holds it is one who first asserts that all "my" expreiences are "had" by a particular body in the sense that they are causally dependent for their occurrence and nature on that body. He then says that the belief in the existence of an ego arises in a confused way out of this contingent fact. Strawson refutes this view quite easily, and it should be clear that there is no connection between it and our view on which we do not have persons at all on the primitive level. We do, of course, introduce persons as defined entities and, at that point, we can ascribe states of consciousness both to our­selves and to others.

Strawson takes the opposite course and ascribes states of conscious­ness both to ourselves and others on the primitive level. Thus he says: What we have to acknowledge, in order to begin to free ourselves from these diffi­culties, is the primitiveness of the concept of a person. What I mean by the concept of a person is the concept of a type of entity such that both predicates ascribing states of consciousness and predicates ascribing corporeal characteristics, a physical situ­ation &c. are equally applicable to a single individual of that single type. 3

This is the essence of Strawson's theory right here: instead of taking the concept of a sensation as primitive he is taking the concept of a person as primitive.

In order to compare Strawson's view with our own with respect to

1 Ibid., p. 100. 2 Ibid., pp. 95-97. 3 Ibid., pp. 101-102.

Page 218: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

212 SOLIPSISM PROPER

economy in a rigorous way we would have to do for his theory what we have done for solipsism. However, we might briefly mention the sorts of concepts which might be taken as primitive in a Strawsonian system. Strawson himself mentions both persons and material bodies as basic particulars and he refers to the concepts of these entities as being primitive in our conceptual scheme. However, there are some other loose ends which appear particularly in the second part of the book. For one thing, there is the concept of a logical subject, or individual, and, while all particulars are individuals, there are some individuals which are not particulars. These include such universals as wisdom which can be talked about as logical subjects. Further, Strawson goes to great lengths to distinguish subjects from predicates and his basic theory is that subjects have a kind of completeness which predicates lack. Thus the name ''John'' can stand on its own feet in a way that "is wise" cannot. A subject is said to typically introduce a particular, and something is presupposed in such an introduction - namely the uniqueness of the particular introduced. Thus a singular term gives us some "news" by the very fact of its use although this news is pre­supposed rather than entailed. Hence it can stand on its own feet because something has been accomplished just by its being mentioned, and it can also figure in some more complex locution and be used to help assert something else. A predicate, on the other hand, presupposes nothing by its very use and it has to be combined with a subject before even a hint of an assertion arises. But, of course, if a subject which introduces a particular makes a presupposition, the statement of that presupposition will very likely introduce another particular and make a further presupposition. Strawson does not think that this goes on endlessly and he suggests a stopping-place even though he qualifies this suggestion and seems very hesitant about it. The suggested stopping­place comes with what Strawson calls feature-concepts. At this point the presupposed sentences cannot involve or introduce any further particulars in any way and he thinks that they should not even involve the sorts of universals whose correct use involves knowing how to dis­tinguish, identify, or re-identify particulars. An example of a "feature­placing statement" is "It is snowing." A feature concept rather than an ordinary universal is involved because we are not told how to split the snow up into grains or expanses that we might re-identify. Strawson then goes on to suggest how a language based on feature concepts rather than particulars might be set up in his chapter on "Language without Particulars. "

Page 219: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ALTERNATIVES TO SOLIPSISM 213

It seems to me that in axiomatizing Strawson's system much more emphasis should be placed on these feature-concepts than is in fact given them in his book. Thus we are left at this point with at least four basic categories: bodies, persons, individuals, and feature concepts. We must now see what might be done to reduce the list. It seems first that one should be able to eliminate the category of bodies as primitive concepts fairly easily. If we are allowed to take the distinction between P-predicates (predicates applicable only to persons) and M-predicates (predicates applicable to both bodies and persons) as more basic we could then define a body in general as any entity to which some M­predicates but no P-predicates are applicable. Most of the predicates which bodies have are predicates which persons can also have, al­though there are some exceptions, such as "weighing 35,000 tons." However, it would seem that predicates such as these are easily de­finable in terms of predicates which do apply to persons. The problem then is how to distinguish M-predicates from P-predicates since there is no single feature of one that we could specify without presupposing the distinction between persons and bodies. However, at the very worst one could simply go through the Oxford English Dictionary and divide all the predicates to be found into the two classes and take only the class of P-predicates as primitive. In fact, one could probably divide these predicates into fairly small subsets and find a distinguishing feature for each subset, but this is not a theoretical requirement. Thus we might eliminate the concept of a body as primitive, but we are pre­supposing the concept of a predicate in doing so. However, we are making progress since what we really need at this point is not the gen­eralized concept of a predicate; features and feature-concepts will do just as well. That is, we could divide the whole class of feature concepts up into M-features and P-features and then take the latter set as primi­tive terms. One might then even be able to get along without being able to refer to particular persons but merely by "placing" the features so carefully and so ingeniously that they turn out to apply only to the persons one might have wanted to refer to. However, it is pretty clear that Strawson does not want to go that far and that, if we were to get a system that we could call Strawsonian, we must keep means of referring to particular persons. Still, we can easily define a predicate wherever we have a feature and we can then apply that predicate to persons. The advantage of taking features as primitive would be that we will have in our language the means of introducing particulars even if we do not attempt to define the terms that refer to particulars.

Page 220: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

214 SOLIPSISM PROPER

Having got this far, the category of individuals should add nothing new to the primitive language. What persuades Strawson to admit non-particular individuals in the first place is the consideration that the locution '( ... is wise), is sometimes true', which can be inferred from 'Socrates is wise,' can be expressed either by 'There exists someone who is wise' or by 'Wisdom exists.'l In the one case the logical subject is a particular and in the other case it is a universal, and there seems no reason for preferring the one to the other. However, if we already have features in our primitive language, and the predicates which are derivable from them, there is no reason why terms referring to features should not be logical subjects and hence individuals. Thus there is no need to admit any more different kinds of individuals.

Our conclusion would then be that a system based on Strawson's philosophy could probably just take persons and features as primitive. Of course, we have this time just given a sketch of a sketch of an axiom­atization, but the point is that if we have erred it has almost certainly been on the side of economy. When one actually came to construct the system one might well find that more primitives are needed, but one would not find that one category or the other is redundant and still have a Strawsonian system. In addition to persons and features one would also need all the logical primitive terminology, including coun­terfactuals, which was introduced in our system. Hence we can fairly safely say that Strawson's system is less economical since we got along with less in the way of non-logical primitives. Of course, we should remember that Strawson is not trying to maximize economy and this relative lack of it is certainly nothing like a fatal defect. But it is some­thing that would have to be borne in mind if the t\-Vo systems are satis­factory in other respects and we came to compare them.

Strawson also gives a refutation of a kind of scepticism. He says of certain sceptical problems: ... Their statement involves the pretended acceptance of a conceptual scheme and at the same time the silent repudiation of one of the conditions of its existence. That is why they are, in the terms in which they are stated, insoluble.2

One might suppose that the kind of scepticism he refutes has some rela­tion to the kind of solipsism we are advocating. However, the scepticism referred to here is that which one gets if one first ascribes states of con­sciousness to oneself in a primitive way and then denies that others have similar states of consciousness, or systematically doubts whether they

1 Ibid., p. 237. S Ibid., p. 106.

Page 221: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ALTERNATIVES TO SOLIPSISM 215

do. This is a kind of scepticism which cannot be formulated in our system. On the primitive level one cannot ascribe states of conscious­ness to anything or anyone and, once we introduce the mechanism whereby we can ascribe states of consciousness to persons, we will be able to prove as theorems statements which ascribe states of conscious­ness to others as well as ourselves. Hence, our solipsism is entirely distinct from this kind of scepticism.

Page 222: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 11

ANTI-SOLIPSISM

It will have been noticed that in our system the basic particulars, sensations, play much the role that sense-data do in the writings of such philosophers as A. J. Ayer and H. H. Price. Many recent philosophers not only find the concept of a sense-datum objectionable but also reject all similar concepts among which they would probably include the concepts represented by the primitive terms of our solipsistic language. Since the whole system depends entirely on the possibility of a language whose terms refer solely to sensations we will take account of these objections in considerable detail. The most important attack on the sort of position we are taking is that of J. L. Austin in his Sense and Sensibilia. Here he is explicitly considering the views to be found in Ayer's Foundations of Empirical Knowledge, but he would undoubtedly think that many of the objections are also applicable to the sort of view presented here. Even where our theory diverges sufficiently from Ayer's so that the objections are not applicable, it is still important to see to what extent we are departing from the traditional sense-datum position and the phenomenalism which Ayer once represented.

In this book Ayer distinguishes between a language in which all the terms refer to sense-data and our ordinary language in which he takes the terms to refer to material objects. Thus he presupposes a sharp distinction between sense-data and material objects almost from the very beginning. It is this that bothers Austin as much as anything and he says (p. 4) "There is no one kind of thing that we 'perceive' but many different kinds, the number being reducible if at all by scientific in­vestigation and not by philosophy: pens are in many ways though not in all ways unlike rainbows, which are in many ways though not in all ways unlike after-images, which in turn are in many ways but not in all ways unlike pictures on the cinema-screen - and so on, without assignable limit. So we are not to look for an answer to the question, what kind of thing we perceive." As Ayer uses them, the terms "sense­datum" and "material object" form a dichotomy which together com­prise the set of all objects which we can be supposed to perceive.

Page 223: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI-SOLIPSISM 217

Austin, however, thinks that this is a bogus dichotomy in that we can­not divide up the objects of perception into two classes such that there is more difference between members of two classes than between members of the same class. While Austin spends relatively little time attacking this dichotomy directly most of his discussion is directed at the notion of being deceived by one's senses and the notion of "direct perception." In Ayer's account these other concepts are closely tied to the dichotomy between a material object and a sense-datum, but since there is no such connection in our theory we will make a con­siderable diversion here in order to examine the dichotomy in itself. Since we have ourselves made use of a similar dichotomy between in­dependent and dependent sensations, we will take this opportunity to clarify that dichotomy with respect to the sorts of questions Austin raIses.

Austin, of course, mentions entities like rainbows, images on a movie screen, mirages, shadows, and other such entities to which we may, or may not, attribute independent existence. Some of these, for instance, have a certain stability and constancy, but do not seem to be physical objects in quite the way that stones and chairs are. Like a rainbow a shadow is not something that can be touched and, unlike almost all physical objects, it is automatically banished when we turn a light on it. Austin might also have mentioned in this connection such events as thunderclaps. On the one hand we want to say that a thunderclap is something that takes place in the physical world and is something that anyone can hear, but, on the other hand, we do not treat sounds as physical objects in themselves but as emanations from physical objects; this case is queer because there is no readily identifiable physical object of which we can say that it is making the noise. Thus, again, we do not know whether to treat it as an object or as a mere sensation which occurs to more than one person with a certain regularity. Since it is really too much of a public and objective phenomenon to be treated just as a sensation, and too ephemeral an object to really be treated as a physical object, Austin's suggestion would presumably be that we put such events into separate categories of their own and not attempt to divide everything into two camps. Any attempt to do that, he thinks, would be largely arbitrary.

It seems that there is another dichotomy which Austin does not mention, since Ayer does not rest his case on it, but which is neverthe­less the traditional basis for this dichotomy between material objects and sense-data, sensations, perceptions, or whatever. This would sim-

Page 224: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

218 SOLIPSISM PROPER

ply be the dichotomy between purely mental events and events which are purely physical or in some sense both mental and physical. Some philosophers, such as Ayer in Language, Truth, and Logic, have consider­ed sense-data to be neither mental nor physical but as neutral entities out of which both minds and physical objects are constructed. Never­theless, it usually turns out that sense-data are purely conscious events and are what we would ordinarily take as mental events despite the fact that material objects may be constructed out of them. There is, of course, a very old philosophical dichotomy between conscious events which are supposed to occur at a time but at no place in the strict sense and material events which occur at both a place and a time. It is also supposed that anything that happens must be either mental or physical and that an event can be both mental and physical only by having a mental component and a physical component. That is, it is supposed that the same part of the event could not be both mental and physical at the same time. Armed with this sort of dichotomy the theorist who has sense-datum inclinations can distinguish three different classes of events. First, there would be individual perceptions or sense-data which would be purely conscious events. Secondly, in order to take care of such entities as mirages, rainbows, shadows, thunderclaps and the like, there would be logical constructions out of sensations. A percep­tion of such an event counts in itself just as a conscious event but it also counts as an observation ofa public object (the rainbow, shadow, etc.) which can be observed by anyone. Thus the existence of the object will consist in nothing more than the fact that sensations of specified sorts could and would be obtained in specified situations. The third class would be the class of genuine material objects such as pens and books which also are observable in certain regular ways but which may be said to have something "behind" the sensations which accounts for them. If the theorist is a phenomenalist he collapses the second and third classes and says that all material objects are logical constructions out of sensations. The distinction between rainbows and pens would then just be that pens can be observed in more different ways than rain­bows; that is, the logical construction representing the one will be more complex than that representing the other. Thus we can see right at the beginning that Austin would be committed to attacking another dicho­tomy - that between purely conscious events and other sorts of events. Here the Austinian would again look to particular cases and might mention such examples as a feeling of anger, a dream, an image of an island, a painful sensation, and a case of seeing a book. It would then

Page 225: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI-SOLIPSISM 219

presumably be argued that we cannot split these cases up between the mental and the physical in any simple way. Actually Austin tends to assimilate the fact that something looks in such-and-such a way to facts about the physical world, and he could be expected to argue that any dichotomy which puts them exclusively in the mental world is a false one. We will subsequently argue that whether one can set up this dichotomy depends upon one's theory of meaning, that Austin more or less unconsciously presupposes a theory of meaning which makes this dichotomy impossible, and that it is a legitimate dichotomy on the basis of the theory of meaning presented earlier in the book.

However, another consequence of Austins' attack on the dichotomy between material objects and sensations should be noticed. It has al­ready been noted that the phenomenalist is one who takes material objects to be simply logical constructions ou t of sensations, and a logical construction out of sensations is just a set of sensations which stand in particular logical relationships to one another. These constructions can be of all degrees of complexity and even a single sensation can be con­sidered as a kind of degenerate logical construction out of sensations. Thus the phenomenalist in the end minimizes the difference between material objects and sensations as opposed to a substance theorist, such as Price, who argues that in some cases there are physical occupants which give off sensations while in cases of hallucination and mirage there are no such occupants. According to his account there would real­ly be a sharp difference between a material object and a sensation - the presence or absence of the physical occupant. However, the basis for any such sharp distinction is lacking in phenomenalism. Thus when Austin argues that there is no basis for a dichotomy between sense-data (or sensations) and physical objects he is actually making the phenom­enalist position more plausible. He does not intend to do this because he probably thinks that it is not possible for the phenomenalist even to set up his language and get started on his analysis, but once started, he can make use of Austin's arguments against those who think that there must be something more to a material object than a logical construc­tion out of sensations.

In order to get started the phenomenalist has to make some sort of distinction between sensations and material objects; according to his own account the distinction will not be an earth-shaking one, and the latter gets constructed out of the former, but he does have to set up a language of sensations which, of course, turns out to be a private lan­guage. In order to do this he has to say what sorts of terms he is going

Page 226: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

220 SOLIPSISM PROPER

to have in his language and to what sorts of entities they will refer. Hence there does have to be some distinction at the beginning between sensations and material objects. It is all very well to say that within our language simple expressions describe sensations and that logical constructions of more than minimal complexity refer to material ob­jects, but this does not illuminate the way we are going to interpret the language unless we can independently make clear what sorts of entities or processes are going to have our simple sensation terminology applied to them. We earlier distinguished between independent and dependent sensings and we have consistently wanted to apply the word "sensation" to the common element in these two different sorts of events. It is in this way that we wish to make clear the usage that our primitive ter­minology would have on its intended interpretation.

Thus our theory presupposes that there is a genuine distinction be­tween independent sensings and dependent sensings. Our dichotomy is founded on the idea that whenever something is perceived it is either believed to be identical with the conscious state involved in perception or not, but this dichotomy might not be as absolute as it sounds. When­ever one is perceiving at all one is perceiving some part of the world even if it be nothing but an empty sky, and we believe that even an expanse of air exists independently of ourselves and is non-identical with our conscious state when we perceive it. The whole question depends on what it is that one thinks one might be perceiving - on what sort of object one's present sensation is characteristic of perceiving. Having decided what sort of object one might be perceiving one can then go on to decide whether it is really there. Thus if! see an airplane­like speck in the sky I am more likely to ask myself whether there is really an airplane there than whether the sky is really there. In those cases where we find out, decide, or know from the start that the sort of object one would expect to be present in such a situation is not present, we tend to say that we are just feeling pain, seeing a mirage or having a certain sort of sensation. Now one can "make" mirages into material objects that more than one person can see in much the way that we can allow that more than one person can see a shadow. Then one might say that one is seeing a mirage and seeing something that goes beyond one's conscious experience in seeing it; that is, one could oneself see it at other times and places, and it might be that others can see it. One could even say that one is, so to speak, having a mirage of a mirage in that what one is seeing is something one knows to be peculiar to oneself at the present time and is not something that one could see if one's

Page 227: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI -SOLIPSISM 221

nerves were in a more normal condition. Thus the question whether one is perceiving something which is non-identical with one's own conscious state depends on what one expects to perceive in the first place. We have said before that this makes the difference between an independent sensation and a dependent sensation, and that the differ­ence is one of belief. This still remains true in that a dependent sensa­tion can always be turned into an independent sensation, or vice versa, by a change of belief, but this is not the whole story. The beliefs we have represent answers to questions, and these questions have been raised, usually unconsciously, by ourselves. In most cases where we think that perception might be delusive there is a fairly standard set of questions to ask. When I look at the horizon and see what looks like an island the questions that are immediately suggested if! think that there is possibility of a mirage are, "Could I land there and feel the beach ?," "If we got close, would it be possible to hear the breakers against the cliffs?", and so on. If I were a nonconformist I could raise different questions such as, "Can I shut my eyes and open them again and still see the same thing?", "Could I back up three feet on the deck and still see the same thing?," or "Can other persons near me see the same thing?". These are questions which people might normally ask, but we would not normally take the answers to these questions as establishing conclusively the reality or lack of reality of what we see. Suppose that I do take them as decisive and come out with the conclusion that any­one in roughly my position can see the same sort of thing that I saw originally. If! take this as establishing the independent reality of what I saw, and look again, then I would have an independent sensation. If someone else standing beside me knows by looking at the chart that there is no island there, but is otherwise in the same position I am in, he will probalby follow the usual conventions for establishing reality and be having a dependent sensation. In fact he may inform me that there is no island there, but since I am taking a mirage as a material object, the upshot would be that we are seeing the same things, that we have the same beliefs, and yet one would be having dependent sensations while the other would be having independent sensations. This is because we identify differently the object we are seeing and would normally expect it to act in different ways; in this case we have discovered that it will act in such a way as to fulfill one set of expecta­tions but not the other. Austin discusses a case where we can look through a telescope and describe what we see as a star, a speck, or an image on one of the mirrors of the telescope. In common with our

Page 228: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

222 SOLIPSISM PROPER

account he distinguishes different objects which we might identify as being what we see, but our example is better suited to our purposes because in our case the disparity between the different sorts of objects is greater.

We can conclude that our dichotomy is relative to the kinds of ex­pectations we have for the objects that we perceive, but that it is a genuine dichotomy as long as we do have some set of expectations. The only intermediate cases between independent and dependent sensa­tions would be those cases where we have suspended belief in the ques­tion of whether some object fulfills these expectations and is conse­quently non-identical with our conscious state. However, these cases do not occasion any difficulty although we can, if we wish, amend our definition of an individual sensation so that a sensation will be what independent sensations, dependent sensations, and intermediate cases have in common. We have also pointed out that the perceiver does not ordinarily formulate these expectations as he goes along and that he does not ordinarily consciously decide what questions to ask and what tests to carry out in order to see whether the object of perception has independent existence. Rather, these questions are largely conventional and the expectations we have for the behavior of the objects we see is pretty well determined in advance. We will look briefly at the general form of these conventions in order to see how widely material objects can differ from one another and still come under our definition which says that they need only be distinct from sensations - that they be logical constructions of more than minimal complexity.

In the case of visual perception we ordinarily expect the object per­ceived, assuming there to be one, to be open to tactual inspection as well. Thus, in the case of the island we expected to be able to feel the beach. In the great majority of cases we also expect that it will make sounds under the right circumstances. In the case of the island we can produce sounds by throwing rocks at it, and almost all visually per­ceived objects can be struck with other objects. In most cases we would expect there to be smells and tastes but it is probably conceivable that there be an actually existing island which contained nothing that had any odor or which could be tasted. While these rules hold for most visually perceived objects they do not hold for all. Suppose that I see a cloud of yellow smoke coming out of the chimney of an oil refinery. My first supposition would be that it is a smelly gaseous sulfur com­pound and that if I could get into position to sniff it, I would smell a certain characteristic odor. It might, nevertheless, turn out that this

Page 229: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI-SOLIPSISM 223

particular sulfur compound is odorless and that my expectation is therefore erroneous. Even here, although I would admit to being mis­taken about the nature of the object, I would still say that I really did see a material object and that what I saw was not identical with my sensation. Further, in this case we have different rules since there is really no way in which one could feel the gas in itself. Of course, one could put it into a container such as a balloon and then feel the balloon, but we would then be feeling the balloon rather than the gas - at least, if there are any tactual expectations, they involve much else besides the gas. In the case of rainbows there does not even seem to be any question of touching, hearing, smelling, or tasting them. Thus one has to admit with Austin that even the physical objects which can be visually perceived have all degrees of "substantiality." As pointed out before, we can admit this point without undermining our fundamental dichotomy. When we see a rainbow we normally expect only that we could see it from other points of view. As long as nothing happens to upset these expectations we are having an independent sensation de­spite the fact that the object is not very "substantial."

When one hears a sound one generally traces it to a physical object and one says that that object or event is what one hears. This would imply that if the object really exists it would be possible to have other sorts of sensations of it. At least this is generally true of our ordinary conceptual framework.1 We might also note as a curiosity that when an object makes a sound it must be open not just to visual observation but also to tactual observation. Thus I may hear the sound made by a gas escaping from a container but in that case I can put my hand on the hole and feel the gas coming out. Similarly objects like rainbows and sha­dows which can be seen but not touched cannot make sounds either. In general it seems that whenever something makes a noise it cannot be an object sitting there passively which can only be seen, but it has to be doing something that creates a commotion and this activity always involves something that can be felt. Thus there are special rules linking auditory sensations not only with visual ones but also with tactual ones.

Of course, as always, one can depart from ordinary usage and pare down a material object until it is only open to inspection by one sense. In our case of a thunderclap we would ordinarily say that we are hear­ing the same thing which we would feel if we were struck by the

1 Strawson constructs an interesting alternative conceptual framework which is based on the possibility of a purely auditory world in chapter 2 of Individuals.

Page 230: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

224 SOLIPSISM PROPER

lightning, and the same thing that we see when we see the flash. This is indicated by the fact that we locate the thunder, that is, the source of the sound, very precisely and I could say, for instance, that the thunder almost deafened me because it was only seven feet away by measure­ment from my position to the burned bark of the tree. If we did treat the noise in itself as a material object which is not open to inspection by the other senses we would not be able to locate it so precisely.

Objects which can be felt can usually be seen and can usually be got to make sounds. However, we do say that we can feel an electric current which we cannot see and which is not powerful enough to make crackl­ing noises or produce the characteristic burning smell of a short circuit. Thus the rules which link tactual sensations to the others are weaker than the other rules, and it may be here that we come the closest to recognizing material objects (which we necessarily take to be distinct from any particular sensation) which are not open to inspection by the other senses. If so, we might say that tactuality has a certain priority over the other senses as far as recognizing material objects goes. One might also suppose this to be true of smells and tastes - we often have tastes and smells which we find it difficult to connect with sensations of other kinds. In the case of taste I may have a mixture of a lot of different foods in my mouth and not be able to trace a taste to a food which could be recognized in other ways. In the case of smell I may simply sniff a funny odor and not be able to trace it to anything in the house. However, in these cases we are much less prone to postulate material objects that are smelled or tasted. In the end if I cannot trace the smell I may well conclude that I just imagined a smell or leave it as a disconnected sensation. In the case of the taste that was difficult to identify I may just give up trying to decide which spice in the pud­ding it was that had that taste and let the question drop without ever isolating any object that was tasted. Sensations of this sort are more likely than the others to be treated like pains in that we are quicker to accept them as dependent sensations. The instructive point here is not really the set of rules which connect one sense with another but the fact that we can formulate such rules so easily. This reflects the fact that, given a sensation of almost any sort, we can immediately say what sort of object, if any, is being perceived and what further sensations are to be expected if the object really does exist. For this reason we do not have to stop and decide what it is that we are going to expect, then find out whether those expectations are true, or at least justified, and finally conclude that we are having an independent, or dependent,

Page 231: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI -SOLIPSISM 225

sensation. In most cases all these steps are unconscious and if one were to be asked whether one is perceiving any material object which exists independently of one's present sensation one could reply immediately without giving the matter any thought.

Our outline of these rules also indicates that where we say that we are perceiving a material object we do have some expectations. Even when we feel an electric current of appropriate voltage we may not expect to have sensations of other kinds, but we do expect that we could have felt the shock even if we had walked up to the wire from a different direction or if we had had a different sort of breakfast. Further we ex­pect that the shock would have occurred even if the wire had been in all sorts of different positions so long as it was not insulated. Thus we never say that we are perceiving a material object simply because we have had a certain sensation and without having other expectations of various kinds. These other expectations are fairly complex and de­lineate the behavior to be expected of the object. Usually they take us from one sense to another even if they do not in this case. Ultimately, it is the existence of expectations such as these which allow us to distin­guish sensations from material objects. Thus we can admit the truth of much of what Austin is saying and still have a dichotomy between some class of material things and sensations. The easiest place to draw the line is between the occurrence of solitary sensations and any logical construction out of sensations which contains at least two sensations which are related in a conditional statement. The dichotomy between our saying that we are just having a sensation and our saying that we are perceiving a material object always involves the presence or absence of some kind of belief; the content of this belief varies according to the context and the sort of object which we think we might be dealing with.

All this assumes that we have some way of talking about purely con­scious events, and in the end I think that this is what Austin would deny. We will discuss this point in detail somewhat later.

Austin, in order to see how the sense-datum theorist (primarily Ayer) tries to make his sense-datum vs. material object dichotomy work, next considers a thesis to which he thinks Ayer commits himself. This thesis is that the ordinary man believes he is not perceiving mate­rial things wherever he is being deceived by his senses, and at no other times. It seems to me that Ayer is not asserting a strict correspondence between being deceived by the senses and not perceiving material things but is implying only a weaker thesis, that the ordinary man believes that we are sometimes deceived by our senses when we are not

Page 232: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

226 SOLIPSISM PROPER

perceiving material things even though we think we are at the time. Ayer certainly wants to leave room for illusions - cases where we are perceiving real things, but not the ones that we think we are perceiving at the time. Hence Ayer would want to admit cases where we are seeing material things but are still deceived by our senses. In any case Austin quite rightly rejects the thesis he states with a couple of counter­examples. "An ordinary man who saw, for example, a rainbow would not, if persuaded that the rainbow is not a material thing, conclude at once that his senses are deceiving him. Nor, for instance, when he knows that a ship at sea on a clear day is much farther away than it looks, does he conclude that he is not seeing a material thing (still less that he is seeing an immaterial ship)." (p. 9) But Austin's main point at this time is that there is no simple dichotomy between being deceived by one's senses and having veridical perception. There are in fact, he thinks, a great many different ways of being deceived by one's senses and these cannot all be lumped together.

Apart from the fact that Ayer presupposes this dichotomy, Austin is concerned with the concept of being deceived by one's senses for two main reasons. First, it figures in the argument from illusion which we will come to shortly, and, secondly, it is connected with the idea that statements about sense-data are incorrigible while statements about material objects are always faillble. Austin will later argue against such a view. Let us then see what else he has to say on the subject.

The first point is that being "deceived by one's senses" is a metaphor which is misleading when applied in this sort of case. It is misleading partly because one's senses do not in themselves tell us anything and, when we combine it with the sense-datum theory, it seems to suggest that there is an intermediate object in between us and the material object which informs us about something. The question would then be whether we can trust what it tells us (whether it is veridical or not). This is the only place in his book where Austin speaks of a sense-datum as an intermediate entity; here he is clearly thinking of the sort of theory held by Moore where there is a three-way distinction between the act of being conscious of the sense-datum, the sense-datum, and the material object. This is a view which Ayer has never held since he has always refused to distinguish the process of being conscious of the sense-datum from the sense-datum itself. From time to time Austin mentions Price and it is not really clear from much of what Price says whether he is like Moore or Ayer in this respect. It is a serious defect in Austin's treatment that he does not take note of such an important

Page 233: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

AN TI -SOLIPSISM 227

difference in the kinds of sense-datum theory that one can hold and apparently treats them all the same. In this case it is a mistake to accuse Ayer of having intermediate entities, but Austin's point holds good that our sensations do not really inform us of anything, and that the meta­phor is misleading in this connection. The question then is whether we know what Ayer has in mind despite his unhappy choice ofterminolo­gy. Austin thinks that "The cases, again, in which a plain man might say he was 'deceived by his senses' are not at all common. In particular, he would not say this when confronted with ordinary cases of perspec­tive, with ordinary mirror-images, or with dreams; in fact, when he dreams, looks down the long straight road, or at his face in the mirror, he is not, or at least is hardly ever, deceived at all." (p. 12) And, more important, he thinks that the cases in which the plain man is willing to say that he is deceived are very different from one another. In particu­lar Austin thinks that cases where a sense organ is deranged or ab­normal, or in some way or other not functioning properly, are very different from cases where the medium or in some way the conditions of perception are abnormal. Still a third sort of case would be where a wrong inference is made or a wrong construction is put on things, such as a sound which is heard. Thus the first case is probably the one where it would be most natural to say that one is deceived by one's senses. In the second case one might conceivably say that it was not one's senses but the abnormalities in the medium that deceived one, and in the third case one might say that one saw or heard perfectly accurately but that one made a false judgment. Hence Austin says that things can go wrong in a lot of different ways which do not have to be, and cannot be assumed to be, classifiable in any general fashion.

