'A Matter of Prolonged Field Work': NotesTowards a Modal Grammar of English1

MICHAEL STUBBS

Univtnity of London Institute of Education

1. INTRODUCTION

What I shall have to say here is neither difficult nor contentious; the onlymerit I should like to claim for it is that of being true, at least in parts. Thephenomenon to be discussed is very widespread and obvious, and it cannotfail to have been already noticed, at least here and there, by others. Yet Ihave not found attention paid to it specifically.

(Austin 1962:1, emphasis added (or subtracted?).)

In this paper I am going to discuss some ways in which language is used in com-munication, to express personal beliefs and adopt positions, to express agreementand disagreement with others, to make personal and social allegiances, contracts, andcommitments, or alternatively to disassociate the speaker from points of view,and to remain vague or uncommitted. I will try and show that such an approachto language has both considerable theoretical interest for linguists, and also con-siderable practical interest.

2. SUMMARY

In ordinary life, a certain laxness in procedures is admitted—otherwise nouniversity business would ever get done. (Austin 1962: 37.)

When we speak or write, we are rarely very clear, precise, or explicit aboutwhat we mean—and perhaps could not be—but are, on the contrary, vague,indirect, and unclear about just what we are committed to. This often appearssuperficially to be an inadequacy of human language: but only for those whohold a rather crude view of what is maximally efficient in communication. Beingvague and indirect can have many uses. Politeness is one obvious reason for deviat-ing from superficially clear or rational behaviour. And claiming precision is onlydone appropriately in certain situations.

On the other hand, we do often warn our hearers that our utterances are vague.My topic is, then, that whenever speakers (or writers) say anything, they encodetheir point of view towards it: whether they think it is a reasonable thing to say, ormight be found to be obvious, questionable, tentative, provisional, controversial,contradictory, irrelevant, impolite, or whatever. The expression of such speakers'attitudes is pervasive in all uses of language. All sentences encode such a point ofview, in a sense which I wfll define more fully below, and the description of themarkers of such points of view and their meanings should therefore be a centraltopic for linguistics.

I wfll start—as a probably inevitable point of reference—from recent workin speech act theory and pragmatics, because I want to discuss questions of illocu-tionary force and indirectness. But I will end up rather a long way from a Searlianview of language, and will argue that a standard speech act view is inadequate,because it ignores the observation of language in use, and therefore ignores the

Applied Linguistics, Vol. 7, No. 1.

2 TOWARDS A MODAL GRAMMAR OF ENGLISH

pervasive indeterminacy of much language. I will try to show just what kinds oflinguistic item speakers may be committed to or detached from. I will show howthe markers of commitment and detachment require to be studied through pro-longed fieldwork, ethnography, and textual analysis. And finally I will try to showthat the concepts of commitment and detachment can explain some central syntac-tic phenomena, and can provide a way of reorganizing a grammar along communica-tive lines.

I think this view of language has significant implications for many uses oflinguistics, including: foreign language teaching; studies of cross-cultural com-munication; and the construction of computer-based 'expert systems'. Becausespace is limited, I will have to spend most of my time explaining the model, andrefer to applications much more briefly.

3. INITIAL EXAMPLESHere is an initial example of the kind of phenomenon I am interested in. The BBCradio news was reporting an explosion in a water-processing plant which hadkilled sixteen people. The newsreader made the following statement.

1 A spokesman from the Water Board refused to speculate on whether methanegas could have caused the explosion. (A)2

An initial observation is that the BBC are declining to commit themselves to theproposition, call it pi, that:

2 pi: Methane gas caused the explosion.

Their position remains as detached as possible, while still contriving to mentionthe proposition in any way at all. First, pi is modalized: methane gas could havecaused the explosion. Or: it is possible that pi. The source of this view is not stated,though someone must have formulated it. As it stands, it is a rumour from anunidentified origin: someone has said that (it is possible that?) pi. The BBC do notthemselves comment directly on even this. They cite a spokesperson refusing tospeculate, and therefore saying, in effect: no comment.

3 BBC say (spokesperson says nothing about the fact that (someone says that(possible (pi))).

It is difficult to imagine a more guarded statement, although just mentioning aproposition, especially on the BBC, might in itself give the proposition somecredence. A BBC reporter seems to have thought pi plausible enough to put it tothe spokesperson. One of the most general interpretative principles is: no smokewithout fire. That is, there is a general assumption that there is method even inapparent madness and that speakers expend the minimum effort: propositions aretherefore not even mentioned (in the technical sense) without reason. Garfinkeland Sacks (1970) discuss in more detail the view that remarks in conversation arenot understood as stating the obvious, but as drawing attention to some unusualfeature of the situation: the features must be literally 'remarkable'. (And cf. Grice1975, on the maxim of quantity: do not make your contribution more informa-tive than required; do not say what is already familiar.)

The following day, the BBC news reported:

4 According to a Water Board official, there had been a sizeable build-up ofmethane in the pipe. (A)

MICHAEL STUBBS 3

Even here, the commitment to pi is far from total. The cause of the explosion isstill not made explicit. This is left to real-world knowledge or to an inference(that methane in enclosed spaces causes explosions) and an implicature that thiscould have been the cause in this case. And the view is attributed to an official: it ispresented as an utterance report.

Several months later, the BBC reported the outcome of the inquest on theincident, a verdict of accidental death.

5 An engineering inspector told the inquest that the explosion was caused bymethane, but that it had not been possible to discover what had ignited thegas. (A)

Other statements were introduced with the prefaces:

6 The court was told h o w . . . (A)7 A Water Authority official told how . . . (A)8 One expert told the inquest it could have been . . . (A)

4. SOURCES OF PROPOSITIONSOne strategy used in the BBC statements is to attribute the views to someoneelse. This does not in itself convey either commitment or detachment. It dependson what credence is given to the source. In an example (from Gilling, in prep.)such as:

9 The noted educationalist A. H. Halsey has claimed that p (A)

we would probably interpret this not simply as an utterance report, but as anindirect assertion. That is, we would interpret it not as objectively reporting anexternal event, but as subjectively expressing a belief. Some people's words arereported as such, and it can be news when some people say 'No comment'. Butsome people are quoted to add credence and commitment to what the speakersays.

Beliefs and commitments to propositions can be attributed to different sources.Consider these examples from classroom discourse, from Torode (1976):

10 Shut up, Alan. You're a distracting member of the class. You know that,don't you? (A)

11 Somebody talking. You know what will happen. No five minute break. (A)

In these examples, p is presented as shared knowledge, known to an individualchild or to the class as a whole. (The maxim of quantity is therefore explicitlycontravened, in order to generate an implicature.) Alternatively p may be presentedas part of some enduring legitimate order:

12 Right now. I think we know the order of events. You've got to get on by your-selves today, and I don't want to see anybody off their seats. (A)

5. TOPICS FOR ANALYSISThe topics which I have so far raised for analysis on the basis of these examplesare: the different degrees of commitment which speakers (or writers) can make toa statement; the fact that propositions can be used or mentioned (quoted); the factthat statements can be attributed to different sources, such as identified and

4 TOWARDS A MODAL GRAMMAR OF ENGLISH

unidentified spokespersons; and that this can be used to add credence to a view,or to hedge a commitment; and that speakers can be explicit or vague about whatthey are saying.

None of these topics is very original in itself. Searle (1975) and others havediscussed indirection in the performance of much language. Brown and Levinson(1978) and Leech (1983) have proposed powerful theories of politeness and tactto account for much indirection in language. Goffman (1981) has deconstructedthe concept of speaker, distinguishing clearly between the animator (who producesthe sounds), the author (who selects and encodes the message), and the principal(who is committed to the beliefs expressed). Lyons (1981) has argued that muchmore attention should be paid to the subjective attitudes which speakers conveytowards the propositions they express. And Bolinger (1980) has argued that linguistsshould devote more time to analysing the way in which propositions aie presentedin the media in potentially misleading ways. The large literature on hedges and dis-claimers is also relevant: Lakoff (1972) raised many of the basic points in his nowclassic paper. (Other interesting papers include Fraser 1975; Katriel and Dascal1984, etc.).

