This article was downloaded by: [University College London]On: 07 March 2014, At: 07:42Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Quarterly Journal of ExperimentalPsychologyPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/pqje19

The recall of affirmative andnegative sentences in an incidentallearning taskElizabeth R. Cornish a & P. C. Wason aa Department of Psychology , West Ham College ofTechnology, the Psycholinguistics Research Unit, UniversityCollege , LondonPublished online: 29 May 2007.

To cite this article: Elizabeth R. Cornish & P. C. Wason (1970) The recall of affirmativeand negative sentences in an incidental learning task, Quarterly Journal of ExperimentalPsychology, 22:2, 109-114

To link to this article: http://dx.doi.org/10.1080/00335557043000032

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information(the “Content”) contained in the publications on our platform. However, Taylor& Francis, our agents, and our licensors make no representations or warrantieswhatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions andviews of the authors, and are not the views of or endorsed by Taylor & Francis. Theaccuracy of the Content should not be relied upon and should be independentlyverified with primary sources of information. Taylor and Francis shall not be liablefor any losses, actions, claims, proceedings, demands, costs, expenses, damages,and other liabilities whatsoever or howsoever caused arising directly or indirectly inconnection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden.Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Q. Jl exp. Psychol. (1970) 22, ~og-I 14

THE RECALL OF AFFIRMATIVE AND NEGATIVE SENTENCES I N AN

INCIDENTAL LEARNING TASK ELIZABETH R. CORNISH AND P. C. WASON

Department of Psychology, West Ham College of Technology, and the Psycholinguistics Research Unit, University College, London

This experiment examined the recall of positive and negative information in an incidental learning task. The two main findings were that a significantly greater number of affirmative than negative clues were correctly recalled and that the majority of errors took the form of conversions from negative to affirmative, independently of meaning. The difficulty associated with the negative clues was explained in terms of their inappropriateness in the situation, namely in the absence of any prior expectations.

Introduction

A considerable number of studies have demonstrated a greater difficulty associated with the use of positive as opposed to negative information in concept attainment (Hovland and Weiss, 1953; Bruner, Goodnow and Austin, 1956), problem solving (Whitfield, 1951; Donaldson, 1959) and evaluation tasks (Wason, 1959, 1961; Eifermann, 1961).

More recently the characteristics of negative sentences have been examined from the point of view of transformational grammar. On the basis of Chomsky’s (1957) syntactic theory, some psychologists hypothesized that (i) grammatical complexity would be reflected in behavioural complexity, and (ii) kernel (simple, active declar- ative) sentences would be psychologically basic. One technique employed for testing these hypotheses was recall. The predictions relating to negative sentences were that (i) they would be recalled less well on account of their greater syntactic complexity, and (ii) there would be a tendency for them to be mis-recalled as their correlative kernel. Thus Mehler (1963), using a rote learning task, involving sentences of various types of grammatical complexity, found evidence for better recall of kernel than transformational sentences and more errors in the direction of a shift towards than away from the kernel (Foa and Schlesinger, 1965, as a result of a reanalysis of the data, question this latter conclusion). On the basis of these results he argued in favour of a “kernel plus transformational‘’ hypothesis, in which he suggested that what is coded in memory is the kernel sentence plus a “footnote” as to its transformational form. For example, the sentence “The door is not open” would be coded approximately as “The door is open” plus a tag to the effect that it was negative.