Austin tends to attack dichotomies in a way one would not expect. That is, the usual way to attack a dichotomy is to show that from one side to the other there is a gradual shading of cases and that between any two cases a third case can be invented. The conclusion would then be that the marking off of the range into two classes must depend on an arbitrary decision where to put the dividing line. This, for instance, is how Quine and others argue in attacking the dichotomy between ana­lytic and synthetic statements. But this is not the way in which Austin actually proceeds. In fact he tries to show that within one side or the other of the dichotomy there are great differences in the kinds of cases. This in itself does not break down the dichotomy because, however much difference there may be in the kinds of cases, there might still be a general feature which all these cases have in common and which no

Page 234: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

228 SOLIPSISM PROPER

cases on the other side have. When someone asserts a dichotomy he is not necessarily asserting that there is any great degree of similarity between the cases on one side or the other but just that there is some specifiable and systematic difference between the cases on one side and those on the other. Thus the conclusion which Austin announces is perfectly consistent with the view he is attacking. But if we look at his examples we can see that a genuine argument against the dichotomy might be constructed. He has given us cases of mistakes and in some cases we are more ready to say that we have been deceived by our senses than in other cases; hence it might be possible to construct a range of cases which varies continuously and which is such that we do not know where to draw the dividing line between being deceived by our senses and not being so deceived. Let us therefore assume that Austin has shown that there is no sharp line which is not artificially created between cases of being deceived by our senses and cases of veridical perception.

We can now look at the dispute between Austin and Ayer. In fact, it is a very queer kind of dispute and it is not even clear that Ayer and Austin are talking about the same issues. In the first place we saw that Ayer is just claiming that we are sometimes deceived by our senses in such a way that we are not perceiving material objects even when we think we are. He can still hold this despite everything Austin has said; Austin has not said that there are no cases of being deceived in this way, so no conclusion which Austin reaches on this issue is logically in­consistent with what Ayer wants to say (we are here excluding ques­tions concerning the corrigibility of statements about sensations and material objects on which there is a genuine dispute and which we will consider later).

Austin spends a good deal of time trying to show, probably correctly, that there are many fewer cases in which we are deceived than Ayer supposes. But this does not really touch Ayer's thesis; at most he might have to throw out a few examples and find more realistic ones, but he could do that easily enough. Austin obviously feels that what Ayer is doing is very wrong, but he seems to place the most emphasis on those points which are of least importance to Ayer's theory. Basically the trouble is this: Ayer is trying to construct a theory in which statements of our ordinary language are analyzed into statements about sensations; Austin, on the other hand, is interested in describing the use of our ordinary language as accurately as possible. Hence some issues are important to the latter which are not important to the former. For

Page 235: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI-SOLIPSISM 229

instance, it does not really matter from Ayer's point of view how often we are deceived by our senses as long as it is something that can occur; the question for him is whether there is such a logical possibility, and the frequency of occurrence has nothing to do with any logical ques­tion. But if someone is trying to describe the customs of our ordinary language and the rules that we habitually follow, he has to look to the frequency of occurrence of certain linguistic phenomena. This also ex­plains why Austin does not attack the dichotomy between being de­ceived by one's senses and veridical perception in the usual way. He seems less concerned with whether there is a genuine logical distinction here that would allow us to use expressions such as "being deceived by one's senses" as technical terms than in seeing whether the concept is useful for descriptive purposes. We have seen that from the logical point of view the dichotomy may be legitimate no matter how many different kinds of cases are subsumed under one side or the other, but in an empirical investigation the concept is useless, even though it may be meaningful, if it covers too many different kinds of cases. For in­stance, I can in a perfectly legitimate way introduce a concept which covers everything that exists except rabbits, but the concept is useless because it covers so many different kinds of entities about which we will want to say different things. Thus someone engaged in a basically empirical study of language may object to a concept just because it is too broad to be useful for his purposes even though the concept may have a use for a system builder like Ayer. I am not suggesting that Austin's study of language is empirical to the degree that the method used by Arne N aess and the Oslo philosophers is, but just that there are empirical elements in it. Further, these elements, such as appeals to the low frequency of certain usages, are not just incidental and they usually playa crucial role when they do occur.

Whatever one may conclude in the end about the dispute between Austin and Ayer, we have nowhere used the concept of being deceived by one's senses and we have based nothing on there being such a dicho­tomy. Our basic dichotomy rests on whether one believes that what one is perceiving is identical with one's present conscious state, and this ultimately gives rise to the dichotomy between sensations and material objects, which turns out to be single sensations vs. logical constructions out of sensations. These dichotomies would remain even if we were never deceived by our senses and even if that concept were meaningless.

The next dispute between Austin and Ayer concerns the notion of "direct perception." Again, this is one that we have not used, but it is

Page 236: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

230 SOLIPSISM PROPER

of considerable historical interest and can also be related to our system. Ayer, in common with many philosophers of the time, held that we never perceive material objects directly, but only indirectly, and that it is sense-data that we directly perceive. This doctrine is traceable back to Moore who said that there are two senses of "see," that in which we see sense-data and that in which we see material objects. One of these senses has subsequently been labeled "direct perception" and the other "indirect perception." Austin points out that the concept of direct per­ception involves another metaphor that has been extended, and he proposes to look at cases of indirect perception to see how the term is used. Thus someone who sees a procession through a periscope might be said to be seeing it indirectly, or perhaps someone who saw an object only in a mirror. He thinks that the term is most at home in the case of vision and that we are there said to see directly if we see in an un­broken line whereas we see indirectly if there is some kink in direction as with the periscope or mirror. Hence Austin says that seeing something with a telescope is not a case of indirect perception at all. When we move to the other senses it is more difficult, Austin thinks, to imagine what distinction would be involved. He asks whether we hear shouts indirectly when we hear echoes and whether we touch someone in­directly when we touch him with a barge-pole; these questions can ob­viously be resolved only arbitrarily. Further, Austin holds that we cannot say that something is perceived indirectly unless it is the kind of thing that could be perceived directly. For instance, we perceive the backs of our own heads only indirectly with mirrors, but others can see the same things directly. Ifwe cannot perceive an object directly then we should say that we do not perceive it at all, even indirectly. Austin concludes as follows: 1

Thus, it is quite plain that the philosopher's use of "directly perceive," whatever it may be, is not the ordinary, or any familiar, use; for in that use it is not only false but simply absurd to say that such objects as pens or cigarettes are never perceived directly. But we are given no explanation or definition of this new use - on the contrary, it is glibly trotted out as if we were all quite familiar with it already. It is clear, too, that the philosopher's use, whatever it may be, offends against several of the canons just mentioned above - no restrictions whatever seem to be envisaged to any special circumstances or to any of the senses in particular, and moreover it seems that what we are to be said to perceive indirectly is never - is not the kind of thing which ever could be - perceived directly.

Here there seems to be another lack of communication between Austin and Ayer, and this time Ayer does more to invite it. Austin has shown pretty conclusively that the ordinary senses of direct and indirect

I J. L. Austin, Sense and Sensibilia, p. 19.

Page 237: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI-SOLIPSISM 231

perception have little or nothing in common with Ayer's use of it and that Ayer must be using these concepts in some other way. However many philosophers have used the concept of direct perception and its function has usually been to mark out the basic units of explanation. Thus a phenomenalist says that we perceive sense-data or sensations directly and infer consciously or unconsciously the existence of material objects which are not perceived directly; the physicalist says that we perceive directly certain physical states of the world, and that any mental events which occur are merely concomitants of this perception. The general rule is that if a philosopher says some entities are directly perceived he holds that these entities are always present in cases of perception. Thus the physicalist will say that we are always perceiving some part of the physical world whenever we perceive at all, and that it is only our descriptions of what we perceive that lead to illusion and hallucination. Ayer, on the contrary, is saying that when we perceive there sometimes is no material object that we perceive, but that there is always the sense-datum. For him it is also true that the sort of object we say we perceive depends on the sort of sense-datum we are having, and Ayer also thinks that we are much less open to mistake when we describe our sense-data. Thus it is fairly easy to see what sort of thing Ayer means when he talks about direct perception.

In retrospect Austin seems to forget from time to time that he is deal­ing with a philosopher who constructs systems and who uses terms such as "direct perception" to mark off distinctions which arise not within ordinary language but within the system itself. Or it may just be that Austin was trying to show beyond any doubt that philosophers such as Ayer do not take their concepts from ordinary language but invent them themselves.

We come next to the reasons Ayer gives for introducing the concept of a sense-datum. These reasons are closely connected with the argu­ment from illusion, which dates back to Berkeley; Austin expounds Ayer's view of it as follows: 1

It is "based on the fact that material things may present different appearances to different observers, or to the same observer in different conditions, and that the character of these appearances is to some extent causally determined by the state of the conditions and the observer." As illustrations of this alleged fact Ayer proceeds to cite perspective ("a coin which looks circular from one point of view may look elliptical from another") ; refraction ("a stick which normally appears straight looks bent when it is seen in water") ; changes in colour-vision produced by drugs ("such

3 Ibid., pp. 20-22.

Page 238: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

232 SOLIPSISM PROPER

as mescal"); mirror-images; double vision; hallucination; apparent variations in tastes; variations in felt warmth ("according as the hand that is feeling it is itself hot or cold") ; variations in felt bulk ("a coin seems larger when it is placed on the tongue than when it is held in the palm of the hand"); and the oft-cited fact that "people who have had limbs amputated may still continue to feel pain in them."

He then selects three of these instances for detailed treatment. First, refraction -the stick which normally "appears straight" but "looks bent" when seen in water. He makes the "assumptions" (a) that the stick does not really change its shape when it is placed in water, and (b) that it cannot be both crooked and straight. He then concludes ("it follows") that "at least one of the visual appearances of the stick is delusive." Nevertheless, even when "what we see is not the real quality of a material thing, it is supposed that we are still seeing something" - and this something is to be called a "sense-datum." A sense-datum is to be "the object of which we are directly aware, in perception, if it is not part of any material thing." (The italics are mine [Austin's] throughout this and the next two paragraphs). [The quotes are all from Ayer's Foundations of Empirical Knowledge]

Next, mirages. A man who sees a mirage, he says, is "not perceiving any material thing; for the oasis which he thinks he is perceiving does not exist." But "his experience is not an experience of nothing"; thus "it is said that he is experiencing sense-data, which are similar in character to what he would be experiencing if he were seeing a real oasis, but are delusive in the sense that the material thing which they appear to present is not really there."

Lastly, reflection. When I look at myself in a mirror "my body appears to be some distance behind the glass"; but it cannot actually be in two places at once; thus, my perceptions in this case "cannot all be veridical." But I do see something; and if "there really is no such material thing as my body in the place where it appears to be, what is it that I am seeing?" Answer - a sense-datum. Ayer adds that the same conclusion may be reached by "taking any other of my examples."

Austin begins by pointing out that Ayer does not distinguish delu­sions from illusions carefully enough, but he does not think that Ayer's view can be refuted just by making that distinction more carefully. In the case of the bent stick Ayer argues, that since the stick looks bent but is not really, it is not what we are seeing (is not a real quality of the material thing). Austin points out that the stick partly immersed in water does not look the way a bent stick out of water would look. This is the way we expect a straight stick to look when it is partly immersed in water, and there is nothing surprising about the way it does look. Hence Austin concludes that there is nothing delusive about the sensa­tion and no reason to say that we are seeing anything other than a real straight stick partly immersed in water. In this case if Austin is correct, and I think he is, the argument does not really get off the ground because a premise of the argument from illusion is that the sensation from which it starts must really be delusive. Of course, this experience might be delusive to a child who does not know about the visual phenomena of immersing things in water, but Ayer could still have chosen a better example.

Page 239: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI -SOLIPSISM 233

Let us then look to the mirage case which is clearly, as Austin admits, a better example for Ayer to use. Here Austin says:l

The mirage case - at least if we take the view, as Ayer does, that the oasis the traveller thinks he can see "does not exist" - is significantly more amenable to the treatment it is given. For here we are supposing the man to be genuinely deluded, he is not "seeing a material thing." We don't actually have to say, however, even here that he is "experiencing sense-data"; for though, as Ayer says above, "it is convenient to give a name" to what he is experiencing, the fact is that it already has a name -mirage. Again, we should be wise not to accept too readily the statement that what he is experiencing is "similar in character to what he would be experiencing if he were seeing a real oasis." For is it at all likely, really, to be very similar? And, looking ahead, if we were to concede this point we should find the concession being used against us at a later stage - namely, at the stage where we shall be invited to agree that we see sense-data always, in normal cases too.

Again we notice a slipping of the gears between Ayer and Austin when Ayer wants to introduce the term "sense-datum" as a convenient name for this sort of phenomenon and Austin says that we already have a word for it. For one thing, Ayer obviously wants to have terms which will apply not just to visual phenomena; just as it is possible to see things that are not there it is also possible to hear things that are not there and it is even possible for a man to feel his faithful old hound dog rubbing against him after the dog is dead. The word "mirage" is limited to visual phenomena so it really is convenient, as Ayer says, to have a wor-d which can be applied to all similar phenomena whichever sense is involved. Further, Ayer is trying to construct a system and in so doing it does not matter what terminology he uses as long as he uses it consistently. In setting up a system of geometry one can discard the term "line" and use "set of points" instead; similarly, in a philosophical system one can discard a term like "mirage" and substitute some other term even if the other term did not have a wider application. Thus this objection of Austin's is another example of the general confusion of purposes. Still, on this and other occasions Ayer does not state his own purposes clearly enough; sometimes it looks as if he is trying to set up a system of the sort contained in the last chapter, and at other times he does talk in a more descriptive and less theoretical manner.

The paragraph quoted above contains another objection which is much more important and more interesting. This is the suggestion that when someone sees a mirage his conscious state may not be exactly what it would be if he were seeing a real oasis. In fact it is on this point that much of the argument against Ayer is really going to hinge. It is also at this point that Austin's views are actually inconsistent with our

1 Ibid., p. 32.

Page 240: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

234 SOLIPSISM PROPER

brand of solipsism. Up to this point Austin has attacked particular examples which Ayer has used and in many cases has been successful in this. However, he has granted Ayer a starting point in the case of mirages and the only reason he has given thus far for rejecting the sense-datum terminology Ayer wants to apply in these cases is that it would proliferate terminology.

Let us now re-state the main argument in such a way as to make the most important premises explicit. The first premise, one that is not stated explicitly, is that we are able to refer to the purely mental or conscious element in perception without at the same time referring to its behavioral correlates. The second premise is that there are some sense experiences which are delusive in that they lead one to suppose that some material object exists where, in fact, there is no such object. The third premise is that delusive perceptions do not always have a distinguishing mark which allows us to tell that they are delusive just by focusing our attention on some part of their content. It then follows that whenever we are perceiving anything we may be having a sensa­tion which will turn out to be delusive. Thus the conclusion is that even if statements about sensations can be certain we can never deductively infer an existential statement about a material object. Although the conclusion does not say anything about sense-data, the legitimacy of introducing them depends almost entirely on the truth of this statement as we will shortly see.

I do not think that Austin would question the validity of the argu­ment as so stated but he would continue to attack the premises. In particular, he denies the first premise and has grave doubts about the other two. We will therefore organize his remaining discussion around the various premises and then go on to discuss his remarks about the corrigibility of statements about sensations and what Ayer would call material objects. This procedure amounts to an elimination of the question, "Is it legitimate to introduce the concept of a sense-datum?" If Ayer is correct and we can talk about the purely conscious compo­nents of perception and we cannot go deductively from sensations to material objects, then there is nothing to prevent Ayer's saying that we have or perceive a sense-datum whenever we perceive at all whether or not we perceive a material object. If, on the other hand, someone were to hold that we can go deductively, or at least with certainty, from the occurrence of a perception to the existence of a material obj ect, then there would be no need for such a tenninology since we could always talk about perceiving material objects instead, and do it without

Page 241: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI-SOLIPSISM 235

having to make any qualifications or ever winding up talking about things which do not exist. If Ayer's premises and conclusion are correct there is some point to the sense-datum language since it can be easily used in such a way that all the terms refer to actually occurring events, and this is how our solipsist uses a similar language. Even on this as­sumption we might be able to construct a language which has similar qualities and which refers only to material objects but which makes use of expressions such as "appears" and "seems to be." However there is no reason to think that such a language would be any simpler and its existence would do nothing to substantiate Austin's claim that the sense-datum language is illegitimate. The most important part of Austin's case against sense-data consists of his denial of the first pre­mise. Whatever else a sense-datum may be it is supposed to be the con­tent of a conscious state. If we cannot refer to the conscious components of perception without at the same time referring to other things, then it is not legitimate to introduce the sense-datum terminology. Thus, as we pointed out earlier, the dichotomy between sensations and material objects tends to reduce to that between the mental and the phsyical -a very old dualism.

We have already begun to talk about the second premise - that we sometimes have delusions. In the end Austin admits three cases of delu­sion: mirages, double vision where we see two objects when there is actually only one, and the seeing of pink rats in the course of D. T. s. In the case of double vision he is willing to say that one does perceive two pieces of paper but not that two pieces of paper really are perceived. He thinks the addition of the word "really" would imply the actual existence of two pieces of paper. Thus Austin admits some cases of delusions in some sense but he also has a tendency to deny that we actually perceive delusions. Part of the trouble here seems to be an ambiguity in the sentence, "Two pieces of paper really are perceived." Ayer thinks that it is all right to say this in a case of double vision be­cause he thinks that there is one sense of perception in which one can refer to the purely conscious aspects of perception without talking about any associated behavior. Then the word "really" has the effect of say­ing that the conscious event really does take place, and, of course, if one is seeing double then it does. But Austin seems to take the concept of perception in general, as well as the particular verbs of perception, to refer not just to the conscious aspects of perception but rather to a complex relationship which involves the person doing the perceiving, his conscious states, and the object which is perceived. Thus, if the

Page 242: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

236 SOLIPSISM PROPER

relationship is really to hold between all these components, they must all exist, and the statement would imply the existence of two pieces of paper. In fact Austin holds that his sense of "perceive" is the only legiti­mate use of the term. He argues against one of Ayer's uses of "perceive" as follows: 1

For one thing, if there really were two senses of "perceive," one would naturally expect that "perceive" might occur in either of these senses in any of its construc­tions.

This argument is not conclusive since even openly ambiguous words may not be ambiguous everywhere they appear. Quite often the imme­diate verbal context effectively eliminates one meaning or the other so that the word is not ambiguous in that context, but it may still be an ambiguous word. Nevertheless, Austin finds these cases of delusions which he has admitted embarrassing because as long as we can say properly even "I am perceiving two pieces of paper" in the full know­ledge that there is only one piece in front of one, it looks as if we could introduce the sense-datum or some other concept to refer to what we think we are perceiving. He goes on to say:2

... double vision is a quite exceptional case, so that we may have to stretch our ordinary usage to accommodate it. Since, in this exceptional situation, though there is only one piece of paper I seem to see two, I may want to say, "I am perceiving two pieces of paper" faute de mieux, knowing quite well that the situation isn't really that in which these words are perfectly appropriate. But the fact that an exceptional situ­ation may thus induce me to use words primarily appropriate for a different, normal situation is nothing like enough to establish that there are, in general, two different, normal ("correct and familiar") senses of the words I use, or of anyone of them. To produce a rather baffling abnormality like double vision could establish only, at most, that ordinary usage sometimes has to be stretched to accommodate ex­ceptional situations. It is not, as Ayer says, that "there is no problem so long as one keeps the two usages distinct"; there is no reason to say that there are two usages; there is "no problem" so long as one is aware of the special circumstances.

One should note first that this is a rather unusual argument for Austin; he is in effect saying that we do not really mean what we say in these circumstances and that it is possible to substitute for the ex­pression we use a different one which better represents our meaning. Thus this is a case of revisionary rather than descriptive analysis and it is what Austin objects to if done on a large scale. His argument here is that such cases are so exceptional that they do not really matter and it is reminiscent of Hume's treatment of the "missing shade of blue."

1 Ibid., p. 90. t Ibid., pp. 90-91.

Page 243: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI-SOLIPSISM 237

A little further on concerning the perception of pink rats he says:l One of the examples - the one about double vision - does suggest, what in any

case is only to be expected, that in exceptional situations ordinary forms of words may be used without being meant in quite the ordinary way; our saying of the D. T.'s sufferer that he "sees pink rats" is a further instance of this, since we don't mean here (as would be meant in a normal situation) that there are real, live pink rats which he sees; but such stretchings of ordinary words in exceptional situations certainly do not constitute special senses, still less "correct and familiar" senses, of the words in question.

We can now also see why Austin is so worried about the frequency of cases of delusion. Unless a usage occurs with a certain requisite fre­quency he is not going to count it as a legitimate usage at all but is going to class it with such cases as that where one accidentally utters the wrong word; the usage in either case gets replaced by some other one which is considered more appropriate. At this point a philosopher such as :'\aess would ask what this requisite frequency is going to be and whether empirical evidence is not relevant in order to see whether this frequency is attained. The Austinian's usual reply to such ques­tions is that what is important is not what most people do say, as deter­mined by questionnaire, but what we could say as determined by our­selves as users of the language. However, it seems here that we could say that we see two pieces of paper or that we see pink rats and that these usages are being discounted on the basis of some other sort of criteria. This case is one that the Oslo philosophers might use to their advantage.

Quite apart from this sort of argument, there is a similar construction which occurs very frequently. 1 can quite easily say that 1 feel a sharp pain in much the way that 1 might say, "I feel something sharp under this sofa cushion." No one questions my right to say that 1 feel a sharp pain in my finger when 1 see that nothing is sticking into it, and the only difference between the sharp pain and the pink rats, apart from the content of the sensations, is a matter of relative frequency of usage. There is nothing delusive about the sharp pain but it is a case of talking in a substantive way about the object of a sensation where no material object is in question. This fails to be delusive just because it happens too often, and this is why we do not put any emphasis on delusions in our account. An intermediate case would be "I have been hearing voices when I am alone again, so 1 had better go back to my psychia­trist"; again, we have the same sort of thing, but this sort of "vision" occurs more frequently than the seeing of pink rats and less frequently

1 Ibid., p. 97.

Page 244: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

238 SOLIPSISM PROPER

than the feelings of sharp pains. Since we can think of such phenomena which have any desired frequency of occurrence, Austin ought to either discount all such usages or none and, since there are too many to dis­count them all, he should discount none, but treat them as he would any other standard usage. It seems in particular that Austin would have done better to have considered some cases other than the visual ones. In his treatment of the premise that there are some cases of delu­sions Austin has ended by not only admitting some cases but also ad­mitting that even in our ordinary language we can speak about the objects of the delusions in a substantive way. Even if one thinks that he is cor­rect in de-emphasizing the importance of this sort of usage the crucial point, that there are some cases of delusion, has been admitted.

Let us now see what Austin has to say about the third premise, that delusions have no distinguishing marks as such. We have already noticed that Austin wonders whether a mirage is really similar in r.ha­racter to what one would see if one saw a real oasis. We said in the first chapter that any particular dependent sensation (and a mirage is usually dependent) may have some peculiar features which distinguish it from other sensations, but that dependent sensations do not have any feature of content which distinguishes them as a class from independent sensations. An argument very similar to the one we gave there could be given in the case of delusions. If delusions in themselves, as opposed to the context in which they appear, have a peculiar feature which marks them off, it must be reflected in some way in the qualities that these sensations have. In the case of visual sensations the delusive qual­ity would have to involve some sort of configuration of colors and shapes which non-delusive sensations would not have. If the sensation were auditory there would have to be some distinguishing feature of pitch, intensity, or timbre. But, of course, if we went to sufficient trouble we could always produce a noise, perhaps by using some sort of organ, which would produce the same sensation but which would not be delu­sive. Again, we could always construct a situation which would produce any required visual sensation in a non-delusive way even though the construction might be laborious and expensive. Thus once one is told what the delusive quality is like, one can with sufficient trouble arrange for someone to have a non-delusive sensation which has that quality. We can ensure that the quality will be non-delusive by showing the subject how the object is constructed and why it looks the way it does from various positions. Austin consistently overlooks this possibility and while he does not argue strictly that such phenomena as delusions,

Page 245: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI-SOLIPSISM 239

dreams, etc. always have a peculiar feature, he comes very close to it.l For instance, he says:2

... descriptions of dreams, for example, plainly can't be taken to have exactly the same force and implications as the same words would have, ifused in the description of ordinary waking experiences. In fact, it is just because we all know that dreams are througlwut unlike waking experiences that we can safely use ordinary expressions in the narration of them; the peculiarity of the dream-context is sufficiently well known for nobody to be misled by the fact that we speak in ordinary terms.

It is true that when we describe our dreams we use ordinary terms and sentences; it is also true that our sentences do not imply the same consequences that they would in another context. But Austin is not saying just that; he is saying that dreams are throughout unlike waking experiences to such an extent that we can let our ordinary expressions have a double meaning and never be confused about which way they are being used. On the contrary, it seems to me that the sentence, "I saw Willie Mays hit a home run" as applied to a dream has the same meaning that it would have when applied to a real situation. In fact many of the implications are the same. If the dream is an orderly one the ball has to leave the playing field between the foul posts and fall into the stands and it will be followed by Mays' circling the bases. On the other hand some of the implications are different. In particular some of the long range implications of the sentence as used of the dream will differ from the long range implications of the sentences as applied to waking life. It is unlikely that one's dream will include read­ing the record book at the end of the season and finding that home run included in the statistics, although even this is conceivable. The differ­ence is, as many philosophers have seen, that dreams tend not to be nearly so coherent and extended as waking life. Thus there will be some difference in implications between the two uses of the sentence, but it is not possible to say, unless we know how coherent and extended the dream in question is, just what this difference in implications will be. When a sentence like this is used without any qualification, or out­side a context which qualifies it, we assume that it is intended to apply to waking life and not to dreams. If it is intended to describe dream experience it has to be placed in a special context such as being pre­ceded by, "Let me tell you about the dream I had last night." Now the implications of the statement used either way are not entirely derived from the statement itself. If someone concludes, "That will give Mays

1 It will be argued somewhat later that Austin goes in this direction. 2 Austin, op. cit., p. 42.

Page 246: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

240 SOLIPSISM PROPER

17 home runs during the month of August," the suppressed premise in that inference is that I am not talking about something I saw in a dream. Other suppressed premises would be that I am sincere, that I can tell Mays from McCovey, and so on. Similarly, if he knows that I am talking about a dream and says, "I think that your dreams are prophetic; I will expect Mays to hit one today," then he is making his inference not from my statement alone, but is also taking account of other statements which classify the context as a dream context. The fact that the sentence has different implications in the two different contexts does not show that it has two different meanings or that dreams are "thoroughly unlike" waking experiences. It just shows that when we assess the implications of a statement either about dreams or waking experiences we take into account not only the statement itself, or the experience itself, but also the context in which it appears. This is what allows us to use ordinary expressions in describing dreams without getting confused. Later Austin goes even farther in this direction: 1

I may have the experience (dubbed "delusive" presumably) of dreaming that I am being presented to the Pope. Could it be seriously suggested that having this dream is "qualitatively indistinguishable" from actually being presented to the Pope? Quite obviously not. Mter all, we have the phrase "a dream-like quality"; some waking experiences are said to have this dream-like quality, and some artists and writers occasionally try to impart it, usually with scant success, to their works. But, of course, if the fact here alleged were a fact, the phrase would be perfectly meaning­less, because applicable to everything. If dreams were not "qualitatively" different from waking experiences, then every waking experience would be like a dream; the dream-like quality would be, not difficult to capture, but impossible to avoid. It is true, to repeat, that dreams are narrated in the same terms as waking experiences: these terms, after all, are the best terms we have; but it would be wildly wrong to conclude from this that what is narrated in the two cases is exactly alike.

On our view that there is no systematic difference between the content of dreams and waking experience the dream-like quality that some ex­periences have is easily explained. It might be the incoherent quality (over short runs of time) which many dreams have and which some waking experiences have, or a sort of blurred look, or some other quality. On the other hand there are some dreams which are, in retro­spect, very vivid and coherent and which are so hard to distinguish from reality that when we are waking up we have trouble deciding whether we are dreaming or not. Thus the dream-like quality, however we describe it, does not systematically distinguish dream experience from waking experience. It is just something which is characteristic of rela­tively few waking experiences and relatively many dreams. In fact,

1 Ibid., pp. 48--49.

Page 247: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI-SOLIPSISM 241

even if all dreams had the dream-like quality and some waking expe­riences have it, as Austin suggests, we would still not get a systematic distinction between dreams and waking experiences but between dreams and a part of waking experience on the one hand and the rest of waking experience on the other. Again, if the dream-like quality is something that every dream experience and no waking experience is supposed to have, we can ask what sort of quality this is, and once we are told what it is like, we can arrange to have a waking experience which will duplicate it. We presumably do not want to say that it is a mysterious indescribable quality which accompanies all dreams but which, for some unknown reason, is never present in waking experience.

