1
Remembering in signs*
URSULA
BELLUGI
The Salk Institute EDWARDS. University of California,
KLIMA San Diego
PATRI...
17 downloads
1559 Views
6MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
1
Remembering in signs*
URSULA
BELLUGI
The Salk Institute EDWARDS. University of California,
KLIMA San Diego
PATRICIA
SIPLE
University of Rochester
Abstract In a short-term memory experiment, signs of Anterican Sign Language in list lengths of three to seven items were presented to deaf college students whose native language is American Sign Language. A comparable short-term memory experiment for words (words representing the English translation-equivalents of the signs) was presented to hearing college students. Recall was written, immediate and ordered. Overall, short-term memory mechanisms in the deaf seem to parallel those found in hearing subjects, even with the modality change. A significant number of multiple intrusion errors made by deaf subjects to signs were based on formational properties of the signs themselves, a result paralleling the phonologically based errors in experiments with hearing subjects. Our results are consistent with a theory that the signs of American Sign Language are actually coded by the deaf in terms of simultaneous formational parameters such as Hand Configuration, Place of Articulation and Movement. Evidence is given that signs are treated by the deaf as consisting of independent parameters -- specific to American Sign Language - which are essentially arbitrary in terms of meaning.
* This research was supported in part by National Institutes of Health Grant # NS09811 and National Science Foundation Grant # GS42927X to The Salk Institute for Biological Studies. Our thanks to G. Sperling, P. Benson and R. Conrad, T. Smith for reading and commenting on
earlier versions of this paper. In addition to the authors, other members of the research group involved in this experiment were: Bonnie Gough, Susan Fischer, Nancy Frishberg, Richard Lacy, Robbin Battison, Carlene Pedersen and Ted Supalla.
Cognition
3(2),
pp. 93 - 125
94
Ursula Bellugi, Edward S. Klirna and Patricia Siple
1. Introduction In the conclusion of his book, Thinking without language, H. Furth (1966) says that when he began working with the deaf some ten years earlier, he hardly foresaw that a certain question often asked by hearing people would lead to such far-reaching implications: ‘What do the deaf think in?’ Furth spends the greater part of his book reworking that question until it begins to yield some tentative answers and, of course, many new questions. In a more recent paper entitled ‘Short-term memory processes in the deaf’, R. Conrad (1970) begins with essentially the same question. His approach is to try to specify one small aspect of the general question; he defines that aspect in terms of memorization processes and limits the study further to the remembering or memorization of visually presented sequences consisting of printed letters of the English alphabet occurring alone and in English words - i.e., significantly, consisting of symbols of written English. With hearing subjects, the results of Conrad’s earlier experiments support the notion that we code and remember on the basis of the sotlnd-form (articulatory and/or acoustic) of words or letters. For the deaf, on the other hand, Conrad suggests that this need not be true. Indeed, for the profoundly deaf, at least in principle, we must allow ourselves to conceive of a memory store for ‘verbal’ material (i.e., for language-symbolic material) in which representations in, visual forms predominate; we must even conceive of a ‘verbal’ store containing visual and/or kinesthetic representations of finger-spelling hands, of signing hands or perhaps even of speaking lips. Conrad, on his part, concludes that ‘The deaf do use “symbols” in memorizing and that the nature of them is open to empirical inquiry . . . That the deaf, with little overt speech, learn and think is self-evident. What they do it in remains a challenge with far-reaching implications’ (p. 194). It is important to bear in mind that in the study cited here, Conrad worked with children who were in a school for the deaf where all the teaching was carried out orally, in the English language. Conrad was exploring the possibility that some of his subjects might be coding in terms of the visual shapes of the printed English letters or words along with, or instead of, articulatory phonological characteristics. He neither ruled out nor explored in this study the possibility that deaf subjects may rely on other types of coding for remembering. Although Conrad limited his subjects in this particular experiment to deaf children whose education was exclusively oral, he did not explicitly specify what further ‘linguistic’ background the deaf children had had or whether any of the deaf children had any knowledge of a sign language. If all the subjects had had native knowledge of a sign language learned ‘naturally’ from deaf parents and used as favored form of everyday communication, then a little reflection will show that this could make a substantial difference.
Remembering
in signs
95
It is the various aspects of this difference that have determined the focus of the research being conducted by our group at the Salk Institute. We ask not how the overall deaf population remembers and processes symbols of written English, but rather how a particular sample of deaf people - those who have learned from deaf parents, as their first language, a purely ‘visual’ sign language - remember and process the visual symbols of such a sign language. At the same time, through such studies as this, we hope to learn more about the internal organization of the ‘linguistic’ symbols used in a primary visualgestural language. Even the grossest observation of signs in the early stages of our research on the structure of American Sign Language (ASL) made it clear that the organizational principle of signs is different from that of spoken words. In the individual sign of ASL there is nothing to suggest an analysis is not a sequence of linearly ordered segments, the way apse, asp, pass, and sup are different segmental orderings (at an abstract level) of the phonological segments /p/, /oe/ and /s/ (Klima, in press). It was apparent from the beginning, however, that certain gestural formatives recurred in different signs. In fact, Stokoe (1960, 1965) had already proposed .that it would be possible for the linguist to catalog signs in terms of a limited number of recurring values on each of three parameters (Hand Configuration, Place of Articulation and Movement) occurring simultaneously rather than sequentially in constituting the individual sign. If, for example, you hold one hand in a tapered ‘0’ (thumb tip touching tips of the fingers), touch the cheek near the mouth, move away in a small arc and touch the upper cheek, it is the sign for HOME. If you keep the same hand configuration and touch on one side of the nose and move away slightly and touch on the other side of the nose, it is the sign for FLOWER (supposedly representing the sme!ling of a flower). If you keep the motion and place of articulation of the sign for HOME constant and change the hand configuration so that the hand is closed and the contact is with the tip of the extended thumb, it is the sign for YESTERDAY. If you return to the tapered ‘0’ handshape and the cheek and make the movement in a sman circle, it is a sign for PEACH. The citation form of an ASL sign can have any one of the following hand arrangements: (a) Some signs are made with one hand only; (b) some are made with two hands both active (in this case the hands will be in symmetrical shapes and with symmetrical movement); and (c) some are made with one hand (dominant) acting on the other as a base hand. Examples of these three different arrangements of the hands are given in Figure 1. l Using some 2,000 ASL signs, we have tabulated the percentage of signs in each category. The figure presents the reader with an initial view of some ASL signs.
Ursula Brllztgi, Edward
96
Figure
1.
Relationships
a. One-handed
S. Klima and Patricia Siple
of hands in ASL
signs (40% of signs)
b. Two hands,
DOLL
BIRD
YESTERDAY
WOMAN
both active (35% of signs)
HEADACHE
MORE
c. One hand acting
on other
EGG
as base (25% of signs)
4 MEDICINE
VOTE
1. Mr. Frank A. Paul made the illustrations of signs for figures throughout the paper.
YEAR
NEWSPAPER
Remembering
in siglzs
91
In addition to considering signs in terms of their physical parameters, one can also consider individual signs as gestalt-like configurational wholes; in this respect, sign language seems indeed different from spoken language. With many signs, when the meaning is given, the signs appear to have a marked iconic2. The sigh for BIRD, for example, is made with the thumb and index finger in front of the mouth supposedly representing the movement of the beak of a bird. In the sign for TREE, the arm and hand supposedly represent the trunk and branches of a tree. However, few signs are so clearly transparent in their iconicity that a non-signer can guess their meaning without some additional cues. In fact, it is usually not possible to guess even the topic of conversation in deaf communication (as would presumably be possible in pantomime). Nonetheless, from the point of view of an outside observer, many of the signs in a visual language like American Sign Language appear not to be totally arbitrary. This presents an interesting question as to the nature and degree of internal structure of what might at first seem to be a set of ‘gestalts’ with some iconic associations. The experiment described below is directed toward this question. More precisely, we want to discover to what extent such parameters as are suggested by Stokoe count as independent, semantically arbitrary units for deaf signers. That is, between the composite configurational whole of a sign and its meaning is there a significant level of language structure essentially independent of meaning and consisting of a limited number of values of a limited number of formational parameters occurring simultaneously to constitute the individual signs? Preliminary to a discussion of our experiments, some information about the subjects of the experiments and about the varieties of sign languages is in order. By one recent census count there are about 450,000 profoundly deaf people in the United States, including those who have become deaf later in life. The number of people who are ‘pre-lingually’ deaf - whose deafness occurred before they could learn a spoken language (i.e., who were either deaf from birth or since the first two years of life) - is far smaller, and most of these deaf were born to hearing parents. Only an estimated 10 to 15% of deaf people have deaf parents, and it is within this small group of deaf families in America that American Sign Language is found as a natural primary language. There are different sign languages just as there are different spoken languages - differing from one another most obviously in the form of the signs which they use but also in their syntax. British Sign language, for example, is quite unrelated to American Sign Language. Since sign languages are without a 2. ‘Iconic’ here is not used in the sense of ‘iconic’ store; rather, the term refers to symbols that are to a significant degree
representational of isomorphism
(i.e., having some degree with, e.g., the denotation).
98
Ursula Bellugi, Edward S. Klima and Putricia Siple
written tradition, there is an expected lack of standardization. There is also some regional ‘dialectal’ variation of ASL in different parts of the country. We must point out that there are several varieties of gestural communication now in current use among the American deaf and that our studies are of one of these only. Beyond the immediate scope of our research are the various methods of rendering the English language in the hands: 1) By fingerspelling each letter of the words of English sentences; 2) by one of a number of sign systems developed in an effort to approximate English word order, morphology and syntax; or 3) by speaking English and signing simultaneously. Our studies, instead, are of American Sign Language which is likely to be the communication pre-lingually deaf people use among themselves and the gestural communication used in deaf families. This again is a language very different from any form of English - printed, written or spoken - as is shown by recent studies (Bellugi and Fischer, 1972; Fischer and Gough, in press; Bellugi and Klima, in press; Klima, in press). Thus, American Sign Language is not at all, as might be assumed, a rendition of English in the hands (this is one major misconception). There are, for example, differences in the lexical structure of the two languages, just as there would be between any two different languages. Some signs require several English words in translation; sometimes it is difficult to find a way to render in English the precise meaning of an ASL sign without resorting to lengthy paraphrase. There are ASL signs which translate into several different English words. This is similar to the difference in dictionary entries between any two languages. English and American Sign Language reveal essential differences not only in the form of their symbols and in the way they are combined (their syntax) but also in the very principles that govern modification of meaning in the English word as contrasted with the sign of American Sign Language.