Fillenbaum (1966) rejects this analysis on the grounds that the semantic content of material is much better recalled than its syntactic form. This was supported

Iog

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

42 0

7 M

arch

201

4

I I0 E. R. CORNISH AND P. C. WASON

by a predominance of meaning-preserving over meaning-changing errors in a recognition task for sentences in which the following two factors were varied: (i) the conceptual nature of antonymous pairs of sentences, i.e. whether they are contraries (which have scalar properties, e.g. hot-cold) or contradictories (which are jointly exhaustive, such that the negation of one term entails the other, e.g. open-closed); and (ii) their phonological shape, i.e. whether they are opaque (phonologically and etymologically distinct, e.g. open-closed) or transparent (possessing the same root but one term bearing a pr& such as un-, dis-, e.g. honest-dishonest). As an alternative theory he proposes that what is stored in memory is a reading of the sentence, probably in reduced form, in which the semantics is preserved, plus a note on the format of the sentence. The sentence “The door is not open” would thus be coded approximately as “The door is closed” plus “negative”. Like Mehler, he believes that the syntactic aspects of a sentence are more subject to interference than are the semantic aspects.

The present experiment used an incidental learning task, rather than a rote learning task, to test the hypothesis that more affirmative than negative sentences would be recalled. The material was presented once as clues for the identification of an object, which in fact did not exist. The subjects were unprepared for r e d which was unconstrained. The material consisted of affirmative and negative predicative sentences of the form “It is x” and “It is not y” where “x” and “y” are attributes. This enabled comparison of the recall of affirmative and negative sentences (as in Mehler’s experiment) and an examination of various combinations of semantic and syntactic error (as in Fillenbaum’s).

Method Task and desafl

The task was to guess an imaginary object. The subjects did not know that the experi- menter had no particular object in mind. Sixteen clues (eight affirmatively stated and eight negatively stated) were presented to the subjects in a random order. Subsequently, the subjects were required to recall these clues. It was predicted that more afiirmative than negative clues would be correctly recalled. Half the subjects were praented with material A and the other half with material B (see

Table I). It was so arranged that clues which were afFirmative in material A were negative in material B and vice versa. The subjects were also divided equally into two further wnditione: one where an hypothesis as to the object’s identity was to be written down after the presentation of each clue (Group W) and one where it was to be written down only after the presentation of the whole 16 clues (Group NW). It was thought that a greater effect might result in Group W, where there was more pressure to process the clues.

The clues were printed on 3 in. x 2.5 in. cards.

Subjects The subjects, 40 non-psychology students, were allocated to the conditions in rotation.

Procedure The subjects were tested individually and sat at a table, separated from the experimenter

by a screen. They were presented with the following instructions: “I have an object in mind. Your task is to discover what it is. You will be shown a number of clues describing it.

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

42 0

7 M

arch

201

4

AFFIRMATIVE AND NEGATIVE RECALL I11

(Croup W: After each clue please write down what you think the object is. If you cannot think of anything leave a blank.) (&oup NW: At the end you will be asked to write down what you think the object is.) Any questions ?’,

When the experimenter was satisfied that the subject had understood the instructions, the subject was told that he would be allowed 8 sec. for each clue and was reminded that all the clues referred to the same object. If further help was required he was told that

TABLE I Material used in conditions A and B

Material A Material B

It is bright It is not beautiful It is wide Lt is tall It is not hard It is not straight It is not strong It is cold It is not clear It is open It is not heavy It is not clean It is deep It is dry It is full It is not old

It is not bright It is beautiful It is not wide It is not tall It is hard It is straight It is strong It is not cold It is clear It is not open It is heavy It is clean It is not deep It is not dry It is not full It is old

the object was an household object. The cards were then placed on the table in turn in front of the subject, 8 sec. being allowed for each. Following this the subject was required to carry out the following instructions:

I.

2.

No time limit was imposed. Introspections were then taken.

“Write down what you think the object is.’’ “Write down as many of the clues as you can remember.”

ReSults

Since 40 subjects were each presented with eight affirmative and eight negative clues, there were a total of 320 affirmative and 320 negative clues. A breakdown of the responses to these is shown in Table 11. The errors were divided into four categories in addition to omissions. These were: (a) semantics preserving and syntax preserving, i.e. synonyms, e.g. “bright” recalled as “shiny” or “not bright” as “not shiny”; (b) semantics preserving but syntax changing, e.g. “bright” recalled as “not dull” or “not bright” as “dull”; (c) semantics changing but syntax preserving, e.g. “bright” recalled as “dull” or “not bright” as “not dull”; (d) semantics changing and syntax changing, e.g. “bright” recalled as “not bright” or “not bright” as “bright”.