Austin seems to think that it is obvious that dreams are in themselves different and consequently gives no argument. It could be that some people's dreams are so fleeting and fragmentary that they always in some sense "know" when they are dreaming, but I think that the ex­planation may be more as follows. Some persons feel that the difference between dreams and waking experience is not just a difference in the context in which the experience occurs, but that there is a phenomen­ological difference as well. I think that there are some grounds for this. Suppose that towards the end of a dream one wins a huge sum of money in a lottery. One may reach a semi-waking stage and still be thinking of the money and all the nice things one is going to do with it; in fact one may even think of the debt one contracted in yesterday's waking experience and decide to use some of the money to payoff that debt. The next moment one is seized with panic and thinks "Did I just dream that I won the money?". Perhaps one thinks over the details of winning the money and sees that one won it in a perfectly real way (as opposed, say, to having a penguin give one a wad of bills) and every­thing seems to be all right. But then all of a sudden one wakes up, the shock of reality hits one, and one suddenly realizes that one is not rich after all. There is a phenomenological factor here, the jolt of waking up. The experiences which precede it may be just as vivid and no more dream-like than the experiences which follow the shock. In fact the shock is likely to be both a mental and a physical one and be accom­panied by such physical operations as suddenly sitting bolt upright in bed. We now have a rule: anything which comes after the shock is waking experience and anything that comes before it is dream expe­rience. We then have no difficulty in distinguishing dream experience from waking experience no matter how similar the two kinds of expe­rience may be phenomenologically. Hence one can say with Austin

Page 248: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

242 SOLIPSISM PROPER

that one could not possibly confuse the two, and there would be a phenomenological element here: the experience of waking up. How­ever, this does not justify Austin's conclusion that there is a systematic difference of quality between dreams and waking experiences. It is still a matter of the context in which the experience occurs that determines which it is: "Does it come before waking up or after waking up?". Thus we do have something to go on in deciding whether something was a dream besides trying to piece it together with other experiences after the dream is over. In that case we have a simpler contextual criterion to use, but it is still a contextual criterion.

There is another argument to the effect that if there were a systematic difference between delusions of all kinds and non-delusive experience we would then never be deluded. We have not used this argument, but I think that it is a good one. Austin counters it as follows: 1

From the fact that I am sometimes "deluded," mistaken, taken in through failing to distinguish A from B, it does not follow at all that A and B must be indistinguishable. Perhaps I should have noticed the difference if I had been more careful or attentive; perhaps I am just bad at distinguishing things of this sort (e.g. vintages); perhaps, again, I have never learned to discriminate between them, or haven't had much practice at it.

A little later he compares this with someone's saying to a professional tea taster that "there cannot be any difference between these two brands of tea for I regularly fail to distinguish them." Something very peculiar has happened here. When Austin talks about delusions here he is talk­ing about the kind of delusions that Ayer has in mind and is thus refer­ring, for the sake of argument, to the purely mental components even though there may be difficulties in doing this. Austin then suggests that even though our delusions may have a peculiar quality which marks them off, we might fail to notice that quality in the way that we fail to notice the difference between two brands of tea. But in the one case we are talking about sensations and in the other case about material objects; while Austin questions that dichotomy, it has to be granted here ifwe are to talk about the issue at all. Then, of course, one cannot fail to notice a feature of one's own conscious state in the way that one can fail to notice a feature of a physical object. The conscious state is the sum total of what one does notice, and there consequently cannot be any unnoticed corners in it. The trouble is that if delusions are system­atically different from waking experiences, to be deluded we then some­how have to fail to be aware of this peculiar quality, which is impossible, or not know how to interpret it. Since we cannot just not notice it, as

1 Ibid., p. 51.

Page 249: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI -SOLIPSISM 243

Austin suggests, the only alternatives would be that we notice it but forget it ever so quickly before attaching any importance to it, or that, while all delusive experiences have a particular feature of which we are aware, we fail to realize that that feature is common and peculiar to delusions. However, neither of these possibilities seems to be sufficiently likely to be worth taking seriously.

On the next page Austin back-tracks a good deal and makes qualifi­cations which have not appeared previously. Here he says:1

I do not, of course, wish to deny that there may be cases in which "delusive and veridical experiences" really are "qualitatively indistinguishable"; but I certainly do wish to deny (a) that such cases are anything like as common as both Ayer and Price seem to suppose, and (b) that there have to be such cases to accommodate the undoubted fact that we are sometimes "deceived by our senses."

His conclusion under "(b)" reflects the previous argument and while it is not logically necessary that there be cases in which delusions and veridical experiences are qualitatively indistinguishable in order to ex­plain our sometimes being deceived by our senses, the denial of the existence of such cases would involve us in some extremely strange em­pirical hypotheses. In the part under" (a)" Austin again appeals to the infrequent occurrence of the phenomenon, but, we have seen before, the important point in the construction of a philosophical system is not the number of cases, but whether there can be any at all. The culmina­tion of this part of Austin's argument is given as follows: 2

But if we are prepared to admit that there may be, even that there are, some cases in which "delusive and veridical perceptions" really are indistinguishable, does this admission require us to drag in, or even to let in, sense-data? No. For even if we were to make the prior admission (which we have so far found no reason to make) that in the "abnormal" cases we perceive sense-data, we should not be obliged to extend this admission to the "normal" cases too. For why on earth should it not be the case that, in some few instances, perceiving one sort of thing is exactly like perceiving another?

The question is, "If we sometimes perceive sense-data (in delusions) and there is no systematic distinction between delusive and normal perception why do we not perceive them all the time?". Austin replies in effect that delusions do not occur very often and perhaps we some­times perceive one kind of thing, sense-data, and at other times per­ceive another kind of thing, material objects, and that the one kind of process of perception is just like the other. But one is then tempted to ask "If the two conscious processes are exactly alike, why not group them together and apply the same name to both?" Of course, later on

1 Ibid., p. 52. • Ibid., p. 52.

Page 250: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

244 SOLIPSISM PROPER

we can find out which perceptions are delusive and which are not and apply different labels to them, but since we cannot distinguish the two conscious processes in themselves, it does seem that we could class them all together, at least for some purposes, and apply a common predicate to all members of the class. Ayer chooses to say that members of this class are cases of "perceiving sense-data." Austin never does give us an answer on this point. In fact, his whole argument heretofore comes down ultimately to assertions about the infrequent occurrence of certain phenomena. He says that there are not many delusions in the first place and, while there is a usage in ordinary language for describing those cases and in terms of which the expression "perceiving sense-data" can be introduced, this usage can be eliminated. There he concluded not only that there is no need of a special terminology based on these cases, but that there can be no legitimate basis for it. Then in the next stage he argues that there are not very many delusive experiences which are indistinguishable from veridical experiences and that there is no need and no legitimate basis for extending the terminology. Of course, in order to establish this latter point he would have had to show that there is some systematic difference between delusions and veridical experien­ces which positively prevents Ayer from extending the terminology. Thus the second and third premises of our re-statement of Ayer's argu­ment have not even been directly attacked and it will be the first premise that is crucial. Similarly, in our solipsistic system we need the first premise in order to introduce our sensation terminology. This is the only part of the argument that we really need and so the dispute over this premise is of particular interest.

Even before Ayer introduces sense-data he talks about the class of "perceptions" and asks whether the veridical are intrinsically different from the delusive. At this point Austin objects:l

... how are we supposed to set about even considering this question, when we are not told what "a perception" is? In particular, how many of the circumstances of a situation, as these would ordinarily be stated, are supposed to be included in "the perception" ?

This is a very basic question; in fact, the most basic question that Austin ever asks. Ultimately, whether we can legitimately talk about sense-data or sensations or anything similar will depend on the answer that we give to it. While Austin does not follow the consequences of this question explicitly, what is at issue is whether we can take a case of perception (which is a complex event in the real world involving not

1 Ibid., p. 53.

Page 251: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI-SOLIPSISM 245

only conscious events but physical events in the brain, events in the ner­vous system, and events external to the body), merely isolate the con­scious events and talk exclusively about them. Can we then describe the content of these conscious events, compare one with another, and form classes of similar events? If we can do this, there is obviously no harm in introducing special senses of "perceive," "see," "hear," etc. in which what is perceived, seen, heard, etc. is simply the conscious state itself. A further question is whether such uses of the perceptual words are already to be found in our ordinary speech. For purposes of setting up a solipsistic system it does not matter very much whether ordinary language contains these uses as long as it would be possible to legiti­mately introduce the terms. Ayer makes use of such a sense of "see," and says that there is a sense of the word in which one can say that an object is perceived but does not exist in any sense at all. This formula­tion may be rather misleading because, of course, even in this sense of "perceive" the object which is perceived is a genuine conscious event which really does take place and one might say that it does exist in some sense; hence Ayer might have done better to say that in such a case the object need not exist in any sense at all in the physical world. I think that this is how Austin interprets him and Austin comments:1

Now perhaps this would be "another sense" of "see," if there were any such sense; but in fact there is no such "sense" as this. If a man says "I see a silvery speck," of course he "implies" that the speck exists, that there is a speck; and if there is no speck in the region of the night sky at which he is looking, if that part of the sky is perfectly blank, then of course he does not see a silvery speck there. It is no use his saying, "Well, that region of the sky may be perfectly blank, but it is still true that I see a silvery speck; for I am using "see" in such a sense that what is seen need not exist."

Further on Austin says:2 The fact is that, just as there is no sense of "see" which is such that what is seen

need "not exist in any sense at all," there is no sense of "see," neither the same sense nor any other, in which it is impossible that what is seen "should seem to have qualities that it does not really have." I am not denying, of course, that we could arbitrarily invent such uses of "see," though I don't know why we should want to; but it must be remembered that Ayer is purporting here to describe "senses" of "see" which are already "correct," and even "familiar."

In trying to establish his special sense of "see" Ayer again uses a poor example, of astronomical bodies looking no bigger than a six­pence, and Austin effectively disposes of it. However, Ayer already has the seeing double and pink rat examples which Austin can cope with only by saying that they do not occur very often. Further, a correspond-

1 Ibid., p. 94. 2 Ibid., pp. 96-97.

Page 252: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

246 SOLIPSISM PROPER

ing sense of "feel" is, as we have already argued, very common since we constantly talk about feeling pains in such a way that there need be no physical object to which the pain is traceable and in such a way that the pain cannot seem to have any qualities that it does not have. Since feeling is at least as important as seeing, it seems that both Austin and Ayer might have paid more attention to it. However, the most interesting remark Austin makes here is that we could arbitrarily invent senses of "see" although he doesn't see why we should want to. The reason why we might want to is to use such terms as primitives in a system of solipsism. It might seem for a moment that there is no real disagreement between the solipsist and Austin or even between Ayer and Austin. If this remark of Austin's is to be taken literally it would imply that he is just arguing that no such sense of "perceive" and "see" is to be found in our ordinary usage, but that we can invent such senses and then base sense-datum or sensation terminology on them. Thus as long as we phrase the first premise not as "We do talk about the mental components of perception exclusively" but as "We could talk about the mental components of perception exclusively" it would seem that Ayer could have his whole argument since Austin has not categorically denied the second and third premises. The solipsist, who is mainly interested in the first premise, could have everything he wants with no further questions asked. However, it seems incredible that Austin would say that there is nothing wrong with the sort of solipsistic system being constructed here, and we must interpret him in such a way as to allow him to deny the validity of such a theory. There are ever so many places in his book some of which we have quoted, that suggest that he does not really mean to allow the invention of these uses and that he has just managed to manoeuver himself into a corner here. It would seem much more like Austin to deny the dichotomy between the mental and the physical and to say that we cannot set up any use oflanguage which refers to mental events without its having physical implications at the same time; we could also expect him to treat these implications as part of the meaning of the expressions themselves. Further, he says in many places that the way a thing looks is a fact about the physical world and not a fact about a person's conscious experience. Moreover, his tone suggests that he thinks that this is the only way we can talk about the way things look, and that we could not just choose to talk in some other way.

One possible way of interpreting Austin would be that we can invent new usages and set up artificial languages as long as we please but that

Page 253: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI-SOLIPSISM 247

we can never derive from such a language an analysis of any part of our ordinary language. Thus, for instance, he might not object to the set­ting up of a solipsistic language as long as one does not claim that one can provide in that language an analysis of statements in our ordinary language. But of course this is just what we do claim so there would really be a conflict, as it seems there ought to be between solipsism and Austin. Thus, having invented or extrapolated this special sense of "see" in which seeing means having visual sensations, the solipsist might say that seeing a certain sort of physical object amounts to having certain sorts of visual sensations under certain sorts of circumstances (which will be specifiable in sensory terms). It is at this point, then, that we would have Austin objecting. On this interpretation he is not only describing our ordinary usage but also asserting that our ordinary sen­tences are irreducible to other kinds of sentences, and it would be at this point that an implicit claim about meaning in general would be invoived. Austin would here be claiming not that two particular sen­tences have different meanings but that no sentence we could produce in the solipsistic language would be synonymous with a sentence such as, "I see a chair over there." Further, it seems that his reason for saying this would have to be that the solipsistic language has been so constructed that all the terms in it refer to purely mental events where­as all the statements in our ordinary language refer to both mental and physical events. Of course, the solipsist would reply that he talks about physical events within his system and that he constructs them logically out of mental events. Austin would not be likely to accept physical events of this sort in lieu of what he would consider genuine physical events, but he does not discuss phenomenalism in detail and offer ob­jections to this kind of analysis of material objects. However, in Austin's discussion of Warnock's views and in the last sentence of the book, "The right policy is to go back to a much earlier stage, and to dis­mantle the whole doctrine before it gets off the ground,"! we get the impression that his approach to anything like phenomenalism or solips­ism is to disallow the setting up of the language in the first place. Thus, this interpretation of Austin is probably too weak and when he says that Ayer's special sense of "see" does not exist he probably also means to assert that there could not be any such sense of "see." Ifhe really meant to say that we could legitimately invent artificial new senses of "see" we would find him giving detailed arguments against phenomenalism, the use of logical constructions, and other such paraphernalia.

2 Ibid., p. 142.

Page 254: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

248 SOLIPSISM PROPER

Why, then, would someone say that it is impossible to talk about the conscious aspects of perception exclusively? The answer has to be found by looking to the theory of meaning that is presupposed. We cannot find out how we could talk meaningfully by looking at the way we do talk exclusively, and we have to have a theory which tells us what ele­ments of our actual modes of speech are essential to intelligible con­versation as such. The theory of meaning we stated was nominalist in the sense that we have to ask what a sentence means for an individual at a particular time as determined by the sorts of instances he would accept at that time. The meaning that a term or sentence has for a whole society is just a generalization over individual meanings and the mean­ing is a consensus. But on a realist theory of meaning the meaning of a term or sentence is not determined, at least in the same way, by what the individuals in the society mean by it; the meaning consists in the society-wide rules and conventions for the use of the expression. On this sort of theory of meaning one can find out what a sentence means by looking at an individual's use of it, but only insofar as one supposes that his use reflects the institutional use. But the rules for the use of the expression must refer to events or objects which are publicly observable by anyone within the society; we cannot have public rules if they refer to private mental events to which others do not have access. Thus, as we would expect, someone who defends the possibility of a private lan­guage, as we have, will be on the nominalist side while someone who denies this possibility will often be on the realist side. Our particular theory of meaning makes images and the mental components of sensa­tions basic to the meaning of any sentence about anything. Since it is a nominalist theory it implies the possibility of private languages and the solipsistic system starts from a private language and constructs every­thing out of it. On a realist theory, we can talk about perceptions only by identifying individual sensations or perceptions in such a way that everyone can talk about the same ones at once; hence one cannot choose to talk about the mental components of perception exclusively but must view the way something looks as a fact about the physical world, and this is what Austin does.

If one holds a realist theory of meaning one does not say just that this is the way we do talk but that this is the only way we can talk about perception and still make sense. In the end one cannot say whether Ayer's basic thesis, apart from his superficial mistakes, is correct or not unless one has a theory of meaning. The same thing is true of our solips­istic system, and this is why we have gone to such trouble to state a

Page 255: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI-SOLIPSISM 249

theory of meaning which supports that system. In fact, there are only four kinds of criticism that can be leveled against such a system: 1) One might find an inconsistency within it; 2) One might find a circularity of analysis; 3) One might construct a different system, based on the same sort of theory of meaning, which would yield more plausible ana­lyses of the statements of ordinary language; 4) One might find an alter­native and more plausible theory of meaning which would imply the impossibility of the solipsistic language. Austin has not succeeded in doing any of these things, and it is not really clear that he is even trying to. His general approach certainly suggests the fourth kind of criticism, but nowhere in his writings is the required sort of theory to be found. The basic trouble is that he mistakes for "facts" about meaning asser­tions which can actually only be derived from a general theory of mean­ing. It is our view that in philosophy there are no facts of this sort to start from; one canjust construct alternative systems and then compare them. Having done this one might be able to make a few judgments to the effect that certain theories have certain advantages over others.

In the latter part of his book Austin talks not so much about percep­tion itself but about the supposed corrigibility and incorrigibility of certain kinds of statements. His thesis here is interesting and deserves some attention. Concerning the incorrigibility of sensations Austin says :1

... it is worth emphasizing, in view of what many philosophers have said, that descriptions oflooks are neither "incorrigible" nor "subjective." Of course, with very familiar words such as "red," it is no doubt pretty unlikely that we should make mis­takes (though what about marginal cases?). But certainly someone might say, "It looks heliotrope," and then have doubts either as to whether "heliotrope" is right for the colour this thing looks, or (taking another look) as to whether this thing really looks heliotrope. There is certainly nothing in principle final, conclusive, irrefutable about anyone's statement that so-and-so looks such-and-such. And even if I say, " .. .looks ... to me now," I may, on being pressed, or after looking at the thing more at­tentively, wish to retract my statement or at least amend it. To rule out other people and other times is not to rule out uncertainty altogether, or every possibility of being challenged and perhaps proved wrong.

The assertion here is that even in the absence of other mistakes one might apply the wrong predicate to one's sensation and Austin says later that one might to this, not through mis-speaking oneself, but through a genuine mistake as to what the predicate means. It is clear that one can only take this view if one holds a realist theory of meaning. On our theory of meaning, the meaning of a predicate is determined by the sorts of instances, both imagined and sensed, to which one would

8 Ibid., pp. 42-43.

Page 256: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

250 SOLIPSISM PROPER

apply the predicate at a given time. Consequently, if one does apply "heliotrope" to a conscious state then it is heliotrope, and if one refuses to apply the predicate, then it is not heliotrope no matter what the term may mean to others at any time or to oneself at some other time. It is only when one says that the meaning of the term is determined not by one's present dispositions but by the way the term is used in the society at large that it is possible to apply it to the wrong sensation by using the term in a deviant way. Austin's view here is not necessarily wrong but it again rests on an unstated theory of meaning and cannot be taken just as a "fact" about language. Even if one's way of deter­mining the meaning of words and sentences involves looking at actual usages very closely, perhaps even in a scientific and systematic way, one still has to have a theory of meaning before one can go from asser­tions about the way expressions are customarily used to assertions about their meaning. One may say, if one likes, that the meaning of an expression consists in the rules for its use, but we have already seen that this is just a slogan which we can even apply to our own image theory of meaning; it is much too vague and ambiguous to constitute a theory in itself.

Austin is not saying that any statement which Ayer would classify as a statement about a sense-datum is corrigible, but he holds that statements of this class (or any such class) are not incorrigible as such. He also holds that statements about what Ayer would call material objects are not necessarily corrigible and that there are some circum­stances in which such statements are incorrigible. Thus he says:1

.. .if I watch for some time an animal a few feet in front of me, in a good light, if I prod it perhaps, sniff, and take note of the noises it makes, I may say, "That's a pig"; and this too will be "incorrigible," nothing could be produced that would show that I had made a mistake.

This is an interesting proposal with which we will agree in the end. However, the first question concerns Austin's use of "could" when he says that nothing could be produced that would show that I had made a mistake. If it is just that nothing can in fact be produced to show that I had made a mistake this is not decisive. When we say that a statement is incorrigible we are using a technical term and we mean not just that it is a statement which we somehow know we will never be forced to take back but that there is no conceivable evidence that would refute it. The traditional philosopher would here bring up as such evidence the possibility that I will later find out that I was in a state of hypnosis and

1 Ibid., p. 114.

Page 257: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

ANTI-SOLIPSISM 251

was convinced that a cow was a pig, or the possibility that the pig was a mechanical imitation, or some such. Thus to say that the statement is incorrigible we would have to say that such evidence would not be relevant. As it turns out, our analysis will have Austin's result since we are adopting a finite form of phenomenalism; this means that a finite number of sensory predictions will be equated with a material object statement. If these were all to be verified then the material object state­ment would be conclusively verified and no more evidence of any sort would be relevant. However, we will not go as far as Austin because, even though our analysis is finite, material object statcments will not be such that they can be verified right on the spot by a few simple obser­vations. They will always extend over a certain period of time and allow for the more obvious kinds of error. Nevertheless, we accept Austin's main thesis that when we make statements such as this we can under certain circumstances say that the question of truth and falsity has been closed. However, part of the reason that we are able to arrive at this conclusion is that in our phenomenalism we substitute for a single rather vague sentence in English a whole range of more precise state­ments in the solipsistic language. These statements are more precise only in the sense that they indicate how they could be verified (what predictions they make) in an explicit way while a statement such as "That's a pig" does not tell us explicitly how we should verify it. Thus on this sort of analysis we can distinguish between the statements about sensations which are actually entailed by the material object statement and those other statements about sensations which are related only because they represent inductive evidence for or against the statements which are entailed. If all the statements which are directly entailed are conclusively verified by observation (having the sensations in question), then the statements which count as evidence for or against those state­ments are no longer relevant. The crux of the matter is being able to mark off the range of statements which are deductively related to the original material object statement from those which are only inductive­ly related. Since Austin does not make such a distinction it is hard to see how he could justify his view even though it seems to be correct. He might appeal to a paradigm case argument and it might also be possible to work out an adequate theory of meaning based on the con­cept of a paradigm, but that would be another story.

Our general conclusion about Austin's criticisms of Ayer, at least as regards their relevance to our purposes, would then be that in ever so many places Austin and the solipsist are not really talking about the

Page 258: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

252 SOLIPSISM PROPER

same things, and are not directly contradicting one another. In view of this one might wonder why we have devoted this much space to a discussion of Austin's criticisms of Ayer. The reason is that this in itself is a significant conclusion and one that clarifies our thesis. We have here been able to make it clear that the solipsist is concerned with or­dinary language only at the end point. In the end he says that he can construct analyses by which our ordinary assertions could be replaced without loss of meaning, but the construction of the system itself need not make use of concepts which come from ordinary language. The only necessity is that the language so constructed be one that we could use, and what sorts of languages could be used is determined by the associated theory of meaning. It is only by confronting the solipsist and the Austinian, as we have tried to do, that this point can be brought out clearly.

Page 259: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 12

FURTHER DEVELOPMENT OF THE SYSTEM: PHENOMENALISM

In this chapter I will attempt to develop phenomenalism further and make the definition sketches of chapter 9 more complete; this will be done by considering certain problems that arise within the theory itself. One of the most interesting of these problems is raised by Wilfrid Sellars in his chapter on phenomenalism in Science, Perception, and Reality. He notes the fact that in any phenomenalist analysis the hypothetical statements belonging to the analysis, which in our case have counter­factual force, will at some point depend on causal generalizations. As he puts it, material objects are analyzed into actual and possible sense­contents (sensations), and he then considers the problems raised by the possible or conditional sense-contents. He next argues:1

For if the presuppositions of statements asserting the existence of conditional sense contents are such as are ordinarily formulated in terms of physical objects, persons, sense organs, etc., as above, then the claim that physical objects are patterns of actual and conditional sense contents which in their tum presuppose generalizations, and if these generalizations are also such as are ordinarily formulated in terms of physical objects, persons, sense organs, etc., then we are faced with the absurdity of generalizations which are such that their own truth is presupposed by the very meaning of their terms.

There are two different replies which the phenomenalist can make here. First, the counter-factual conditionals certainly do presuppose generalizations and these generalizations are usually stated in material object language. However, these generalizations, while they are pre­supposed bythecounterfactuals, are not explicitly part of those counter­factuals. There would be an overt circularity if we analyzed the mate­rial object statements in terms of counterfactuals about sensations and then these counterfactuals about sensations turned out to refer to material objects themselves. But since these principles are not part of the counterfactuals there is no overt circularity although there may still be a covert one. The question then depends on whether one is going to interpret counterfactuals in such a way that when they are fully stated their presuppositions become part of the statements themselves.

1 Wilfrid Sellars, Science, Perception, and Reality, p. 80.

Page 260: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

254 SOLIPSISM PROPER

In fact, our interpretation of counterfactual statements comes very close to doing this so we cannot give this reply. Further, in many cases the material object statements will be deducible from our axioms, which is to say that the analyses of those statements will be deducible, and any presuppositions of the statements contained in these analyses would ultimately have to be in the axioms. Hence we are committed to stating any presuppositions of our counterfactuals in the language of the system, and Sellars' objection so far carries weight.

The second reply is, granted that these presuppositions are ordinarily formulated in material object language, that they can then be re­formulated in the sensation language. Before Sellars can establish the circularity he has to show that they cannot be reformulated in this way. However, instead of following this argument directly, he makes a different objection against phenomenalism with the unfortunate conse­quence that the two arguments get mixed together in his discussion. His second point is that the generalizations which are presupposed and needed could never be supported by instantial inductions, the premises of which referred to actual sensations only. As he points out, the terms of the generalizations refer to actual and possible sensations so the pre­mises from which the generalizations are inferred would also have to refer to possible sensations (conditional sense contents). This is essen­tially the point that we made in setting up the axioms of the system: we need to assume more about the world than can be garnered from present and easily remembered sensations in order to come out with our common sense beliefs. For this reason we added the second group of axioms, which assert more than the occurrence of these sensations, to our original group. We further stated that the second group not only corresponds to memories which go beyond sensations which can be easily remembered but also contains generalizations about actual and possible sensations which go beyond the evidence which can be wrung out of our actual sensations. Hence, we are in complete agree­ment with Sellars on this point, and this was our reason for adding the second group of axioms to the first.

Sellars, however, seems to assume that once the phenomenalist adds as premises principles which are not inductively supported by our actual sensations he cannot formulate these principles in terms of sensa­tions at all. Our view is that the axioms of group II, while they are not dependent on our actual sensations for their truth, still talk solely about sensations: that is, they talk not only about our present sensations but about ones we might have in the future or might have had in the past.

Page 261: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PHENOMENALISM 255

One has to separate questions of truth and meaning at this point. We cannot substantiate the truth of these principles by looking to our present sensations, or even to all the sensations that we have ever had; as a matter of fact, we cannot substantiate them at all, but must just assume them. But it does not follow from this that these axioms refer to anything other than sensations.

Having emphasized the role of "conditional sense contents" in phen­omenalism, Sellars asks whether there are independent general truths about sense contents which involve no reference to conditional sense contents and which could be arrived at inductively taking our present and past sensations as evidence. He here seems to be thinking of these generalizations not just as reformulations of material object generaliza­tions (mere reformulations would involve conditional sense contents) but as entirely separate laws whose antecedents would refer only to sets of sen­sations which we might actually have all at the same time. Sellars rejects this possibility and it does seem an unlikely one. Our position is just that there are generalizations which can be stated without referring to anything but sensations, but that these generalizations are reformulations of generalizations about material objects and that their use involves reference to sensations which have not yet occurred. In fact, our earlier assertion that even in the presence of strong inductive principles we cannot infer our set of common sense beliefs from a description of present and past sense experience implies that there are no independent generalizations about sensations of the sort that Sellars has in mind. However, when he rejects the idea that there are independent general­izations over sensation which will allow us to bridge the gap from our actual sensations to all our common sense beliefs, he also rejects the idea that the generalizations we actually use can be reformulated so that they refer only to actual and possible sensations. The first rejection is justified but the second is not.

There is another factor which tends to distinguish our view from the kind of classical phenomenalism which Sellars is discussing. The classi­cal phenomenalists usually talked about actual sensations (sense con­tents, sense-data, etc.) which could be given proper names if one desired and possible sensations to which we might refer with definite descrip­tions. In our system, however, we will have very few constants which refer to particular sensations. Rather the constants, represented by the sets of dots, almost always refer to properties of sensations and allow us also to refer to kinds of sensations. We then say that if sensations of one sort occur sensations of another sort will occur simultaneously. Thus

Page 262: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

256 SOLIPSISM PROPER

one cannot really make the distinction between actual and possible sensations that the classical phenomenalists did. Of course, some kinds of sensations will be exemplified in our present and past experience while other kinds will not be so exemplified and we can produce images of many kinds of sensations that we cannot remember having had. But we do not get mysterious particulars called "possible sensations" or "conditional sense contents" which have to be described as individuals in peculiar ways. This point is not directly relevant to Sellars' objec­tions since they can still be applied to our system, but one has to keep it in mind in interpreting his arguments.