2. Experiment:
Short-term
memory
in a language
in another
mode
2.1 Background What William James called ‘primary memory’, psychologists in recent years have refined and re-refined into a complex series of different memories. Commonly agreed upon are the characteristics of what is now generally called which is in Don Norman’s words ‘. . . a true working ‘short-term memory’, memory . . .’ (1972). Short-term memory is a transient memory of events of the present. Material can be inserted within it at will, and its contents are readily accessible. Presumably there is a sensory information storage system for each sensory modality which would feed directly into short-term memory.
Remembering
in signs
99
The properties of short-term memory are generally thought to include: A very limited temporary storage of approximately five to seven unrelated items; a preferred form of encoding based on some sort of phonological representation; and a loss from short-term memory within a few seconds if rehearsal is prevented. The types of errors that are made when information is retrieved from memory have led toward an agreement that for hearing people short-term memory has a phonological or phonetic basis (perhaps acoustic, perhaps articulatory, perhaps a combination of the two). Conrad (1962) was the first to show that intrusion errors on a short-term memory task that involved remembering visuaEZy presented printed letters correlated with errors in perception of spoken letters under noise. When given as a stimulus, the printed letter C is not ‘mis-remembered’ as the visually similar closed 0 but rather as the phonologically similar voiced Z. Since then, Conrad (1972) has shown that hearing children appear to code pictures pictorially up to a certain age and then appear spontaneously to abandon pictorial short-term memory coding in favor of word-based short-term memory coding. Indeed, by now there is abundant experimental evidence that, for most people, coding in short-term memory involves phonological characteristics of the material stored. This may well be the case for most people, but what happens with that special subgroup of the profoundly deaf who have native knowledge of a sign language learned ‘naturally’ from deaf parents? To explore this question, we have so far completed three experiments on shortterm memory for signs in the deaf, one of which is described here. 2.2 Method Signs of American Sign Language were presented on videotape at the rate of one sign per second by a native signer in list lengths of from three to seven signs. Subjects were deaf college students of deaf parents who learned ASL as a primary natural language. Recall was immediate, ordered and written. The experiment was preceded by a naming task in which subjects saw and identified the signs one at a time. In a parallel control experiment, English words corresponding to the ASL signs were presented on audiotape at one per second in comparable lists to hearing college students with no knowledge of sign language. There was a prior naming task corresponding to that for the deaf, and recall of spoken words was immediate, ordered and written. 2.2.1 Notation We have developed English words and
some notational devices to distinguish ASL to distinguish different levels of representation
signs from for both
100
Ursula Bellugi,
Edward
S. Klima and Patriciu Siple
signs and words. Consider, for example, the word ‘name’. In spoken language, we must distinguish between the actual sound (or, for example, some spectrographic representation of it) from a phonological representation such as /n e y m/. In ASL we must likewise distinguish the original gesture (or some pictorial representation of it) from an attempt to provide a notational description of it, e.g., the Stokoe” notation for the corresponding ASL sign: /Fi, H, :/ (see Stokoe et al., 1965). In this paper, we will further distinguish the orthographic representation of the spoken English word, which will be printed in small letters and underlined as in name. (This may be, but need not be, isomorphic to more systematic representations.) Since there is no written form for ASL, signs will be presented in this paper by the letter form of the English word in capitals which is a translation equivalent (or gloss) of the sign, as in NAME. (The form of the spelled English word has, of course, no relationship whatever to the form of the ASL sign.) The written responses of the subjects in our experiments, both hearing and deaf, are represented in this paper by the letter form of the English word in italics, as in name. The response name when written by hearing subjects represents a word of English. The same response name when written by deaf subjects represents an English ‘translation’ of a sign (without any necessity for sounds to be involved). That is, any time a form like nm occurs, a certain English sound form is referred to. Any time a form like NAME occurs, a certain ASL gesture is referred to. Any time a form like lzame occurs, a written response in English orthography is referred to. These distinctions are made in Figure 2. 2.2.2 S1rbjects The subjects were two groups of eight college students, one deaf group and one hearing group. The eight deaf subjects were students at Gallaudet College for the Deaf in Washington, D.C. All were deaf children of deaf parents and grew up in a home where the form of communication primarily used was American Sign Language. All of the subjects learned ASL as a primary natural language at home, just as the hearing subjects learned spoken English. The deaf subjects were, .of course, highly fluent signers. (Since they were college students one can also presume that they had some knowledge of written English; their command of other aspects of verbal codes was not determined.) 3. In Stokoe’s notational system the presence of two identical capital letters indicates that the hand arrangement is that of a symmetrical two-handed sign; the line above the first capital letter indicates that the dominant hand is active and above the non-dominant hand; the subscript to the
capital indicates the pointing of the hands generally inward in orientation; the particular capital letters indicate a hand configuration with the index and middle fingers extended from a fist; the collection of symbols to the left of the capitals indicates the movement of the sign.
Remembering
Figure 2.
Notation forms Language
Sound: Spoken word in American English
:I
Phonological
representation:
in signs
and variants for English and American
10 1
Sign
Gesture: Sign of American Sign Language (pictorial representation) \
Notational
representation:
/Fi> H< ;-/
‘Translation’ into spelled form of English word(s)
rOrthographic form of spoken word represented by underlined standard spelling: name
)
Translation equivalent of sign in English word(s), represented by capital letters: NAME /
II;n English response .by SubJects represented m italics:
1
102
Ursula Bellugi,
Edward
S. Klima and Patricia Siple
The hearing subjects were eight undergraduate students at the University of California at San Diego, and their participation in the experiment was a part of their course requirements. The eight subjects reported on here were randomly selected from ten who completed the experiment: None of these hearing subjects had any knowledge of a sign language.4 2.2.3 On choosing signs for the memory experiment We have already emphasized that ASL is a language very different from English and that there are differences between the lexical structures of the two languages just as there are between any other two languages. Nonetheless, just as any two languages used by people in more or less similar cultures will have some pairs of lexical items which are near semantic equivalents across the two languages (i.e., the two forms will commonly be inter-translatable), there are ASL signs and English words which form a fairly good match, as the ASL sign represented by the English gloss GIRL and the English word ‘girl’. Thus, although there are differences in the lexical correspondence between ASL and English, we can make use of the Dictionary of ASL and of the intuitions of native deaf signers with fluency also in English to compose lists of signs with fairly direct translation-equivalents in English. 5 It is these signs which we used, as far as possible, in our experimental studies in sign language. Thus, for this experiment we used some 135 common signs of ASL which had noun-like functions in ASL. Our first memory study (‘Remembering with and without words’, from now on RWW, 1973) helped in choosing the signs of ASL to be made into lists of random signs. There is, of course, no word frequency index in ASL, and so we had to rely on other criteria for the selection of signs. For the present experiment we chose signs according to the following criteria: a) Signs which occurred in the vocabulary of the young children in our language acquisition study (ages 18 months to 5 years); thus we included the signs for GIRL, MILK, COOKIE, TREE, CANDY, etc.; b) signs which were judged by several deaf signers as ‘commonly known’; thus we included the signs for CALIFORNIA, LAW, VOTE, ARMY, DIAMOND, etc., although these had not occurred in the signs made by the children in our study; c) signs which were easily 4. Special
thanks to Professor William Stokoe, Jr., of the Linguistics Research Laboratory of Gallaudet College and to David McKee for their help in arranging and administering the study with the deaf subjects. 5. In a study of the learning of English words by deaf subjects, it was already as-
certained that the presence of a corresponding sign facilitated the learning of the English word, suggesting that the English words are mediated, for some deaf subjects, by ASL signs (Odom, Blanton, and McIntyre, 1970). Our studies, to our knowledge, are the first to present signs themselves for remembering.
Remembering
in signs
103
translated by deaf signers into single English words; thus we excluded signs for which we then had evidence that there was disagreement in translation and signs which required more than one English word in translation, like HEARTBEAT or SIGHT-SEEING or ASSEMBLY-LINE6 ; d) we eliminated from consideration signs which seemed to us relatively long or large, or exceeded the usual signing space, in order to fit the requirement of presentation at a standard rate and to allow a close-up of the signer in making the test tape; thus we excluded some compound signs in ASL and some signs like DOG, SKUNK and LION, which exceed the usual signing space. In our preliminary explorations, we had tried different rates of presentation, making some lists at 1% seconds and some at 1 second per sign. We decided to use the rate of one sign per second, so that we could present stimuli at the same rate for deaf and hearing subjects. (We did find, during analysis of the results, that although the rate of presentation was one per second for signs of ASL and for spoken words of English, there still was a measurable difference between the two; since the signs took longer to produce than the words, more of the one-second interval of presentation was taken up in signing than in speaking.) We constructed lists of three, four, five, six and seven different signs, using each sign once for each list length, so that we used approximately 100 signs at each length. Other signs were selected from the remaining pool of signs in order to fill out an even number of lists for each list length. ‘l’hese lists were divided into two sets with half the lists for each list length selected for each set. In addition, ten extra three-item lists were added as a warm-up task. 2.2.4 Procedure The lists of ASL signs were presented to the deaf subjects on a Sony %” AV 3600 Videotape Recorder. They were made on videotape by a native signer, with approximately 12 seconds between lists. We framed the signs so that they would be maximally visible on videotape, including an area from forehead to just above the waist of the signer. The signs were made with facial expression suppressed. (This is rather unnatural for ASL, since facial expression ordinarily occurs concurrently with signing in the deaf. The absence of the usual facial expression allowed us to study the processing and remembering of ASL hand gestures without other confounding factors.) The signs were made with what may be considered the equivalent of a ‘list’ intonation. After completing the test-tape, another tape was made of each of the 135 signs presented slowly and clearly in random order with seven seconds between signs. This was for an identification and naming task which pre6. If more than one English word is required to translate an ASL sign, the words
are connected
by hyphens.