What seems to be missing in much work, however, is a combination of: (1) aconceptual analysis of commitment and detachment; (2) a linguistic description ofa corpus of data; (3) an ethnographic description of the contexts in which differentlinguistic forms are used; (4) an analysis of the effects of such pragmatic factors onthe grammar; and (5) a discussion of the practical importance of the questions raised.

In summary, I am going to propose in this paper that the expression of com-mitment and detachment, or of modality in all its senses, can be seen as a centralorganizing principle in language.

6. PROPOSITIONAL, ILLOCUTIONARY AND LEXICAL COMMITMENTSnoopy (in a 'Peanuts' cartoon by Charles Schulz) is writing a story:

'Our love will last forever,' he said. 'Oh, yes, yes, yes!' she cried. 'Forever beinga relative term, however,' he said. She hit him with a ski pole.

My initial example concerned the guarded expression of propositional informa-tion. In fact, it is possible to indicate degrees of commitment to just three kinds oflinguistic item: not only (1) to propositions but also (2) to illocutionary forces and(3) to individual lexical items. This is perhaps most visible for illustrative purposes,in utterances where speakers first commit themselves to something, and thenwithdraw the commitment The fact that speakers can be seen to shift in theircommitment motivates two layers of meaning, the content of utterance in somesense, and the speaker's attitude to this content. For example,

13 We move in on Thursday—all being well. (A)14 I promise I'll come—if I can. (I)15 . . . consumer durables—as the Economist calls

them, whatever that means. (A:L)

It is possible in such examples to see speakers commit themselves to a proposition,an illocutionary force, and a lexical item: but also to observe surface markers ofdetachment This therefore opens up questions such as: What markers are available?And how and when are they used? The answer could be found only by both textualand ethnographic study.

MICHAEL STUBBS 5

7. PROBLEMS WITH SPEECH ACT THEORY

Marcie: You bought a wig? . . . No, I promise not to laugh. . . . No, I promise,I really do. I promise, yes, I really promise—yes, I promise not to laugh . . .I really, really promise.

Patty: Say it one more time. (Charles Schultz: Peanuts)

On the face of it, speech act theory should be a good starting point for work inthis area, not only because it has a lot to say about indirect speech acts, but alsobecause it has a lot to say about explicit illocutionary verbs as markers of com-mitment. The main problems are a lack of observation of how such markers areactually used; and the fact that illocutionary verbs are seen as a special category.

Speech act theory has been, historically speaking, very ambivalent in its attitudetowards naturally occurring language in use. Austin (1962: 148) suggested that amain task was to collect a list of explicit performative verbs (as a guide to possibleillocutionary forces) and that this was a 'matter of prolonged fieldwork'—hencethe title of this paper. However, what he seems to have had in mind by 'field-work' was looking through a dictionary—'a concise one should do', he says.Despite Austin's odd view of fieldwork, his theorizing is deliberately urbane andunrigid, based on observations about everyday language at least to some extent, andvery laid-back in the way it deconstructs its own argument as it goes along. Searle's(1969, 1979) systematization of Austin is much more rigid, and moves furtheraway from actual behaviour, leaving speech act theory in the odd position ofdemanding a study of language as social action within a theory of social institutions,but studying hardly anything but invented data. For example, Austin has a built-in interactional emphasis which is lost in Searle and only recovered in work suchas Handler's (1979). Lyons (1981: 172) has actually pointed out that the termlanguage act would be better than speech act, therefore emphasizing the abstract,decontextualized view of language in Searle's work. Speech act theory also clearlyapplies to written language, as well as spoken, but is rather unconcerned with thedifferences between spoken and written language. It appears to be based on thedubious view that language can be studied independently of its medium of trans-mission. (Chomskyan linguistics is at least explicit about this: speech act theory isnot.) But written and spoken language have different possibilities for commit-ment: consider requests to 'let someone have it in writing', or the particular typeof commitment made possible by the use of a signature—for which writing is anobvious prerequisite (Stubbs 1983b). And explicit illocutionary verbs are muchmore likely to occur in formal writing than in casual speech (see below).

In this paper I propose to return to Austin's suggestion of prolonged fieldwork:but by observing instances of usage of illocutionary verbs, not by looking for themin a dictionary; and by looking also at other related markers within a more generalframework of commitment and detachment to propositions, illocutions, and words.

One of the oddest features of Searlian speech act theory is that the use ofexplicit illocutionary verbs is regarded as normal, or 'canonical', in Searle's ownterm; whereas a more common-sense point of view is that non-performatives aresimpler and more usual. Consider two utterances such as

16 111 go tomorrow.17 I promise I'll go tomorrow.

Clearly, both can be promises, since both can commit the speaker. But Searle

6 TOWARDS A MODAL GRAMMAR OF ENGLISH

appears to regard such pairs as equivalent. And in addition he regards the explicitperformative as the typical, paradigm, canonical case. I think he is wrong on bothcounts. They are not equivalent, since the first could have vague or indeterminateforce: it could be a prediction, a suggestion, a warning, or some combination ofthese, as well as a promise. And the second is marked stylistically, as being lessambiguous, more determinate. (The commitment can still be withdrawn by thespeaker saying, for example, . . . if I can, or by the addressee saying... no, don'tbother, I'll go).

A better question is: why should speakers ever bother to use the explicit form?Why should they ever refer metalinguistically to the act they are performing, andtherefore to their own commitment? (These are questions posed by Leech 1983;though he does not propose the kind of answer I do here.) An obvious answer,very brief for the moment, is that there are some circumstances which requirespeaken to be explicit and unambiguous and to go on record about what is beingsaid. There may, for example, be reasons of politeness or reasons to do with thespeakers being in some official situation. The distinction between object languageand metalanguage, and the ability of language to refer to itself, is often taken as atopic for philosophy and artificial intelligence. But in fact metalanguage is aneveryday phenomenon, though commoner in some situations than others: it is,for example, very common in school classrooms (Stubbs 1983a: chapter 3). Itsinvestigation therefore requires ethnographic study. It is the job of pragmaticsto explain the relationship and difference between explicitly performative andnon-performative utterances, and not just to assume their equivalence, as Searledoes.

All of this means that any version of the performative hypothesis (£ la Ross),which places a performative verb in the highest node of all sentences, must bewrong, since it is the performative sentence which is atypical. Transformationsare usually assumed not to change meaning, and deletion of a performative insurface structure would therefore have to be assumed not to change meaning:but I have just argued that the presence or absence of an explicit performative verbin surface structure does change meaning. If nothing else, I would hold to thegeneral principle that a difference in form must involve some difference in mean-ing—and certainly when it involves the presence or absence of something asobvious as a lexical verb.

8. COMMITMENT AND DETACHMLNT

Performative verbs are just one of a set of ways of being explicit or remainingvague. Part of my argument is therefore that this topic within speech act theoryshould be just part of a more general theory of commitment and detachment.

So far, I have simply used these terms and illustrated them. We now requirea more detailed definition. I am assuming that there is a continuum of commitment,whose end points are complete commitment and complete detachment. Withreference to propositional information, full commitment could be made by thecategorical assertion that p is the case. By complete detachment from p, I do notmean an assertion that not-p: this could involve full commitment to not-p. Com-plete detachment involves, rather, some kinds of quotation or mention. There areeveryday expressions for this, as when we talk of just 'entertaining' a proposition,

MICHAEL STUBBS 1

'for the sake of argument'. I do not think that the number of points on the degreescale is an important issue, since, as I have already started to indicate, I think thatmany utterances are by their nature indeterminate and could not be placedat some particular point on the scale. (It may be possible, in principle, just tomention a proposition, 'for the sake of argument'. But in practice, in a particularcontext, a general interpretative principle will search for the reason for themention: cf. example (1) above. The extreme end point of the detachment scaleis elusive.)