It is unfortunate that Sellars' two objections become inextricably intertwined in his argument, but it is no accident that this happens and, in order to see why, let us continue briefly his hypothetical debate between Mr. Realist and Mr. Phenomenalist:1

Mr. Phenomenalist: So you see, Mr. Realist, that while I am trying to translate all your statements

about material objects into statements about sensations I have never promised to help you infer all your common sense beliefs from statements about your present sensations or even from a sensuous description of your whole past.

Mr. Realist: But I thought the whole point of phenomenalism was to analyse our statements

about material objects and persons in such a way as to put them on a very firm basis. In fact, I thought you wanted to analyse these statements into statements about sensations only because we can be certain about the sorts of sensations we are having and could then infer all our other beliefs in a legitimate way. But now you tell me that our common sense beliefs depend on the occurrence of certain sensations which have not occurred in fact. You then provide me with no independent generalizations which will allow me to bridge the gap, so I don't see the point of the whole enter­prise. Our statements about material objects are no better off than they were before.

Thus Mr. Realist wants from Mr. Phenomenalist something that he cannot provide and something that we never supposed that he could provide. However, it is quite natural for Mr. Realist and Sellars to expect this since traditional phenomenalists have attempted to provide exactly that and a common motive behind phenomenalism has been to shore up our common sense beliefs by basing them ultimately on certainties. Here one must remember that in using our quasi-axiomatic approach we followed the modern axiomatic method and not that of Euclid. Both Euclid and the traditional phenomenalist were trying to shore up their respective subject matters, geometry and common sense, by providing axioms which are self-evident or certain in some sense. But the modern mathematician uses the axiomatic method only to

1 Ibid., p. 82.

Page 263: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PHENOMENALISM 257

organize his subject matter in a convenient way and we use it only to organize our common sense beliefs in such a way that we can see what ontological presuppositions are made. Thus we are not trying to per­suade anyone that the axioms of group II are true but simply that they are presupposed even after we have arranged our subject matter in the most economical and efficient way. Hence we have no compunctions about providing axioms for which we can state no evidence. While Sellars does make one objection which would be damaging if sustained, it gets subsumed under another objection which is not applicable to our kind of phenomenalism, and the former eventually gets lost in the shuffle.

In developing his objections Sellars goes so far as to claim that one cannot obtain any generalizations which hold for the world generally from one's own sense experience. Further, in this claim he includes in one's sense experience not just present sensations and those that can be easily remembered but all one's past sensations as well. This is a reason­able extension when dealing with generalizations because one might be able to remember that a generalization holds for sensations that have themselves been forgotten. As noted before, we are not committed to holding that we can induce all our common sense beliefs even from this enlarged class of sensations since the axioms of group II may go beyond any sensations we have ever had (although they do not go beyond sensations as such). Nevertheless, Sellars' claim here seems to be too strong and the reasons he gives for it involve what I think is a misinterpretation of classical phenomenalism. In any case his inter­pretation is common enough to be worth examining.

Sellars starts with a distinction between accidentally autobiographi­cal generalizations and essentially autobiographical generalizations. Both kinds of generalizations hold universally for one's past sense ex­perience but the essentially autobiographical generalizations cannot be projected to cover other people's sense experiences while the acci­dentally autobiographical generalizations can be so projected. One must avoid being confused by the fact that Sellars is using "accidental" in an unusual way. Ordinarily one would speak of the essentially auto­biographical generalization, which holds only for a limited range, as the accidental constant conjunction; the other kind of generalization, which Sellars calls accidental, would then be regarded as a genuine causal principle. Still, the point is that we have generalizations which hold for our own experience, and the question is whether we can pro­ject them. Ordinarily in such cases some of the generalizations could be

Page 264: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

258 SOLIPSISM PROPER

projected while others could not, and we would look to our principles of induction to tell us which ones. Sellars, however, thinks that any generalizations which we arrive at by going over our past sense ex­perience are bound to be such that we cannot project them over the experience of others. (The fact that Sellars talks about different persons here causes no problem even though our system is solipsistic. Since persons are defined concepts in the system it just means that we are not talking in primitive notation.) The reason he gives for thinking this is as follows: 1

For the uniformities each of us finds are not only autobiographical, they are ex­pressions of the fact that each of us lives among just these individual physical objects. The uniformities I find are bound up with the fact that my environment has in­cluded wallpaper of such and such a pattern, a squeaky chair, this stone fireplace, etc., etc.

That is, the fact that Jones has the sensations he does in the order that he has them depends upon his living in a particular environment. Ifhe were someone else living in a different environment these patterns would automatically be changed and the generalizations would no longer hold. However, other persons are not really essential to the argu­ment. There are, of course, individual differences between persons which affect perception: I must have my glasses on to see what Jones sees from a particular location, etc. But we can still formulate general­izations which hold despite these differences by putting into the ante­cedents the necessary conditions in sensation terminology. Thus we can have a condition which says that if one has abnormal vision one must have the proper sorts of glasses and so on; in this way it should be possible to initially state a generalization and then doctor it suffi­ciently so that it allows for individual peculiarities and consequently holds for everyone. The point here is not that the hypothetical state­ments belonging to our analysis of material object statements are generalizations of this type. They are intended to hold only for one person - the person who is never mentioned in primitive notation but from whose point of view the system is being constructed. However, we can easily start from these hypothetical statements, generalize them, and then put conditions into the antecedents so as to "protect" the generalizations against individual differences of perception. Still, this does not seem to be what bothers Sellars; it is rather that different persons live in very different environments as far as perception is con­cerned. But if that is the difficulty it should be remembered that a

1 Ibid., p. 82.

Page 265: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PHENOMENALISM 259

single person changes his environment almost constantly. One might ask why, if the generalizations about our past sense experience hold true when we change environments, they should not remain true when we substitute another person's environment for our own. That is, if the generalizations that hold for my sensations do not hold for Jones' sensations because of his different physical environment, then one would expect them not to hold for me any longer if I went and stood beside Jones. But a generalization which could be upset just by taking up Jones' position would not even hold for me since I have taken up such a variety of positions. Thus Sellars is really committed to ques­tioning whether there could be any generalizations about sensations which hold even for one person as he changes his physical environment.

We have already seen that in his analysis of material object state­ments the phenomenalist, in the antecedents of the hypotheticals, refers to location sensations which serve to locate the observer. If I get up from my chair and walk straight ahead I will come upon a fireplace whereas Jones would, in his differing house, run upon a vase of flowers. Our consequent sensations would differ but our initial location sensa­tions would also differ. If in making generalizations we borrowed this feature from the phenomenalist analysis and put location clauses (referring to the kinds oflocation sensations we have and might have) into the antecedents along with the other conditions, then Jones' ex­perience would not constitute a counter-example to a generalization inspired by my patterns of experience or vice versa. The point is simply that a generalization which holds only over a limited range can be transformed into a universally true generalization if we put into the antecedent a condition describing that limited range. If all swans on the Thames are white but there are black swans elsewhere we can form the universally true generalization, "If anything is a swan and is on the Thames, then it is white." Similarly, we can form universally true generalizations about sensations if we put into the antecedents of those generalizations conditions which only apply the consequents to certain limited ranges. Such a generalization would be accidentally auto­biographical according to Sellars' definition since the generalization would hold for any person as long as he is in the right environment. Putting it briefly, what is important is not who you are but where you are. Sellars recognizes correctly that the generalizations one might derive from one's own sensations hold only over very limited ranges, but he has mistakenly identified the range with the person as opposed to the physical environment. He himself in making his objection points

Page 266: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

260 SOLIPSISM PROPER

to the things constituting the environment rather than to the person as the essential factor.

However, Sellars then goes on to say:l In other words, what the phenomenalist wants are generalizations, in sense con­

tent terms, which are accidentally autobiographical, generalizations in which the antecedent serves to guarantee not that I am in the presence of this individual thing, e.g. my fireplace, but rather that my circumstances of perception are of a certain kind. What he wants for his antecedents are patterns of sense contents which are the actual sense content counterparts of the kinds of perceptual circumstances which common sense expresses in the language of persons, sense organs, and physical things. The best he can get, however, are essentially autobiographical uniformities in which the antecedents, however complex, are the actual sense content counter­parts of the presence to this perceiver of these individual things.

There are three points to be made here. First Sellars is repeating an earlier point, that the set of actual sensations is not a sufficient basis for scientific induction, and we have granted that. Secondly, he may here be implicitly accepting the possibility of the kind of universal general­ization we have just mentioned. At least it is not clear whether he does or not; if not, it would be because he has not taken sufficient account of the possibility of putting locating conditions into the antecedents of these generalizations. These location sensations might not be actual ones but that has nothing to do with whether such a generalization could hold universally; the important thing there is simply that there be no actual pattern of sensations which constitutes a counter-example. The third point, and the most interesting one, has to do with the ques­tion of how "general" these generalizations might be. Suppose, for instance, that we can form the kind of universal generalizations we have mentioned but that the antecedents of such generalizations are very restrictive and only described a very limited context or environ­ment. They may be universally true but they are not very interesting because their instances are all so local. At one point Sellars seems to be suggesting that these are the only kinds of generalizations that one can form and, if that is SO, one might wonder how one could ever work up to the more important and interesting generalizations presupposed by science. This is something about which one might well wonder and it reinforces Sellars' first point. The important thing, then, is not whether one can form any universally true generalizations at all but whether one can form any important generalizations starting just from one's actual sensations. We have not wanted to suppose that one can and we added the second group of axioms for that reason. This is also Sel­lars' position and the arguments contained in the last two quotes turn

1 Ibid., p. 83.

Page 267: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PHENOMENALISM 261

out just to reinforce that position further. Thus while Sellars sets out to attack classical phenomenalism, what he says turns out to add more to the plausibility of our theory than it detracts.

Another instructive objection to phenomenalism is stated by R. Chis­holm in his book, Perceiving: A Philosophical Study, and amounts to the following. 1 If the phenomenalist program is to be carried out, any material object sentence must be capable of being analyzed in terms of a number of hypothetical statements about sensations, or observation statements, as he calls them. Thus the material object sentence will entail a conjunction of observation statements, and will consequently entail the observation statements individually. However, we can always conceive of circumstances under which the material object sentence will be true, but any given observation statement will be false. As we have seen, the phenomenalist might suppose that when he has sensa­tions characteristic of going over and reaching out to touch the table, he will then have other sensations characteristic of feeling the table, and hypothetical statements to this effect would then be included in the analyses of sentences about the table. However, if he is suffering from some abnormality of the tactual nerves, the hypothetical observation statement might turn out to be false, as he might still have the sensa­tions mentioned in the antecedent, but not those mentioned in the consequent. He would nevertheless still be justified in believing in the table's existence in that he would still observe the table in various other ways, and he might even find out about the abnormality of his nerves which caused him not to feel the table. Thus, since there is a possible situation in which the material object sentence would remain true while the observation statement is shown to be false, it follows that the former does not entail the latter. Similarly, one could always conceive of some sensory abnormality which would falsify any observation state­ment included in the analysis of a material object sentence without refuting the latter. Further, as Chisholm points out, it does the phenom­enalist no good to put into the antecedents of his hypothetical observa­tion statements tests for the normality of the senses in purely sensory terms. One could conceive of further abnormalities involved in the application and interpretation of these tests, which means that no purely sensory test or tests could be devised which would detect all possible sensory abnormalities.

In his book, The Problem of Knowledge, A. J. Ayer states an objection similar to that of Chisholm and then goes on to make a converse objec-

1 R. Chisholm, Perceiving: A Philosophical Study, pp. 189-197.

Page 268: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

262 SOLIPSISM PROPER

tion to phenomenalism.1 He here argues that the existence of a material object, and more generally the truth of a material object sentence, can­not be deduced from any set of statements about sense-data or sensa­tions. From his argument it is quite clear that he means that the mate­rial object sentence cannot be deduced from a finite number of state­ments about sense-data, and he seems to have assumed that an analysis in terms of an infinite number of statements is really not an analysis at all. Most of his argument is then directed to showing that there is no point where the phenomenalist analysis can leave off, and that the analysis will not entail the material object sentence at any given point. This is because it is always possible to imagine further ways of testing the truth of a material object sentence no matter how well it may have been verified. If the sentence could be conclusively verified finitely we would fall into self-contradiction if we were to suggest that further tests are relevant to its truth-value, but this point never seems to be reached. And it does appear that even though I may be quite sure that I am now holding a pen, and even though it would be foolish to attempt to test assertions to this effect any further, one can still think of further tests which would be relevant to the truth of the same assertions.

Hence Chisholm is arguing that the material object sentence does not entail the observation statements, and Ayer also argues that the latter, even taken together, do not entail the former. If either of these objections can be sustained, phenomenalism must be rejected.

I will discuss Ayer's objection first. The traditional phe:1omenalist would seek to avoid this by resorting to an infinite analysis, but I will argue first that this leads to fatal difficuldes, and second trat Ayer's objections to a finite analysis can be met.

On the first score it seems absurd to hold that when someone utters a material object sentence he means to say that certain observations could be made when he is not aware of their possibility, and would not recognize them as being relevant to the truth of the original sentence even if he were aware of them. However, the usual phenomenalist in­finite analysis does include references to ways ofveruying the statement in question in the very remote future. For example, let us suppose that A is a passenger on a ship docking in N ew York and that he writes in a letter that the ship is docked, giving the date and time. If asked how this statement could be verified in a hundred years, he would probably say that it could be done only by looking at records as the ship would long since be gone from the dock and into the hands of the shipbreakers.

3 A. J. Ayer, The Problem of Knowledge, pp. 138-144.

Page 269: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PHENOMENALISM 263

However, it might also be that a nuclear attack took place an hour later sinking the ship at the dock and contaminating the whole area so thoroughly that no great changes in the area were made in the next hundred years. Thus A's original sentence could be partially verified (its probability raised) by someone's simply coming back to the dock and observing the ship's presence there. But surely we cannot interpret A as meaning to assert that the ship could be observed at the dock one hundred years later as nothing could be further from his beliefs at the time of writing the sentence. Hence, we must conclude that it is a nec­essary, but perhaps not sufficient, condition for the inclusion of a given observation statement in the analysis of a material object sentence that the person uttering the sentence would admit the relevance of the observation to the truth of his statement at the time of making it.

It is also clear that the "temporal range" of the material object sen­tence, as well as that of the observation statements, will be finite in the sense that if they are to be conclusively verified, this process will have to take place within a finite period of time from the time of making the statement. It seems obvious that no one making a material object state­ment would admit as relevant to its truth-value any particular observa­tion which is supposed to be made indefinitely far into the future. He would have no idea what to expect on the basis of his statement's being true, or for that matter, false.

Having rejected infinite phenomenalism, the question then is whether Ayer's objections to finite phenomenalism can be met; here it is highly desirable, and perhaps essential, that the finite phenomenalist be able to set up necessary conditions for the inclusion of observation state­ments in the analysis.

The finite analysis is not, of course, committed to the view that material object sentences can be conclusively verified all at one time; the number of tests which are relevant would in the majority of cases still be very great and stretch out over many years. Nevertheless, the point of the finite analysis is that no matter how long and thorough the process of conclusive verification may be, it can be brought to an end by a careful enough observer, and it is just then that Ayer holds that further tests are conceivable without falling into self-contradiction.1 But, supposing someone to have gone through this long and tortuous process of conclusive verification, the question Ayer does not raise is whether the English sentence still has the same meaning for the utterer that it had in the beginning. If new tests are always conceivable, it may

I Ibid., p. 141.

Page 270: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

264 SOLIPSISM PROPER

be that we always project the range of the material object sentence ahead of us before we reach the end of it; but it will still be finite at any given time. These new tests would have been considered irrelevant at the time the sentence was originally uttered, according to our criterion, and their possibility would most likely have been discovered during the process of verification. In such a case it seems much more accurate to say that the meaning of the original sentence has been extended to include the new test than to say that the new test was somehow all along included in what was meant, even when it was thought to have no bearing. Consequently, if the meaning of the original sentence has been extended there will be no contradiction in saying that the original statement has been conclusively verified while the truth of a slightly stronger, but very similar, statement is still open to question. The appearance of a contradiction is generated by the fact that the same English sentence is used all along even though it undergoes gradual changes of meaning when analyzed into the language of the system. This is, of course, what we would expect, as there are not nearly enough sentences in English, or any other natural language, for there to be a separate sentence to correspond to all the logically distinct assertions we can make in a sensation language.

It is also sometimes objected that if the phenomenalist analysis is to be finite, there must be a sharp criterion for determining just which observation statements are to be included in the analysis. If this is an objection, it applies just as much to infinite phenomenalism; the in­finite phenomenalist does not include all possible observation state­ments in his analysis, and he must also make a selection according to some principle. However, in view of the fact that the phenomenalist does not supply any actual translations of material object sentences, it does seem desirable that he make the instructions for so doing as ex­plicit as possible. As will be seen shortly, the criterion stated above is not strong enough in itself, and I will now add other conditions to it in order to make it more nearly adequate.

In our previous example we saw that a certain possible observation was not believed to be relevant to the truth or falsity of the material object sentence because it was not supposed that the circumstances which would make it relevant, a nuclear attack on New York, would come about. Hence no hypothetical statement about sensations which asserted the possibility of this observation was included in the analysis of the material object sentence. However, a new observation statement might be formed in the following way. It would be a hypothetical, and

Page 271: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PHENOMENALISM 265

its consequent would simply be the original observation statement asserting that the ship could be observed at the dock in one hundred years (this would, of course, be in terms of sensations and would itself be hypothetical). The antecedent of this new observation statement would assert in terms of sensations that the circumstances which would make the observation described in the consequent relevant, in this case the occurrence of a nuclear attack, will take place; this will also be a hypothetical, so the result will be a large hypothetical statement of which both the antecedent and the consequent are themselves hypo­thetical. It will say in the sensation language that if an atomic attack is to take place sinking the ship etc., then it will be possible to observe the ship in the same place in one hundred years. It is not obvious that this latter statement is excluded by the criterion we have above; if such a test were suggested to the man making the original statement at the time he made it, he would certainly regard it as a very roundabout way of testing a simple statement, but he might admit that its outcome would be relevant to the truth of his original statement, if that state­ment had not already been conclusively verified or refuted. However, this observation statement describes just the sort of remote possible consequence of the ship's being where it is that would never occur to the man who wrote down the statement. In admitting that this is a possible method of verifying the ship's present position (sinking it where it is and observing it there later) he is not necessarily admitting that he means to assert the possibility of this test or other similar tests; rather he means only to assert the possibility of those tests which are simple and obvious to him. In admitting the relevancy of one of these obscure observations to the truth of the material object sentence he is only ad­mitting that, given the truth of his original assertion, it is probable that certain other observations can be made in addition to the ones origin­ally envisaged. If the original statement is still in doubt these extra observations might raise or lower its probability, but if the original assertion has been established or refuted the extra observations will help us only to discover the consequences of its being true or false. The meaning of the statement might, of course, be extended to include them, but then they would not be included in what was originally meant. Hence a criterion for what a speaker means by a material object sentence when he utters it must be a stronger one. This might be the old criterion of supposed relevance at the time of utterance plus the additional condition that the speaker also supposes that he could voluntarily bring about the truth of the antecedents of the hypothetical

Page 272: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

266 SOLIPSISM PROPER

observation statements. In case the original statement is about the past or the remote future, the condition will be that the speaker supposes that he could bring about the truth of the antecedents voluntarily if the event were to take place at the present time but be unchanged otherwise. If the events which the statement describes or predicts are distant in time the analysis will also be affected in other ways, but we are not concerned with that problem here. It must be remembered that in the ordinary phenomenalist analysis the antecedents of the ,hypothetical statements about sensations describe in sensory terms the position in which the observer must put himself in order to observe the phenomenon as well as describing the somatic sensations, those charac­teristic of reaching out the hand, focusing the eyes, etc., which he must also have. It is normally possible to produce these sensations by making the appropriate bodily movements. On the other hand, one does not suppose that one can bring about sensations characteristic of a nuclear attack unless one is standing at a missile control panel, or something of the sort. Thus the obscure tests mentioned above are again elimin­ated from the analysis of what the speaker originally meant. These conditions also have a pragmatic consequence; any hypothetical re­maining in the analysis will be relevant to the individual making the statement in that it describes a test which he supposes himself capable of carrying out. Thus the finite phenomenalist can at least provide rough directions for carrying out the analysis of any particular material object sentence even though it may still be too complex to carry out in prac­tice. We have said that the observer supposes himself capable of making anyone of the observations described by the observation statements, but, of course, he does not necessarily suppose himself capable of making them all together. Further, a material object sentence may turn out to be at least partially unverifiable in practice if some of these suppositions turn out to be false.

Thus far we have outlined a finite phenomenalist analysis and ar­gued that the observation statements belonging to it will, if correctly chosen, entail the assertion which is being made when someone uses a material object sentence. We will now go back to Chisholm's problem and try to establish the converse entailment. In my opinion the best attempted solution of this problem has been put forward by R. Firth.1 The essence of his view is that a given material object sentence stands ambiguously for a number of sets of statements about sensations; each

1 R. Firth, "Radical Empiricism and Perceptual Relativity 11," Philosophical Review July, 1950.

Page 273: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PHENOMENALISM 267

set of statements about sensations, which he calls a physical statement, describes a number of ways of verifying the material object sentence, and since the observations described by each physical statement are roughly similar to each other, they all form a family. "If the phenomen­alist analysis is correct, in other words, the sentence 'This is a door­knob' would symbolize ambiguously each member of a large family of physical statements; and, apart from the inferences which could be drawn from the context, therefore, nobody who heard us utter this sentence could know just which statement in the P-family we actually intended to assert."l Hence, if the speaker is suffering from some senso­ry abnormality which causes one of the physical statements to be refuted, this does not mean that the material object statement is false. There are still a number of other physical statements belonging to the family which might serve as translations for the material object state­ment; these describe slightly different observations which may not be affected by the abnormality, and in fact some of them will be protected against that very abnormality. Thus we may still be able to find a sense of the material object sentence which will not be refuted.

Chisholm mentions this proposed solution and says that he believes that it leads to implausible consequences, but does not say what they are'! In fact there is a possible ambiguity at one point in Firth's posi­tion. When he says that the material object sentence is ambiguous, this might mean that its meaning is itself indeterminate, or it might mean that the speaker is asserting something perfectly determinate, but his words do not succeed in conveying to his audience what he does in fact mean to assert as distinct from certain other similar assertions which he might have made instead. If we took the former case, however, Chisholm's objection would not be met. However well the various physical statements signified might be protected against sensory abnor­malities, a set of abnormalities could still be imagined which would render all the physical statements false even though the material object sentence which is supposed to entail them remains true. The point of Firth's analysis is that the material object sentence can be re-inter­preted after these sensory abnormalities have been discovered. Hence, ifhis analysis is to work and the material object sentence is only finitely ambiguous, which I think we must assume, then we must be allowed to change the meaning of our original assertion in the light of new facts, even though the new meaning will be included under the same material

1 Ibid., p. 322. 2 Chisholm, op. cit., p. I 94n.

Page 274: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

268 SOLIPSISM PROPER

object sentence. Having posited the ambiguity of the material object sentence in this sense, there is no reason to hold that it is ambiguous in the other sense as well, and this seems to be Firth's position. Thus in cases of illusion the situation will be that our original assertion is refuted and another one substituted for it, which latter is protected against the illusion that is thought to obtain; if another illusion is also found to be operating, the second assertion will be refuted and another one sub­stituted for it, and so on until the speaker decides that there is no similar assertion which is true.

There are some important consequences of this interpretation. For one thing, the phenomenalist will be committed to saying that the meaning of a material object sentence changes as new information comes in, but we have already attempted to justify this on other grounds, so there does not seem to be any difficulty here. However, it must be remembered that this process will not occur in any case of sensory abnormality. As Austin has pointed out, there are many cases which philosophers have traditionally labeled "illusions" where we are not deceived or surprised at all. Thus we do not expect a straight ruler held half underwater to look "straight." Consequently this fact would be taken account of in the analysis of the original statement. If I say in material object language that a particular ruler is straight when it is partly under water we analyze the statement in such a way that it is implied that I will have "crooked stick sensations" if I look in the right place. Hence there is no deception and nothing gets refuted. There might, of course, be some persons who would really be surprised at this or a similar phenomenon. If a person really does expect sticks held half under water to look straight, then we would have to analyze his mate­rial object statements accordingly and, if the stick he is talking about is in or near water, then one of the hypotheticals belonging to the ana­lysis would get refuted and another one would have to be substituted for it if the material object statement is to be retained. Still, in order to want to retain the material object statement, the speaker would, under the circumstances, find out about this illusion and his new dis­covery would be reflected by the revised analysis. The general rule of analysis is to follow the beliefs and expectations of the person who makes the material object statement; since we are constructing a solip­sistic system one would typically analyze material object statements in such a way as to reflect one's own beliefs and expectations. Thus no part of the analysis is falsified in standard cases of "illusion"; this happens only if one really is taken in. If one then finds out the mistake

Page 275: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PHENOMENALISM 269

it is plausible to assume that this will affect what one means in the future when one makes similar statements. Hence Firth's view is not contrived as Chisholm suggests but has the consequences one would expect.

The phenomenalist now has a way of solving any problems which arise from the possibility of illusion, and we saw before that the adop­tion of a finite analysis avoids a host of other problems. We have also attempted to give some rules which tell us what to put in the phenomen­alist analysis of a material object statement and what to leave out. This amounts to the filling out of the definition-sketches given earlier and, while we have still not given the definitions themselves, we can now claim to have given a reasonable approximation of them.

We next come to the problem of translating names and other singular terms, as they appear in our ordinary language, into our solipsistic language. This turns out to be easier than one might suppose. Let us first consider proper names and other singular terms such as "that thing" which are meaningless within the context unless they denote uniquely within that context. The function of such terms in English is to pick one object or person out of the context and allow us to talk about that object or person and none other, and the use of a name is closely tied to the existence of features which differentiate that object from all others in the context. That is, knowing how to use the name involves being aware of those features, or others which do the samejob, and it also involves knowing that the name is properly applicable only to the object which has those features. Thus one has to have some knowledge about the world in general and the object named in particu­lar in order to use a proper name. When we translate a sentence in­volving a proper name into our solipsistic language the analysis reflects this knowledge. We can see how this knowledge will be mirrored in the analysis by comparing the analyses of the sentences, "There is a ship a mile up ahead" and "The Honshu Maru is a mile up ahead." Any­thing that is included in the analysis of the former will also be in the analysis of the latter, but the latter goes beyond the former in two main ways. First, the descriptions of the object sensations in the second case will usually be more determinate than those of the former case. In the former case we want the descriptions of the sensations to be indeterminate enough so that they can "cover" the sorts of sensations we would have in viewing almost any kind of ship. Of course, we do not in practice use the term "ship" to cover such things as nuclear-powered carriers, submarines and other weird craft which are sui generis, but

Page 276: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

270 SOLIPSISM PROPER

we still want the descriptions of sensations to be indeterminate enough to cover any sensations which might be obtained by observing any ordinary merchant or passenger ship. However, in analyzing the state­ment about the Honshu Maru we can specify those kinds of sensations which are characteristic of seeing a very rusty, very old tramp steamer which has a tall straight stack, two masts, and steam leaking from the engine room. Secondly, there will be components of the second analysis which are entirely lacking from the first. For instance, the Honshu Maru will be flying a Japanese flag and this fact is reflected in the analysis; further, the analysis will say, in sensory terms, that if we get out our binoculars and train them on the stern of the ship, we will be able to make out the letters Honshu Maru and below them Yokohama. Thus the set of statements comprising the second analysis refers to a particular ship while those comprising the first analysis do not so refer. We never need to talk about particular sensations in order to talk about the Honshu Maru, again for two reasons. First, even a particular object turns out not to be ultimately particular when we move to the sensation level. That is, even a specified object can look differently in different circumstances and we can never say in advance exactly how it is going to look, feel, etc. in any future circumstances. vVe can narrow down the range of sensations which are characteristic of perceiving the object, but I will never be in a position to say that when I see the object I will have sensations exactly like the ones I had on a particular occasion yester­day. One can put this by saying that even proper names have a certain degree of qualitative indeterminacy and that one can translate them without referring to particular sensations but only to kinds of sensations. Secondly, it is obvious why proper names are not absolutely determin­ate: we do not need absolute determinacy in order to refer uniquely to an object in the context and this kind of determinacy would be cumber­some and practically unusable if we had it. In the context above we do not need to know everything that there is to know about the Honshu Maru in order to identify her at sea, and the name "Honshu Maru" is not packed with that kind of meaning. In order to identify her we need only know that she is a battered old tramp ship with the appropriate name on her stern. Consequently, these being the distinguishing features that are used in identification, no more than that need be implied by the use of the name.