104
Ursula Bellugi,
Table
1.
Diagram
Edward
S. Klima
of experimen
Prelingually deaj subjects who learned ASL as a native language from deaf parents
Presentation mode
Response
task mode
Written Eng lish word translation equivalent of sign
Random list of signs used as stimuli in memory task -13 5 signs signed one at a time on videotape, 7 seconds between signs
Siple
tal design
I Naming and identification
Subjects
and Patricia
I
Short-term
memory
j Response
Presentation
t mode List lengths of 3,4,5,6 and 7 items Signs of AX made by native signer and presented on videotape in lists at rate of one sign per second, i.e., signed representation of Ir\H~q’\ / uF; /, 10’ F0 /
task mode
(Immediate, ordered written response)
Written English translation equivalents of remembered ASL signs, i.e., horse, cat, tea
:I
+ Hearing subjects, native Americans with no knowledge of sign language
135 spoken words presented on audiotape, 7 seconds between words
Written Eng
lish alphabet spelling of word
English words spoken on audiotape in lists at the rate of one word per second, i.e., spoken versions lhgrsl, lk=ti, itiyl
Written English alphabetical spelling of remembered spoken word, i.e., horse, cat, tea
ceeded the memory experiment. (See Diagram of Experimental Design, Table 1). The test materials were prepared for hearing subjects in an analogous manner. As stimuli for the hearing subjects, the words (i.e., the English translationequivalent of the signs) were spoken by a native speaker of American English and presented on an audiotape recorder. The same words in the same order were used for list lengths three through seven, and a tape was made for identification and naming. The procedures on the tests are analogous (ASL signs for deaf; English spoken words for hearing), except that the deaf responded by writing down what is for them the English translation-equivalent of the sign whereas the hearing responded by writing the alphabetic representation of the spoken word. (Justification for this will be discussed in a later section.)
Remembering
2.2.5 Identification
and naming
in signs
105
task
Instructions were presented for each part of the task for the deaf in ASL by a deaf person on videotape. Before beginning the memory experiment, subjects were asked to watch each sign on the videotape and to write down an English word translation for that sign. This naming task served two important functions. First, the English translations of ASL signs gave us some objective basis for scoring correct and incorrect responses for each individual subject. Second, the opportunity to identify and name each sign apart from the particular pressures of rapid identification and remembering in the memory task itself reduced the possibility that intrusion errors resulted from misperception of the sign rather than from some loss in the process of remembering. Analogously, the hearing subjects also began by identifying 135 spoken words from an audiotape recording. 2.2.6 Short-term memory task The experiment was divided into two sessions. In the first session, the naming task was followed by the warm-up lists of three signs, and then by one set of memory lists. Two weeks later, the subjects were given the second set of memory lists. In each session, all lists of three items were presented first, followed by lists of four items, up through lists of seven items. Each list was recalled immediately after presentation. Subjects were instructed to write the recalled items in their answer books in the correct serial positions indicated by columns on the page. They were allowed the freedom of writing the last presented items first, as long as they ended up with a correctly ordered sequence. Presentation for the deaf was monitored by a deaf experimenter who turned off the tape recorder after each list until all subjects had completed recall. 2.2.7 Scoring In the naming task, each deaf subject had given an English translationequivalent (a written English word) for the 135 signs. We catalogued for each subject the instances in which the name he gave for a sign differed from the word which we had chosen as a gloss for that sign. While ordered recall was required by the instructions, we scored the results in two ways. In position scoring, an item was scored as correct if it was recalled correctly in the correct position. In item scoring, an item was considered correct if it had appeared anywhere in the presented list. A written word was considered as a correct response for hearing subjects if it was the same as the word listed by that subject for naming or some homophone. For the hearing subjects, the kind of variations which occurred in naming in the preliminary naming task were for the most part homophones: bare and bear; knight and night; tea and tee.
106
Ursula Bellugi, Edward S. Klima and Patricia Siple
There were occasional real ‘errors’: One hearing subject wrote purple for the word pronounced turtle. In the memory experiment with hearing subjects, not only the word actually presented but also the subject’s response on the naming task was considered as a correct response. For the deaf subjects, the same principle of scoring was used. A response was considered as correct if it matched either the English translation given by the subject himself in the naming task or appropriate translations provided by native deaf signers or by the Dictionary of ASL. As is expected in translating between two languages, subjects sometimes differed in the word chosen as a translation-equivalent for a sign. For 108 of the 13.5 signs used, at least seven of the eight deaf subjects gave the same word as their translation-equivalent for that sign. When greater variation occurred, the alternate translations for the most part were (a) synonyms of the chosen name (cop and policeman); (b) the verb rather than the noun form for the name chosen (eat and food); or (c) homonyms in ASL (gold and California). We found that in nearly all cases, subjects consistently used the names they had given on the identification task as translation-equivalents for the correctly remembered sign. Again, either the name the subjects gave on the naming task or the Dictionary listing for that gesture was scored as correct for that subject. Thus, if a subject wrote city for the sign we glossed as TOWN, either response was scored as correct for him. We scored separately several classes of responses; and most significant among the responses scored as errors were the intrusion errors. By ‘intrusion error’ in this study, we mean a response that was incorrect for a position on the specific list in question but was: (a) Not a misplaced item from that list; (b) not a repetition of a response made to another item on that list; (c) not an occurrence of an item from the immediately prior list. Among the intrusion errors that were in fact made by our subjects were not only items that occurred somewhere else on the test (i.e., among the signs that had been identified in the naming task) but also numerous new items added by the subjects which did not appear anywhere on any list. 2.3. Results and discussion
2.3.1 Mean number correct: Memory span The mean number of items correctly recalled in the correct position are presented for each list length in Figure 3 for the two experimental groups. Memory span, defined as the threshold, or list length, at which exactly half the lists were recalled correctly, was also computed (Woodworth, 1938, p. 202). Both these measures are indicators of the capacity of short-term memory. Memory span for the deaf in signs is 4.9 items and is approximately one item shorter than for the hearing subjects in words (5.9 items). This
Remembering
Figure 3.
Mean number subjects
correct
scm~G
7-
I-
I
I 4
I 5 LIST
Figure 4.
I
I
6
7
LENGTH
Serial position curves based on percent HEARlNG 9.
107
at each list length for hearing and deaf
POSITION
3
in signs
correct recall DEAF si
108
Ursula Bellugi, Edward S. Klima and Patricia Siple
difference was found to be significant by Newman-Keuls (p < .Ol). An ANOVA performed on the data for the deaf and hearing groups (position scoring, mean number correct) was consistent with the difference in memory span. The hearing subjects recalled more items overall (F = 12.16, p < .Ol >. An interaction between experimental groups and list lengths (F = 5.30, p < .05> was also found since the deaf reached asymptotic performance before the hearing subjects. The same pattern of results was found using item scoring. Why does this difference in overall performance exist for the two groups? To answer this question, we must take a closer look at the data. 2.3.2 Serial position Serial position curves for deaf and hearing subjects are shown in Figure 4. Except for the differences in overall level of performance, no gross differences were found between the groups. Both produced bowed serial position curves with large primacy and recency effects. This similarity in serial position curves can be taken as evidence for similar processing mechanisms in our deaf and hearing populations. For hearing subjects, the recency portion of the curve is usually attributed to an ‘echo box’, or precategorical acoustic store, for auditorily presented words (e.g., Crowder and Morton, 1969). While this store is short-lived, a few recent items can be retrieved from it if they are recalled immediately, producing the recency effect. A similar precategorical store is thought to exist for visual presentation. However, its shorter duration leaves less time for the reading off of visually presented words. Thus, visual presentation results in a slightly reduced, but still strong, recency effect. Other authors (e.g., Sperling and Speelman, 1970) argue that incoming words are immediately encoded into a phonemic representation whether presented visually or auditorily and that the entire recency effect is due to an auditory store. Our data clearly show strong recency effects for both deaf and hearing subjects. However, it would be inappropriate, as we will describe later, to assume that our deaf subjects were encoding the signs as auditory words in order to store them. Since our deaf subjects clearly show what has been called ‘the echo box effect’, we must conclude that the echo can be in some other modality - that there is more than one kind of precategorical store. Clearly, a similar mechanism operates for both groups. For hearing subjects, the recency effect seems consistently to cover the last two items in list lengths 5, 6 and 7, which is a common result at this rate of presentation. For the deaf, there is also a recency effect, but the recency portions of the curves for comparable list lengths 5, 6 and 7 are not the same (one, two and three items, respectively). These differences are most likely due to changes in recall strategies as the
Remembering
in signs
109
deaf subjects discover that they can recall more by starting with the last few items. While hearing subjects seem to do this almost automatically, we have often observed that our deaf subjects seem to ‘light up’ as they begin to use this strategy. Since this strategy is only necessary once memory span is exceeded, we would expect our deaf subjects to discover it only at longer list lengths, producing the greater recency effects observed. The primacy portions of the curves are consistent in their differences between deaf and hearing. At the longer list lengths, the primacy effect for the deaf subjects covers the first two items b_efore the curve begins to level off or rise. For the hearing subjects, the primacy effect extends one item further into the list. The primacy effect is considered to be a product of the rehearsal mechanism which keeps items available until they can be transferred into a more permanent memory system (Atkinson and Shiffrin, 1968). Items at the beginning of the list get more rehearsals and, thus, have a greater probability of being stored in memory. If we allow ourselves to consider rehearsal to be the same as implicit speech (or, for the deaf, the same as implicit signing), then we can measure its rate. Landauer (1962) has shown that the rate of implicit speech is no different from overt speech. We have likewise found that the rate of overt signing does not differ from the rate of implicit signing. Thus, the rate of rehearsal and the rate of transferring items into long-term memory should be proportional to the rate at which the items can be produced. In a comparison of rate of signing and rate of speech, we have shown that the rate of word production is nearly twice that of sign production. (Bellugi and Fischer, 1972.) We would then expect that, if the deaf are rehearsing in signs, they should have a smaller primacy effect than the hearing, since they would have fewer opportunities for rehearsal, and they do. Thus, over all, short-term memory mechanisms, like those producing rehearsal and the ‘echo box’ effect, seem to parallel those found in hearing subjects, regardless of the modality change. We have seen that the memory span for our deaf subjects was about one item less than that for the hearing subjects. We now see that this difference is most probably due to the nature of the rehearsal process for signs and words for these two groups.