What I have said above implies that an utterance such as

18 It could be that p.

will leave the speaker only partially committed to p, rather than fully committedto possibly p. This is the difference between subjective and objective epistemicmodality (see Lyons 1977, 1981). Since subjective modality is much commonerin language use, I have defined the terms as I have done.

Degree of commitment is distinct from manner of commitment. For example,a speaker is committed to the same degree to a proposition p whether p is assertedor presupposed:

19 (I assert that) p.20 I realize that p.

Both convey full commitment to p. The test is simply that the speaker could notthen deny p without logical contradiction. But in one case p is presented within theillocutionary act of assertion. In the other case, p is presupposed by being embeddedunder a factive verb. Other manners of commitment include implicatures asopposed to assertions, or using many different kinds of lexical or syntactic markersof commitment and detachment. Some manners of commitment are, by theirnature, deniable and therefore less than full. A great deal of recent work in prag-matics has been concerned with establishing the difference between propositionswhich are logically entailed by what is said, versus those which are defeasible incontext. (Levinson 1983 gives a thorough review.) I am taking an implicature to bea proposition which is not asserted, but is inferred and therefore deniable.

It might be objected that just asserting p does not involve full commitment,since a speaker might say:

21 I'm absolutely sure that p. (I)22 I really do honestly believe that p. (I)

However, such forms cannot increase the commitment beyond the logical maxi-mum. What they do is perform the discourse function of responding to a previousutterance, by countering a previously expressed uncertainty.

Linguists have no special insight into what propositions are true, or even whatpropositions speakers believe are true. Apart from anything else, speakers mayexpress an argument which they believe to be false, in order to persuade an audiencefor some rhetorical purpose. This is a common pedagogic tactic: to follow througha line of argument, to some logical conclusion, in order to show that it is false.There will generally be the implicature that if a proposition is presented as beingtrue, then the speaker believes it to be true (Leech 1983: 34). All that can be

8 TOWARDS A MODAL GRAMMAR OF ENGLISH

studied, therefore, is how speakers present their attitudes: what public display theyprovide of their commitment; and how the status or role of speakers affects theircredibility.

Commitment has to do therefore with whether a proposition is presented astrue, false, self-evident, a matter of objective fact or of personal opinion, sharedknowledge, taken for granted or debatable, controversial, precise or vague, contra-dictory to what others have said, and so on. We are dealing with what Lyons(1981: 240) refers to as self-expression, in the literal sense of the speaker's expres-sion of him or herself: the subjectivity of utterances—how speakers report theirown beliefs, attitudes, and so on. Lyons claims that relatively few utterances makeunqualified assertions (certainly this one does not). (And he points out that in somelanguages, it is not even possible to do so.) But to discover just how many utter-ances are qualified, and to what degree and in what manner in different contexts,is a matter of prolonged fieldwork.

There is another major aspect of commitment which I will not deal with here:namely, that commitment may not be the result of one proposition from onespeaker, but the outcome of several utterances from one or more speakers(cf. Handier 1979; Edmondson 1981).

The above points apply pari passu to illocutionary commitment. For example,if an illocutionary force is indirect or off the record to some degree, it will bepossible to claim, if challenged, that it was never issued. And similarly with lexicalcommitment

9 . EXPLICITNESSWe need one more piece of conceptual apparatus and then we can discuss somemore data. The concept of exph'citness is closely related to the concept of com-mitment Explicitness implies commitment, since if you are explicit, then you statesomething openly and publicly and go on record; whereas inexplicitness impliesvagueness and therefore deniability. A standard starting point for analysing utter-ances as speech acts is the formula F\p), where F is the illocutionary force whichmay be marked by an illocutionary force indicating device, such as an explicitillocutionary preface, and p is a proposition. Austin (1962: 61-2) talks, forexample, of a performative being expandable into a form with a first-person singularpresent-tense verb, and of such expansions making explicit both that an utteranceis performative and also which act is being performed; and again (1962: 103) of anillocutionary act being made explicit by a performative formula.

There have been two related themes in speech act theory: the indirection argu-ment, that the surface lexical, syntactic form of an utterance often does not makeexplicit the illocutionary intent of the speaker; and the expressibility principle,that the illocutionaiy force of an utterance can always be made explicit:

Wherever the illocutionary force of an utterance is not explicit it can alwaysbe made explicit . . . whatever can be meant can be said. (Searle 1969: 68;cf. 1979: be.)

Despite his emphasis on indirect speech acts, Searle's view of language often seemssimplistic and unsubtle, and he seems to go badly astray in his formulation of theexpressibility principle. First, it is untrue that all speech acts can be made explicit:there cannot be explicit hints, for example. In general, speakers cannot be explicit

MICHAEL STUBBS 9

if they want the commitment to be deniable. Second, being explicit is hot a merestylistic preference: it is something which is done in only some social settings, andthis needs to be specified by much detailed ethnographic work. It is chajracteristic,for example, of written rather than spoken English. Anyway, as Garfinkel (1967)argues in detail, it is impossible ever to say in so many words exactly what is meantby any utterance. The fact that we can never say explicitly all that we mean impliesthat much always has to be taken for granted, and this leaves the way open to deceitby omission. I do not mean to adopt a mystical position that there are things ofwhich one cannot speak, but only to point out that being explicit changes themeaning.

Explicitness is a difficult concept It does not mean saying all that can be said(which is impossible), but finding the right balance between what is said andwhat can be assumed, and therefore not giving more information than is neededor wanted (cf. Grice 1975 on the maxim of quantity). It can therefore only bedefined relative to particular contexts. The distinction between what has to bemade explicit and what can be assumed therefore has implications for what isconveyed about group membership. (This was clearly recognized in Bernstein'swork on restricted and elaborated codes, although this insight often gets lostamongst others: for example, social class is not the only group or context whichis relevant)

A standard example of explicitness is a computer program: everything must bespelt out, with no steps missing. But exactly how explicit you have to be dependson which language you are using. Fortran requires variables to be declared inadvance, while other languages allow them to be taken for granted. Furthermore,any high-level language, such as Fortran, takes for granted a low-level machinelanguage which makes things explicit to the hardware. It follows that explicitness,clarity, ambiguity, etc. are not inherent properties of texts, but are a function oftexts in contexts (Nystrand, in prep.). And some contexts (e.g. the law) are lesstolerant of ambiguity than others.

10. EXPLICITILLOCUTIONARYPREFACES'.Snoopy has received a rejection letter for one of his stories:

Dear contributor,We regret to inform you that your story does not suit our present require-

ments. On second thoughts . . . Actually we don't regret it at all.

One point which I made above was that performative verbs are used in somecontexts rather than others, and that their use is actually rather restricted. Wetherefore have to consider why and where they are used. One context whichproduces a large number of explicit performatives is the formal written languageof business correspondence. (It may also be that the fprms discussed are recog-nizably British rather than American.) One small corpus I have studied consistsof about 500 explicit illocutionary prefaces from such letters, including such forms,all (A:B), as:

23 Further to my letter of 25 February, I would advise you that p.24 With reference to my letter of 19 May, I am authorized to inform you that p.25 In relation to x, I was merely seeking to point out that p.26 I would however draw your attention now to the following regulation.

10 TOWARDS A MODAL GRAMMAR OF ENGLISH

27 May I wish you a successful and interesting conference.28 I emphasize that p.29 Let me say again how sorry I was that p.30 X, I'm sorry to say, died several months ago.31 A quick note to tell you that p.32 I would suggest that p.33 You have my consent for p.

Perhaps the most striking feature of this corpus is that the first-person singularsimple present-tense form is rare, being restricted to certain verbs and/or to veryformal contexts. / apologize is quite common, but this is the only verb whichoccurs commonly in this form. / thank you co-occurs with Dear Sir in a letterfrom a bureaucracy. / hereby certify occurs on a legal form. We announce formally(not singular, of course) occurs in a letter about a company merger; and the actualword formally occurs.