One can, of course, think of other contexts in which more would be implied. Suppose that we pay a visit to the ship-breaker's yard and find a dozen old tramp ships moored to one another; all these ships are now

Page 277: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PHENOMENALISM 271

under Panamanian registry, are named for relatives of a certain Greek, and all look about the same to a casual observer. Suppose one now overhears the ship-breaker saying, "We will break up the old Honshu Maru next week," obviously referring to one of the ships in the row. Here one might have to know a great many facts about the original ship and her subsequent history in order to distinguish her from the others. Alternatively, one might distinguish her from the others by knowing that, say, the third ship in the row was bought from a company which bought her from another company which in turn bought her from another company which in turn bought her from a certain Japan­ese concern as the Honshu Maru. There might even be as many differ­ent ways of picking the ship out as there are persons to talk about her. As we have seen earlier, we have to take the statement as uttered by some particular person and then analyze it so as to reflect the particular distinguishing features which he knows about and uses, or would be inclined to use. Thus, so long as he knows how to discover which ship is the Honshu Maru, we can analyze his statement without any diffi­culty. If the speaker himself does not know how to pick out the ship then he either does not know what he is talking about or the term "Honshu Maru" is being used in the way that one might ordinarily use the phrase "one of the ships in the line." In the latter case we would simply analyze the statement in the way that we would if it contained no proper name. Hence, it can be seen that we can analyze any state­ment in English containing a proper name into our solipsistic language. Actually there is a continuing debate between Quine and Strawson over whether singular terms can be eliminated in the natural language itself or something close to it.1 Our thesis, however, implies no commit­ment in this controversy and we will remain neutral to it.

We will now briefly consider definite descriptions. The difference between the standard definite description of the form "the object such that .... " and a proper name is that the former makes explicit some of the identifying features which are implicit in the proper name. Both, however, are supposed to refer uniquely to one object, and a definite description need not make explicit all the features that we might need to identify the object in any given context. For instance, the definite description, "the desk in the next room" does not tell us how to pick out the desk without taking note of the context in which a sentence con­taining the description would be uttered. When someone makes a

1 This dispute pops up in widely scattered places in most of the writing of Quint" and Strawson.

Page 278: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

272 SOLIPSISM PROPER

statement that contains a definite description the principle of transla­tion is basically the same that it would be in the case of a sentence con­taining a proper name. In both cases we take the identifying features of the object which the speaker means to refer to and put these features into the analysis in an explicit way; the definite description just men­tions some or all of the identifying features which the speaker would ordinarily have in mind.

We now come to the classic problem of what to do when a definite description does not denote or does not denote uniquely. Suppose, for instance, someone in a Monaco hotel remarks, "The king in the next roomjust flew in from Sweden yesterday," and let us suppose that there is either no king in the next room or more than one king. This really presents no very grave problems because even though the definite description, taken literally, does not denote uniquely, we would or­dinarily have no difficulty in discovering which king the speaker meant to refer to. The speaker himself, when informed of the situation in the next room, might be able to supply the king's name or point him out on another occasion. The classis case, "The present king of France is bald" presents peculiar problems because no one believes that there is a present king of France and it is almost impossible to imagine what someone who made such a statement might mean. But in any ordinary case where natural mistakes might be made this difficulty is not present. If the mistake is a natural one and the speaker could supply identifying features, we could then translate, "The king in the next room flew in from Sweden yesterday" in the same way that "King Gustav flew in etc." might be translated. The identifying features might be different but they might also be different if the proper name were used at two different times. The only problem is that some of the identifying features which the speaker has in mind do not turn out to be possessed by the object of reference if there is no king in the next room, but we would ordinarily still have enough features left to allow him to be identified. The situation for the phenomenalist is then exactly as it is when we encounter a perceptual illusion which surprises us: the object we are talking about exists, but it turns out not to have some of the observa­tional consequences that we expected it to have. Consequently we apply Firth's analysis to the Swedish king and when the speaker finds out about the absence of the king he "cuts down" the meaning of his original statement so as to eliminate this feature and substitutes some other mode of reference such as "King Gustav" or "the king I saw walking towards the door of the next room." If there is more than one

Page 279: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PHENOMENALISM 273

king in the next room there may be no problem at all. The other iden­tifying features which the speaker has in mind, when made explicit in the analysis, will probably give us unique reference. If the speaker, when apprised of the situation, cannot say which king he wants to make his assertion about, then we can substitute "one of the kings in the next room" for his original definite description and capture his meaning. Our whole theory of meaning tells us to translate sentences as used by particular speakers on particular occasions and it tells us to take the context not only into account but to treat it as an integral part of the meaning. The important thing is to capture the intention of the speaker on the particular occasion. Consequently we have no problem in trans­lating sentences which contain ordinary definite descriptions even if they do not denote uniquely.

There are still a few cases where one might meaningfully use a de­finite description without knowing whether it denotes uniquely and, at the same time, realize that one does not know how to pick out the denoted object if there is one. Suppose that in the lobby of the same hotel in Monaco a woman who looks like an ex-actress walks by loaded down with jewelry. Someone might remark, "The husband of that woman may be very old but he is also very rich." Here the statement is made in the spirit of a prediction and the bystander does not really know whether she has a husband or not; for all he knows she might be the divorced Duchess of Sussex. We cannot translate this statement as we did the other ones because the speaker could supply only the sketch­iest identifying features for the possible husband and these, even when made completely explicit, might not be sufficient to pick out that worthy. Further, we cannot follow our previous recipe in such cases and substitute "some husband" or even "all husbands" for the definite description. The bystander does not mean to assert that there is a hus­band in a categorical way, so the substitution of "some husband" can­not be made since it would have him making such an assertion. Even the substitution of "all husbands" is improper because the speaker has a picture of a particular husband in mind and he is not making a general assertion of the form "any person who is related to the woman in such-and-such a way will have the following features: ...... " Rather, the particular husband pictured, but not known to exist, is supposed to have traded money for good looks, and a causal factor is implied to be present. Thus we can best render the statement as if it were a counter­factual in English and say "If there were (is) a man who is married to that woman, then he is old and rich." This allows us to talk about a

Page 280: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

274 SOLIPSISM PROPER

particular silly old man who temporarily lost his equilibrium, but we allow that he may inhabit a hypothetical world rather than the actual world. The definite description has now been eliminated and there is no problem of translation into the language of our system. There is already one counterfactual in the English version but the same counter­factual connective appears in our object language and the rest of the English sentence can be translated in the usual way. Similarly, if one were to make any sense at all out of "The present king of France is bald," it would have to be as "If there were a present king of France he would be bald." It seems, however, that most definite descriptions that do not denote uniquely are used as a result of honest mistakes and this mode of analysis is not then necessary.

The next topic concerns the relationships which hold between the object language and the metalanguage of our system. This discussion might logically have appeared three chapters earlier but, because of the abstrusity of the issues, we have delayed until they could be more easily understood. The most important metalinguistic concept was that of meaning itself; that concept, as developed in the first part of the book, was intended as an analysis of our ordinary notion of meaning, and the same thing is true of such other concepts as synonymy and understanding. One might ask at this point why we do this instead of inventing some concept of meaning of our own. The reason has to do with our assumption that we can use the language meaningfully in the first place. That is, the language that we ordinarily use is assumed to have meaning in our ordinary sense of "meaning." One might even number this assumption among our common sense beliefs, but it is not one that will be found among the axioms or theorems of our system; we are there trying only to systematize our common sense beliefs about the world, and since this is a belief about our language rather than the world, it will not appear there. Nevertheless, it is a real assumption and without it we could not do philosophy at all. Suppose that we sub­stituted some other sense of "meaning" for our ordinary one. Could we then be sure that the language we use has meaning in this new sense? We could if the new sense of meaning were a special case of the ordinary sense; at least, if we knew exactly how this special case of meaning were related to meaning in general, we would then say what part of our ordinary language would have meaning in this new sense. That is, if the new sense of meaning is a special kind of meaning, we might then decide that part or all of our ordinary language has this kind of mean­ing. Suppose, however, that the new sense of meaning does not take off

Page 281: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PHENOMENALISM 275

from the old one but is defined independently and goes beyond the ordinary sense of meaning in some way. We would then have no reason to think that we could talk meaningfully in this new sense of meaning. In short, one has to first talk and then find some concept of meaning which is believed to fit or apply to one's actual conversation - then one knows that one can talk meaningfully because one already has. But if one says, for instance, that a meaningful term is one that denotes an "eternal object," then we do not know whether or not we talk meaning­fully until we find out what an eternal object is. If it turns out that we can use our ordinary language to talk about eternal objects by extend­ing the language definition ally then the new concept of meaning will have to be reducible to the old one. If, on the other hand, it takes a radically different kind oflanguage, of which we are not now capable, to talk about these eternal objects then we have no reason to think that we could ever talk meaningfully in the new sense. The new concept of meaning is therefore useless at best. This is where the "no exit from language" comes in; it is permissible to talk in a way that we do not ordinarily talk only as long as our ordinary concept of meaning sanc­tions the kind of talk indicated. Then the basic assumption that we can talk meaningfully in the way that we do talk "covers" that other way of talking. But if we try to think up a new concept of meaning to cover some new kind of language that we don't use we then become com­pletely lost and anything that one can say is as good as anything else.

In addition to developing our system we want to say that, on its intended interpretation, the analyses it contains are correct on the basis of our theory of meaning. In short, we want to say that our philosophy is at least adequate, but in what language do we make that statement and how do we make it? What we want to say is that all statements about physical objects and other persons are reducible to statements about sensations but, since there are no persons at the primitive level, one wonders where and how this statement will fit in. A consequence of our reduction is that someone observing a person using the language of our system would note that when the speaker is presumed to be having a sensation and, when by his behavior he is presumed to be talking about his own sensation, the way he talks is logically much simpler than the way he talks when he is presumed to be talking about someone else's sensation. In other words, when a person using such a language is talking about what we, in the metalanguage, would call his own sensations he makes simple categorical statements; when, on the contrary, he is presumed to be talking about the sensations of other

Page 282: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

276 SOLIPSISM PROPER

persons he makes very complex hypothetical statements (their logical structure will be delineated in the next chapter) and he may make use of defined terms as abbreviations (he has no way of abbreviating state­ments about his own present sensations). The distinction here is some­what over-simplified since one can, in primitive notation make hypo­thetical statements and just be talking about oneself (as we would say in the metalanguage). However, one could never make categorical statements about anyone else and we will eventually see how we could, in the metalanguage, always distinguish the speaker's statements about "himself" from his statements about "other persons."

At this point the metalanguage is being used for two different things. First, it allows us to say that statements in the object language, con­structed according to our directions, are synonymous with correspond­ing material object statements of our ordinary language. Secondly, it is being used to say that it is possible to tell when the solipsist is using his language to talk about himself and when he is using it to talk about someone else even though persons are not primitive in his language. We can say in the metalanguage that the solipsistic language makes a distinction parallel to the distinction between persons in our ordinary language and, further, that this distinction is so drawn that we can get relations of synonymy between statements of the object language and statements of the ordinary language. The first assertion amounts to saying that the system, on the intended interpretation, yields a correct analysis of any statement of ordinary language; the second assertion is more specialized and says that, although the system does not refer ex­plicitly to persons on the primitive level, it is still rich enough to allow us to say anything that we want to say about persons. Of course, we knew from the start that we would have to be able to say these things in some language or other and, since the metalanguage is all-inclusive, we can obviously make these statements in that language. The question is now whether the metalanguage can be collapsed into the language of the system by a series of definitions; it would then be possible for the object language to talk about itself and to say that on its intended interpretation it yields a correct analysis of the statements of our ordi­nary language.

If this is to be accomplished the language of the system will obviously have to be enriched but less than one would think. First, since we want to say within the system that statements of the system are synonymous with statements of ordinary language, we will have to have in the system some way of referring to th e statements of ordinary language; the system

Page 283: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

PHENOMENALISM 277

will also have to be able to refer to its own statements, and the language is rich enough to do this. It will be sufficient for this purpose to treat the sentences of ordinary language as utterances or physical events - as spoken sounds or patterns of ink on paper. These physical objects or events will then be analyzable into the solipsistic language just as are other kinds of physical objects and events. We can then form the various sentence and word tokens into sentence and word types in the usual way. The next thing that we need in the object language is the concept of meaning; this concept was defined in the first part of the book in terms of sentence types (via preliminary statements) ,images, sensations, the attitude of looking for a basic instance, and feelings of satisfaction and dissatisfaction. This means that we would first have to add the concept of an image to our system and this could be done very easily. We already have strings of dots denoting various kinds of sensations, and, since images and sensations correspond qualitatively, we could add some mark, say a star, after the dots to indicate some kind or class of images rather than sensations. Similarly, the word, "sensation," occurs in the object language and the word, "image" could be added to parallel it. When using variables we could then specify in any par­ticular case whether the variables range over images or sensations or both. These conventions would not complicate the object language in any very serious way. Further, since the having of the requisite attitude and the feelings of satisfaction and dissatisfaction are all mental events, we could simply invent names for these kinds of mental occurrences and add them to the object language as well. We already have in the lan­guage the temporal relationships of simultaneity and temporal pre­cedence so no more apparatus would be needed to handle these new primitives. We would then be in a position to add the predicates, "is synonymous with," "is meaningful" and "understands" to the system as defined terms in the way that we have previously added other terms. Of course, the definition of synonymy would stipulate that the senten­ces which are asserted to be synonymous have to be specified as to context of utterance but this is something that we have wanted to do anyway and, once we have added the concept of a person to the system, there will be nothing to prevent us from making those speci­fications and talking about synonymy within the object language. We will then be able to say that certain statements in primitive notation are synonymous with certain material object statements, uttered in specified contexts, and we will be able to say this within the object language itself. In fact it now appears that anything we can say within

Page 284: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

278 SOLIPSISM PROPER

the metalanguage can also be said in the language of the system. The metalanguage has all along just been our ordinary language plus the semantical concepts such as that of meaning; we have already reduced the concepts of ordinary language to concepts about sensa­tions and we have just seen how the concept of meaning, as previously analyzed, could also be added to the system. Thus the metalanguage collapses into the object language.

The important point about this collapse is that while it is necessary to add some new primitives to the object language in order to accom­plish it, the new primitives still refer to mental events and no new kind of ontological commitment arises. The fact that such a collapse is possible will be useful shortly when we analyze the concept of past ness.

Page 285: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 13

STATEMENTS ABOUT THE PAST

We have now discussed at some length many of the problems which arise out of the phenomenalist analysis ofstatcments about the material world. However, there is a special sub-class of statements about the material world, statements about the past, which present special problems, and we will deal with them in this chapter. The problem is basically that of analyzing tensed statements in ordinary language which refer specifically to the past, the present, or the future. It turns out that in order to construct an adequate analysis we will have to add slightly to the primitive language of our quasi-system. In particular, we will have to have constants, which will be viewed as proper names, and which will refer to particular sensations.

Up to this point we have always referred to kinds of sensations, as in " .... sensation occurs," or classes of sensations, as in "x E •••• ". We will add constants referring to particular sensations by saying that the letter "k" with any integral subscript is well-formed; this will allow us to say such things as "k5 occurs" in the primitive language of the quasi­system.

We have just said that these constants are to be construed as proper names referring to sensations, but much more explanation is needed. In our ordinary language proper names do not always have the same role, and we must specify exactly which role we have in mind before we can make it clear how these constants are to be used in the object language. Further, the proper names in our ordinary language are almost always used to denote objects or persons as opposed to sensations. We must therefore show that proper names can still play what we take to be their characteristic role when they are used in a sensation language.

One obvious condition for the correct use of a proper name is that there should be a unique object or person which is denoted, and that the user of the name know which object or person it is. While we ordinarily have no reason to apply proper names to sensations, this condition does not make it impossible to so use them. If I want to apply

Page 286: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

280 SOLIPSISM PROPER

a proper name to a certain tickling sensation which I am now having, there is certainly exactly one sensation that is denoted, and I certainly know which one it is that I am talking about. The fact that a sensation has a relatively short duration does not in itself keep it from having a proper name. Nothing which bears such a name lasts forever, and there is no minimum time that a baby, for example, must live in order to have a name. Moreover, the fact that sensations are private does not make them ineligible as bearers of proper names, at least in the context of the system being discussed. For the solipsist all terms which refer to anything ultimately refer to private entities. Thus, it follows that proper names which refer to material objects refer ultimately to one's own sensations. In fact, we saw in the last chapter how to translate material object statements containing proper names into the primitive language of our quasi-system. Hence, there is no a priori reason for thinking that we cannot apply proper names to sensations. However, it will turn out that we can do this only under certain very restricted circumstances. Let us therefore look more closely at the ordinary use of proper names in order to see what these circumstances are.

Since a proper name denotes a particular object rather than a kind of object, the main pre-condition of its use was that we be able to distin­guish the object denoted from other similar objects belonging to the same class. For example, when a proper name is used to refer to a person, we can pick out the person concerned even if he is with other persons of similar appearance, manners, etc. In the physical world this may ultimately have to be done by tracing a person's spatio-temporal course back to a baptismal situation. If, for instance, two children on a playground look very much alike, we can discover which one is Johnny Jones by discovering where he lives, who his parents are, and so on. Physical objects which have proper names are ultimately identified in similar ways. However, when we come to name sensations, the situa­tion is different. In our discussion of private languages we argued that one can refer to sensations which have occurred in the past, and that one can verify to some degree statements asserting that one had one kind of sensation as opposed to another at such-and-such a time. But it does not follow from the fact that one can talk sensibly about the kinds of sensations one has had in the past that one can apply proper names to them. One can uniquely describe a past sensation with such a phrase as "the sensation I had when I fell down my front steps," but even this does not mean that a proper name can be legitimately applied to that sensation. It would be like saying "I name the first person to

Page 287: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

STATEMENTS ABOUT THE PAST 281

board TWA Flight 348 on Dec. 28, 1974 'George Adderly', whoever he may turn out to be." In both cases we can uniquely identify the object named, but there is something more involved in naming than this. The something more is closely tied in with the use of indicator words such as "this" and "here," and with the act of pointing at things. It not only involves picking an object out from similar ones, but picking it out in a particular way. One does not pick the object out by specify­ing the relations that it bears to other things, but by taking note of the most important properties that the object itself possesses.

We could conceivably have a language in which all the names were introduced in a relational way, and one might then be able to make certain analytically true statements relating the named objects. How­ever, one would never know what the names named, until some objects central to the system of names were identified in terms of their proper­ties. In this language definite descriptions might serve that purpose and might consequently serve to anchor the system of names. But this is very definitely not the role that names have in our ordinary language. In our ordinary language proper names do not need to be anchored, and, in fact, we often identify new objects by relating them to objects which bear proper names.

We saw before that the user ofa proper name must know how to pick out the object named. We can now go further and say that the person giving a proper name must be aware of those features or properties of the object which distinguish it from other similar objects. In the case of the man boarding the future airplane flight we had available what we will call "tracing procedures" for picking out the person concerned. How­ever, we were not familar with any of the features which would be in any way essential to that man. Thus, we were not in an appropriate position to name him. Actually, of course, there may be only a fine line between properties and relations. Thus, we cannot say categorically that knowing the relation an object bears to other objects never puts us in a position to name that object. However, this is a fairly safe general rule, and, given any object, there will always be "core" features or properties which we will have to know the object to possess before being in a position to name it. Of course, we do not need to have this kind of knowledge about an object in order to refer to it by means of a definite description. There all that is needed is that we succeed in isolating the particular object in any way that we can.

Since we have not supposed that there is any absolute way of distin­guishing properties from relations, or distinguishing essential features

Page 288: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

282 SOLIPSISM PROPER

of an object from non-essential features, it will probably also be that there is no sharp line between the roles that definite descriptions and proper names have in ordinary language. The line becomes particular­ly difficult to draw when we consider, not the conditions under which it is appropriate to introduce a proper name, but the conditions under which previously introduced proper names can be properly used. Let us return to our earlier example where we had Johnny Jones on a play­ground amid a group of other children. We can properly say such things as "I am supposed to take Johnny Jones home" if we know which boy he is, and in what features his individuality lies. Even if someone cannot say for the moment which child Johnny Jones is, he can use such a sentence properly if he is aware of the essential features which distin­guish Johnny Jones from the others and can pick him out on closer examination of the group. He could then say correctly that he knows Johnny but did not recognize him at first. If, however, he has to ask each child its name until he comes to Johnny, no one would, in practice, ob­ject to the use of the proper name, but a definite description could have been used just as naturally. In this case the speaker could have said instead, "I am supposed to take home the boy named 'Johnny Jones'." If the speaker does not even know the name of the boy, but is merely supposed to take home the one with blue jeans, it would then be very odd for him to give the boy a proper name of his own for the occasion. In this way it becomes more and more natural to use a definite descrip­tion as it becomes less and less natural to use a proper name. Still, there is probably no place where we can draw a line separating the two usages.

Further, in ordinary English there are cases where proper names have been used over a very long period of time. In these cases it often happens that the users of the language are not familiar with the prin­cipal individuating features of the object named, and, in fact, they may not know how to pick the named objects out from other similar objects. For instance, most people who nowadays use the proper name "Mont­calm" know practically nothing about Montcalm except that he was the general who lost the battle of Quebec. In such a case they could use the definite description, "the general who lost the battle of Quebec" interchangeably with the proper name, and nothing would be lost. This is because most present-day users of the proper name would not be aware of those features of appearance, personality, etc. which distin­guished Montcalm from other persons. His associates would have been aware of these features, and they would have played a vital part in

Page 289: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

STATEMENTS ABOUT THE PAST 283

their use of the proper name. If Montcalm had been an ordinary per­son, his name would have fallen into disuse as people forgot what was characteristic about him since they would then have had no reason to talk about him at all. However, since he was a historical figure, the proper name has appeared in certain contexts ever since, and is still used even under conditions where a proper name would not ordinarily be used. Hence, if a proper name stays in use long enough after the disappearance of the object named, its role tends to relapse into that of a definite description.

In the case of sensations this happens much more quickly. We can use a proper name or constant in connection with a present sensation, but once it is past, we very quickly find it impossible to remember those features of the sensation which distinguish it from other similar sensa­tions. One is first tempted to say that all toothaches are pretty much the same, but on reflection one realizes that this is not true. Each pain will be intertwined with varying other sensations and tones of feeling depending on its location in the mouth, and there will always be various secondary pains, throbbings, or even tickles mixed in. However, these secondary elements are very quickly forgotten and it is virtually im­possible to say how any two toothaches we have had in the past differ from one another. In fact, those nuances of feeling which give a par­ticular toothache its individuality may be forgotten even a few moments after the pain has ended. We may, at some later time, be able to infer that we probably had a pain of one sort rather than another by re­constructing the causes of the toothache with the help of X-rays, etc. However, it is unlikely that we could do this with sufficient precision to allow us to distinguish that particular toothache from similar ones. Further, even if we could, we would be using a proper name in an ex­tended way if we were to apply it to that reconstructed toothache. It would again be somewhat like naming the first passenger to board some remote airplane. Even if we can, by some chain of inferences, say rough­ly what he is like, or what the past toothache was like, we would not be familiar with the relevant features in a firsthand way when using the proper name. Thus, if we use proper names in the strictest possible way and disallow their extended usages, we will only be able to apply proper names to sensations which are either present or which lie within the very recent past.

We will now introduce constants into the primitive language of our quasi-system with the understanding that they will function as proper names used in this very strict way. Since there is no sharp line between

Page 290: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

284 SOLIPSISM PROPER

what we have called the "strict" use of proper names and their extend­ed or vestigial uses, we cannot say exactly how strict this usage will be without begging the question by saying that proper names or constants are applicable only to present sensations. However, from the examples that we have given, it should be relatively clear what sort of role we have in mind for proper names when used strictly. It will then be a consequence of this usage, as applied to sensations, that proper names will denote either present sensations or sensations that lie within what some philosophers have called the "specious present." Again, it is not possible to say exactly how much ti~ne the specious present will take in. Nevertheless, it is clear that we will have no problem in locating sensa­tions temporally within the specious present. However much time it takes in, all the sensations that lie within it will be directly remembered, and the primitive relation of "precedes" can be directly applied to any two sensations that lie within this specious present if one does come before the other. Thus, we will be able to set up the temporal order of sensations within the specious present, and we will dismiss as useless the question of what part of the specious present is really present. Such a question would be analogous to asking whether a book I am holding is as "here" as the hot dog I have just eaten.

It is also a consequence of this restricted use of proper names in connection with sensations that we will not be able to use a constant to "refer back" to a sensation that has occurred even a fairly short time in the past. We can always refer to a sensation of such-and-such a kind that occurred in the past, but individual constants will be almost use­less in this respect because we will always have to bring in new con­stants to refer to recurring sensations even of the same sort. While these constants will therefore be of little practical use, they will be very useful indeed from a theoretical point of view, as will soon be shown. Their lack of practical usefulness is immaterial since the system as a whole is designed for purely theoretical purposes.

As Ayer points out, it is almost as important a feature of sensations that they are momentary as it is that they are private. The fact is that they become blurred almost immediately after they occur, and our own past sensations are lost to us almost as thoroughly in practice as are other people's sensations. Hence there is a very strong connection be­tween the presentness of a sensation and our being aware of its indivi­dual features. Consequently, any sensation referred to by a constant in our language must lie within the specious present, and we therefore have no need to say of such a sensation that it is present. It already has

Page 291: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

STATEMENTS ABOUT THE PAST 285

to be present to occupy that position in our language, and we therefore have no need for such an indicator word as "now" in our primitive no­tation. It will be recalled that "now" did appear as a yet-to-be-defined term in the second group of axioms for our quasi-system. We will shortly say exactly how it can be eliminated in favor of purely primitive no­tation.

\Ve have now seen that we can talk about present sensations in such a way as to indicate that they are present. Next, we want to discover a way of saying, in the language of our quasi-system, that a sensation is past and a way of saying that a sensation is future. It should be remembered that we have in the language of our system the primitive relation of "precedes", and this will be used as one of our building blocks. For example, suppose that we want to say that a sensation of x sort has occurred in the past. To this end we might say that it precedes some sensation named by a constant, which latter would, of necessity, be present. However, this is not sufficient. The "precedes" relation is only a phenomenal relation which holds between sensations which can both be dearly remembered. Like the other primitive relational predicates, it has to be taught by example, and can be used only when one is directly aware of a temporal difference in the way that one might be aware of a color difference between two sensations. On the other hand, we want to build up to the point where we can say that a sensation of a particular kind occurred in the past even if it occurred in the remote past, and even if one can no longer remember the sensation, much less compare it directly to a present one.

In order to say that a sensation ofx kind occurred in the past, we will first say, of course, that it occurs (it will be remembered that the use of the present tense here is arbitrary and does not signify anything), and we will then use another primitive relational predicate to say that it is not simultaneous with some sensation named by a constant. Finally, we want to deny that there is a chain of sensations starting with the present one denoted by a constant and terminating with the one asserted to be past, such that each member of the chain precedes (temporally) the succeeding one. In this way we will be asserting that the sensation occurs but will be denying that it is either present or future; it will therefore have to be past. Unfortunately, this is not quite enough. It might be that in one's sensuous experience there are gaps between sensations which are sufficiently great so that the last sensation before the gap does not precede (again in this special sense) the first sensation after the gap. In such a case the sensation could lie in the future and there could still

Page 292: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

286 SOLIPSISM PROPER

fail to be the kind of chain described above. Nevertheless, it seems that even if there are such gaps in our C0nscious experience (images will do as well here as sensations) it will always be that at any time one would have sensations under certain circumstances. Even if one is asleep, there is always something that can be done to produce a sensation or image of some sort, be it a waking experience or a dream experience. Further, it does not matter for our purposes what kind of sensation or image occurs. All it must do is enter into the "precedes" relation with other mental phenomena. Thus we want to amend the condition stated above so as to say that under no circumstances would there be a chain of mental phenomena starting with the one denoted by a constant and ending with a sensation ofx kind such that each "precedes" the next one in the sequence. Of course, in our language a "circumstance" is represented by the truth of some set of hypothetical counterfactual statements. When we say that "under no circumstances will there be such a chain of sensations", we are actually quantifying over these circumstances, and hence over counterfactual statements. Thus, we are operating in the metalanguage as opposed to the language of the quasi-system. Nevertheless, we saw before that it is theoretically possible to collapse the metalanguage into the object language and express all the meta­linguistic concepts within the object language itself. Hence, it is seen that we can use metalinguistic notions in analyzing the concept of pastness, and we have therefore indicated a way of translating state­ments about the past, at least of this sort, into the language of the quasi­system itself. The actual mechanics of the translation would be incredi­bly complex, but since the way is clear and our system is only quasi­axiomatic anyway, it does not seem necessary to go into further detail. We now have a way of saying that a sensation is present, and a way of saying that a sensation of any given kind is past; we can then say that a sensation is future just by denying that it is past or present.

Now that our basic temporal mechanism has been set up, we will consider some objections by Arthur Danto to the phenomenalist treat­ment of time. He has in mind the traditional phenomenalist analysis where the translation is comprised entirely of counterfactuals which are tenseless in the sense that they can refer equally to the past, the present, or the future. Danto then argues that the tensed statements of ordinary language cannot be translated into such a tenseless phenomenalist idiom. He says:l

It is not easy to see how we should render tenses in experiential terms. One might

1 A. C. Danto, Analytical Philosophy of History, pp. 51-52.

Page 293: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

STATEMENTS ABOUT THE PAST 287

of course, propose some such stratagem as this: giving an experiential equivalent for moving through time. Thus: we arrive at 44 B. C by traversing a series of event­stages, and each of these can be rendered phenomenalistically. True, we cannot occupy these positions, but it is possible that we might have, for the reasons already considered. The difficulty, however, would be in making thejirst step between here and 44 B. C. For the first step must be to an event which is past if our trip is to be in the right direction, and the question is how we indicate that the first step is in the direction of the past, or differs from a first step in a future time journey. One could say: the first step in the direction of 44 B. C. But then we have to express somehow that 44 B. C. is in the past, and this then begs the question. Certainly you cannot hope to do the trick in a tenseless idiom. For suppose we say that 44 B. C. is 2007 years before now. But "now" indicates the use of a present tense, and would have to be replaced with a date, that isA. D. 1963. We might then say that the statement that 44 B. C. is 2007 years before A. D. 1963 is true, and, for that matter, analytic­ally true. But this does not tell us that 44 B. C. is past. For someone could have uttered this truism at any time, including 43 B. C., when the years referred to were future. We have to know when the sentence is uttered, and then whether this time is before or after or concurrent with the time at which we raise the question. So we cannot readily eliminate the sort of information tense gives us. But then, if we can­not incorporate this information into our phenomenalistic translations, phenomen­alismco llapses as a programme for expressing all that is meaningful in our ord­inary language.