2.4 Intrusion errors For us as psycholinguists, the critically interesting questions revolve around what we can learn about coding in short-term memory from the evidence of that special set of errors called intrusion errors that were made by deaf subjects - what we can learn about the encoding processes of deaf subjects attempting to remember signs in such an experimental situation and what we
1 10
Ursula Bellugi, Edward
S. Klima and Patricia
Siple
can infer therefrom about the nature of the internal structure of the signs of a primary gestural-visual language. There is now a tradition in the use of intrusion errors to investigate the nature of encoding in short-term memory. Baddely (1966), Conrad (1962), Hintzman (1967), Wickelgren (196.5) and Sperling (1963) have shown that when verbal items (words or letters) are visually presented for immediaterecall or for recall delayed by several seconds, the material is encoded in phonological form by hearing subjects. This is an important finding and makes an interesting claim. Our studies show that the intrusion errors in short-term memory for signs by deaf native signers also give clues to the nature of processing in a language in a different modality. An intrusion error, as previously stated, was considered to be any response given in a particular serial position which was different from any of the set of items considered correct for the item presented in that position, excluding repetition errors and errors which were items from the immediately previous list. From the point of view of the concerns of this paper, a most critical fact is that there was zero overlap between the intrusion errors made by deaf subjects and by hearing subjects to the same items: In other words, the errors made by the deaf subjects were in y10 instances the same as the errors made by the hearing subjects. As is by now a familiar result in short-term memory experiments, the intrusion errors made by our hearing subjects were often words which have some phonological resemblance to a word that was on the original list. The error made by the deaf subjects were of a very different sort. In the preliminary memory study (RWW), we compared immediate ordered recall on the same lists of ASL signs when the responses were in written form (English words) as contrasted to when they were in signed form (ASL signs). That is, we asked for two types of responses: Subjects viewed the lists of signs on videotape and, depending on prior instruction, responded either in the written form of the translation-equivalent of the sign remembered or responded by actually signing ASL sign which were videotaped. At the end o? transcribed (using English translationthe experiment, the subjects equivalents) their own signed responses from videotape. Of the multiple intrusion errors which occurred to a single stimulus in RWW, in over half of the instances the same error occurred both when the response was actually signed in ASL by a deaf subject and when the response was written down by a deaf subject in the form of an English word serving as translation-equivalent of the sign remembered. This suggested to us that the deaf subjects were using some of the same strategies for remembering when the response required was in ASL signs as when the response required was in a different language and a different mode. The types of intrusion errors made in signing - and we shall see that these were ASL-based (not English-based) errors - were carried over
Remembering
in signs
1 11
into the written English responses. This was very convenient for the current experiment because it allowed us the advantage of group presentation which would not have been possible if each subject had to be videotaped while responding in signs. Again in RWW, we found that the intrusion errors made by deaf subjects had little or no overlap with the intrusion errors made by hearing subjects (although a correct written response for both might be the same). This suggests that the deaf subjects were using different strategies for encoding than were hearing subjects. Consider some specific examples of intrusion errors for hearing and for deaf subjects in this experiment: For a word that a hearing person heard as horse, the intrusion error given was the written response house. For the American Sign Language gesture that a deaf person named as HORSE, the intrusion error was the written response uncle. There is similarity in phonological form between the spoken words horse and house: They differ only in their internal segments’. Equally, there is a close relationship in form between the signs represented by the English translation equivalents HORSE and UNCLE; that is, the signs are highly similar visually. The written response tree was given as an intrusion error for the spoken word tea - more than once by our hearing subjects. The written translation equivalent tree was given as an intrusion error more than once by our deaf subjects for the sign which they named NOON. In both cases, the errors bear a resemblance to the original item: tree and tea are auditorily similar; the sign for TREE and the sign for NOON are visually similar. 2.4.1 Errors made by hearing subjects Let us examine some of these item-and-error pairs in more detail. Again, the errors made by our hearing subjects by and large were sound based, paralleling the results from short-term memory experiments with hearing subjects previously cited. There were hearing errors that differed from the word presented in initial segment only: Vote, e.g., was mis-recalled as boat, peas as knees, soap as hope. There were errors that differed from the presented word in medial segments only: Wood mis-recalled as word, horse as house. There were errors that differed from the word presented in final segment only: Coke mis-recalled as coat, bath as bat, word as work. And there were some errors that were not phonologically very similar to the original word presented (car and fat>. There were also a very few errors that could be said to share category membership at some level, which is one indication of similarity of meaning: turtle, and cattle, for example (see Table 2 below). The predominant impression overall, however, is that a significant number of the errors made by the hearing subjects were sound-based and of such a
1 12
Ursula Bellugi, Edward S. Klima and Patricia Siple
nature as supports an analysis of words into phonological segments. From this experiment one might indeed infer that the hearing subjects were coding and remembering words presented on the list in terms of these phonological properties. We should recall here that we would expect this result from hearing subjects regardless of the mode of presentation of the material. That is, had we presented words in their printed form (i.e., as visual stimuli), we could anticipate similar results - errors based on the sound-form of words. For that reason, with hearing subjects, short-term memory is claimed to have a phonological basis. 2.4.2 Errors made by deaf subjects Now let us turn to the errors made by deaf subjects to ASL signs, remembering that the responses of both groups of subjects were made in written English words. Of the intrusion errors made by the deaf, 26 occurred more than once (multiple errors). It is these multiple errors that we used as the basis for more intensive analysis. We first note that the errors are not at all similar to the errors made by hearing subjects and that there was zero overlap in actual responses. The deaf errors were clearly not based on the sound of the words. Furthermore, an examination of the lists yields no evidence that the deaf errors as written down were related to the English word translation of the original sign presented in terms of some similarity in the graphic forms themselves (i.e., the letters, their shape or their number). We have mentioned that some signs could be presumed to have an iconic origin. While it is difficult to specify precisely what kind of error would give clues that the signer is coding in terms of the ‘iconic’ (or representational) origin of a sign, there were certainly no examples of error pairs that would suggest that this even might be the case. The error for the sign CAT, for example, which supposedly represents the whiskers of a cat, was never purr, paws, meow, claws, fur or other things that might be included in the delineation of a cat along with its whiskers; rather, one error for the sign CAT was Indian. One must thus seek other bases for the errors that occur. In fact we did find a general type of error which occurred frequently. If we re-convert back into ASL signs the English w’ords written down as responses, we find that in a striking number of cases, the sign presented as stimulus and the sign which is an ASL translation-equivalent of the ‘error’ are highly similar visually. Not only are the two signs similar, but they are similar in form in specific and predictable ways. The errors tended to preserve the basic hand arrangement of the original sign presented: That is, if the sign presented was a one-handed sign, or with two hands active, or with one hand acting on the other as a base, the sign recalled in error tended to preserve this characteristic. But more significant than that, the majority of the multiple errors preserved
Remembering
Figure 5. SIGN
HOME
in signs
113
Pairs differing only in Hand Configuration ERROR
YESTERDAY
SIGN
CANDY
ERROR
APPLE
all but one of the parameters of the original sign and differed from the sign presented in only one prime7 of one parameter. Let us illustrate this by some examples. (See Figures 5-8.) For each item-and-error pair, the original sign presented is shown in the drawing on the left and the translation-equivalent in ASL of the written word response which counted as an error is shown in the drawing on the right. There were sign-and-error pairs which differed in Hand Configuration only (Figure 5). That is, the sign which is the ASL translationequivalent of the written intrusion error preserved the basic form and also preserved the Place of Articulation and Movement of the original sign, so that the only difference between the sign presented and the error is a difference in the particular configuration in which the hand is held during the sign. Among these are the sign CANDY mis-recalled as apple (the written response), HOME as yesterday, ROLL as who, PENNY as idea, WEEK as nice and CANDY as jealous.
There were a number of sign-and-error pairs which differed only in Movement (Figure 6). That is, the sign which is the ASL translation-equivalent of the written intrusion error preserved the basic form of the original sign and furthermore preserved the Handshape and Location of the original sign, so that the only difference between the sign presented and the error is a difference in the Movement of the two signs. There were very many examples of such sign-and-error pairs, among them: The sign BEER was mis-recalled as brown, CAT as Indian, GIRL as aunt, KNIFE as peas, NAME as egg, NOON as tree, VOTE as tea, MONTH as temperature and PENNY as for. 7. We refer to the particular values of the major parameters as the primes of each of these parameters; thus, the ‘tapered 0’, for
example, is one prime Hand Configuration.
of the parameter
1 14
Ursula Bellugi, Edward S. Klima and Patricia Siple
Figure 6.
Pairs differing only in Movement
SIGN
ERROR
SIGN
ERROF
NAME
EGG
NOON
TREE
Figure 7.
Pairs differing only in Orientation ERROR
SIGN
STAR
SOCKS
Figure 8.