One major point is therefore the stylistic implication of these forms. They arenot a 'paradigm device' as Searle (1969) calls them, but are largely restricted toformal settings. The word hereby, for example, is almost entirely restricted towritten legal settings: it does not occur once in the London-Lund corpus of halfa million words of spoken British English. (This throws doubt on its use as a testfor performative verbs, since its use will therefore disturb reliable intuitive judge-ments.)

The commonest surface form is modal phis lexical illocutionary verb, oftenreferred to as a hedged performative (Fraser 1975):

34 I would advise you that p.35 I think we should decline your offer.36 X and I would like to extend to you our thanks and very good wishes.

But there are very many others, and the next striking feature of the data is thesurface variation, what Brown and Levinson (1978) refer to as a 'baroque ensemble'of forms for performing indirect speech acts. This in itself makes the use of intui-tive data very dubious, since intuitions about linguistic variation are notoriouslyunreliable.

One often ignored feature of this variation is that there is a tendency to talk ofillocutionary verbs, although illocutionary nouns and other parts of speech alsooccur.

37 This is our suggestion . . . (A:B)38 I'm only putting it forward as a suggestion. (A:L)39 Well, look, honestly Mrs Finney, my suggestion to you would be that if... (A:L)40 In answer to your second question . . . (A:B)41 Congratulations! (AS)

Another observation is that speakers explicitly distinguish between differentdegrees and sources of commitment.

42 I would like to thank you, officially for the Association and personally forme. (A:B)

MICHAEL STUBBS 11

43 This is very much a preliminary letter . . . (A:B)

And they speak and write on behalf of other people.

44 I am writing at the request of the Board to invite you to . . . (A:B)45 I am writing on behalf of x to thank you very much f o r . . . (A:B)

Nor do illocutionary forces have to be conveyed directly to the addressee. Theycan be passed around. The following examples are from spoken business settings:

46 Perhaps you would pass on my apologies. (A:L)

That is, A asks B to pass apologies to C.

47 Could you give Professor Worth's apologies for the Mathematical Societymeeting on Friday. (A:L)

That is, A asks B to pass C's apologies to D. When exactly do such illocutionaryforces come into operation? Perhaps not till months later, when they appear inwriting in the formal minutes.

A study of such forms is part of the theory of: (1) what can be referred to inconversation: linguistic forms, propositions, illocutionary forces, topics (Lyons1977: 667 ff.); (2) utterance reports and their truth conditions (Gazdar 1981);and (3) the distinction to be made between the person who actually speaks and theperson who is thereby committed (Goffman 1981).

The linguistic style of business letters is also a topic of practical importance,for example in courses of training in business English, possibly in an EFL context,or as part of campaigns for plain English. Sir Ernest Cowers' (1962) very influentialbook The Complete Plain Words contains what is in effect an analysis of hedgedperformatives. He criticizes such forms as: / would inform you that... or / have toinform you that . . . on the grounds that they are 'crushingly stiff, and, centralto the topic of my paper, that they obscure the source of the commitment, giving,as they do, the impression of a remote bureaucracy staffed by robots. There is aclear recognition by Gowers that illocutionary verbs occur in particular settings,that they do not always occur in simple present-tense forms, that they can beused both to put on record and to obscure commitment, and that language cantherefore both convey and obscure messages.

The question of who is committed to what is also unclear, because of differentdegrees of commitment. One might get the impression from reading Searle that apromise is a promise. However, illocutionary forces are not categorical, but scalarand often indeterminate (Leech 1983 discusses this in detail). For example, it ispossible to refer to less than fully committed acts, such as half-promises:

48 . . . having more or less promised . . . (A:L)49 He'd sort of had half promises. (A:L)

A tentative promise might be made by uttering:

50 I may come—but don't count on it. (I)

Or consider the following tentative illocutions:

51 . . . so is it possible to say provisionally yes and that I will confirm as soon asI can . . . I'll ring again to definitely confirm it. (A:L)

12 TOWARDS A MODAL GRAMMAR OF ENGLISH

Are the following utterances offers or not?

52 We'll be offering the course subject to the availability of staff and facilities. (A:L)53 I would be interested to offer a course of lectures next session. (A:L)54 The cheapest I can offer you at the moment is Sigma, which is priced at

forty-nine fifty. (A:L)

Are the following invitations?

55 If you are ever in this part of the world, I'd be delighted to invite you to give alecture. (A:B)

56 I hope you will be able to attend this weekend, for which you will be receiv-ing an official invitation soon. (A:B)

I do not think that these questions are answerable. The whole point is oftento be negotiable, deniable, indeterminate. Speech act theory has ignored suchexamples, due to reliance on introspective data which do not reflect such indeter-minacy. It is therefore being currently undermined by the study of actual languagein use.

11. TWO TYPES OF SPEECH ACTHowever, it is not possible to prevaricate or give less than full commitment toall types of speech acts. Consider the following data, all (I):

57 He almost excommunicated me.58 He almost promised to come.59 He did sort of christen the child.60 He did sort of promise to come.

(57) means that he did not excommunicate me: perhaps he changed his mind atthe last minute. (58) is ambiguous: perhaps he changed his mind at the last minute,or perhaps he entered into some kind of commitment. (59) seems to imply anunconventional ceremony. (60) again could mean that he entered into some com-mitment.

These data motivate a distinction between two types of act. Type 1 is institu-tional and conventional, and therefore not illocutionary at all: because they couldnot be performed by any speaker of the language, but only by someone by virtueof occupying some social role. That is, they are not really part of the language assuch. For example, one must be specially authorized in order to christen or ex-communicate people, appoint them to or fire them from jobs, name ships, sentenceoffenders, declare war, and so on. These are all declarations in Searle's (1976)sense, in which saying really does make it so. They bring about a correspondencebetween words and the world, due to consciously formulated (and therefore notlinguistic) conventions.

Type 2 can be performed by anyone whose English is good enough to conveytheir intention: that is, anyone can make promises, requests, complaints, etc.

Although Searle (1969: 71) claims to be setting up an 'institutional theory ofcommunication', and distinguishes between brute and institutional facts, he doesnot make this distinction between two types of act (and therefore misclassifiesdeclarations as a type of illocutionary act).

MICHAELSTUBBS 13

12. LEXICAL COMMITMENTAs well as the cases of prepositional and illocutionary commitment discussed sofar, there are also cases of lexical commitment versus detachment. Items such asso called, as I would call it, so to speak, and quote unquote provide obvious surfacemarkers of detachment. I have in mind examples (all A:L) such as:

61 one of the so-called binary star systems62 this so<alled improvement in the standard of living63 hospital of tropical diseases or whatever it's called64 the old idea of a certain code of behaviour, what I would call being a gentle-

man65 what they call musicology66 I was prepared to push the others under the carpet so to speak in order to

make the marriage work67 we came to Minsk and there we dug in so to speak for the winter68 not the person who needs to be quotes treated69 people who were in inverted commas distress

The meaning of such items is complex, but the various cases do seem to havesomething in common. The use of such markers of detachment means that a lexicalitem is being mentioned rather than, or as well as, being used. They indicate insome way that the meaning of a lexical item is problematic: there may for examplebe no consensus about the meaning of a term, or its meaning may be technical,or unknown to the hearer. In some way, there is an admission that the meaning orinterpretation of a lexical item differs in different groups or with differentspeakers. The speaker may be disassociating him or herself from some group, ornot assuming that the hearer is a member of some group. They are among theinnumerable ways of conveying in-group membership. Though a practical problemis often: who is called x by whom? who is the source of the description? And thereare many different ways of using such terms.

The form

(as)y[what"! x

Las J"calldescriberefer

is very common in the Lund corpus: over fifty instances in about half a millionwords of 'adult British educated spoken usage'. The forms so-called and so to speakare also common: over twenty instances. (These figures are high relative to thefrequency of lexical items in a corpus of this size.)