Our answer to Danto's objection can be stated fairly easily in terms of the account given above. He talks of the phenomenalist giving an experiential account of moving through time, and, while we would not want to talk this way, our primitive relational predicate of "precedes" does much the same job. We then defined another relation of temporal priority which is more general and which can hold between events that are far enough separated in time so that we cannot apply a phenomenal relation to both at once. Danto then argues that without tenses one cannot talk about "now", and that one consequently has no way of saying that an event is either present or past. Our answer here is, of course, that whenever we use a constant we are, of necessity, referring to a present sensation, and that we can talk about past sensations as past by saying that they are temporally prior to such a sensation. We can refer to such a present sensation in all kinds of contexts, and such refer­ence either presupposes or entails (depending upon whether we prefer a Strawsonian or Russellian account) the occurrence of such a sen­sation. There will always be such a sensation, or at least an image, since uttering a sentence, or even thinking it to oneself, involves the oc­currence of a mental phenomenon of some sort. It turns out that the assertion of the occurrence of a present sensation differs in logical form from the assertion of the occurrence of a past or future sensation, and, indeed, from all other assertions. To say that a present sensation, named by a constant, is occurring is just to say "kn occurs" where kn is the

Page 294: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

288 SOLIPSISM PROPER

constant in question. This is a categorical statement, and a logically simple one. In fact, all such logically simple statements refer to our own present sensations. The difference between talking about one's own sensations and those of other persons is also a difference oflogical form, although, as we will see in the next chapter, it is a more subtle differ­ence. Further, any statement about the physical world will be analyzed in a hypothetical way, so no categorical statement such as this could be about the physical world. Moreover, we saw before that assertions that sensations of such-and-such a kind have occurred or will occur involve not only a categorical statement of the occurrence of a present sensation, but also other very complex components which talk about chains of sensations in a hypothetical way. Thus, any statement about past or future sensations will be logically complex and will contain hypothetical components. Of course, we may want to say not just that kn occurs, but that it is of such-and-such a kind. In that case we still need say only "kn occurs and kn B ••• " where the number of dots represents the class in question. Again, we have no hypothetical component. Thus, it is characteristic of all and only statements about our own present sen­sations that they contain no hypothetical components whatever.

The second group of axioms for our quasi-system was constituted by such statements as " ... sensation occurs now". The term "now" was a defined term, and those axioms can now be replaced by statements such as "k6 occurs and k6 B ••• ". In this way those axioms can be put into primitive notation.

It hasjust been seen that we have a way of talking about the past and the present in phenomenalist terminology; moreover, this has been done without making explicit use of tenses. Of course, given any language in which we want to make a distinction between the past and the present, there has to be some convention for talking about the one as opposed to the other, at least in certain contexts or with respect to certain kinds of objects. In our language we have decided to use proper names very restrictively, and this has the consequence that when we use proper names to refer to sensations, they have to refer to present ones. At this point one begins to wonder whether there is any great difference between adopting a convention of this sort and adopting a convention which involves the use of tenses. At least, the whole question seems merely to hinge on maximum convenience and economy of expression. It consequently makes relatively little metaphysical difference whether one ultimately winds up with a tenseless language. The important point to notice is that in our language the difference between the past and the

Page 295: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

STATEMENTS ABOUT THE PAST 289

present is traceable ultimately to this usage of proper names and to the use of the primitive relational predicate of "precedes". If one wants to know what ontological commitment is involved in speaking about the past, our answer is that to speak about the past we have to be in a position to learn and use this primitive relational predicate, and to learn and use proper names in the way prescribed.

Up to this point we have been talking about temporal relations which hold between sensations. However, we will in the end want to talk about temporal relations that hold between physical events, and we must now see what is involved in bridging this gap. But before we translate tensed statements about physical events into the language of our quasi-system, there are some problems which will arise on any theory, and which concern the dating of a physical event and the determination of its temporal boundaries.

At times C. I. Lewis said that physical events are extended sufficient­ly far in time to include within their time span all the evidence which anyone might later find for the occurrence of those events. Thus, the physical event of a dinosaur walking across a rock would not end when the dinosaur has crossed the rock, or even when he is dead. Rather, it would extend even to include the present-day scratches on the rock which would now offer evidence for the dinosaur having walked across it. Danto criticizes this position as follows:1

Thus the Battle of Hastings, plus the Bayeux Tapestry, plus all the other effects of the Battle of Hastings, go to form a single time-spread object. Let this be O. And since, as he has said, "at every date after the happening of an event, there is always something ... possible of experience ... ," there is at this very moment something, which is an effect of the Battle of Hastings, which it is temporally possible for me to experience. Hence I can experience O. To be sure, I could just call 0 "The Battle of Hastings" and so say that I can experience the Battle of Hastings. But it would surely startle students of English history to learn that the Battle of Hastings is still going on.

Surely, Danto's objection is a good one, but there is still some point in Lewis' temporally extended objects. The point is that most events have ever-widening circles of consequences. There is no general criterion which will tell us where to draw the line between the event itself on one hand and the consequences on the other. Lewis seems to have gone from this acknowledged fact to the conclusion that any demarcation would be arbitrary and should not be made. He therefore extends the object to include the consequences. Danto's contrary position rests on the fact that in ordinary language we certainly do make a distinction

1 Ibid., p. 38.

Page 296: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

290 SOLIPSISM PROPER

between the event proper or "core event" and the "extended object". We will argue here that even though there is no perfectly general criterion which tells us how to make such a distinction, it is nevertheless not an arbitrary distinction. The fact is that we distinguish between the core event and the extended object in different ways when we are dealing with different kinds of events and objects. Still, given any kind of event or object, we may make the distinction in a fairly systematic and regular way.

As an example, let us consider the class of persons and speak of them from a phenomenalist point of view. We count sensations which are characteristic of perceiving the person's bodily movements or speech as sensations of the core object. Most other sensations relating to the person would count only as sensations of the extended object. Thus, seeing a letter that the person has written or seeing afterwards the mess that he made when trying to hang wallpaper are not cases of seeing the person; they are rather cases of seeing something that he has made or left be­hind. There may be some exceptions to rules such as this. For instance, we would not count it as a case of seeing the person if he is dead and someone is moving his arm up and down. Thus, seeing the person move may not be a sufficient condition for seeing the person, and we might have to add the condition that he be alive. Perhaps a large number of similar additional qualifications would have to be made, but it is clear that the distinction between seeing the object or event proper and merely seeing something which relates to it is not an arbitrary one. The rules which tell us when we are perceiving a person as opposed to the effects he leaves behind would probably also apply to animals, but would not apply to most other sorts of events or objects. For example, seeing scratches on rocks is not a case of seeing a dinosaur move, but we can also talk about the scratches left by dinosaurs instead of talking about the dinosaurs. In this case, the same sensations would count as cases of perceiving the scratches themselves. What belongs only to the extended object in one case belongs to the core object in the other case, and it is obvious that very different criteria are operating in the two cases.

Now that we have a distinction between the core event and the ex­tended object, we can say what it means for an event to be past. An event is past if and only if the core event is past. But the core event is past if and only if sensations of that core event would have been past if they had occurred, or were past if they did occur. We already have a way of saying that any given actual or possible sensation is past, and so

Page 297: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

STATEMENTS ABOUT THE PAST 291

we now have the wherewithal to say of any physical event that it is past. We will shortly explain the analysis of statements about past events in greater detail. It can already be seen, however, that we will have to know which sensations are sensations of the core event and which ones are only of the extended object before we can construct our analysis. This is because it is only in the former case that we can and must in­clude conditions which will insure the pastness of the sensations con­cerned. This means that the rules which tell us how to construct the phenomenalist translation of statements about the past will vary when we are concerned with varying sorts of past events. However, this is nothing new to our system, and we have never claimed to be giving rules which would tell someone in a purely mechanical way how to translate any English sentence into the language of our quasi-system. We have insisted only that it can be done in any particular case, and we have given rules which are designed to solve the more difficult problems that would be encountered. In this case there is no very seri­ous problem since anyone who knows the English language is able to say within fairly narrow limits what counts as the core event for a given kind of event.

Suppose now that we are analyzing the material object statement, "Nelson died at Trafalgar". Most of the counterfactuals belonging to the analysis will be concerned with what one would have perceived had one been at Trafalgar and been in a suitable position to observe Nelson's unfortunate experiences. The antecedents of these counterfactual state­ments will describe these necessary conditions in sensation terminology. Since this is a statement about the past, as indicated by the past tense, one of the conditions which these location sensations would have to ful­fil is that they be past sensations. Consequently, we would build into the description of each of these sensations the condition that there is no chain of mental phenomena, of the sort described above, leading from a present sensation named by a constant to the sensation which is sup­posed to be past. We would not have to build these temporal conditions into the consequents of the counterfactuals since they already contain an assertion of simultaneity with the location sensations, and this will insure that they would be past as well.

All this has the effect of insuring that the Battle of Trafalgar is not a future event. If it were, I would still be having the same sensation I am hav­ing at present, but there would be the sorts of chains the existence of which is being denied. But have we eliminated the possibility that the battle is occurring right now? If we suppose ourselves not to be right there

Page 298: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

292 SOLIPSISM PROPER

watching it, then we would have to get into position to observe it, and the relevant observations would have to take place at least in the very near future. Hence that possibility is also excluded since the relevant chains could occur in these circumstances. But suppose that the battle is now taking place and that I am aboard the Victory watching Nelson at the very moment that he is departing this world. In that case I might not be having the sensation I am in fact now having (if the sensation is not somatic, it would almost certainly be replaced by another one) and which is named by a constant. Hence, the sensations I would be having would be non-simultaneous with my present one. Further, there would fail to exist the kinds of chains mentioned above. Hence, this possibility still remains open and we must add another component to the analysis in order to exclude it. This is easily done since, if! were observing Nel­son in his death throes, I would be having the sensations mentioned in the antecedents of the relevant counterfactuals. We can therefore insure that I am not observing Nelson's death by asserting that that particular conjunction of sensations is not occurring. We have now excluded the future and present possibilities, and so the translation asserts that Nel­son died in the past.

Besides the counterfactuals asserting what would have been perceived at Trafalgar, there would probably be some counterfactuals dealing with present evidence, and here, of course, these temporal conditions would be omitted. Thus, since we can use constants to refer to present sen­sations, there is no great difficulty in adapting our analysis of material object statements so as to take account of past tenses.

It can now be seen clearly that our analysis is not subject to another of the objections which Danto uses against Lewis. Danto interprets Lewis as saying that statements about the past will always be reducible to predictions about the future. In particular, these predictions will always concern evidence that could be found in the future for the past event. This, of course, has some very odd consequences. However, on our account, a statement about the past will, like all other statements about the physical world, be analyzed in terms of counterfactual state­ments. But a counterfactual statement is not a prediction about the future. As we noted before, it itself is quite timeless and can refer to something that might have happened, but did not happen, in either the past or the present, or to something that might happen in the future. Ifwe want to make a counterfactual refer to the past we simply add the appropriate components to its antecedent as indicated above. It is just because counterfactuals are in themselves untensed that we can make

Page 299: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

STATEMENTS ABOUT THE PAST 293

them refer to either the past, the present, or the future, as we may choose.

There is still another objection which Danto aims at pragmatic theories with respect to their handling of statements about the past, but which the present theory eludes. This as follows: 1

One reason why a reasonable man might wish to retreat from this position, his taste for paradox notwithstanding, is that it entails a revision of what a sentence like "The Battle of Hastings took place in A. D. 1066" means each time that senence is verified. That is to say, most of us might allow that in some sense "The Battle of Hastings will take place in 1066" differs in meaning from "The Battle of Hastings did take place in 1066." Perhaps we would say this because the former could have been, while the latter cannot be, verified by experiencing the Battle of Hastings (though in fact few of us would give this as a reason for saying they differ in meaning). But who would wish to say that "The Battle of Hastings took place in 1066" differs in meaning from "The Battle of Hastings took place in 1066?'' Yet this is what we might be committed to say on the original verificationist analysis: the sentence changes its meaning each time it is verified. For suppose it is taken at one time as prediction that a certain experience will be had, and this experience is had. Then it can no longer predict that experience, but another one and so its meaning changes. We can give in to our prejudice that it has always the same meaning, only by the artificial means of using it to predict an experience to take place after the absolutely final utterance of the sentence. But in many cases it is too late for that. Thus "Caesar died" no longer means what it once did, partly as a consequence of the meddling inquests of Marc Anthony. So the theory induces a radical instability in the meaning of most sentences about the past, or at least all of them which ever have been verified. Indeed, in a kind of Heraclitian way, we could never verify the same sen­tence twice. We should always, instead, be verifying a different sentence, in case difference of meaning means difference of sentence. And this would entail that "Caesar died" and "Caesar died" are not the same sentence in the case where one of them has been verified. Yet we surely want to say that these are both statements of the same sentence, and that this same sentence has always the same meaning. Nor would it help much to say that these are different uses of the same sentence to make different statements. For these different statements could never mean the same if one of them were ever in fact verified, or if they were verified by different experiences.

Now, we have argued that certain sentences do change meaning under certain rather unusual conditions of verification. We have also argued that the meaning of a sentence may change as time passes and people find new ways of verifying it, or if they change their views as to the nature of the objects involved - that is, if the concept of the object changes. However, we have nowhere said that the meaning of a sentence automatically changes as time passes, or that its meaning changes every time it is verified. Suppose that a counterfactual statement belonging to the analysis of a material object statement (either one about the past or not) is verified by my having the sensations specified. There is then no reason to remove that statement from the analysis as Danto seems to

1 Ibid., pp. 45-46.

Page 300: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

294 SOLIPSISM PROPER

think. We may now be convinced that the statement is true, but that does not imply that it is no longer what we mean by the material object statement. We would even allow the theoretical possibility that all the counterfactuals belonging to such an analysis be verified. If we were to delete each counterfactual as it is verified, we would then have emptied the material object statement of all factual content. However, it is quite clear that the meaning of such a sentence could remain constant throughout the process of verification, and that verifying a sentence does not automatically change its meaning. The only possible justifi­cation for the view that verification always changes the meaning of a sentence would be that the sentence must always somehow imply only predictions about the future. However, we have just seen that it is not part of our view to convert statements about the past into statements about the future.

Let us now consider still another position concerning the analysis of tensed statements which is similar to ours but which also runs into trouble that we can avoid. Danto attributes this position to Ayer, al­though it is not certain that Ayer would have subscribed to the view exactly as it is stated by Danto. In any case, the position is that any tensed statement can be broken down into a conjunction of two logically independent tenseless statements. One conjunct says something about the event concerned and the other conjunct says something about the temporal position of the speaker relative to that event. Thus, "Caesar died" can be analyzed as a conjunction of "Caesar dies (where 'dies' is tenseless)" and "the death of Caesar is prior to the date of utterance." This position is like ours to the extent that a tensed statement is analyzed in terms of tenseless statements. Our main disclaimer will be that the analysis need not be this simple, and, in particular, that it need not have this logical form. Let us now see how Danto argues against it.

Danto first points out that tensed statements depend upon the time of their utterance for their truth value while tenseless statements do not. He then says: "It follows, then, either that we cannot give a tenseless rendering of tensed sentences, or that some tenseless sentences very much depend, for their truth value, upon the time of their utterance."l It is, of course, true that a tensed sentence is never logically equivalent to the corresponding tenseless sentence where we simply neutralize the tense of the verb in some way or other. But Danto is saying much more than this: he is saying that no tensed sentence can ever be logically equivalent to any logical construction of tenseless sentences whatever.

1 Ibid., p. 56.

Page 301: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

STATEMENTS ABOUT THE PAST 295

As it turns out, the difficulty of translating a tensed sentence into a phenomenalist language is very similar to the difficulty of translating a sentence which contains an indicator word such as "this." In the latter case one has to find out what the "this" refers to in a particular context before one can give a translation. This constitutes no difficulty for us since we have consistently held that the meaning of particular utter­ances precedes the meaning of sentence types. Consequently, we will always be translating what someone says on some particular occasion in some particular context. The point, then, is simply that we cannot, and should not be able to, translate sentences into our phenomenalist language until we know the context in which the sentence is uttered. Much the same thing is true for tensed sentences. "Caesar died" can be uttered at any time, and it may have different implications depending upon the time of its utterance. Thus, one cannot give a complete analy­sis of such a sentence until we know when it was uttered in just the way that we cannot give a complete analysis of "This chair is red" until we know where this sentence is uttered. But, once again, there is no theoreti­cal difficulty about the analysis of a tensed sentence as uttered on a particular occasion.

Danto then goes on to point out what I think are more important difficulties in Ayer's position. He here argues that if the timeless com­ponent in Ayer's analysis (in the previous example, "Caesar dies") is false then the event in question (Caesar's death) would never have occurred. But, he continues, if the event never occurred, then it can never bear any temporal relationship to the present (or, Danto ap­parently thinks, to any other time). He then concludes that the two components of Ayer's analysis are not logically independent as alleged. Rather, Danto says, the second component is dependent on the truth of the first.

In our analysis we have also spoken of temporal relations involving events which have never actually occurred. In our case, of course, these events are sensations which would have occurred under certain circum­stances, but which may not have occurred. In the antecedents of our counterfactuals we have laid down very strict conditions which sen­sations must satisfy if they are going to count as instances. There seems a priori to be no reason why we should not insist that they satisfy certain temporal conditions along with the other necessary conditions. Further, given our analysis of what it is for a sensation to be past, there is nothing to prevent us from speaking of the possibility that a sensation which did not occur might have occurred, and that it might have fulfilled the

Page 302: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

296 SOLIPSISM PROPER

conditions there specified. Hence, we can talk about sensations which might have occurred in the past, but did not.

Nevertheless, Danto asks rhetorically "How can I stand in any temporal relationship with a non-existent event?" (p. 57). On in­spection, it can be seen quite clearly that one can stand in temporal relationships with non-existent events. For example, I might have had a glass of beer yesterday at lunch even though I did not, but it is obvi­ous that this event, which did not occur, lies in the past as opposed to the present or future. The relationship here is probably not the same as that between a past event that did occur and my present self. But even though our view commits us to saying only that such an event would have been past if it had occurred, there seems to be no harm in saying directly that certain possible events are past events. If the past event did not occur, more counterfactuals are going to be involved in saying that it is past, but the same relationship of pastness (as applied to sen­sations) will ultimately be involved whether or not the event actually occurred. Thus, one of the premises on which Danto bases his objection is false, but it still seems that a valid objection can be made along these general lines.

The main point is that, while we can talk about the temporal re­lations we bear to non-existent events, we do this only as long as we are speaking of these events in a counterfactual way as events which would have occurred if .. " or as possible events which might have occurred, or as fictional events. But the statement "Caesar dies" is not counter­factual, nor is there any suggestion that we are talking about a possible event or a fictional event. If this statement is false, then Caesar is still alive, and, since it is a tenseless statement, Caesar will be eternal. There is then no such actual event for us to talk about, and while we could still talk about any number of possible deaths that Caesar might have had, the sentence gives us no hint as to which of these possible deaths is in question, and which one it is that the second component is asserting to be past. The trouble with Ayer's position is that the first component is supposed to pick out an event which the second component then asserts to be past. But suppose that the first component states some putative facts which have to be the case in order for us to know what event is being talked about. Suppose, for example, that we start with the sentence, "There was a day on which exactly six gray cats died," which would come out as "There is a day on which exactly six gray cats die" and "That day is prior to the date of utterance." If the first component is false and there is never a day on which exactly six gray cats die, then

Page 303: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

STATEMENTS ABOUT THE PAST 297

the event in question has been insufficiently identified. No actual event has been picked out and we are not told enough about any possible or fictional event to be able to relate it temporally to other events. In this case the second component does not make sense because we do not know what actual or possible event it is that is supposed to be prior to the date of utterance. Similarly, if Caesar never dies, we do not know what event it is that is supposed to be prior to the date of utterance in the other example.

Of course, if the event is a possible one, it may be indeterminate with respect to date of occurrence in the way that it may be indeterminate with respect to color or exact shape. If we want to talk about the castle Mozart might have built if he had had the money, and do not specify its color, we cannot then compare it in color to certain existing castles. Similarly if Caesar is eternal and we want to talk about his possible death we have to say something about that death which lets us locate it temporally before we can say whether that date is prior to the date of utterance. This condition would be satisfied if we were to talk about his possible death at the hands of Brutus, or at the hands of Napoleon, etc. There is no difficulty in giving such a possible event a determinate temporal position which will allow us to talk about its temporal re­lations. Most of the possible sensations which we talk about in the antecedents of our counterfactuals are temporally located in this way. But, on the other hand, we can talk about a possible event without so locating it. Thus, it seems that there is something wrong with the view that a tensed sentence can be rendered in the way that Ayer suggests. Such an analysis is far too simple.

It will be remembered that in our analysis many counterfactuals are involved and that, within each counterfactual, the main connective will be quantification over the counterfactual "if... then." In fact, the conjunctions between the temporal and non-temporal conditions that sensations must satisfy will be buried within the antecedents of these counterfactuals. Consequently, our analysis will have a totally different logical form from that of Ayer.

We have now sketched out our analysis of statements both about past sensations and about past physical events. In order to tie all this more closely to our theory of meaning, we will now see how the state­ments, "A sensation of x kind is now occurring" and "A sensation of x kind has occurred," differ in meaning according to that theory. Let us consider the former statement first and suppose that there is no difficulty in imagining a sensation of the required sort. Such an image would,

Page 304: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

298 SOLIPSISM PROPER

however, only be a partial instance of the associated preliminary because it could as easily be an instance of a preliminary which refers to a past or future sensation. Thus, for that image to be an image of a present sen­sation it has to be simultaneous with some present sensation which can be named by a constant. We have in mind here some sensation, proba­bly somatic and probably having to do with a body tone, which would not be replaced by a sensation of x sort but which would occur along with it. This condition would be satisfied if the image actually contained that other sensation as a part of itself. However, this notion of an image containing a sensation is a new concept which must now be investigated. Actually, this phenomenon is not an all unusual. Suppose that I have a raging toothache and I imagine that I am in the dentist's chair, and that he is going at my tooth with his various drills, augers, etc. Insuch a case I would certainly be having an image, but part of that image, the pain of the toothache, would actually be a sensation. In fact, it is quite likely that a great many, or perhaps most, of the images we have contain sensations, particularly of the somatic kind, as parts. This is also in line with what we said earlier about the difficulty of drawing a sharp line between sensations and images. It may be that in order to have a "pure" image with no sensation elements at all, we have to make an effort of abstraction in the way that we might have to make a special effort in order to have a purely visual image that has no tactual or other elements mixed in. Thus, it is quite possible to imagine instances in such a way that these images will be simultaneous with a present sensation.

However, such an image would not be any kind of instance of the latter sentence where an image of a past sensation would be required. Again, an image which is unconnected with any present sensation and which is ofx kind will be a partial instance of the preliminary associated with this latter sentence: this is because such an image would be neutral as regards past, present, or future. Imagining a complete instance here would be much more difficult. According to our previous analysis, we would have to imagine not only an instance of x, but also a chain of sensations starting from the one originally imagined and ending with an image which is simultaneous with a present sensation such that each precedes the next one in the chain. Again, the present sensation in­volved would probably be of a body tone and would be unaffected (relative to our counterfactual assumptions) by a change in our previ­ous sensuous history. This would, nevertheless, be very difficult to imagine, particularly if the past sensation is supposed to lie very far back in time. Still, there is again no theoretical impossibility in such a

Page 305: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

STATEMENTS ABOUT THE PAST 299

project of imagination. At each step we would know how to proceed next, and it is only a question of how long we could keep going before tiring of the project. It can now be seen that there is nothing about our theory of meaning which militates against an adequate analysis of statements about the past.

Page 306: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 14

FURTHER DEVELOPMENT OF THE SYSTEM: OTHER MINDS

The next move in developing our system will be to sketch out a way of translating statements about other minds into the primitive language of the system. This amounts to doing for statements about other minds what was done for statements about material objects in our chapters on phenomenalism, but it may antecedently appear to be more difficult since the language of the system is one in which all the key terms refer to one's own sensations as opposed to the sensations and feelings of others. As it turns out, this translation is no more difficult from a theoretical point of view but the complexity of analysis becomes even more staggering, and the directions given for the translation of state­ments about other minds into the system will have to be even more meager. In fact, we will deal only with the statement, "Jones is in pain" and the reader will have to see for himself how to alter the analysis to cope with any of the great number of other statements that could be made about the mental states of other persons.

In theories of other minds there are several difficulties which are often met with, and, contrary to what one might think, the solipsist turns out to be in a favorable position to avoid them. In the first place, it is undesirable to interpret statements about other minds in such a way that they could not conceivably be verified. If this were the case it would imply that these statements have a very different status from such other commonsensical statements as those about material objects and those about one's own feelings. Whenever we make distinctions between the factual statements of common sense and ethical state­ments, or between factual statements and performative statements, statements about other minds seem to belong with the factual state­ments in that they impart information and they do not have some special function which is not primarily informative, as do the other sorts of statements I have mentioned. As regards verifiability, statements about other minds seem to have about the same status as statements about the past. Neither can be verified directly in practice, but we feel that there are events which make them more or less plausible.

Page 307: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

OTHER MINDS 301

This leads some philosophers to give an analysis of statements such as these solely in terms of the actual and possible behavior of the other person. However, it is a consequence of these analyses that it will be logically impossible that another person should have some feeling, thought, etc. without at least having a disposition to show it by ap­propriate behavior under some circumstances. As an example let us take the case of a very small baby drinking milk. We may assume on the one hand that he has some sort of a biological tendency to drink milk, and that he would respond positively to certain sorts of tests. The question now is whether he is deriving pleasure from his milk on this particular occasion. The child is not smiling or showing any other signs of pleasure, but it may be that he never has displayed these signs; in any case, there does not seem to be any reason to suppose that pleasure in such a small baby is correlated with the same sorts of symptoms with which it is correlated in adults. It may also be the case that if the milk were suddenly removed the baby would cry (although this need not be so). However, we also know that whenever any of the baby's natural tendencies are thwarted he cries. This need not involve pain; it might be purely instinctive and even unconscious. Further, even if the baby would be pained ifhis milk were to be taken away, we cannot conclude that he is feeling pleasure at the moment. It even seems possible that babies of this age only feel pain and never pleasure. Thus nothing that does or might happen at the moment is inconsistent with the hypothesis that he is feeling pleasure. It also seems likely that the baby will have forgotten the whole episode in a very short time, so that there will be no conscious effect of it. Thus it will do no good to define the baby's pleasure in terms of an indefinitely long series of behavioral events, since all the evidence indicates that his future behavior will not be affected by his present pleasure or lack of it.

It is sometimes argued that events of this kind have behavioral effects which can be observed only by means of scientific procedures, either physiological or psychological. That is, physiologists may at some time discover a very close correlation between what we call pleasure and certain states of the brain, nervous system, etc. It is also possible that psychologists may discover an unconscious correlation between this sort of pleasure and future behavior. However, as far as our present knowledge goes, there may not be any such correlation in some cases, and even if one is discovered, the connection will not be a necessary one. It could be argued that even though certain particular effects of the pleasure or pain have not been manifested, the disposition for these

Page 308: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

302 SOLIPSISM PROPER

effects to occur still remains. However, the crucial point is that we can imagine a case where none of these events ever takes place under any circumstances; on the behaviorist analysis this should be a logical impossibility. Thus it will be incorrect to analyze statements about other minds in purely behavioral terms given the present meaning of the sentences involved. The solipsist will therefore look for another ap­proach to the problem.

It might again be objected that at the very outset we have begged the important question. Thus, it might be argued that before we can set up a language in which mental descriptions refer only to one's own mental states, we must have a criterion for distinguishing one's own mental states from those of other people, or a criterion of personal identity. While we have discussed this sort of objection before, let us again give our reasons for thinking that there can be a perfectly usable language in which there is no reference to persons at the primitive level. One basic fact is that one can never be mistaken as to whether a certain present sensation is one's own or someone else's. Hence the statement "I am in pain" as uttered by me can never be false, provided I am using the terms in the usual way, unless I am deliberately lying. It is, of course, possible to be mistaken about a past experience and to suppose oneself to have had a sensation which someone else in fact had. It seems to me to be implicit in some theories of personal identity that the meaning of "I" is determined by a large set of experiences which I have had as well as certain other characteristics. Thus if I were mistaken about nearly all my past experiences, supposed myself to be Napoleon, etc., my statements about my own present experiences could be false, since on my meaning of "I," I would in effect be asserting not that I am having a certain experience but that someone else is having it. On such a view the term "I" itself would already presuppose a criterion of personal identity. It seems to me that any view having this consequence is mistaken; however deluded I may be as to who I am, or as to my past history, this does not affect the meaning of first person reports of present sensations in the least. The "I" functions here not as a descriptive term, but as an indicator word. One might say that the notion of the present self is logically prior to the notion of the self that continues over time. Similarly, one can never be mistaken as to whether two sensations be­long to the same self as long as they are present. In that case they must both belong to one self. Since it is a convention of the language we are setting up to refer directly only to one's own sensations, the use of the terms "I" and "my" become superfluous at the primitive level. In

Page 309: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

OTHER MINDS 303

stating the analysis of statements about other minds we will continue to use these terms in the usual way, but it will be possible to eliminate them in favor of language which does not identify sensations as be­longing to particular people.