BIRD
Pairs differing only in Place of Articulation
NEWSPAPER
RUBBER
DOLL
Remembering
ilz signs
1 15
In this experiment, there were some pairs that differed only in Orientation of the hands (Figure 7), as when the sign SOCKS was mis-recalled as sf~r, but there were no sign-and-error pairs which differed only in Place of Articulation. With a larger sample of subjects, we assume some errors of this type would have occured since in other memory experiments we have completed there were some sign-and-error pairs differing only in Place of Articulation (Figure 8): As examples, the sign RUBBER was mis-recalled as doll, BIRD as newspaper, and DOCTOR as duty (RWW and forthcoming).8 In addition, as with the errors from hearing subjects, there were instances of sign-and-error pairs which had no discernible relation and some instances which could be said to share category membership. (See Table 3 below.) A significant number of the errors by the deaf subjects, however, were visually based. 2.5 Visual and auditory
similarity
ratings
Having experimentally supported the supposition of the existence of distinct formational parameters in the structure of ASL signs, we next turned to questions of degree: To what degree are the errors made by deaf subjects visually based (i.e., based on ASL signs) ? And how does that compare with the degree to which the errors made by the hearing are sound-based? For these purposes, we took an equal number of item-and-intrusion-error pairs made by hearing and by deaf subjects, presented them at one time as pairs of signs for visual similarity ratings and at another time as pairs of spoken English words for auditory similarity ratings. For quantitative analysis, we selected all the intrusion errors which were made more than once to a single stimulus item (the multiple errors previously referred to). There were 26 such multiple intrusion errors made by deaf subjects when ASL signs were signed and eight made by hearing subjects when English words were spoken. In order to obtain lists of equal length from the intrusion errors of the deaf and hearing subjects, 18 other errors were randomly selected from those which occurred once among the hearing subjects. The printed form of the English translation-equivalent of the 26 item-and-intrusion-error pairs made by the deaf and the printed form of the 26 item-and-intrusion-error pairs made by the hearing subjects were combined into one list and randomized. These 52 pairs were then prepared in 8. We have completed in short-term memory strutted three types (a) Signs which share of Hand Configuration the other two major
another experiment in which we conof lists of signs: the particular prime but are different on parameters; (b) signs
which share a particular Location but are different on the other major parameters; and (c) signs which share a particular Movement prime but are different on the other major parameters.
1 16
Ursula Bellugi,
Edward
S. Klima and Patricia Siple
the two different modes; the pairs of stimulus items and intrusion errors (combined and randomized for the two groups) were recorded on audiotape in spoken form by a hearing person; the same pairs of items and intrusion errors were translated into signs of ASL and made on videotape by a native signer. a. Hearing raters. Ten hearing subjects with no knowledge of sign language were asked to rate the 52 pairs of signs for visual similarity. (The subjects were not told the source of the pairs or the meaning of the signs.) A scale of 1 to 5 was used with a 1 rating meaning most similar and a 5 rating meaning most different. Ten other hearing subjects were asked to rate the set of items from the audiotape presentation in terms of acoustic similarity along the same scale. In each case, we gave examples of pairs not on the list that could be considered as most similar or most different. The results are shown in Figure 9. It can be seen that there is a clear separation for the two groups of subjects. The errors made by the deaf had a mean rating of 2.63 when rated as pairs of signs and a mean rating of 4.20 when rated as pairs of spoken words, while the errors made by the hearing had a mean rating of 4.19 when rated as pairs of signs and a mean rating of 2.58 when rated as pairs of words. That is, when all item-and-error pairs were presented as spoken words and rated on the basis of auditory similarity, the figure shows that a large proportion of the errors made by deaf subjects were rated as low in auditory similarity, while the majority of the errors made by hearing subjects were rated toward the ‘similar’ end of the scale. Conversely when all the item-and-error pairs were presented as signs, and hearing subjects with no knowledge of sign language were asked to rate them for visual similarity, a large proportion of deaf item-and-error pairs were rated as visually similar. In fact, according to the mean ratings, the errors made by the deaf are visually similar to the original item presented to the same degree that the errors made by the hearing subjects were auditorily similar to the original item presented. Not only was there no overlap in terms of actual intrusion responses made by the deaf and by the hearing, but we can now give evidence that the very nature of the errors made by hearing and deaf subjects is fundamentally different. The multiple intrusion errors made by the deaf to signs of ASL are demonstrably errors which are visually similar to the sign presented. b. Deaf raters. That the errors are not visually similar in just some general way is suggested by another rating result. It is important to bear in mind that what we have presented above are similarity ratings made by ten hearing subjects with no knowledge of any sign language. Next we asked deaf people skilled in ASL to rate the item-and-error pairs for visual similarity. The results pairs show, for both groups, a clear separation between item-and-error generated by the deaf and those generated by the hearing. However, in
Remembering
Figure
,
9.
OF
,TEM
Visual and auditory and-error pairs VISUAL RATINGS
AN0
INTRUSION
ERROR
similarity
ratings by hearing
PAIRS
I
OF
ITEM
in signs
1 17
raters of item-
AUDITORY RATINGS ANO INTRUSION ERROR
PAlRS
II -
10 -
e-
2
8-
iz,_ FJ = 6-
1
15432,-
i
addition we have found an important and interesting difference between the visual similarity ratings made by naive hearing subjects and the same ratings made by deaf subjects fluent in ASL. Overall, there was considerably greater variance in the ratings made by the hearing subjects than by the deaf subjects, who had a greater degree of agreement. The mean of the standard deviations of the ratings for deaf raters was .41 and for hearing raters .84. The difference between matched pairs of S.D.‘s of deaf raters and hearing raters was found to be significant by the Wilcoxon test (z < .Ol). The difference is particularly striking with respect to the item-and-error pairs generated by the hearing subjects: For these pairs, the mean S.D. for hearing raters was .84, while for deaf raters it was only .19. For the deaf raters, the mean rating of these errors was 4.88, almost the maximally dissimilar rating of (5). The hearing raters also rated these errors as low, but not as low, and their scores varied significantly more. Let us take a typical example. For the signed item-and-error pair DRESS and TOWN, the ten hearing raters scored it as follows: Three hearing subjects
1 18
Ursula Bellugi,
Edward
S. Klima and Patricia Siple
gave the pair a rating of (2); one subject gave it a rating of (3); two a rating of (4); and four a rating of (5). In contrast, all ten deaf raters agreed unanimously in assigning the pair a rating of (5). A hearing person, when afterwards asked to describe the similarity between the two signs, noted the following. Each sign uses two hands. In each sign the hands come together and do the same action. Both signs are in the same area in front of the chest. Both signs use the whole hand to make the sign. Thus this subject rated the signs as similar, or (2). Now let us examine the way in which a deaf person analyzed the signs. He noted that the movement of the two signs differs: It is a contact in TOWN and a downward brushing in DRESS. The hands in the two signs are different: TOWN uses an open flat non-spread hand, and DRESS uses an open spread hand with the middle finger bent forwards. (These are, in fact, two distinct recurring formational primes.) The signs are made in two different places: TOWN is made in the space in front of the signer, and DRESS is made by touching the chest. The deaf person, therefore, rated the pair as maximally dissimilar, or (5). For a deaf person, then, none of the primes of the major parameters are the same for this item-and-error pair, and the general vague similarity noted by a hearing person does not count. For the deaf raters overall, the item-and-error pairs generated by the hearing subjects had virtually nothing in common-nothing, according to our interpretation, significant from the point of view of the special formational parameters of American Sign Language. 2.6 Semantic
similarity
ratings
The evidence presented so far already indicates that the deaf subjects are coding and rehearsing in signs to the same extent that the hearing subjects are in words. The next question we asked is to what extent the item-and-error pairs are semantically related. In order to investigate the degree of semantic similarity of item-and-error pairs of the deaf and of the hearing, we presented the same randomized list of 52 pairs (this time in printed form) to another group of ten hearing subjects. The subjects were instructed to rate the pairs of words only on the basis of similarity of meaning, again on a five-point scale with (1) meaning most similar and (5) meaning most different. The item-anderror pairs of both the hearing and the deaf (the items of the latter, of course, in English translation-equivalent form) were rated toward the dissimilar end of the scale: The mean of the deaf item-and-error pairs for semantic similarity was 3.77, and the mean of the hearing item-and-error pairs was 4.33. The slightly lower mean rating for the deaf (still, however, showing the item-and-error pairs to be semantically dissimilar) is accounted for by a few
Remembering
in signs
119
examples of pairs which are related both in formational properties and in meaning. Nearly all of the errors by deaf subjects have no discernible semantic relationship to the item presented (i.e., VOTE and tea, NAME and egg, NOON and tree, etc.). Among the few pairs which are both visually and semantically similar are GIRL and aunt, BOY and man. The signs GIRL and AUNT share Hand Configuration and also Place of Articulation (on or near the lower cheek). Similarly, the signs BOY and MAN share Hand Configuration and also Place of Articulation (on or near the forehead). In fact, signs for several terms denoting human males are made on or near the forehead, including not only BOY and MAN, but also UNCLE, FATHER, GRANDFATHER, HUSBAND, BROTHER; and signs for several terms denoting human females are made on or near the lower cheek, including not only GIRL and AUNT, but also MOTHER, GRANDMOTHER, WIFE, WOMAN and SISTER. Thus a few Figure 10.
Semantic similarity ratings by hearing raters for printed item-anderror pairs SEMANTIC OF
I3 I2 -
ITEM
AND
m
DEAF
0
HEARING
RATINGS
INTRUSION
ERROR
ERRORS
PAIRS
ERROR
PAIRS
120
Table
Ursula Bellugi, Edward S. Klima and Patricia Siple
2.
Hearirlg subjects:
item presented in spoken words
In trusiorz errors
Hearing subjects’ errors
Mean similarity Auditory similarity of words ~~~_
* vote __~ * * *
tea coke -_ father -_ seat 3 wood peas bath * work coffee word horse gravy train water * b& turtle book friend * yeste* m x *shoe movie *
boat tree coat brother eat hope word knees bat word COPY work house baby tree mother fat cattle body fish day day fat coat day
town Mean rating
ratings
Visual similarity of signs
Semantic similarity of pairs
1.0 1.0 1.1 1.1 1.2 I .3 1.6 1.6 1.6 1.7 1.8 1.8 2.1 2.2 2.8 3.0 3.0 3.1 3.4 3.5 3.5 4.4 4.1 4.8 4.9 4.9
4.0 4.5 4.8 4.6 4.1 4.2 4.2 3.9 4.4 3.7 3.9 3.6 4.8 4.4 4.4 2.6 4.7 4.8 4.4 3.9 4.5 4.5 4.1 3.0 4.6 3.7
4.9 4.3 4.8 3.9 3.7 4.8 5.0 4.9 5.0 4.4 4.9 4.4 4.4 4.6 4.3 4.9 4.8 2.8 4.8 4.5 1.6 4.8 4.8 2.9 4.7 3.7
2.58
4.19
4.33
Intrusion errors made by hearing subjects used for similarity judgments by hearing raters. The starred errors are multiple errors; the others were selected randomly from the full list of hearing errors. The pairs were rated as signs for visual similarity, as spoken words for auditory similarity and as printed words for semantic similarity.