Note also that such forms can achieve reference without being committed tothe truth of the referring expression. Referring expressions allow a hearer toidentify a referent: that is, the hearer must be able to infer that the expressionapplies to the referent (Searle 1969: 88; Lyons 1977: 177). Speakers can obviouslyuse many different referring expressions to refer to the same referent. It followsthat speakers can use different forms in order to convey other information, forexample, to pick out some feature of the referent rather than another, to conveythe speaker's attitude to the referent, to convey group membership by choosinga description that the hearer does or does not know, and so on. Since reference isutterance-dependent, the actual referring expressions which speakers use can bestudied only by fieldwork and textual analysis (Schegloff 1972).

14 TOWARDS A MODAL GRAMMAR OF ENGLISH

Consider the following more complex examples (from Gilling, in prep.):

70 Today's so-called teachers are themselves frequently uneducated. (A)71 . . . the anti-social behaviour of these so-called women. (A)

It is clear that the words teachers and women are not being used purely descrip-tively. The referents are teachers and women. But some of the normal commit-ments to these words are suspended. A moral point of view is being conveyed.(Cf. Lakoff 1972: 197 ff. for other expressions which can suspend part of themeaning of words, e.g. technically, strictly speaking!)

The strategy has been called the no-true-Scotsman move (Flew 1975: 47).Imagine a Scot who reads in his newspaper about the dastardly exploits of anEnglish sex maniac. He might say to himself: 'No Scotsman would ever do sucha thing'. But then suppose he reads, the next week, of Angus McSporran, whohas committed even more dastardly deeds in Aberdeen. He might then argue:'No true Scotsman would ever do such a thing'—thus converting his Initial state-ment to one which is irrefutable, since it is now true by arbitrary definition.

13. VAGUE LANGUAGEOther cases of extreme denotational vagueness are provided by examples such asthese (again all A:L):

72 I'm going to be in and out of libraries and things today.73 Don't get yourself worked up into a state and run into lamp-posts and things.74 . . . discussions of world food problems and things like that.75 . . . a great horsehair sofa and that kind of thing.76 The boys aren't left to do the washing up and that kind of thing.

Channell (1983) discusses what would be a suitable semantics for vague expres-sions. She points out that such examples pose problems for truth conditionalsemantics, since it becomes impossible to specify when utterances containing themcease to be true. The vagueness does not disappear even in context: it is an inherentproperty of language. Since the denotational range of all lexical items is arguablyvague, the same point could be made of all utterances: it is just more striking incases containing markers of lexical detachment (vague category identifiers, inChannell's term). Channell's observations are also based on a prototype theory oflexical meaning in which category membership is a matter of degree from typicalto marginal (cf. Lakoff 1972; Rosch 1975).

Channell points out further that the interpretation of such utterances dependson the discourse context: an informal discourse context is likely to demand lessabsolute accuracy in denotation, although a discussion of, say, the performance ofmotorcycles may demand precision even in informal contexts. That is, socioUnguis-tic conventions affect semantic representations. Austin (1958: 12) pointed out thattruth depends on the use and purpose of utterances. For example77 Italy is shaped like a boot and France is hexagonal. (I)

is accurate enough as a mnemonic for schoolchildren, but not for geographers andtour operators.

The same problems for semantic representation are raised by number approxima-tions (such as there were about half a dozen or so people there). These are discussed

MICHAEL STUBBS 15

by Lakoff (1972) in his article on hedges and fuzzy logic. (Cf. Channell 1980 forfurther discussion.)

Such examples appear to argue for degrees of truth (this is LakofTs view), andtherefore a many-valued logic, which is dependent on different sociolinguisticconventions. But I have argued further that speakers can themselves specify thestandards against which they wish the truth of their utterances to be judged.Markers of commitment and detachment are instructions to interpret utterances inmore or less rigorous ways. Speakers project their own definition of the situation.I have also argued that truth value is not the only concept we require. We alsorequire a concept of state of information, which gives not only the truth of pro-positions, but also their source (cf. McCawley 1981: 385; Belnap 1977), andconsequently the degree of confidence we can have in them.

14. PRAGMATICS AND GRAMMARSo far, I have tried to show that all utterances express not only content, but alsothe speaker's attitude towards that content. This claim may seem so general that itis self-evident or true by definition. The claim would be much stronger if I couldprovide independent motivation for it, by showing that the expression of commit-ment and detachment can interact with syntactic and morphological form. Ideally,one would be able to show that the kind of pragmatic concept which I havediscussed can provide a unified functional motivation for many otherwise dis-parate features of surface syntax. The strongest type of argument would, in addi-tion, provide an explanation for previously unexplained syntactic phenomena.

15. SIMPLE VERSUS -ING FORMS OF VERBSIn fact, I think it is possible to show that matters of truth, or degree of confidencein truth, can interact with syntactic form.

Probably the most semantically complex area of English syntax is the verbalgroup. One of the biggest puzzles in English syntax is the meaning of pairs such as:

78 I go/1 am going; I warn/1 am warning..

I will use the terms simple form and -ing form to refer to the surface morphologyor syntax. We have to distinguish clearly between surface forms and meanings,since these forms can be used as markers of tense, aspect, and, as I will argue here,of other things as well. It would thus be prejudging the issue to use a term such as'simple present-tense form'.

It is well known that performative verbs differ from other verbs in the relationbetween the two forms (Austin 1962: 64). Thus, with many verbs, reference to themoment of utterance is made using the -ing form (I'm working, go away), but per-formative verbs use the simple form (/ promise). A common view is that performa-tive verbs are odd in this respect. It is also well known that stative verbs often donot admit the -ing form. But they are regarded as an exception among verbs(e.g. Quirk and Greenbaum 1973: 15, 21). However, these observations leaveunexplained the relation between performative and stative verbs. And in additionthey disguise the fact that many types of verbs take simple forms, either exclusivelyor regularly; and this therefore leaves us with no explanation of the forms.

The following are main categories (all examples are (I) except where stated).

16 TOWARDS A MODAL GRAMMAR OF ENGLISH

15.1 Verbs of cognition

79 I know/*am knowing he's right. (Cf. believe, realize; suspect, think)

Note that an -ing form can indicate change of state:

80 He's understanding maths much better these days.

15.2 Verbs of perception

81 I love/? am loving it.

82 That tastes/*is tasting funny. (Cf. feels, looks, smells, sounds.)

A form such as I'm loving it probably implies recent change (cf. 15.5):

83 I've just started teaching and I'm loving it.

15.3 Verbs of conveying and receiving information

84 I hear/*am hearing that Harry's dead. (Cf. see, understand.)

Note that the simple past-tense form would imply that the information is no longerassumed to be correct:

85 *I hear that Harry is dead, but he isn't.86 I heard that Harry was dead (but he isn't).

15.4 Definitions, external truths, habitual states, permanent and timeless states

87 A triangle has/*is having three sides.88 Ice floats/*is floating on water.89 Harry smokes/*is smoking after dinner.

15.5 Permanent versus impermanent or recently changed states

90 He is/is being silly.91 We live/are living in Scotland.

15.6 Relational verbs and verbs of permanent state

92 I own/*am owning six cars.

93 It contains/*is containing arsenic. (Cf. belong, depend, deserve, matter.)

Again note that an -ing form can indicate changed circumstances:

94 More people are owning their own houses nowadays. (A)

15.7 Future states of affairs which are predictable, possibly because they are partof some scheme

95 The exams start/?are starting on Thursday.96 We leave/? are leaving on holiday next week.

A case where a simple form would seem absolutely impossible is where it refers toa future event which is inherently undecidable by its nature, such as a sports event;hence the oddity of:

MICHAEL STUBBS 17

97 'Scotland beat France tomorrow.

But the oddity of even such cases disappears if the context makes it clear that afixed schema is involved:

98 I've fixed everything: bribed the referee and the linesmen—Scotland beatFrance tomorrow and lose to Germany next week.

And note how an -ing form can be used to indicate a hypothetical statementabout a future event:

99 We should be seeing the results soon.

15.8 Verbs in demonstrations and commentaries

100 I take/?am taking six eggs . . .101 I place/?am placing the rabbit in the hat.102 Matthews passes/?is passing to Lumley.