In developing the analysis I will first attempt to state necessary and sufficient conditions for the truth of the statement, "Jones is in pain," uttered on the occasion of Jones' being stuck with a pin, in terms of statements the speaker, who is observing Jones, might make about him­self and the physical world. We might start with the condition, "If I (the speaker) were in Jones' position, I would be in pain." Being in Jones' position is construed in such a way that we can describe it with­out making any reference to Jones' mental phenomena, but only to his body and his environment; thus, if the speaker were in Jones' position, he would, among other things, be stuck by the pin. It is highly probable then that if "Jones is in pain" is true, then "If I were in Jones' position, I would be in pain" is true as well, and vice versa. However, it is easy to imagine cases where one sentence would be true and the other false; this would happen if certain of my nerves had been damaged in such a way that even if I were stuck with a pin in the way that Jones is, I would still feel no pain. In fact it becomes clear that any bodily or environmental difference at all could conceivably cause one such state­ment to be true and the other not, so we will have to take a very strict interpretation of "being in Jones' position." One would have to be in exactly the same relationship to the world and one would have to have a body exactly similar to that of Jones. The question then naturally arises whether one would in fact be Jones. One would then be called Jones and there would be no one called by one's present name. How­ever, in order to satisfy the antecedent one need not be identical with the existingJones: one might be in Jones' position in the strict sense but be different from the present Jones due to having a different present sensation or mental experience of any kind. Hence "If I were in Jones' position (in a strict sense) I would be having his present experience" is never a tautology. The statement, "If I were Jones, I would be having his present experience," is, of course, a tautology; being Jones implies having all his experiences, including his present one. However, I am using the phrase, "being in Jones' position," in such a way that it does not include having his present experience, although, as we shall see, it implies many other things. If my argument against behaviorism is correct, someone's having a given experience will not follow logically from any other fact about him, and the first statement quoted cannot be

Page 310: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

304 SOLIPSISM PROPER

a tautology. The important point is that I know what it would be like to be fat instead of thin, to have a short nose instead of a long one, and to have a whole different set of bodily characteristics; one can even imagine oneself in the position of Kafka's salesman who wakes up with the body of a cockroach. This would also involve our having had different sensations from the ones we have had, but there is no difficulty in this. Anyone can think roughly how he would have felt if he had fallen down a flight of stairs, or one can imagine the feeling one would have had if one had won a game instead of having lost it. Thus, since it is conceivable that one should have had any of these characteristics and experiences separately, and since their co-presence violates no logical or semantical rules, it is then logically possible that one should have had all these contrary-to-fact attributes and experiences instead of the act­ualones.

It turns out that it is not even enough that one should have the same body and the same environment as Jones. If the speaker has been raised in an extremely stoical society and taught to ignore pain, he might not always feel pain when Jones does despite the presence of the other conditions. Hence, even if two people have exactly similar bodies and are placed in the same environment, their past experiences might cause them to have different present experiences. This would not be true if every experience left a distinctive bodily mark, but we do not want to presuppose that. We can now qualify our conditions further and say that any person would have had the same mental experience asJones at time t if he had an exactly similar body, had been placed in the same environment, and had had all the same experiences up to time t. Of course, this statement of necessary and sufficient conditions does not yet constitute an analysis because it talks about Jones' past experiences and would be circular. The counterfactual conditions would, if fulfilled, make the speaker identical with Jones up to a certain point in time, un­less one supposed him to inhabit another universe which might be qualitatively exactly similar to this one; in any case, there is no logical impossibility in one's having had the experiences and body which would cause one to be called ''Jones'' as long as one doesn't have one's actual characteristics at the same time. It is logically possible that I should have any of Jones' properties or experiences singly, and since Jones exists there is clearly no incompatibility between them.

We have seen that Jones will have a given experience at time t + I if and only if I would have an experience of the same kind at t + 1 provided that I had had the same experiences through time t and the

Page 311: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

OTHER MINDS 305

other conditions had been fulfilled. This presupposes only that like causes will produce like effects, and that principle follows from our inductive assumptions which we placed in the axioms. We now have part of an analysis since we are analyzing a statement about Jones' present mental state into statements about the physical world and statements about his past mental states. These statements about his past mental states can be analyzed in the same way into statements about other mental events which lie still further back in Jones' past. Thus we will eventually arrive at the point where only statements about Jones' first mental event are left unanalyzed. However, Jones' first conscious experience will not be dependent on any previous ones, and will be determined completely by the non-mental world in the form of his body, environment, heredity, etc. We can then give an analysis of statements about his first mental event which will not mention any other mental events, but only physical objects. Having given a long recursive definition for Jones being in pain, we can now shorten it.

The intermediate steps are not really necessary; if someone were to have the same physical characteristics and environment as Jones prior to his first experience he would have the same first experience, and given this, the same second experience, and so on. Thus, assuming the needed causal principle, it will follow that if any person had had the same physical characteristics as some other person prior to any experi­ence, he would then have the same experiences as well. In view of this fact, it will always be possible to translate statements about another mind into counterfactual statements about one's own mental states. Thus, "Jones is in pain" will be equivalent to "If I had had Jones' physical and environmental characteristics (prior to experience), I would be in pain." Of course, the analysis here offered is not within the primitive language of the system, but is a metalinguistic abbreviation of a rather complex statement within the object language. In moving from the metalanguage to the object language the phrase "I would be in pain" would be replaced by "pain would occur," or rather by "If... then ...... occurs" (assuming that six dots denote the particular kind of painful sensation involved). In the original metalinguistic statement there is also a reference in the antecedent to "physical and environ­mental characteristics that I might have." Here the "I" appears again, but it can again easily be eliminated. To the extent that the "I" refers to my body and its circumstances we can use the general phenomenalist analysis outlined in previous chapters and substitute a logical con-

Page 312: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

306 SOLIPSISM PROPER

struction out of object language expressions for the "I"; in short my body can be treated like any other material object. To the extent that the "I" refers to sensations or feelings that occur or might occur, we can again simply delete the "I" since, as pointed out repeatedly before, all sensation terminology in the object language ultimately refers to one's own sensations.

It is conceivable that even though Jones is in pain and even though I had the same physical characteristics, the resulting mental state would not be what I would now call "pain" because Jones and I mean differ­ent things by "pain." In this case "Jones is in pain," uttered by me, would be false. However, if I were Jones (up to time t), I would also share his linguistic habits so that I would call the experience "pain" if I satisfied the antecedent. However, we would usually want to assert more than this when we say that someone is in pain; we want to assert that he is having the sort of experience we sometimes have. Thus, we must add to the analysis the statement that this experience would be similar to those which I have had on certain specific occasions, or certain exemplars of pain which have occurred in my experience. This will insure that the translation of "Jones is in pain" will also be false in this case.

Despite the above amendation to the analysis, something still seems to be left out. We feel that the original statement is not just hypothetical, but has some categorical force. It says not just that pain would result under certain circumstances, but that someone is in those circumstances even though I am not. It will be useful here to introduce the concept of a "behavioral person." The behavioral person corresponding to Jones consists of his body, all its patterns of behavior, and all its past history, but involves no peculiarly mental states; in fact, it amounts to the extreme behaviorist's analysis of Jones himself, but we are not asserting the behavioral person and the actual person to be identical. We can now add a categorical element to our analysis and say that the be­havioral person whom we identify as Jones is in fact in a certain situation which we believe to be conducive to pain. The real Jones is more than a behavioral person in that we believe him to be similar in mental respects to ourselves, and this amounts to saying that in his situation we would have the same mental states, and that in similar but not identical situations we would have similar but not identical mental states.

It might also be objected that the antecedents of these conditionals which run "If my mental states were dependent on the Jonesian be-

Page 313: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

OTHER MINDS 307

havioral person instead of my own behavioral person then ... " are just a way of talking about Jones instead of myself. But it must be re­membered that we have adopted first person phenomenalism for physical objects. On this view it is contradictory to assert that anything can exist or happen while supposing that I can have no experiences (that there are never any experiences). The resultant language in which one can refer directly only to his own experiences makes it impossible to refer directly to those of Jones instead of one's own. Consequently, the above antecedent is comparable to "If Caesar had not crossed the Rubicon ... " in that it mentions possible events which might really have taken place; it is not comparable to "Speaking of Jones' mental states ... "

So far we have been taking statements about other minds out of context and producing equivalent statements in another language. We must also ask what people might mean by such statements as "Jones is in pain" when they are used for practical purposes and without re­flection. The practical purpose of such a statement will always be that a person must be treated in a certain way, that his conduct must be allowed for (if, for instance, he is angry or drunk), or that we should feel sympathy, antipathy, etc. In any of these cases, we are in effect telling the listener, "Put yourself in Jones' place for a moment and then act accordingly, taking into account your actual feelings and needs." When I assert Jones to be in pain I am not necessarily suggesting that my listeners comfort Jones, but I require that however they treatJones, they do it intentionally. Nevertheless, when we say tis, we are suggest­ing that they put themselves in his position in their imagination only in a very rough and approximate way. It is not suggested that they take account of his childhood experiences, for instance. The context may demand only that the listener imagine accidentally hitting his finger with a hammer or act towards Jones as he would like to be treated in such a situation. Thus, in an actual context a much simpler analysis can be given for statements about other minds.

Let us now take an extraordinary case and suppose that Jones' fingers are abnormal in such a way that it does not hurt to hit them with a hammer. If I then assert that he is in pain my statement will be false, but on the above short analysis, it would seem to be true because I would be in pain in that situation. In order to resolve this issue we must look somewhat more closely at the notion of a behavioral person. When we speak of Jones it is not clear exactly what is included in our concept of him, excluding for the time being his mental states. Part of our

Page 314: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

308 SOLIPSISM PROPER

concept of Jones involves, of course, his body, even though it may change a great deal and still be Jones' body. Further, we must re­member that, from a practical point of view, we are most interested in Jones' reactions and personality traits (speaking of these from a purely behavioral point of view). It seems possible that the essential elements of our concept of Jones are not so much his present bodily states, but his dispositions. These would include his way of walking and speaking, his readiness to smile, and his characteristic movements. On the other hand, it is not really possible to say exactly which of these dispositions must remain unchanged for Jones to continue to be Jones. It is likely that on different occasions we use different criteria to identify Jones (the behavioral person), and the more we know about him, the more our concept is likely to be filled out. If I have just met Jones, I will accept as Jones anyone who has that name and looks similar, and if I identify the wrong man, I need not be making a factual mistake about that man; it may be that I am not subject to any illusions, and I may have any amount of correct information about the man I am confronted with. It is just that my concept of Jones is so limited and I have so few criteria for J oneshood that this information does not do me any good. I am forced to accept as Jones anyone who fulfills my limited criteria. Let us now suppose that someone points out to me that the man I am talking with is not the same man I was introduced to before. I may now admit that my belief that this was Jones is false, yet it is not false in the way that a poor description of either man would have been false, provided that the referent of the description was unambiguously identi­fied. It is more accurate to say that my original concept of Jones has been discarded now that I have more complete criteria for the identifi­cation of him. Thus, my first statement was literally true in a sense, but not very useful since the concept involved did not apply to a single man, but to a large range of men. My new belief involves a concept with a more precise and smaller range ,but it too may be replaced in the same way.

The same sort of thing happens with the queer case mentioned above. We include in the concept of any behavioral person the supposition that he has normal reactions to standard stimuli unless something causes us to alter our concept. Thus our original analysis is a true state­ment about the sort of man we think we are dealing with, and about whom we intend to talk. The statement is replaced by another when we learn about the real Jones and our concept of him changes.

We have now given an analysis of statements about other minds

Page 315: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

OTHER MINDS 309

which is consistent with solipsism and which avoids the difficulties mentioned at the beginning of the chapter. Thus, our solipsistic language is powerful enough to allow us to refer not only to physical objects but also to refer to other minds. Only one step remains to com­plete the reduction.

Page 316: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CHAPTER 15

BELIEF

It will have been noticed that up to this point no analysis has been given of the key concept of belief. Anyone who is trying to analyze the statements of ordinary language must take account of statements about belief and it has been presupposed at many points that the solipsist could give an account of belief. However, many philosophers hold that it is not possible to analyze statements of this sort into statements about sensations, much less into statements about one's own sensations. The solipsist must therefore show that this can be done in order to make a case for the plausibility of his position. Actually, he has not just his sensations to work with but terms referring to his feelings and emotions can easily be added to the primitive language of the system. It is one of these, the feeling or emotion of surprise, which will play the central role, and the basic analysis will be very simple: to believe p is to have a dis­position to feel the emotion of surprise upon discovering that p is not the case. Thus we need only add a primitive term which will refer to the feeling of surprise on the intended interpretation of the system. Since we have counterfactual conditionals within the system we can say that A believes that p if and only if A would be surprised at not p. In case A refers to another mind the term can be dealt with in the ways indicated in the last chapter, and if it refers to oneself the term can be eliminated entirely. We must next deal with the phrase "be surprised at not-p."

It seems clear that when someone believes that some event will take place, or already has taken place, he must feel some degree of surprise upon finding out that he is mistaken. It might be argued here that we can discover our false beliefs without actually feeling any emotion. Our finding out that Rome is virtually as far north as Vladivostock when we had believed the contrary is unlikely to upset our peace of mind. When I classify surprise as an emotion I do not mean to assert that it must al­ways be either pleasant or unpleasant; in fact, it will not be, where the belief concerned is not of immediate practical importance, as is the case with Rome and Vladivostock. At the very least, however, it will always involve curiosity; such questions as, "But isn't Vladivostock north of

Page 317: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

BELIEF 311

Manchuria, and isn't Manchuria even with northern Europe?" and "How can it be so cold there?" immediately occur to one and are asked with a certain degree of animation. We seem to have enough of an emotional stake even in beliefs like this so that when one is discredited we feel enough anxiety to investigate related beliefs in order to make sure the belief really was false (more exactly, to see why the grounds of our old belief do not really imply it). It causes us to feel a new anxiety about some of our related beliefs.

There is another class of cases where it might be argued that we have beliefs even though we do not feel surprise. For instance, I would not be surprised to learn that the average camel has 12, 14, or 18 teeth. Yet this is not the sort of case where I have no beliefs at all; such a case might be the presence or absence of pimples on Fortinbras. Thus the former might appear to be a case contrary to this analysis, since I have beliefs on the subject, but none of several different outcomes occasions surprise. However, it should be noted that I cannot here be said to be­lieve that a camel has any particular number of teeth. What I in fact believe is that the average camel has some teeth and that he does not have any great number of them. I believe very strongly that he does not have more than forty, somewhat less strongly that he does not have more than thirty, and I believe quite hesitantly that he does not have more than twenty. From the foregoing facts it does not follow that I be­lieve him to have any particular number of teeth. Thus if I am told that a camel has 14 teeth this confirms my belief that he has between two and twenty teeth even though I had not believed him to have exactly 14 (I had not disbelieved it either); hence I feel no surprise.

We must now ask whether there are any cases where people feel surprise even though no beliefs are upset. Perhaps the most likely case is that where we receive startling and exciting news on a subject con­cerning which we have no particular beliefs. Suppose that the Alle­gheny River suddenly rises and inundates Pittsburgh washing away the inhabitants. The inhabitants would feel surprise as well as indignation, but few people have ever considered this possibility and rejected it, and it might be said that people neither believed that it would happen nor disbelieved it; this situation might be analogous to Fortinbras' pimples. We have already seen that in order to believe the proposition pone does not have to be actively considering it and one is ordinarily said to believe p ifit follows immediately and obviously from some proposition q which one believes. Now the proposition, "The Allegheny will not

Page 318: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

312 SOLIPSISM PROPER

inundate Pittsburgh" follows quickly and immediately from a number of propositions which we clearly believe: "Water does not run up hill," "The Allegheny is a smallish river," etc. Thus the people of Pittsburgh do believe that Pittsburgh will not be inundated. In fact we are often said to believe in much more extreme cases than this; most people are willing to say that they believe scientific theories which they do not understand, and many Christians recognize contradictions in the Bible but insist that they believe every word of it just the same.

Another possible objection is that there are some cases where we would be surprised whatever happens, even if nothing happens. It seems possible that a person would be surprised if Russia attacked the United States, and would at the same time be surprised if after twenty years no war had broken out and the United States had not been at­tacked. It is, of course, tautologous to say that the United States either will or won't be attacked within any given period; one might then argue that a person cannot believe both, as he would under the present analysis of belief. In the first place, it does not follow from the fact that both p and not-p cannot both be true that someone cannot believe both p and not-po On the contrary, it is common for people to believe two contradictory statements without realizing the contradiction, and, as we have seen, it is even possible for them to recognize the contradiction and still believe both statements. To analyze "belief" in such a way that this becomes impossible would be to distort the ordinary meaning of the term. It is therefore one of the advantages of this theory that it makes it logically possible for us to believe contradictory statements. The example cited above, however, is probably not a case of believing contradictory statements. It is more likely that the person in question believes that there will be a war within twenty years, but not at any specified time. Thus he will be surprised whenever the attack comes be­cause he had not expected it at that particular time, and he will also be surprised if the attack does not come within twenty years since his belief that there will be a war in that time will have been shown to be false. Thus I think that any case where we seem to be surprised at all, possible alternatives can be handled in one of these two ways. Whether it turns out that we believe inconsistent statements or not, an account can be given which is consistent with this analysis of belief.

As has been suggested in an earlier chapter, saying that a person believes x may be consistent with his not expecting x when dreaming, having imaginative reveries, etc. It might be that one would normally be surprised at having a sensation of Cleopatra, but one is at present so

Page 319: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

BELIEF 313

immersed in imagining her that one would not be surprised at having an actual sensation (as opposed to an image) of her. We could then both say that this person believes in the existence of Cleopatra and that he does not believe in her in quite different senses. "Belief" as ordinarily used is ambiguous on this point, and we would have to rely on the context to tell us which is meant. I think that the analysis will be closer to ordinary usage if we exclude dreams, reveries, etc. from the range of the disposition to be surprised.

In the analysis of belief stated here there has been no mention of images, lively or otherwise. This is not to deny that a lively image may accompany a belief. In fact it does seem likely that certain lively images accompany our stronger beliefs; it also seems likely that when we have a very strong image we expect some event similar to it to occur. Thus there may be some correlation between having very vivid images and being surprised when it is found that there is nothing similar to those images in actual fact. However, it is also clear that the connection is synthetic rather than analytic; it is possible to conceive of someone's having a strong image without having what is commonly called a belief and vice versa. Thus we can admit that having a strong image may be a sign of belief without building it into the analysis of belief. The analyses of belief in terms of surprise and lively images have, nonetheless, a great deal in common since they both explicate belief in terms of purely mental phenomena, and they both have the advantage that no entities are introduced that are not admitted on most theories anyway.

It is often argued against this sort of view that when we assert that a person believes a proposition, we want to assert something about his behavior. If someone tells us that he thinks a certain horse is going to win and then bets on another horse we are likely to conclude that he does not really believe that the first horse will win and is trying to deceive us. In fact, our assertions about other people's beliefs are al­ways founded on their behavior in some way or other. There are even cases where we take our own past behavior to be indicative of our past beliefs. All this is perfectly consistent with the analysis of belief given above; we can admit that certain beliefs are always, or almost always, accompanied by certain sorts of behavior. Because of this we can in practice infer a man's beliefs from his behavior. But again it does not follow that there is an analytic connection between certain beliefs and certain behavior, since we could still make this inference with a high degree of probability if there is only a synthetic connection which holds in the great majority of cases.

Page 320: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

314 SOLIPSISM PROPER

In the first place it seems impossible to analyze belief in terms of any actual patterns of behavior. It is logically possible for a person who has been completely paralyzed his whole life to have beliefs. It is presumed that he will have sensations and that his brain will act in much the usual way, and it is possible that physiologists will sometime discover strong correlations between behavior of certain parts of the brain and beliefs. It is clear, however, that even if there were such a correlation, there would still be only a synthetic connection between these actions of the brain and what we now mean by "belief." We might still analyze the paralyzed person's beliefs in terms of dispositions to behave: "if this person became unparalyzed and were put in a certain position, he would behave in a certain way," etc. The difficulty here is that given any belief, there is no action and no set of physical circumstances under which the action would be inconsistent with the belief. On the other hand, if belief is to be analyzed in terms of dispositions to behave, we would expect to be able to deduce from certain pieces of behavior the conclusion that certain beliefs are not being held (and perhaps that other beliefs are being held). We can always make any action and any belief consistent by supposing the person performing the action to have certain motives. As we have seen before, believing that one horse will win is consistent with betting on another if the bettor feels an over­whelming loyalty to the latter horse. Thus there are good reasons for not referring to behavior in any way in the analysis of belief. All that is necessary is to preserve a synthetic probabilistic connection between certain beliefs and certain patterns of behavior.

The most important objection against this sort of theory is that of Chisholm. Although he is talking about a somewhat different theory of belief we will quote from Perceiving - A Philosophical Study at some length.1

"If we apply 'expects,' as defined, to human behavior, then we must say that the appropriate fulfillments or disruptions must be caused by the occurrence, or nonoccurrence, of the 'intentional object' - of what it is that is expected. But it is easy to think of situations which, antecedent­ly, we should want to describe as instances of expectation, but in which the fulfillments or disruptions do not occur in the manner required. And to accommodate our definition to such cases, we must make qualifi­cations which can be expressed only be reintroducing the intentional concepts we are trying to eliminate.

This difficulty may be illustrated as follows: Jones, let us suppose,

1 Chisholm, op. cit., pp. 182-183.

Page 321: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

BELIEF 315

expects to meet his aunt at the railroad station within twenty-five minutes. Our formulation, as applied to this situation, would yield: 'Jones is in a bodily state which would be fulfilled if he were to meet his aunt at the station within twenty-five minutes or which would be dis­rupted if he were not to meet her there within that time.' But what if he were to meet his aunt and yet take her to be someone else? Or if he were to meet someone else and yet take her to be his aunt? In such cases, the fulfillments and disruptions would not occur in the manner required by our definition."

Chisholm would presumably argue in the case of our analysis that being surprised at not-x presupposes the taking of some state of affairs to be an instance of not-x. He would then say, as before, that this "taking to be" is intentional. Thus our analysis would be circular since this "taking to be" would be a form of belief, and that is what we were trying to analyze in the first place. It should be noted that Chisholm's whole argument depends on the possibility of Jones being mistaken when he takes someone to be his aunt. If this "taking to be" were infallible it would not be intentional in Chisholm's sense since the truth of "Jones takes this to be his aunt" would entail that this is his aunt. Similarly, the sort of belief that cannot possibly be mistaken is not a kind of belief at all.

We have already seen that this is the case with beliefs about presently occurring sensations. When I think that I am having a sensation it is impossible that I should ever be mistaken. We also argued that it is impossible to unconsciously misdescribe a presently occurring sensation. This suggests that there is a certain category of beliefs to which Chis­holm's argument does not apply. These would be beliefs to the effect that a certain set of sensations will or will not occur. According to the analysis, such beliefs would amount to being surprised at the non­occurrence or occurrence of these sensations. This would involve taking a certain sensation to be an example of a certain description - perhaps as a case of a "green speedboat-like sensation." But this would not be Chisholm's intentional sense of "taking" because it is impossible to mis­takenly take something else to be a green speedboat-like sensation ifit is not one. We might with a slip of the tongue call something else a green speedboat-like sensation, but we can always conclusively decide the question on the spot by comparing the sensation with images which are instances of the qualities in question. It should also be noted that the analysis of "Jones believes x at time t" is "Jones would be surprised at not-x at time t"; thus if he should change his linguistic habits between

Page 322: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

316 SOLIPSISM PROPER

the time of initially having the belief and the time that he verifies it, the belief changes as well.

The whole import of Chapter 12 was that statements about material objects can be reduced to statements about sensations. The conse­quence of this is that all our beliefs about material objects will amount to beliefs about the occurrence and non-occurrence of certain sorts of sensations. Chapter 14 extended the same basic account to statements about other minds. Thus Chisholm's objection will be groundless as long as this sort of theory of belief is combined with solipsistic reduction­ism.

Page 323: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

CONCLUSION

We have now stated a system of analyses which satisfies our original definition of solipsism: it interprets the statements of common sense in such a way that one can assent to them without ever talking about any­thing except one's own mental states. Further, it is logically solipsistic in that it reduces assertions oflogical necessity to statements about one's own imaginative capabilities, and says that this is all we ever meant in the first place. It would be hard to imagine a more extreme thesis and, even though it does not involve doubting the existence of the world, or living in a tub, it is a legitimate heir to the title of solipsism. Further, this system is hopefully free of any contradiction or circularity of analysis. It is at least as economical as its competitors even though there are some assumptions which must be granted.

Very little has been said about the weaknesses of this sort of solipsism. This is not because it is believed that it has no weaknesses, but for a reason which will appear shortly. Most of the difficulties which have appeared can be put under two main headings. First, it was necessary to make some very strong presuppositions in order to support counter­factual induction, and counterfactuals were subsequently used in al­most every analysis. A system which accomplished all of the same things without these presuppositions would be much more economical. Second, it is an undesirable feature of our system that almost all the basic concepts such as 'image,' 'sensation,' 'comprehension,' 'analytic,' etc. are surrounded by a certain degree of vagueness. Most of the key distinctions have turned out to be distinctions of degree, and it was thus impossible, in many cases, to draw sharp lines where they would have been desirable.

However, on the basis of the conception of philosophy which under­lies this book, we are, even now, not in a position where we can say exactly how serious these difficulties are. I do not think that whole philosophies are shown to be true and false, or adequate and inade­quate. But we can and should develop as many kinds of theories as far as they can be developed. It is at that point, and only at that point,

Page 324: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

318 CONCLUSION

that we can say that one philosophy has certain advantages over an­other. Of course, the advantage that one system has over another will be in some particular respect, and it may be difficult to weigh accurately advantages and disadvantages in different respects. But even if sub­jective elements are involved there will still be some rational basis for the choice of one system over another.

Thus, progress in philosophy is not constituted by refuting one system and vindicating another. Rather, it is the developing of all the com­peting theories to the point where we can make an intelligent choice among them.

Page 325: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

BIBLIOGRAPHY

J. L. Austin, Sense and Sensibilia, Clarendon Press, Oxford, 1962. A. J. Ayer, Language, Truth and Logic, 2nd ed., Dover, New York, 1946. -, The Foundations of Empirical Knowledge, MacMillan, London, 1953. -, The Problem of Knowledge, MacMillan, London, 1956. G. Berkeley, Principles of Human Knowledge, Library of Liberal Arts, New York, 1957. 1. Berlin, "Empirical Propositions and Hypothetical Statements," Mind, July, 1950. Max Black, ed., Philosophy in America, Cornell Univ. Press, Ithaca, 1965. A. W. Burks, "The Logic of Causal Propositions," Mind, 1951. -, "On the Presuppositions of Induction," Rev. of Metaphysics, June, 1955. R. Carnap, Der Logische AuJbau der Welt. R. Chisholm, Perceiving: A Philosophical Study, Cornell Univ. Press, Ithaca, 1957. -, "The Contrary-to-Fact Conditional," Mind, 1946. A. C. Danto, Analytical Philosophy of History, Cambridge Univ. Press, London, 1965. R. Descartes, A Discourse on Method and Selected Writings, Dutton, New York, 1951. R. Firth, "Sense-Data and the Percept Theory I and II," Mind, Oct. 1949 and Jan.

1950. -, "Radical Empiricism and Perceptual Relativity II" Philosophical Review, July,

1950. N. Goodman, The Structure of Appearance, Harvard Univ. Press, Cambridge, 1951. -, Fact, Fiction, and Forecast, Harvard Univ. Press, Cambridge, 1955. C. 1. Lewis, An Analysis of Knowledge and Valuation, Open Court, LaSalle, 1948. L. Linsky, ed., Semantics and the Philosophy of Language, Univ. of Illinois Press, Urbana,

1952. N. Malcolm, "Wittgenstein's Philosophical Investigations," Philosophical Review, Oct.,

1954. G. E. Moore, Philosophical Studies, Harcourt, Brace, and Co., New York, 1922. P. H. Nowell-Smith, Ethics, Penguin Books, Harmondsworth, Middx., 1954. H. H. Price, Perception, Methuen, London, 1932. -, Thinking and Experience, Hutchinson's Univ. Library, London, 1953. W. V. Quine, From a Logical Point of View, Harvard Univ. Press, Cambridge, 1953. -, Word and Object, MIT Press, Cambridge, 1960. G. Ry1e, The Concept of Mind, Barnes and Noble, New York, 1949. G. Santayana, Scepticism and Animal Faith, Dover, New York, 1955. W. Sellars, Science, Perception and Reality, Humanities Press, New York, 1963. P. F. Strawson, Individuals, Methuen, London, 1959. J. Taylor, The Behavioral Basis of Perception, Yale Univ. Press, New Haven, 1963. W. Todd, "The Theory of Meaning and the Learning of Language, Inquiry, Winter,

1965. -, "Causal Laws and Accidents," Theoria, No.2, 1965. R. L. Wilder, Introduction to the Foundations of Mathematics, Wiley, New York, 1952. P. Wilson, "Quine on Translation," Inquiry, Summer, 1965. L. Wittgenstein, The Blue and Brown Books, B. Blackwell, Oxford, 1958. -, Philosophical Investigations, B. Blackwell, Oxford, 1953.