Remembering
Table 3.
Deaf subjects:
Multiple
Translation equivalent of signed item
Deaf subjects’ errors
GIRL CANDY
aunt peas apple
CAKE
CUP
VOTE
tea gum brown grease Indian jealous sit man house town tree glass egg owl tea penny shoe town milk gum color cow
KNIFE
CANDY BEER LEAK CAT CANDY SALT BOY TOWN HOUSE NOON CAKE NAME SUN KEY LETTER SALT TEA MEAT TURTLE SOAP ANIMAL
Mean rating
intrusion
in signs
12 1
errors
Mean similarity
ratings
Visual similarity of signs
Auditory similarity of words
Semantic similarity of pairs
1.3 1.4 1.5 1.5 1.8 1.8 1.9 1.9 1.9 1.9 2.1 2.1 2.2 2.2 2.3 2.4 2.6 2.7 3.4 3.1 3.8 4.1 4.2 4.3 4.7 4.8
4.9 4.7 4.9 3.0 4.8 5.0 3.1 3.2 5.0 4.9 2.4 4.9 4.4 4.5 4.9 4.7 4.7 4.9 1.1 4.4 3.6 3.3 3.3 5.0 4.4 5.0
2.3 4.2 2.3 3.7 4.9 1.9 4.2 3.6 4.1 4.8 4.9 1.3 2.6 2.7 4.7 4.5 4.7 4.1 4.9 4.8 4.9 4.8 2.0 4.8 4.5 1.7
2.63
4.20
3.77
Multiple intrusion errors made by deaf subjects were used for similarity judgments hearing raters. The pairs were rated as signs for visual similarity, as spoken words auditory similarity and as printed words for semantic similarity.
by for
I22
Ursula Bellugi, Edward S. Klima and Patricia Siple
errors may well reflect the combination of semantic and formational properties. Some of these formational and semantic relationships in ASL are described in Frishberg and Gough ( 1973). Despite these minor occurrences of item-and-error pairs which are related both semantically and in their form, it is clear from our results that the major source of intrusion errors is from signs which are exclusively similar in form to the original sign presented on the test. (See Figure 10 for mean semantic ratings and Tables 2 and 3 for data from three similarity ratings. made by hearing raters.)
3. Conclusions We have argued that for one group of deaf subjects - deaf people whose native primary language is the visual gestural system called American Sign Language - intrusion errors in short-term memory may indicate significant aspects of the nature of the coding of a special restricted set of visual symbols: ASL signs. For these subjects, even when the responses are in written English words, multiple intrusion errors do not at all reflect, as they do for hearing subjects, the phonological structure of the words; nor, rather surprisingly, do they reflect the visual form of those words in terms of the letters, their shape or number. Nor do the errors seem to reflect, as might be supposed, the somewhat more iconic (or representational) character of the basic meaningful units of sign language; nor, finally, do those errors reflect an essentially semantic organization in the processing and remembering of signs in the short-term memory paradigm. Rather, the multiple sign-and-intrusion error pairs reflect special organizational principles of American Sign Language. The results of this experiment show that the errors made by the deaf in a short-term memory task in ASL can be described in terms of a specific, limited set of simuZtaneousZy occurring formational parameters, along the lines of the parameters suggested by Stokoe. More significantly, the results are consistent with a theory that the signs of ASL are actually coded by the deaf in terms of such parameters. A similar claim has already been made about spoken language - in terms, of course, of the altogether different types of parameters characteristic of spoken language, involving phonological segments occurring sequentially; our experiment in short-term memory for hearing subjects is consistent with that claim. According to our memory experiment with a special group of deaf subjects, thus, there was no reason to suppose a priori - as had been supposed - that signs are processed as unitary wholes without systematic internal analysis. The
Remembering
in signs
123
results of our experiment are suggestive of a high degree of systematicity in the form of recurring formational parameters which combine simultaneously to constitute the individual signs of ASL. These results also suggest that, whatever force iconicity may have had in the past with respect to the origin of ASL signs, or may now exert in a different context of language usage (for example, in teaching signs to a hearing person), the signs are treated by the deaf in this experiment as consisting of parameters which, at a significant level of structure, are essentially arbitrary in terms of meaning. This result is consistent with others of our studies, including an analysis of the historical development of ASL signs, in which a general trend from comparatively more iconic to more arbitrary was noted. The sign for SWEETHEART, for example, some fifty years ago, was much more a representation of a beating heart (the hands held together like the shape of the bottom of a valentine heart, directly over the signer’s heart). Currently, the sign has moved in location, following a general trend toward centering, and has lost the ‘heart’ shape, replacing it by appropriate arbitrary parameters. (See Frishberg, in press for detailed discussion of historical change from iconicity to arbitrariness.) As Conrad observed, by now there is abundent evidence that short-term memory ‘thrives on a speechlike input’ (Conrad, 1972, p. 231). Our studies show that, in a broadened context, it would be advisable to amend this to ‘language-like input’, so that the characterization is not restricted to the speech mode.
REFERENCES Atkinson, R. C., and Shriffrin, R. M. (1968) Human memory: A proposed system and its control processes. In K. W. Spence and J. T. Spence (Eds.), The psychology of learning and motivation: Advances in research and theory. New York, Academic
Press. Baddely, A. D. (1966) Short-term memory for word sequences as a function of acoustic, semantic and formal similarity. Q. J. exp. Psychol., 18, 362-365.
Bellugi, U., and Fischer, S. (1972) A comparison of sign language and spoken language. Cog., 1, 173-200. and Klima, E. (in press) Aspects of sign language and its structure. In J. Kavanagh and J. Cutting (Eds.),
The role of speech in language. Cambridge, M.I.T. Press. and Siple, P. (1974) Remembering with and without words. In F. Bresson (Ed.), Current problems in psycholinguistics. Paris, Centre National de la Recherche Scientifique, Pp. 215-236. Conrad, R. (1962) An association between memory errors and errors due to acoustic masking of speech. Nature, 193,1314-1315. (1970) Short-term memory processes in the deaf. Brit. J. Psychol., 61, 179-195.
124
Ursula Bellugi,
Edward
S. Klima and Patricia Sipk
-~
(1972) Speech and reading. In 3. Kavanagh and I. Mattingly (Eds.), Language by ear and by eye: The relationships between speech and reading. Cambridge, M.I.T. Press. Crowder, R. G., and Morton, J. (1969) Precategorical acoustic storage (PAS). Preception and Psychophysics, 5, 365-373. Fischer, S., and Gough, B. (in press) Verbs in American Sign Language. In E. Klima and U. Bellugi (Eds.), The Cambridge, language. signs of Harvard University Press. Frishberg, N. (in press) Arbitrariness and iconicity: Historical change in American Sign Language. In E. Klima and U. Bellugi (Eds.), The signs of lanHarvard Univerguage. Cambridge, sity Press. and Gough, B. (1973) Morphology in American Sign Language. Unpublished manuscript. The Salk Institute for Biological Studies. Furth, H. G. (1966) Thinking without language. New York, The Free Press. D. L. (1967) Articulatory Hintzman, coding in short-term memory. J. verb. Learn. verb. Beh., 6, 312-316. Klima, E. (in press) Sound and its absence in the linguistic symbol. In J. Kavanagh and J. Cutting (Eds.), The role of speech in language. Cambridge, M.I.T. Press. Landauer, T. K. (1962) Rate of implicit
speech. Pert. motor Skills, 15, 646. Norman, D. (1972) The role of memory in the understanding of language. In J. Kavanagh and I. Mattingly (Eds.), Language by ear and by eye: The relationship between speech and reading. Cambridge, M.I.T. Press. Odom, P. B., Blanton, R., and McIntyre, C. K. (1970) Coding medium and word recall by deaf and hearing subjects. J. sp. hear. Res., 13, 54-58. Sperling, G. (1963) A model for visual memory tasks. Hum. Factors, 5, 19-31. and Speelman, R. G. (1970) Acoustic similarity and auditory short-term memory experiments and a model. In D. A. Norman (Ed.), Models of human memory. New York, Academic Press. Stokoe, W., Jr. (1960) Sign language structure: An outline of the visual communication system of the American deaf. In Studies in linguistics. (Reissued Washington, D.C., Gallaudet College Press). and Croneberg, C. Casterline, D., (1965) A dictionary of American Sign Language. Washington, D.C., Gallaudet College Press. Wickelgren, W. A. (1965) Distinctive feaerrors in short-term tures and memory for English vowels. J. acoust. Sot. Amer., 38, 583-588. Woodworth, R. F. (1938) Experimental psychology. New York, Henry Holt.
R&me’ Dans une experience de memoire a court terme, on a present& des listes de 3 a 7 items de signes gestuels de PA. S. L., a des sujets, etudiants universitaires sourds, dont la langue maternelle etait 1’A S. L. Une experience similaire de memoire a court terme a Bti: effectuee parallelement, avec des mots representant l’equivalent des signes gestuels traduits en anglais, sur des Btudiants qui entendaient. Immediatement aprts la presentation
des items, un rappel ordonnt et tcrit Btait demand& On a remarqui: d’abord que les mecanismes de la memoire B court terme chez les sourds sont comparables a ceux des sujets qui entendent mdme avec le changement de modalitt. Un nombre significatif d’erreurs de type intrusions multiples fait par les sourds en signant est lie a des proprietes de formation des signes. Ceci se compare aux erreurs de type phonologique que l’on constate chez Ies sujets qui
Remembering
entendent. Les resultats viennent confirmer la theorie selon laquelle les gestes de 1’A. S. L., seraient codes par les sourds en termes de parametres situationnels simultan& tels que Configuration de la Main,
in signs
125
Lieu de l’Articulation, Mouvement. Ainsi, les sourds traiteraient les gestes comme des paramittres independants, specifiques a 1’A. S. L., et parfaitement arbitraires en termes de signification.