15.9 Headlines, captions below pictures, etc.

103 Cook lands/?is landing in Australia.

15.10 Stage directions

104 Howard exits/*is exiting stage left..

15.11 Summaries of stories, narratives in 'historic present'

105 So she comes/*is coming in and hits/*is hitting him over the head.

15.12 Performatives

106 I promise/*am promising to come.

But see the large literature on performative -ing forms (e.g. Edmondson 1981).There are other parallels between performatives and some of the above catego-

ries. For example, several categories can take non-literal can.

107 I can promise you.108 I can smell gas.109 I can understand what you mean.110 Ice can float on water.

The question is therefore: what do all these forms have in common? The answeris that they all report events which are necessarily, habitually, or eternally true,which are certain or predictable or presupposed to be true, or which are unchallenge-able in some way. For example, some report A-events (Labov and Fanshel 1977):that is, events to which the speaker has privileged access. If I claim that I feelhappy, you may accuse me of lying, but you have no way of checking on thetruth of my claim. Therefore some scholars have used the terms 'private verbs'or 'mistake-proof sentences' (Ljung 1980: 50 ff.). Other events are unchallengeablebecause the speaker has some special expertise: in the case of radio commentary,the hearers may not be able to see the original events themselves. This is essentiallyPalmer's (1974: 60 ff.) argument, that the simple form merely reports in those

18 TOWARDS A MODAL GRAMMAR OF ENGLISH

special cases where we need to report present activity: normally we do not have to,since present activities are observable. However, Palmer does not relate his analysisto the debate within speech act theory over how performatives can best be analysed.Within the approach which I am arguing here, performatives can be analysednaturally as reporting propositions which are true by virtue of being uttered.If I say that I promise, then it is true that I have promised even if I have no inten-tion of keeping my promise: the commitment has been made (cf. Kempson 1977for a truth conditional analysis of performatives).

In summary, the simple form in all these cases conveys that speakers havespecial reasons to be confident in the truth of the proposition they are expressing.In some cases, they could not, logically speaking, be wrong. In all cases, theirconfidence derives from more than publicly accessible observations.

(Note that these points mean that performative verbs are not a special class,and are especially difficult to distinguish from 'private' verbs. I do not think thatthe hereby test is reliable: see Stubbs 1983b. The only test I know which distin-guishes is that performative verbs can take optional you: /promise (you) III come;*I deplore you what he's done)

16. PSYCHOLOGICAL AND MODAL INTERPRETATIONS OF 'PRIVATE' VERBSThere is a second case of a relation between pragmatics and syntax which I willmention much more briefly (but see Gilling, in prep.). A superficially obviousmarker of tentativeness is 'private' verbs such as: believe, think, suspect, expect,and so on. However, sentences containing such verbs are ambiguous. The basicproblem is evident in such sentences as:

111 I think he's gone.

This is unlikely to be a fully committed statement about the speaker's personalbeliefs, but more likely to be a tentative assertion that 'he has probably gone'.Similarly,

112 That looks red.

could be a dogmatic assertion about an appearance, but could also be a qualifiedassertion that 'it is probably red'. Thus 'private' verbs can be used to make state-ments about internal psychological states. But they can also have modal uses whichrelease speakers from total commitment to propositions. It can be difficult torecognize which use is intended on some occasions, and this can depend on thecontent of the proposition. For example,

113 I believe the situation in South Africa is now in a state of ferment. (A:L)114 I believe that murderers should be hanged—provided they're sane, of

course. (A:L)

would normally be taken as psychological reports, though there is nothing in thesurface syntax which signals this. In fact, one test is that in (113) and (114) thedeletion of I believe alters the meaning very little, if at all. Contrast (111), wheredeleting / think does alter the meaning. One rule of interpretation is that if theproposition is not empirically verifiable, then the utterance will be given a psycho-logical interpretation.

However, on other occasions, the syntactic behaviour of these verbs does reflectthese different interpretations. For example, the structure

MICHAEL STUBBS 19

115 I believe in NP

can only be used as a psychological report. Whereas the modal use is signalled bythe use of so or not as a dummy clause in structures such as:

116 I believeLn°otJthink

expectguesshopeimaginesuppose

It seems

Modal use is also signalled by utterance-final / believe, etc. Hence the oddity of:

117 Does God exist??I believe so.

118 ?God exists, I believe.

17. LOGICAL AND PRAGMATIC CONNECTORSA third case is provided by the so-called logical connectors (e.g. and, but, or, if,because). Their uses in everyday English are not reducible to their logical func-tions in the propositional calculus, but have to do with speakers giving reasons forhaving confidence in the truth of assertions, or otherwise justifying speech acts.The main notions are evident in the behaviour of because.

Because has two uses, logical and pragmatic, as in these examples:

119 He was drowned because he fell off the pier.120 He was drunk, because he fell off the pier.

The first has the structure effect phis cause. The second has the structure assertionphis justification. The second, pragmatic, use is often signalled by epistemic must:

121 He must have been drunk, because he fell off the pier.

There are several syntactic tests which distinguish between these two uses. Prag-matic because does not allow reversal of the clause sequence, clefting of the becauseclause, or yes-no interrogation of the whole sentence.

122 *Because he fell off the pier he was drunk.123 *It is because he fell off the pier that he was drunk.124 *Was he drunk because he fell off the pier?

(I am assuming that the last example is spoken as a single tone group.) The London-Lund corpus contains a large number of examples of utterances with pragmaticbecause, e.g.:

125 . . . but wait till you have a baby cos you'll find sort of dirty nappies in everycomer and sort of banana skins . . . (A:L)

Similar points hold for pragmatic if, or, but, and and. Consider these examples:

126 There's some food in the fridge, if you're hungry. (I)127 There's some food in the fridge, or aren't you hungry? (I)

20 TOWARDS A MODAL GRAMMAR OF ENGLISH

where the if- and or-clauses provide reasons for making the statement. Or:

128 Let's eat.But I'm not hungry.

where the 6ur-clause questions the justification of the preceding utterance.There is, of course, a very large literature on connectives. Johnson-Laird (1983)

conveniently summarizes much psycholinguistic material. Davison (1975), Van Dijk(1979), Morreal (1979), McTear (1980), Stubbs (1983a) provide many furtherexamples and discussion. Channell (1983) has also pointed to the non-logical use ofor in vague expressions, such as Could you get some apples or oranges or something.The items linked by or must be recognizable as members of the same set.

18. COMMUNICATIVE GRAMMARIdeally, grammars should be organized in such a way as to reflect the communica-tively important things which speakers do with their language. Only in this way canwe have functional explanations of the vast range of surface syntactic phenomena.In fact, I think it is possible to show that many features of surface syntax have thefunction of presenting speakers' attitudes to propositions, illocutions, and words.Individual cases are, of course, widely discussed, but they are seldom if everbrought together into a unified description, in what could be called a modal gram-mar of English.

An obvious case of items which function to express speakers' attitudes to thetruth of what is said is provided by sentence adverbs such as frankly, obviously, andso on (see Greenbaum 1969, on attitudinal and style disjuncts). But many othersyntactic features have this function too. Passivization allows the deletion of theagent, and therefore avoidance of commitment to certain prepositional informa-tion. The uses of some and any (as R. Lakoff, 1969, points out) can signal speakerexpectations: in questions some signals the expectation of a positive answer (Hassomeone come?), whereas any remains neutral or signals the expectation of a nega-tive answer (Has anyone come?). A large number of constructions can trigger pre-suppositions: including factive verbs, cleft sentences, and non-restrictive relativeclauses (Levinson 1983: 181-4, after Karttunen). Nominalization can turn actionsinto static 'things' and therefore attribute objective reality to states of affairs.