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

42 0

7 M

arch

201

4

I12 E. R. CORNISH AND P. C. WASON

An analysis of variance applied to the data showed no significant effect due to the two forms of material, nor to the two conditions of writing down the hypothesis; nor was there an interaction between these effects. The effect of type of presenta- tion was highly significant (Fl,ss = 2403, P < 0.001): 177 a h a t i v e clues and only 124 negative clues were correctly recalled. None of the interactions between type of presentation and other effects was significant.

Since the error frequencies were small, sign tests were used to test frequency differences between the types of error. Significantly more errors were made to

TABLE I1 Total and mean number of each type of response ma& under each type of

presentation by 40 subjects

Type of response

Correctly Semantics Semantics Omitted Grand recalled preserving changing total

Incorrectly recalled

syn. syn. syn. syn. pr. ch. pr. ch. (4 (b) (4 (4

Total 177.00 2-00 o 2.00 8-00 131.00 320 Affirma- Mean/

Type tive subject of (n = 40) 4'42 0.05 0 0.05 0'20 3'27

presenta- tion Total 12400 2-00 20'00 4'00 13.00 157mo 320

Nee - Mean/ tive subject

(n = 40) 3.10 0.05 0.50 0.10 0-32 3'92

In the four cases where two responses were apparently given to the same clue, the first was taken.

negative than to affirmative clues (P = 0-002, one-tailed: 18 cases for, 4 against, 18 equal). Significantly more of these errors to negative clues were syntax changing than syntax preserving (P = 0.001, two-tailed: 16 cases for, I against, 23 equal.) There was no significant difference between the number of semantics preserving and semantics changing errors to negative clues (10 cases for, 9 against, 21 equal.)

Discussion The hypothesis that more a h a t i v e than negative clues would be correctly

recalled was strongly confirmed. This extends the number of areas where diffi- culty is found with negative information to include an incidental learning task.

An analysis of the types of error that occurred showed that the majority of these involved conversions from negative to affirmative, independently of meaning. High frequencies occurred for type (b) and type (d) errors on negative clues (20

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

42 0

7 M

arch

201

4

AFFIRMATIVE AND NEGATIVE RECALL “3

and 13, respectively). I t is strong evidence for the need to convert negative to positive information if this occurs at the expense of meaning.

The greater number of type (d) errors for negative than for aflirmative clues, although insignificant on a sign test, is in the direction of Mehler’s finding of a greater number of shifts in the direction of the kernel than away from it. He found that 36 negative sentences were recalled as kernels whereas 14 kernel sentences were recalled as negative.

The results are also of interest in relation to Fillenbaum’s results. He found that meaning preserving errors predominated significantly over meaning changing errors except for unprefixed contraries. But the material used in the present experiment does consist of unprefixed contraries, and hence comparison between the two experiments is of considerable interest. The classes of error corresponding to Fillenbaum’s analysis are type (b) (meaning preserving) and type (d) (meaning changing). Sign tests applied to these confirmed FiUenbaum’s failure to obtain a difference between meaning preserving and meaning changing errors when the results from affirmative and negative clues were combined (9 cases for, I I against, 20 equal), and also when the results from negative clues alone were treated (10 cases for, 7 against, 23 equal). However, when the results from affirmative clues alone were analysed, a significant difference (P = 0.016, two-tailed: o cases for, 7 against, 33 equal) was obtained in the opposite direction to Fillenbaum’s general hypothesis, i.e. there were significantly more meaning changing than meaning preserving errors. It is not clear why this should have been so.