2
The effects of ventriloquism on the right-side advantage for verbal material*
J0.d
MORAIS
Universite’ libre de Bruxelles
Abstract Subjects were asked to recall one of two simultaneous messages coming from loudspeakers situated either at 90” or at 45” from the median plane to the left and to the right. They were told that the messages were coming from two visible dummy loudspeakers which were also situated either at 90” or at 4.5”. Pre-stimulus cueing of the side to be recalled was given. Significant right-side advantage was obtained in the 90” real-fictitious condition, not in the other conditions. These results show that right-side advantage can be obtained with presentation over loudspeakers and unilateral recall, and dismiss a purely structural or purely cognitive view of lateral asymmetries in audition. Role of structural and cognitive factors is discussed. hidden
Introduction
Kimura (1961) has found that when different verbal stimuli are presented simultaneously to the two ears of subjects with left hemispheric dominance, the material presented to the right ear is better recalled than that presented to the left ear. The reverse effect was found for a small group of subjects with right hemispheric dominance. Subsequent work, controlling for order of report and biases of voluntary attention (Bryden, 1963; Myers, 1970; Gerber and Goldman, 1971), or using a single syllable pair for identification (Shankweiler and Studdert-Kennedy, 1967), showed that this asymmetry is a true perceptual asymmetry. * This work has been partially supported by the Belgian ‘Fonds de la Recherche fondamentale collective’ under contract number 10.152. I wish to thank Prof. Paul
Bertelson for his advice. Reprint requests should be sent to JosC Morais, Laboratoire de Psychologie expkrimentale, Av. Adolphe Buyl, 117, 1OS0 Bruxelles, Belgium. Cognition
3(2/, pp. 127 ~ 139
128
Jos+ Morais
Kimura (196 1, 1964, 1967) attributed the laterality effect favoring the ear opposite to the dominant hemisphere to functional prepotency of the contralateral over the ipsilateral auditory pathway. With monaural listening, the slightly greater effectiveness of the contralateral connections would not be great enough to pefmit a difference in performance. But when different stimuli are presented to the two ears, as in dichotic listening, the impulses arriving along the ipsilateral pathway would be partially occluded in the common neural units which, according to Rosenzweig (I 9.5 I), respond mainly to contralateral stimulation; a greater loss of information from the ipsilateral path would give rise to a greater number of recognition errors for the message presented in the ipsilateral ear. Suppression of the ipsilateral input by the contralateral input in dichotic listening has been demonstrated by work on split-brain patients (Sparks and Geschwind, 1968; Milner, Taylor and Sperry, 1968). Sounds entering the ear ipsilateral to the dominant hemisphere reach the processing centers through the minor hemisphere and the Corvus callosum, therefore, with greater loss of information, weaker or later than sounds which simultaneously enter the ear more directly connected with the dominant temporal lobe. Thompson, Samson, Cullen and Hughes (1973) have suggested a model which considers (1) the two ears as independent channels, (2) some form of ‘gating out’ of the information transmitted via the ipsilateral pathway and (3) the addition of noise to the left ear channel information as a consequence of the neural processing necessary to accomplish interhemispheric relay. The interpretation et al. and others for the ear put forward by Kimura, Sparks et al., Thompson differences is based on structural properties of the auditory system and assumes that the critical factor in obtaining a laterality effect is whether sounds do or do not stimulate the ear privileged for the kind of information to be extracted. The well-established crossed relation between ‘dominant ear’ in dichotic listening experiments and dominant hemisphere have inspired a considerable body of experimental work on speech perception with the use of the ‘right-ear advantage’ effect (Studdert-Kennedy and Shankweiler, 1970; Haggard, 197 1; Darwin, 1971). In order to get relevant and interpretable data on speech perception processes it was not indispensable to.deal with the mechanism of the dominant ear-dominant hemisphere crossed relation. This may partly explain why the model put forward by Kimura remains largely accepted, although modern developments in cognitive psychology would normally lead to eyeing it somewhat suspiciously. The right and left auditory nerves are obvious channels. But channels in the mind may not correspond exactly to channels in the brain. Simon, Small, Ziglar and Craft (1970) found that reaction times to a high- or a low-pitched
The effects of ventriloquism
129
tone were slower when the tone came in the ear opposite to the responding hand than when it came in the ear ipsilateral to it; reaction times were also slower when, by presenting the tone binaurally and manipulating the interaural phase, it seemed to come from the opposite side than when it seemed to come from the ipsilateral side. This means that the interference effect must be attributed to direction of the apparent source, not to the ear stimulated per se. Morton, Crowder and Prussin (197 1) showed the effect of a binaural suffix on the recall of an auditorily presented sequence of items to be much less than that of an ipsilateral suffix and to be comparable to the effect of a contralateral suffix. Rejection of the suffix does not involve turning off one ear but rather segregating its trace on the basis of apparent spatial origin. These results suggest that signals are labelled according to the region of auditory space from which they originate and not according to the ear stimulated. Experiments on the ‘right-ear advantage’ effect have used dichotic and (more rarely) monaural presentations, so in both cases each message has been presented to one ear and to one ear only. However a message presented to one ear alone also appears to come from the corresponding half of the external space. Right-ear advantage might thus reflect better processing of messages coming from the right as well as better neural connections of the right ear. In a first attempt to separate the effects of spatial position from those of ear of entry, Morais and Bertelson (1973) used diotic presentations of two simultaneous verbal messages over loudspeakers. Three pairs of simultaneous consonant-vowel syllables were presented on each trial, and the subject attempted total recall. With the loudspeakers situated to the left and to the right of the subject, the message from the right was found to be better reproduced. In that situation, the speech sounds reached both ears, but with natural time and intensity differences. Thus an ear-of-entry interpretation was not necessarily invalidated. If time differences could not be relevant because of larger onset asynchronies between the two signals of each pair, an effect of intensity differences by themselves and not by their role in localization became a more serious possibility. In order to clarify this point, Morais and Bertelson (1975) tested subjects with stereophonic presentations of two simultaneous messages using only time differences between stimulation of the two ears by the same message. Pre-stimulus cueing of the side to be recalled in each trial was given. A right-side advantage was obtained, showing that laterality effects can arise as a consequence of the spatial position of the speech sounds. The effect observed was, however, significantly smaller than with dichotic presentation. This discrepancy was tentatively accounted for by differences in apparent localization of the sources; in the stereophonic condition using time differences only, sounds seemed to come frequently in
130
Jo& Morais
the frontal areas of the head, whereas in the dichotic situation they were almost always located near the ears. The foregoing experiments show that Kimura’s interpretation of ear differences in dichotic listening is too simple. However, a spatial position phenomenon is not inconsistent with another form of structural interpretation if the terms contralateral and ipsilateral pathway are thought of as relating halves of auditory space to the two cerebral hemispheres instead of ears. The shift from ear to region in auditory space is supported by electrophysiological data. Rosenzweig (1954) showed that the magnitude of the electrical response at either hemisphere was greater on the side opposite to the location of a sound, this localization being achieved by time or intensity differences between two clicks delivered independently to the two ears. The further over to the side the sound was heard, the greater the difference between the activity at the two hemispheres. An interpretation of lateral asymmetries in audition which would integrate these data would consider the link between dominant side and dominant hemisphere as proceeding entirely from the constraints of the auditory system. Alternatively, one might consider a cognitive interpretation which would stress control processes of sensory stimulation by the central structures, like the one put forward by Kinsbourne (1970, 1974). This interpretation attributes asymmetries to the fact that activation of one cerebral hemisphere by tasks calling for its specialized cognitive skills directs attention to the contralateral side of space. Conversely, the hemisphere contralateral to the direction of attention is temporarily more efficient than the hemisphere ipsilateral to that direction. One way of assessing the relative importance of structural and cognitive factors may be provided by separating the effects of believed and actual lateral separation of sound sources on the lateral asymmetry phenomenon. In the experiment reported here, subjects were presented with two simultaneous sequences of three consonant-vowel syllables coming from hidden loudspeakers situated either at 90” or at 45” from the median plane to the left and to the right, They were told that the messages were coming from two visible dummy loudspeakers which were also situated either at 90” or at 45”. The four combinations of the two real and the two fictitious sources were used. Thus, in two conditions, sound origin and visual cues were separated by 45”. This angular separation was unnoticed (ventriloquism effect), although it exceeded the maximum value generally indicated for the perception of a single unitary event (Witkin, Wapner and Leventhal, 1952; Jackson, 1953). This is probably because not one but two simultaneous ‘unitary’ events were occurring in this experiment, and each was difficult to identify in the presence of the other. The subject had to recall only one side on each trial, and pre-stimulus cueing
The effects of ventriloquism
13 1
was given. This procedure is advantageous because it controls for biases of voluntary attention. Broadbent (1974) reported an experiment he conducted with M. Gregory in which two simultaneous messages were presented over loudspeakers; instructions were given for half the messages to be recalled with the right-side message first and for half in the opposite order, and no lateral asymmetry was found. As in Morais and Bertelson (1973) subjects attempted total recall in any order; one accessory aim of the present experiment was to ensure that a right-side advantage could be obtained with presentation over loudspeakers and unilateral recall. Right-side advantage was thus predicted when both real and fictitious loudspeakers were situated at 90” from the median plane. No lateral asymmetry, or a smaller right-side advantage, was expected when both real and fictitious loudspeakers were situated at 45”. Information about the relative importance of structural and cognitive factors would be provided by comparing results of the misleading and non-misleading conditions corresponding to each position of real loudspeakers. Rightside advantage when both real and fictitious loudspeakers were at 90” contrasting with no lateral asymmetry, or a smaller one, when real loudspeakers were at 90” and fictitious ones at 45” would indicate the intervention of some cognitive factor and would be inconsistent with a purely structural interpretation. On the other hand, right-side advantage when real loudspeakers were at 4.5” and fictitious ones at 90” contrasting with no lateral asymmetry, or a smaller one, when both real and fictitious loudspeakers were at 45” would indicate that cognitive factors may generate, or increase, by themselves, the lateral asymmetry phenomenon.