Past tense forms signal unreal or hypothetical states of affairs, for example, incounter-factual conditionals which commit the speaker to the falsity of a proposi-tion (// only he was here . . .). Past tense forms are also of course used to signalpoliteness, in utterances such as:

129 I did wonder if I might ask you a favour.

The relation between these various uses is remoteness. That is, tense is deictic, sinceit locates the speaker relative to a deictic present. A past tense form, as in (129),shifts the speaker back in time, thus distancing speaker from hearer, and puttinga hedge on illocutionary force. In general, what are traditionally known as past,present, and future tenses have more to do with expressing modality than with timereference (Lyons 1981: 239). And as Lyons (1977: 817) also points out, it is noaccident that the so-called future tense in English uses will, which also expressesinference (cf. That will be the postman). References to future time are necessarilyhypothetical and predictive. (See also Levinson 1983 for a detailed argument thatalmost every utterance encodes the speaker's point of view in the sense of deixis,

MICHAEL STUBBS 21

and that the pervasiveness of deixis has also been greatly underestimated in linguis-tic description.)

Tag questions are a complex syntactic and semantic topic, but, very briefly,like the other cases discussed, they do seem explicable only with reference tocontexts of usage in discourse. They allow statements to be presented as obvious,dubious, or open to challenge. Tags with the same polarity as the main clause(He is, is he?) refer to propositions whose source is the addressee. Tags withreversed polarity (He is, isn't he?, He isn't, is he?) refer to the speaker's own beliefs.One major function is to implicate: 'I am certain/not certain thatp; I want to checkif you believe thatp' (cf. Bublitz 1978: 140-61).

Lyons (1981: 241) argues that there is much in the structure of languages thatcannot be explained without reference to the notion of subjectivity of utterance;and (pp. 235-6) that much work in semantics and pragmatics is seriously flawed,because it has not given sufficient prominence to the concepts of modality, sub-jectivity, and locutionary agency. I have tried in this paper to provide some data toillustrate these claims, and to begin to show what a modal grammar of Englishmight look like.

19. A MATTER OF PROLONGED FIELDWORKA relatively simple, but extremely time-consuming method of analysis, would be totake texts from different settings; to 'strip off' all the surface markers of speaker'spoint of view; and to classify these according to the degree and manner of com-mitment conveyed. This would be a very crude method of analysis, but we do nothave even the simplest taxonomic descriptive information for some of the topicsI have discussed in this paper. Such a task would keep armies of doctoral studentsbusy for years, although convenient computational techniques are now availablefor searching large corpora for particular forms, and so on. Only in this way wouldwe begin to see just how pervasive the expression of such points of view is, and howit differs across different contexts. (Brown and Levinson 1978 provide a taxonomyof cases which could be useful as an initial classification.)

A start has been made on such studies. Bublitz (1978) provides a detaileddiscussion of several indicators of speaker attitude in German and English, using(for English) data from the London-Lund corpus and elsewhere. Coates (1982),in another corpus-based study, has shown that epistemic modals are more frequentin informal spoken and- personal written language than in formal impersonallanguage. And Holmes (1983), on the basis of a corpus of 50,000 words, has esti-mated that lexical items expressing degree of certainty make up on average 3.5 percent of any text, but are twice as frequent in speech as in writing. Much more workrequires to be done, however, to study their occurrence and reasons for use toindicate, among other things, ignorance, deference, social uncertainty, informality,and shared group membership.

20. APPLIED LINGUISTICSIn this paper I have been concerned essentially with an approach to language whichplaces as central goal-directed and evaluative behaviour, and the communicativevalues of speakers. I will conclude by being more explicit about some of the moreimmediate applications of the kind of modal grammar which I have proposed. Thisshould provide further external motivation for the approach.

First, and most obviously perhaps, the kind of study I have proposed fits into

22 TOWARDS A MODAL GRAMMAR OF ENGLISH

a general study of style and variation in style in different social contexts. I havemade the point that intuitions about stylistic variation are simply not accessible orreliable, and therefore demand a corpus-based study.

Second, it can contribute to teaching EFL. I have mentioned a rather specializedkind of teaching with reference to business English. However, I have also discussedmore general aspects of the sotioHnguistic competence which is involved in express-ing polite, tentative, tactful statements, about controversial subjects about whichone has reservations. These aspects of language are a notorious problem for foreignlearners, depending, as I have noted, on assumptions about group membership.For example, it is notoriously difficult to translate modal particles from onelanguage to another. And it is well known that speakers of English as a foreignlanguage can sound rude, brusque, or tactless to native speakers if they makemistakes in this area. One problem is that their mistakes are not recognized aslinguistic mistakes at all, but as social ineptitude. It is also widely recognized thatforeign learners have comprehension problems with indirect forms. (Holmes,1983, discusses such data from this point of view at length.)

Third, Gumperz (1982a, b) has carried out detailed ethnographic work on cross-cultural differences in the expression and perception of credibility and trustworthi-ness. He shows that in modern bureaucratic industrial societies, where there isunprecedented cultural diversity, an increasing amount of communication is not insmall face-to-face groups where speakers know each other well, but betweenstrangers, who are interacting not as individuals but as members of social roles. Theinterpretation of indirect utterances depends on shared, taken-for-granted know-ledge. And there are cultural differences in the expression and use of indirect-ness. He documents legal cases where prosecutions hinged on whether defendantswere perceived as being convincing or credible or trustworthy, and argues that,owing to linguistic problems, they were not.

Fourth, as Bolinger (1980) has argued, linguists have the social responsibility toreveal the kinds of rhetorical tricks which underlie much propaganda, advertising,manipulative and biased language. There is nothing wrong per se with makingpresuppositions, for example. In fact, there is no way to avoid it. But as a manner ofmaking propositional commitments it can be awkward to challenge, and can there-fore be used manipulatively. This type of practical semantics can open up ideologi-cal questions. Any study of institutional discourse, between professionals andclients of whatever kind, becomes a way of studying social power and deference,manipulation and evasion. This is therefore a modem study of rhetoric, which canhelp professionals and others to understand how they can be affected by language.

One aspect of this type of study is the study of the language used in the massmedia to discover just what implicit points of view are being conveyed along withthe news. Much of the work in this area has been done by sociologists, and thelinguistic analysis has sometimes been less than systematic, though importantobservations are made. One interesting sociolinguistic study is by Bell (1984).He studied the way in which news stories arrive in New Zealand via agencies suchas Reuters, and in a form designed essentially for printed transmission in news-papers; and are then edited into much shorter forms for transmission on radio news.He shows that the radical shortening which is involved can mean editors reducingdetails, rounding specific figures, deleting hedges, omitting attributions to spokes-persons, and so on. And he shows how the degree and manner of commitment canbe altered between the original text and the broadcast version.

MICHAEL STUBBS 23

Fifth, the kind of analysis I have discussed here is very relevant to the design ofexpert systems: that is, computer-based systems designed to understand naturallanguage input, to make decisions in some complex area (such as medical diagno-sis), and to give these decisions in natural language. Attempts to program computersto understand discourse have revealed the ubiquity of implicit inferences in languageuse. But attempts to set up expert systems have not yet started to cope with theubiquity of modal qualifications in all uses of language.

21. CONCLUSIONSThere are, of course, many aspects of vague and indirect language which I have noteven mentioned. For example, I have not dealt with metaphorical language, or themany non-literal or non-serious uses of language in lying, irony, exaggeration,teasing, and joking. Some of these aspects of language in use, which have previouslybeen swept under the carpet, are now being taken on board, so to speak, by somelinguists. In some ways, the pendulum has swung full circle. Aspects of languagewhich always seemed to linguists to be far from the bread-and-butter side oflanguage study are now being seen as the backbone of the enterprise. Fieldworkcan be an uphill grind. But if you can sense which way the wind is blowing andswim with the tide, then sometimes it seems possible to grasp the nettle instead ofclutching at straws or planting primroses in a gale. We are not necessarily on thehorns of a dilemma, between theory and data: one linguist's Scylla is another'sCharybdis. If we can get our foot in the door, then the little acorn may grow intoa mighty oak, as it were, as it begins to snowball downhill with the theoretical windin its sails.