Page 326: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

INDEX

Abnormality of vision, 11 color mis-learned, 26

Acceptance, dispositions and feelings in, 67-9, 81

in comprehension, 75-8 in synonymy, 90 in verification, 102-3 of images, no rules for, 61-2 of sensations and images, 48, 67, 69,

72-3, 75 Acts of apprehension,

and direct apprehension, 3 in relation to a sensible, 4-6

Ambiguity, basic, 101 can vary with context, 236, 313 in "knowing the rules for use", 1 05 and material object sentences, 267-8

Analytical hypotheses, about synonymy of word pairs in two

languages (Quine), 194-5 Analytical statements, 96-100

Class I, true by arbitrary convention, 88

Class II, true by well-established ar-bitrary convention, 89

Class III, empirical hypotheses, 89-90 Class III, not really analytical, 90 Class III, distinguished from I and II,

92 Class II and synthetic statements, 94 factual in special way, 91 and notational abbreviation, 93 three main uses of, I, II and III, 88 truth of, dependent on synonymy, 90 ultimately rests on verbal definitions,

90 Assertions, 54, 56, 69, 76, 78-9, 86, 95-6

Chs. 3-5, 10-11 and accidental connections, 128 act of, and what is asserted, 51-2 and consideration, difference of at-

titudes, 71 and considering, in our own thinking,

70 contrasted with preliminary, 85 identified, 87 neutralized in preliminary, 52 of axioms, 201 of past or future sensations, 287-8 of preliminary, certain abilities, 70 our, about other people's beliefs, 313 our ordinary, in solipsistic language,

209, 252, 264 and phenomenon of belief, 85 and preliminaries differing in context,

52-3 Association, 25-7, 30, 34, 39, 49, 59-61,

67, 78, 89, 115 dissociation, 40-1 in ostension, 26-7, 29 in private language, 40-1 traditional view of (language), 26

Austin,]. L., 5, 268 criticism of Ayer and its meaning for

solipsism, Ch. II Axiomatized scientific theory,

notion of explanation, 160 vs. philosophical systems, 159-61

Axioms, about present "now", 174, 288 assumptions placed in, 305 basis for common sense belief, 163 causal law added, 190-1 classes of, 174-191 Class I, 174 Class II, 176, 180, 254-5 Class III, 179, 187-191 concept of completeness, 154 dependent, 176 different set for each person, 158 explanations of, 160 for sensations, some temporal latitude,

169

Page 327: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

INDEX 321

in quasi-axiomatic solipsistic system, 305

in scientific theory, 160 least possible number of primitives

in solipsism, 184 not "true" but "applicable", 153-4 primitive terms and interpretation,

154 and rules of inference, 174 presuppositions of material object

statements, 254 tenaciously held beliefs, 176 and their theorems, 155-8

Ayer, A. J., 101-4, 177,294-7 criticized by Austin, Ch. II passim on material object sentences, 262-3 on public and private languages, 42 phenomenalism and objections to

finite analysis, 262 sensations correctly recognized, 43 sensations momentary and private, 284

Behaviorism, assumption of, 204 as to other minds, 301-2 and the "behavioral person", 306-8 and beliefs of other people, 313 and certain beliefs; synthetic, pro-

balistic connection, 314 and comprehension, 80 in analysis of belief, 314-15 knowing the rules for use, 105 physicalist-behaviorist system, 205 and the truth of"IfI were Jones", 303 and understanding words, 41

Beliefs, 162-5, 176-7, 185-9 about material objects, 316 affected by context, 225 analysis of, and instancehood, 87 any system must satisfy, 193 asserting things to ourselves, 70 common sense beliefs, 199-200, 209,

211,255-7,268,274 common sense and science added to

object language, 158 concept of (solipsistic language), 310 curiosity involved, 311 and dependent-independent sansa-

tions, 10, 12, 14,221-2 and determinacy, 140 Humean account of, 122 and inconsistency, 314 ordinary, 139

and sensations of others, 40 and surprise and lack of surprise,311-12

Berkeley, G., 46, 112, 115, 231 Berlin, I., 165 Brentano, F., 115-16 Burks, A. W.,

assigning probabilities, Carnap's, 189-90

Carnap, R., 188-191 Causal connection,

can deduce counterfactuals, 134-5 and constant conjunctions, 148 difference one of degree, 133 difficulties involved, 132-3 durable properties of, 138-9 formation and clarification of, 127,

129-30 and generalizations, 134 and rule for similar properties, 137 vs. accidental connection, 128, 131-2

Causal laws, 127, 143, 191 causal possibility, 136, 148 and counterfactuals, 139-41, 145-8 defined and clarified, 136 and hypothetical world, 144 in Humean analysis, 137 inviolable, but can apply different

causal laws, 135 involving durable properties, 138-9 many relatively specific, independent,

190 less basic, more numerous on another

view of world, 145 non-legislative but predictive, 134 statements of, 130

Causal principle, 257, 305 Chisholm, R., 115-16,261-2,266-269

behaviorism and belief involving surprise, 314-15

Comprehension analysis of, regarding preliminaries,

84-5 -behavior, 80 cognitive, 108 and disposition to recognize instances, 76 improbable maximum of, 77 individual words, 87 in verifiability and image theories, 102-4, 106

Page 328: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

322 INDEX

and logically complex statements, 100 meaning of, 72 more basic than understanding, 71,83 no minimal degree of, 78 of counterfactuals, 139, 147 of preliminaries, sorts of instances

accepted, 75-6 of statements, analytically and syn­

thetically, 97-9 and verification, 102, 104

Concept of existence, 171-3 "Conditional sense contents" (sensa-

tions), 254-5 Conjunction-disjunction, 99, 180 Conscious event, 225 Conscious resolve, preceding images, 16 Conscious states, 221, 225, 242, 245, 250

in form of scepticism, 214-15 Strawson, 211

Consciousness, 5-10, 12, 16-17, 19, 119, 214-15

Consistency, 62, 114 in use of words, 38 of terminology in private languages,

43-5 Constant connections, 127, 129, 138, 148 Constants, 298

construed as proper names referring to particular sensations, 279

functioning as strictly-used proper names, 283-4

introduced into primitive language of quasi-system (as "k with any inte­gral subscript is wff"), 279

making a present basis from which to locate events in the past (using "kn"), 287, 292

referring to particular sensations in specious present only, 284-5

use mainly theoretical, 284 Contextual implication, 95, 97

and analytic sentences, 96 and linguistic rules, 96 Nowell-Smith on, 96 relation to meaning, 94

Contradiction, 312 and tautologies, 98

Correlation, 138-9 between brain-states and behavior, 30 1 between brain-states and belief, 314 between vivid image and surprise, 313 in causal connection, 130 instead of causation, 135

Corrigibility, 70, 228 in public conversation, 33 and ostensive definition, 28 and sensation, Austin's thesis, 249-50

Counterfactual statements, 173, 214, 253, 317

analyzed, 139 and belief, 310 in phenomenalist treatment of time,

286, 291, 296 in quasi-axiomatic system, 190-1 and material objects, 258 on causal laws and prior chains of

events, 132, 134-5, 140-8,254,273 and other minds, 305 and personal identity, 304

Danto, Arthur 289, 292-3, 296 objections to phenomenalist treat­ment of time, 286-7 on tensed statements, 294-5

Definite descriptions, 281-3 denoting or not denoting, 272-3 when substituted for proper names,

284 Definitions, 73, 93, 156, 211, 275

altering definitions of key terms, 154-as notational abbreviation, 93-4 as correlate of synonymy, 93 crux of quasi-system, 158, 176-7, 180-

209 for durable properties, 138-9 and genuine analytic statements, 89-

91 in finite phenomenalism, 182 in imagining instances, 49-50 in learning language, 24-8, 50 in ostension, 25-6, 89 and logically complex statements, 99-

100 operational, for each variable, Tay­

lor, 202-3 and Quine, 173 reasonable approximation for (quasi­

system), 183, 269 series of, to collapse metalanguage

into object language, 276 very complex, in system economical

ofaxions, primitives, 184 Dependent and independent sensations,

basic difference between, 10-11 and beliefs about existence of ma­

terial objects, 1O-11

Page 329: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

INDEX 323

correspondence, 9 and delusive beliefs and expectation,

221-3 dichotomy between, 217, 224-5, 238 intermediate cases between, suspend­

ed belief, 222 and "sensings" contrasted, 8-9, 12

Descartes, Rene, his example of chiliagons, 109

Determinism, and beliefs of common sense, 140-1 and causal possibility and causallaw,

136 definition of (determinacy), 145 determinacy-indeterminacy, 73, 108,

112, 192, 194-5, 198, 200, 270 Dispositions, 62, 81, 214, 250, 301

holding in varying degrees, 68 in analysis of meaning, 148 in attitude of acceptance, 67 in identity, 308 psychological, 92 to produce images dependent on con­

text, 48 to recognize instances, in compre­

hension, 76 to recognize instances ("darker than")

60-1 Dreams, 4-6, 13, 18,218,227,239-42

excluded in disposition to be sur­prised, 312-13

"Durable properties", 138-9, 145

Economy, 161, 184, 186, 191 basis of preferring one system to

another, 162 in language, Carnap and Goodman,

191 in solipsism, 46,127,139,214,288 in Strawson's theory, 210-11 and language of analysis, 148-9 metaphysical importance to a system,

174 Emotive tone, 15-16 Euclid, 153-4, 256 Extensionality thesis, 148-9

Firth, Roderick, 272 material object sentence III pheno­

menalism, 266-9 percept theory, 19-20

Focusing attention, 25, 41, 234 in learning ostensively, 28-31

Formal systems, 88, 93

Goodman, N., 174, 188, 198 accidental constant conjunction, 131 and present quasi-system, 191

"guessing" , at child's focus of attention, 30 compared with "hit upon", 29 in interpretation in os tension, 28-9

Hallucination and Illusion, 4-5, 11, 13, 33,44,64,74,226,232,235,272,308

in dependent sensation, 9 in material object assertions, 268-9 in physicalist theory, 231 in substance theory, 219 "seeing", Ryle, 18

Hempel, C. G., footnote, 104; 198 Hilbert, D.,

formalist program, 154 Hume, David, 14,22,46,105, 122, 127-

8, 130, 134-5, 137-9, 145,236 Humean theory of causation, 133

Illusion, see Hallucination Images, see also Belief, Comprehension

and Synonymy, 16, 18,39,55,67,69, 74,76-7,81-3,313

basic, 50, 60-1, 64 corresponding to sensations, 21-2 in imagining instances, 47-51, 56,

58-61, 63-6, 75 in Moore, 3 in object language, 277 in random imagining, 61 in Ryle, 17 -instances, contrasted with thing­

instances, 73 involved in remembering by

"matching", 64 not reducible to behavior, 24 partial, as incomplete instances, 72 and theory of meaning, 46, 87, 89,

101-6, 108-12, 114-122 vs. sensation, contrasted and com­

pared, 9-10, 13, 19-23, 75 Imaginability,21 Imagining, see also Images, 16-23,47, 56,

62,66,70-1,73,76,78,82,85-7,90-1, 98, 103-4, 108-14, 119, 147-8

basic instances, 50, 61 chiliagons, 109 "darker than", 65

Page 330: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

324 INDEX

deficiencies in, 22-3, 112 definition of, 48 in analytic statement II, 90-1 in remembering, 63-4 an instance, 49, 58-60, 72, 75, 77,

116 instances of preliminaries, 65, 68,70 powers of, and their scope, 82, 110,

112-13 random, 50, 58, 61 range of instances, 67, 72 similar objects, sensations, images,

58 testing powers of, 85-6 theory of, 1 7 and understanding images produc­

ed, 118 Imagists, 117 Indeterminism,

and accidents, 128-9 and beliefs of common sense, 140-1 and chance events, unpredictable and

inexplicable, 128 definition of, 140 determinacy-indeterminacy, 73, 108,

112, 192, 194-5, 198, 200, 270 Inference, 190, 196, 267

and axioms, 176, 188 inductive, principles of, 175 inductive, principles of, for truth of

scientific theories, 186 particular, examples of, by contextual

implication, 97 rules of, as rules of deductive logic,

174 Instances, see also Comprehension,

Meaning, and Verification, 60, 147 concept of, 46-7, 55, 66 imagining, Chs. 3-4, esp. 48-9, 66,

98, 112 in analytic statements, 89-100 in generalizations, 115-16, 134-5 of predicates, Ch. 3, 68-9 of preliminaries, Chs. 3-5, esp. 55 range of, 67, 71

various sorts of: basic, def. 50; com­plete and partial esp. 55, 72, 102; natural and contrived, esp. 61-3; and random, 50, 58

Interpretation, 19, 28, 39 93r 96, 105, 165-6, 193,204-7,220,243,247,254, 256-7, 263, 268, 276 in mathematical systems, 158

in scientific system, 159 is finding a "model" for a system,

154, 156 a model for philosophical-mathe­

matical system, 157 models from working mathematics

and variant, 156-7 model, special arrangement of

points, 156 Wilder: can have non-geometric

model,156 Introspection, 15, 68, 90 Invariant connections, 129, 134, 138

in causal and accidental connections, 131

over very limited range of space and time, 130

"Is sp. part of", 1 70 "Is tem. part of", 1 70 Isomorphic logical structure, of complex

statements, 100

Kant, I., on empirical investigation, 140

Kinaesthetic phenomena, as to images and sensations, 14-15

Lewis, C. I., 289, 292 traditional phenomenalists, 177 traditional theory, footnote, 25

Malcolm, N., 37-42 "Matching", see also Remembering, 206

a sensation with an image, 64 Material object language, 177-8, 190,

280 and exioms of Group II, 185 eliminable by primitives, 190 part of ordinary language, 276 and relation to object language, 276

Materialism, 4 Mathematical systems,

as paradigms for axiomatic systems here, 184

language for, 172 often need defined terms to state all

axioms, 176 vs. a philosophical system, 157-9

Meaning, see also Comprehension, Sy-nonymy, Understanding, 248-50

answering questions about, 87-8 cognitive, criterion for, 101-8 concept of, and metalanguage, 266,

274-8

Page 331: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

INDEX 325

importance of context in, 95-6, 313 in analytical sentences, 89-91

in imagining instances, 66 in a private language, 37

and meaningful expressions, 57-8, 101-2

natural and contrived, 61-5 range of, 83-4 stimulus, 197-9 and thought, theories of, esp. 120-3

Memory, consistency of, 43 image of past event, 63 in correct usage of words, 42 in individual's body of beliefs, 175-6 and sensation lacks, 174-5

Metalanguage, 166, 173, 176, 178-9, 181 -2, 201, 305

abbreviations, 180 all-inclusive, 276 and object language, 276 and theory of meaning, 274-8 vs. ordinary language, 276-8 vs. solipsistic language, 174, 276,

286 Metaphysics, descriptive vs. revisionary,

209,236 Mill, J. S.,

inductive principles, 190 Mirages, 9, 11, 14, 18, 217-21, 233-5,

238 Misunderstanding, 44, 52, 72, 86, 96,

III defined, 71 degree of, 84 and os tension, 27-8, 41 and a private language, 34-6, 40

Models, see also Interpretation, definition, 154, 156 from working mathematics, 156-7 in mathematical systems, 158 in scientific system, 159 non-geometric possible, Wilder, 156

Modern geometry, R. L. Wilder, 155

Moore, G. E., 3-8, 12, 226, 230

Naess, Arne, and the Oslo philosophers, 229,237

Nominalism, 248 Nowell-Smith, P. H.,

doctrine of contextual implication, 96-7

Object language, 179 about other minds, 305 can supercede metalanguage, 286 complexities abbreviated by metal-

anguage, 286, 305 conjunction-disjunction in, 180 enriched, 276-7 expression for "I", 306 in which physical space constructed, 181 in which primitives refer to sensations,

158 Ostension, 57, 59, 88-9, 91

basic in language learning, 29, 31 classic case of, 25 definition of, 25-9, 31-45 focusing attention in, 28-31, 41 "guessing", 28-30 in quasi-system, 167-8 psychological association in, 26 and sensation language, 31-45 traditional view of, 25-6 Wittgenstein's discussion of, 27-32

Other minds, 39, 83, 88, 106 disposition to show feeling, thought,

in behavior, 301 finitist analysis of, 99, 177 and identification criteria, 308 "J ones is in pain" analysis, 303-5 and the metalanguage, 276 and reactions, personality traits, 306-8 referred to in solipsistic language, 309 statements about, and verification,

302-3 statements about, into primitive

language, 300

Parmenides, quotation, 105 Past, statements about, 103-4, 279

about events that have never occurred 295-6

hypothetical components in, 288 and image of past sensation, 297-9 in causal possibility, 136 and phenomenalism, 286-7 physical events in, 289-9 I and possible event, 297 temporal mechanism for, 285 and tensed sentences and temporal re­

lations, 291-7 translated to phenomenalist language,

286,294-5 two components of tensed sentence,

Ayer, 294-7

Page 332: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

326 INDEX

and use of proper names and "pre-cedes", 288-9

Perception, 4, 11, 19-20, 231 Personal identity, eliminating "I", 305-6 Personal identity,

eliminating "I", 305-6 "I am in pain" cannot be false if used

truthfully, and as usual, 302 "I" and "my" superfiouous at primi­

tive level, 302-3 "I", an indicator word, 302

Personal sensation language, see Private language, 254, 265

Phenomenalism, 162, 193, 204-7, 216, 247,253,269-79,290-1,295,305-7 classical, 255-6 and definite description not denoting

uniquely, 272 a definition of, 185 dicussion of Ayer, 262-6 discussion of Chisholm, 261-2, 266-9 finitist form of, 99, 182, 251, 269 Firth on material object sentences,

266-7 infinite analysis in, and remote

future, 262-6 logical constructions of, 33-4 on material objects, 218-19, 231 past and present terminology of,

288 Realist vs. Phenomenalist: Sellars,

254-7 traditional, C. 1. Lewis and A. J.

Ayer, 177 traditional, infinite number of

statements, 182 treatment of time in, 286

Physicalism, 193, 201-3, 207, 231 physicalist-behaviorist system, 202,

204-6 variation of, present theories, 208

Positivist, 105 Predicates, 4, 51, 60, 63, 69,71,84,249-

50,285 contrasted with preliminary ex-

pressions, 53-6, 59, 68 in assertions, 53 instance of, 46-7, 59 primitive, 206 primitive, -P, in Strawson's theory,

212-13 primitive, and "sensation", re­

interpreted to apply to "stimu-

lations" instead, 204 primitive, temporal, 170

Preliminaries, 13, 74-8,81-2, 115, 119, 298

abbreviations for, 87 basic, 60, 65, 67, 72 definition of, 52, 76 having contrived meaning, 62-3, 65 imagining instances of, 65-7, 69 in analytical sentences, 88-91, 98-9 in cognitive meaning, 108 in comprehension, 78-80 in understanding sentences, 84-7 in verification, 102-3 and meaning, 87 vs. assertions, 57-8, 69-71, 73, 85 vs. predicate expressions, 53-5

Present, and constants, q.v., 279, 283-4, 287 in statements about own present

sensations, 288 Presuppositions, 121, 135, 143, 185,190,

219, 310 in belief, 70, 87 in counterfactuals, 143,253-4,317 in determinism-indeterminism, 140 in physicalist system, 202 in selecting similar images, 113-61 of generalizations, 253 of image theory, and understand­

ing, 116 of quasi-system, assumptions to sup­

port counterfactual induction, 148

of theory of meaning, 219, 248 ontological, for quasi-system, 257

Price, H. H., 216, 219, 226, 243 image theory of thought, 117, 121-2 imagination, 22-3 word-thinking, 117-22

Primitive language, 154-7, 160, 164, 177, 190, 192-3,280

axioms, theorems, definitions of, 158-9

Class II axioms, defined terms ad­ded, 176

definition of, 158, 164, 191 definitions in, translate subject

matter, 158-9, 184 economy in, 184, 191 feelings, emotions, especially sur­

prise in, 310 no "I" or "my", 214

Page 333: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

INDEX 327

primitive terms refer only to sen­sations, 158, 279

and Strawson, 213-14 whether possible to use, 184-191

Primitive terms, 39, 164,201-3,216,305 feelings, emotions, and esp. surprise,

310 in classes of primitives, 165-73 in quasi-system refer only to sen-

sations, 154, 158, 279 meaningless until interpreted, 154 no persons in, 214 and notations, object language, 181,

209, 274, 277 and other systems, development of,

154-191 Strawson, 209, 212 usage of, on their intended interpre­

tation, 220 various classes of, and interpretations,

165-73 Private language, 139

according to Malcolm, 37-42 according to Wittgenstein, 37-42 as understood by others, 32-6 discussion of and definition, 31-45 limitations of, 39-40 phenomenalist's view of, 34, 36 private ostensive definition, 39, 40-5 "same", 37-8 and Strawson, 208 translations from material object

sentences, 177 vs. public languages, 32-45

Proper names, 284-5, 287 added in primitive language to sen-

sations, 279 or constants, q.v. and definite descriptions, 281-3 denoting particular sensations, 279-83 and the future, 286 past and present located by proper

names and "precedes", 288-9 Psychological disposition, in imagining

instances, 59-61 pure chance

and accidents, 129 matter of, or causal law, 140

Pure instances and verbal definition, 24

Quality words, applied directly to sensations, 167 and ostension, 167

Quantification, 171-3, 297 universal and existential, 99

Quasi-axiomatic solipsistic system, 155-6, 159-171, 173-190, 285-6, 305

crux of, definition and defined terms, 158

economy of, 174, 191 following modern axiomatic me­

thod,256 interpretation, 154, 157-8, 279 language and common sense beliefs

about world, science, and ma­thematicsin, 172

method of, 208 reasons for using, 153

Quine, W. V., 173, 197-201, 227 analytic-synthetic distinction, 92-4 controversy with Strawson, 271 formal systems, 93 notational abbreviation, 93 on translation, 192-5 referential opacity, 94

Realist theory, 248 Recognition, 19, 59, 79, 83, 104-5

imagining of instances, 74 in synonymy, 91 in tactual sensation, 224 involved in comprehension, 74, 76-8,

81-2 of instances of preliminaries, 78, 98 of instances, process, 48, 58,72,74-6 sensations crucial for, 42-4

Reductionism, 148 Referential opacity,

of quotation, belief, etc., Quine, 94 Referring quality, see also Intentionality,

8-10 Remembering, 49, 63, 175

consistency in, 43 criteria always relevant, 65 criteria for, 42 "matching", 64

Representation, 115-16 Rules, 49,56,66,91,96, 101, 118, 147,

222-4, 291 application in comprehension, 79-80 for constructing a definition, 183 for generalization of accidental

connection, 134 and image instances, 62 in causal-accidental connections,

131

Page 334: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

328 INDEX

knowing, for use of expression, 105 "knowing the rules" ambiguous,

105 none for imagining basic instances,

50,55 none for random imagining, 50 of a private language, 37-8, 44

Russell, B., 39, 287 Ryle, Gilbert, 17-19

about imagining, 70

Scepticism, in comparison with solipsism, 214-15 Strawson's refutation of, 214

Scientific theory, 161, 172, 185-7, 195-200

vs. philosophical system, 160 Searle, John, 51 "seeing", 3-4, 18, 205, 230, 245-7 Sellars, Wilfred,

on phenomenalism, 253-61 Sensations, 87, 112, 181-5,217-29,253,

277, 315, and Chs. 1, 9, and 12-13 as "stimulations", 201-7, 211 denoted by proper names, con­

stants, 279-84 dependent-independent, 166 describing other minds, bodies, Ch.

14, esp. 302 in classes of primitives, 168-73 in ostension, 26, 88, 167 in primitive language, 191, esp. 279 in a private language, 31-45 in somatic, location, object classes,

178-80 sequences of, 165, 173-4, 177-80 temporal relations between, 285-9

Sense-data, 3, 5, 7, 17-18, 20, 216-19, 225-7,231,233-6,243-4,246,250,262

theory of, 12 Sensible, 3-7, 12 Set membership, 156-8, 171-2, 202

in an example from Wilder, 155 Solipsism, 91, 176,234,245-6,248

alternatives to, 192-216 assumptions in, 139, 147 Austin, 251-2 Ayer, 244-9 beliefs and expectations in, 268-9, 310,

316 causation in, 135-6, 139 conclusions about, 317-18 counterfactuals in, 147-8

criticism of, 249, 317 economy of primitives and axioms in,

184 and extensionality, 148-9 language of, q.v., esp. 24 and object language, q.v., esp. 276 other people in, 276, 300, 302-9 and other theories, esp. 317-18 past and specious present in, 169 proper names in, 280 quasi-axiomatic approach in, 153 Sellars, 258 theory of meaning in, 13, 46, 122-3 verifiability in, 105

Solipsistic language, 171, 174,216,235, 245-7,249

axioms about sensations in, 158 explained in metalanguage, 276 in translations, 193-5,200,251,276,

305 material object sentences into, 200 and metalanguage, q.v., 178 and object language, q.v., 179-80 objects analyzable into, 277 of quasi-axiomatic solipsistic sy­

stem, 168 and other minds, 300, 309 and primitive language, q.v., and

158, 164 and private language, 45, 248 and proper names, definite de­

scriptions, 269-74 Spinoza, 153 "Stimulation" ,

in discussion of Taylor, 201-8 in perceptual variables, 202 in somatic, location and object classes,

201 and "sensation", 20 I

Strawson, P. F., 171,223,287 controversy with Quine, 271 his theory, 208-15 persons as primitives, 211

Substance philosophy vs. phenomenal­ism,

basic difference of meaning, 162 theorems and concepts, 161-2

Synonymy, 99, 102,247 between sentences, 90-1, 198-9,276-7 criterion for, 87-8, 90, 159 in logically complex statements, 100 Quine, in translation, 194 resting on notational abbreviations, 93

Page 335: William Todd (Auth.)-Analytical Solipsism-Springer Netherlands (1969)

INDEX 329

Wilson, in translation, 195 within a context, 94 within object language, 277

Tautology, 98, 303-4, 312 truth table, 93

Taylor, James, perceptual variables, 201-6 psychology of perception, 201-8 "stimulation" as primitive, 201-8

Theorems, 153-5, 160, 176-7, 184, 209 as beliefs of ordinary language, 189 axiomatizing science and beliefs, 158 difficulties of, in physicalism, 206 of quasi-axiomatic solipsistic. system,

186 and probability qualification, 190 reflecting definitions and their re­

lations to axioms, 158 Theory of Meaning,

and concept of image, 13 for past sensations, events, 297, 299 and how to talk about sensations, 167 imagining instances basic to, 66 and criterion for synonymy, 88 of Carnap vs. quasi-system, 191 vs. theory of thought, defined, 120

Theory of Thought, see Thought Thought, 70

image theories of thought and of meaning, logically independent, 122

image theory of thought, 120 in Price, 117 possibility of thinking about nothing,

118 word and image thinking, 117-122

Time, 289 "going back in time", 104 in quasi-system, 286 in phenomenalism, 286-7 reversal in, 104 and specious present, 284 and true for all time, 107, 134

Truth functional, concepts, 99 connectives, 96, 100, 143 language of analysis, 148 statements, 139, 148

Understanding, 28, 32, 35-6,40-1, 75-6, 88

an analysis of, 106-7 analytic sentences, 98-9 counterfactuals, 139 degrees of, 83-4 imagining instances for, 66-7, 112-

13 and imperatives, 107 knowing rules of, 104-5 meaning and synonymy, 71, 87, 91,

100, 106 misunderstanding, q.v., 84 of assertions and preliminaries, 69-

70, 85-7 vs. comprehension, 83-4, 71-2

Universe of discourse, 55-6, 69-70, 99

Variables, 99 in quasi-axiomatic system, 172,277 partial, 113 perceptual, Taylor's example, 201-3

Verifiability, 44-5, 80-3 Ayer, 101-4 criterion of, for cognitive compre-

hension, 108 in principle and practice, 103 and "knowing how to verify", 105 of beliefs, 70 of metaphysical and theological state-

ments, 101 of other minds, 300, 303-4 of past, 104 theory, compared to image theory,

101-6 theory of determining meaning, 10 1 and universal statements, 103

Vividness, 313 of images and sensations, 14

Warnock, G., discussed by Austin, 247 Well-formed formulas 170-1 White, M., on analytic-synthetic dis­

tinction, 93 Wilder, R. L., 155, 157, 176

geometric and non-geometric models, 156

Wilson, Patrick, 193-5 Wittgenstein, L., 68, 114

and Ayer, 42 criticism of traditional ostension, 27-

32 interpreting ostension, 27-29 and Malcolm; "same", 3 7 and manometer, 44