It is suggested that an important reason for the difficulty associated with the handling of negative sentences in this and other previous experiments is their inappropriateness in the particular experimental situations involved. Negatives are frequently used as qualifiers; they assume and depend on a prior state of affairs, either existent or supposed. It is unlikely that the sentence “It is not x” would be uttered unless there were good reason to suppose that it might have been “x” or that someone thought that it might. In the abstract task of the present experiment the subjects had no prior expectations about the object whose identity they were required to guess. A statement of the form “It is not x” was inappropriate be- cause the subjects had no preconceived ideas about the object, nor did they have any other objects with which they might have compared it.

One method they adopted to utilize these clues was to convert them to affirma- tive form (sometimes incorrectly). The need to transform negative to positive information is often cited (for example, in some of the studies mentioned in the introduction) as a reason for the difficulty in handling negative information. In order to evaluate a negative statement efficiently one must be in possession of an affirmative one, usually a statement that provides the grounds for the negation. In a previous experiment (Wason, 1965) it was shown that under similar circumstances the latency for understanding a negative sentence was significantly reduced.

An attempt was made to examine whether ease of such conversion from negative to affirmative had a differential effect for the individual words used, by conducting a subsidiary experiment. Ten subjects were asked to give affirmative synonyms for the 16 negative clues used in the main experiment. The mean response times and the variability of the responses were calculated for each word. However,

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

42 0

7 M

arch

201

4

1’4 E. R. CORNISH AND P. C. WASON

no clear relationships were found between either of these measures and the number of negative to afFirmative conversions for each word in the main experiment. (Spearman’s rank correlation coefficient was not significant in either case.)

Another factor which may have inAuenced the results is the fact that the ranges of the affirmative and negative clues were not equal. For example, “not bright” includes not only regions of the continuum to which “dull” applies but also regions of uncertainty to which neither “bright” nor “dull” apply. The size of these ranges would have been relevant to the use to which a word could be put in delimiting the number of possible objects, and presumably played a part in determining the number of h a t i v e rather than negative clues that were recalled. It might also account for some of the conversions from negative to affirmative. The problem of measuring the size of ranges of words is a more intractable one.

This paper is based on part of an unpublished University of London Ph.D. thesis by the first author.

We are grateful to Mr. J. W. Whitfield for statistical advice.

References BRUNW, J. S., GOODNOW, G. G. and AUSTIN, G. A. (1956). A Study of Thinking.

CHOMSKY, N. (1957). Syntactic Stnrctures. The Hague: Mouton. DONALDSON, M. (1959). Positive and negative information in matching problems. BY.

EIFERMA”, R. (1961). Negation: a linguistic variable. Actu Psychol. 18, 25873. FILLENBAUM, S. (1966). Memory for gist: some relevant variables. Lung. &? Speech

FOA, U. G. and SCHLESINCER, I. M. (1965). Syntactical errors in sentence recall: a re- analysis of Mehler’s data. (Unpublished manuscript.)

HOVLAND, C. I. and WEISS, W. (1953). Transmission of information concerning concepts through positive and negative instances. JZ ex@. Psychol. 45,175-82.

MBHLBR, J. (1963). Some effects of grammatical transformations on the recall of English sentences. Jl verb. Learn. verb. Behav. 2, 346-56.

WASON, P. C. (1959). Q. Jl e e . Psychol. 11, 92-107.

WASON, P. C. (1961). BY. J1 Psychol. 52, 133-42.

WASON, P. C. (1965). JZ verb. Lernn. verb. Behav. 4, 7-11.

WHITFIELD, J. W. (I 95 I). An experiment in problem solving. Q. JZ exp. Psychol. 3, I 84-97. Received 18 February 1969

New York: Wiley.

Jl Psychol. 50, 253-62.

9, 217-7-

The processing of positive and negative information.

Response to affirmative and negative binary statements.

The contexts of plausible denial.

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

42 0

7 M

arch

201

4