I hope that at least some of what I have said will seem to you to have been—basically, in essence—true. And I hope that the core of what I have said was clearenough for the purpose, even if its edges were fuzzy, tattered, or in shreds. I amlogically committed to my assertions and entailments, but I can, without logicalcontradiction, deny my implicatures.

(Received May 1985)

NOTES1 An abbreviated version of this paper was given as an invited lecture to the Annual Meeting,American Association of Applied Linguistics, Baltimore, USA, 27-30 December 1984 (jointmeeting with Linguistics Society of America).

Several of the ideas in this paper have been developed in conversation over four years withAndrew Gilling, and some of the topics are treated in more detail in Gilling (in prep.). I am alsograteful to Joanna Channell, Walter Grauberg, Gabi Keck, and John Sinclair, and to seminaraudiences at the Universities of Birmingham and Sussex, England, for comments on earlierversions of this paper.1 tfotational conventions: This paper uses the following conventions to distinguish betweendifferent sources of data:(A) attested: an example of language which has been recorded as it actually occurred in a

real social context without the intervention of the analyst.

(A:L) attested in the Lund corpus of half a million words of spoken British English: see Svartvikand Quirk 1979.

(A:B) attested in a corpus of about 500 business letters collected by the author.

(I) invented: intuitive or introspective data which has been invented purely to illustrate an

argument

"p": double quotes are used to indicate proporitional or lexical meanings.

24 TOWARDS A MODAL GRAMMAR OF ENGLISH

REFERENCESAustin, J. L. 1958. 'Performative-constative' in Searie (ed.) 1971.Austin, J. L. 1962. How to Do Things with Words. Oxford: Oxford University

Press.Bell, A. 1984. 'Good news - bad copy: the syntax and semantics of news editing'

in P. Trudgill (ed.) Applied Sociolinguistics. London: Academic Press.Belnap, N. D. 1977. 'A useful four-valued logic' in M. Dunn and G. Epstein (eds.)

Modern Uses of Multiple-Valued Logic. Dordrecht: Reidel.Bolinger, D. 1980. Language the Loaded Weapon. London: Longman.Brown, P. and S. Levinson. 1978 'Universals in language usage' in E. N. Goody

(ed.) Questions and Politeness. London: Cambridge University Press.Bublitz, W. 1978. Ausdrucksweisen der Sprechereinstellung im Deutschen und im

Enlgischen. Tuebingen: Niemeyer.ChanneD, J. M. 1980. 'More on approximations.' Journal of Pragmatics 4: 461-76.Channell, J. M. 1983. 'Vague Language: Some Vague Expressions in English.'

Unpublished PhD Thesis, University of York.Coates, J. 1982. The Semantics of the Modal Auxiliaries. London: Croom Helm.Cole, P. and J. L. Morgan (eds.). 1975. Syntax and Semantics, Vol. 3, Speech Acts.

New York: Academic Press.Davison, A. 1975. 'Indirect speech acts and what to do with them' in Cole and

Morgan (eds.) 1975.vanDijk, T. A. 1979. 'Pragmatic connectives.' Journal of Pragmatics, 3/5: 447-57.Edmondson, W. 1981. Spoken Discourse. London: Longman.Flew, A. 1975. Thinking about Thinking. London: Fontana.Frazer, B. 1975. 'Hedged performatives' in Cole and Morgan (eds.) 1975.Garfinkel, H. 1967. Studies in Ethnomethodology. Englewood Cuffs, NJ: Prentice

Hall.Garfinkel, H. and H. Sacks. 1970. 'On the formal structures of practical action'

in J. McKinney and E. Tiryakian (eds.) Theoretical Sociology. New York:Apple ton-Century-Crofts.

Gazdar, G. 1981. 'Speech act assignment' in Joshi et al. (eds.) 1981.Gilling, A. (in prep.) Unpublished PhD thesis. University of Nottingham.Goffman, E. 1981. Forms of Talk. Oxford: Blackwell.Goody, E. N. (ed.). 1978. Questions and Politeness. London: Cambridge University

Press.Gowers, E. 1962. The Complete Plain Words. Harmondsworth: Penguin. (1954).Greenbaum, S. 1969. Studies in English Adverbial Usage. London: Longman.Grice, H. P. 1975. 'Logic and conversation' in P. Cole and J. L. Morgan (eds.)

1975.Gumperz, J. J. 1982a. Discourse Strategies. London: Cambridge University Press.Gumperz, J. J. (ed.) 1982b. Language and Social Identity. London: Cambridge

University Press.Hancher, M. 1979. 'The classification of co-operative illocutionary acts.' Language

in Society 8/1:1-14.Holmes, J. 1983. 'Speaking English with the appropriate degree of conviction' in

C. J. Brumfit (?d.) Learning and Teaching of Languages for Communication.London: CILT.

Johnson-Laird, P. N. 1983. Mental Models. London: Cambridge University Press.Joshi, A. K., B. L. Webber, and I. Sag (eds.) 1981. Elements of Discourse Under-

standing. London: Cambridge University Press.Katriel, T. and M. DascaL 1984. 'What do indicating devices indicate?' Philosophy

andRhetoric 17/1:1-15.Kempson, R. 1977. Semantic Theory. London: Cambridge University Press.Labov, W. and D. Fanshel. 1977. Therapeutic Discourse. New York: Academic Press.

MICHAEL STUBBS 25

Lakoff, G. 1972. 'Hedges: a study in meaning criteria and the logic of fuzzyconcepts.' Paper: of the Chicago Linguistic Society 8:183-228.

Lakoff, R. 1969. 'Some reasons why there can't be any "some-any" rule.' Language45:608-15.

Leech, G. N. 1983. Principles of Pragmatics. London: Longman.Levinson, S. C. 1983. Pragmatics. Cambridge: Cambridge University Press.Ljimg, M. 1980. Reflections on the English Progressive. Goteborg: Acta Universi-

tatis Gothoburgensis.Lyons, J. 1977. Semantics. Vols. 1 and 2. Cambridge: Cambridge University Press.Lyons, J. 1981. Language, Meaning and Context. London: Fontana.McCtwley, J. D. 1981. Everything that Linguists Have Always Wanted to Know

about Logic. Oxford: Blackwell.McTear, M. F. 1980. "The Pragmatics of Because: Mimeo: Ulster Polytechnic.MorreaL J. 1979. The evidential use of because.' Papers in Linguistics 12, 1/2:

231-8.Nystrand, M. (in prep.) Reciprocity as a Principle of Written Description. New

York': Academic Press.Palmer, F. R. 1974. The English Verb. London: Longman.Quirk, R. and S. Greenbaum. 1973. A University Grammar of English. London:

Longman.Rosen, E. H. 1975. 'Cognitive reference points.' Cognitive Psychology, 7:532-47.Schegloff, E. A. 1972. "Notes on a conversational practice: formulating place' in

D. Sudnow (ed.) 1972.Searle, J. R. 1969. Speech Acts. London: Cambridge University Press.Searle, J. R. (ed.) 1971. The Philosophy of Language. London: Oxford University

Press.Searle, J. R. 1975. 'Indirect speech acts' in P. Cole and J. L. Morgan (eds.) 1975.Searle, J. R. 1976. 'A classification of illocutionary acts.' Language in Society,

5:1-23. (Also in Searle 1979.)Searle, J. R. 1979. Expression and Meaning. London: Cambridge University Press.Stubbs, M. 1983a. Discourse Analysis: the So do linguistic Analysis of Natural

Language. Oxford: Blackwell.Stubbs, M. 1983b. 'Can I have that in writing, please? Some neglected topics in

speech act theory.' Journal of Pragmatics 7:479-94.Stubbs, M. and S. Delamont (eds.) 1976. Explorations in Classroom Observation.

London: Wiley.Sudnow, D. (ed.) 1972. Studies in Social Interaction. New York: Free Press.Svartvik, J. and R. Quirk (eds.) 1979. A Corpus of English Conversation. Lund:

Gleerup.Torode, B. 1976. Teachers' talk and classroom discipline' in M. Stubbs and

S. Delamont (eds.) 1976.