Method Material and experimental
situation
The tape used by Morais and Bertelson (1973) provided the material for the present experiment. On each trial three pairs of nearly simultaneous consonant-vowel syllables were presented at the rate of two pairs per second. Each train of syllables was preceded immediately by a 150 ms burst of a 1000 Hz tone, itself preceded at a 3 set interval by an announcement of the serial number of the trial. CV syllables, recorded by one male speaker, were formed by pairing each of the six stop consonants /b,d,g,p,t,k/ with each of the six vowels /a,e,i,o,u,y/. No syllable was presented twice on any trial, and the same pairing did not occur more than once in the entire test. The subject sat on an adjustable chair with his back to the experimenter
132
Josh Morais
and the tape recorder, facing 0” and looking at a small colored circle on the wall. The room was quiet though not sound-proof, and walls and ceiling were black. The subject wore a head-light with a narrow beam. He was instructed to keep the beam on the colored circle during trials. Two tables were placed obliquely and symmetrically to the sagittal plane passing through the fixation point-subject axis. Two loudspeakers (Isophon, HSB 1 S/8) were placed on each table, in such a position that one was situated at 45” from the latter axis, the other at 90”, at a 1.30 m distance from the subject at about shoulder height. These four loudspeakers were connected to a Revox A 77 tape recorder and were the real sources, two on the left and two on the right of the subject. On each side, a horizontal board was placed on top of the loudspeakers, and the whole was covered with dark cloth. The real loudspeakers were thus invisible. Two dummy loudspeakers were placed on the boards, one on the left side and one on the right side, and were visible to the subject. According to the experimental condition, the experimenter placed them at 90” or at 45” from the median plane. The experimenter sat at a table behind the subject, slightly to the left for half the subjects and slightly to the right for the other half.
Subjects Thirty-four right-handed students who reported no hearing defect were tested. After the experimental session they were questioned about their impressions concerning source location. Two of these subjects noticed at times the spatial discordance between sound origin and visual cues, and their data were discarded. Ages of the subjects whose data were kept were in the range 15-28. Six were male and twenty-six female.
Procedure Pilot testing, in which the tape was presented twice, showed that sessions should be short and without trial repetition in order to prevent awareness of the spatial discordance. Thus, in the experiment reported here, each subject participated in one session which lasted about 30 minutes during which the tape was played once for 12 practice trials and 96 experimental trials. At the beginning the subject was shown a list of all possible syllables. Trials were presented under four conditions: (1) real and fictitious loudspeakers at 90” on either side of the median plane (condition R 90-F 90); loudspeakers at 45” (condition (2) real loudspeakers at 90”, and fictitious R 90-F 45); (3) fictitious and real loudspeakers at 45” (condition R 45-F 45);
The effects of ventriloquism
133
(4) real loudspeakers at 45” and fictitious loudspeakers at 90” (condition R 45-F 90). Half the subjects received the first 48 experimental trials with fictitious loudspeakers at 90”, then the experimenter placed them at 45” for the last 48 experimental trials; the reverse order was used for the other half. For each of the F conditions, there were eight blocks of six trials. For half the subjects, the first and the fourth block (as the fifth and the eight) were presented with real loudspeakers at 90” and the second and the third block (as the sixth and the seventh) with real loudspeakers at 45” ; the reverse order was used for the other half. The practice trials were presented with fictitious loudspeakers in the position used during the first half of experimental trials; they were presented in four blocks of three trials with R conditions in the same order as in the experimental trials. The subject was told before each block of six trials which side he should listen to. Half the subjects had these instructions in the order LRLR in each group of four blocks, the other half in the order RLRL. Each subject listened to blocks of practice trials in the same order as he listened to blocks of experimental trials. The subject was instructed to write down immediately after each trial, on a response sheet, the three syllables he had heard from the side previously indicated. Instructions for ordered recall and for leaving no blanks (i.e., ‘Guess when uncertain’) were given. The subject returned to the listening position, facing 0” and looking at the fixation point, after he had written down his response and before announcement of the following trial. Subjects always received the same channel through the loudspeaker they were supposed to listen to. Real loudspeakers which were on one side for half the subjects were put on the corresponding positions of the other side for the other half.
Results The percent of errors scoring procedure, put forward by Krashen ( 1972) as the feast biased by guessing measure of relative degree of lateralization was used in this experiment. The percent of errors score expresses the error score for a particular side as a percentage of the total number of errors. The error score for one side is the number of syllables presented to that side which were not recalled correctly. The left side percent of errors score (left side errors/total errors), averaged over the 32 subjects, is shown in Table 1 for each of the four conditions. The hypothesis that left side percent of errors score is greater than 50 was tested for each condition by a one-tailed t test. Right-side advantage was clearly
134
Jose Morais
Table
1.
Distribution of subjects according to side differences in number of correct syllables, left-side percent of errors* scores and t tests on these scores
Condition
Side giving better performance (score: Number of subjects) Right
Left
None
23 17 17 17
7 15 11 13
2 0 4 2
t test (df=
Mean left-side percent of errors
_____-
_ R R R R
90-F 90-F 45-F 45-F
90 4.5 45 90
* Left-side percent
Table
2.
Condition
R R R R
90-F 90-F 45-F 45-F
90 45 45 90
3 1)
of errors =
3.50 p < 0.005 -0.40 ns. 0.80 ns. 0.92 ns.
52.91 49.39 50.60 50.92
left-side errors ~ x 100. left-side errors t right-side errors
Distribution of subjects according to side differences in number of intrusions, mean differences left-right in number of intrusions and t tests on differences between intrusions in the left and intrusions in the right ~___ Side admitting more intrusions (score: Number of subjects) Left
Right
None
19 14 12 20
6 13 13 10
7 5 7 2
Mean difference left-right in number of intrusions
t test (df= 3 1)
1.13 0.19 0.56 0.56
2.50 0.50 1.47 1.14 ____
p < 0.001 ns. ns. ns.
____
significant in condition R 90-F 90. ‘l’he means of the other three conditions were not significantly different from 50. Table 1 shows also the distribution of subjects according to the side for which they obtained the better score under each condition. Table 2 shows the difference between the number of intrusions from the right-side message in the left-side message and the number of intrusions from the left-side message in the right-side message, averaged over the 32 subjects for each of the four conditions. This difference was significant on a one-tailed t test in the condition R 90-F 90, and non-significant in the other three conditions. Table 2 also shows the distribution of subjects according to the side in which fewer intrusions were made.
The effects of ventriloquism
Table 3.
Proportion of syllables correctly recalied (%), averaged over two sides under each condition Condition R 90-F R 90-F R 45-F R 45-F
90 45 45 90
135
the
Percent of correct syllables 44.2 45.1 38.9 39.0
Questions asked previously about the structural and cognitive interpretations of lateral asymmetry were framed as two independent planned comparisons. These were performed both on the left side percent of errors scores and on the differences between the number of intrustions in the left-side message and the number of intrusions in the right-side message. For both types of scores, the mean of the R 90-F 90 condition was significantly greater than the mean of the R 90-F 45 condition on a one-tailed test (respectively, t = 1.68, df = 93, p < 0.05, and t = 3.48, df = 93, p < 0.0005); the mean of the R 45-F 90 condition was not significantly different from the mean of the R 45-F 45 condition (t = 0.15 and t = 0.00). Percent of syllables correctly recalled averaged over the two sides under each condition is shown in Table 3. Performance averaged over the two R 90 conditions was better for 27 subjects, and worse for 5 subjects, than performance averaged over the two R 45 conditions. This difference was significant beyond /I = 0.001 by a sign test. Discussion When simultaneous messages came from 90” to the left and 90” to the right without misleading the subject (who was asked to reproduce only one particular message) performance was better for the right-side message. This result replicates one previously found by Morais and Bertelson (1973); furthermore, it clearly demonstrates that such an effect is not a mere artifact of the order of report or of biases of voluntary attention. With a procedure controlling for these additional sources of variance, both stereophonic presentation through headphones (Morais and Bertelson, 1975) and presentation over loudspeakers (the present experiment) have now yielded right-side advantage. When messages came from 45” to the left and 45” to the right with dummy loudspeakers in the same positions, no significant right-side advantage was observed. This finding might be considered as supporting the view held by korais and Bertelson (1975) that the differences in the size of the effect between their different presentation conditions (in particular, between
136
Jest
Morais
dichotic presentation and stereophonic presentation using time differences only) were related to the degree of source lateralization. The novelty of the present experiment is the use of misleading conditions regarding the spatial origin of the sounds. It must be noted that performance averaged over the two sides seems to be a function of the lateral separation of real sources and to be unaffected by misleading the subjects or not. Focusing attention on a point 45” away from the real source did not impair perforof the mance when real loudspeakers were at 45” ; it only changed distribution correct responses between the sides when they were at 90”. The fact that the overall accuracy for the misleading and the corresponding non-misleading conditions were of the same magnitude warrants comparison of degrees of lateral asymmetry between these conditions. Comparison between the R 90-F 90 condition and the R 90-F 45 condition showed that right-side advantage vanished if sources were believed to be situated at 45” from the median plane rather than at 90”. This outcome was observed with two types of scores, percent of errors and number of intrusions. It argues of course against an interpretation based exclusively on structural factors. It seems impossible to explain the mechanisms of the dominant side-dominant hemisphere crossed relation only in terms of a constant feature of the auditory system, for instance, the greater activity evoked in one particular temporal lobe by sounds coming from the opposite side. On the other hand, for real sources at 45” no right-side advantage was created or increased by putting the dummy loudspeakers at 90” from the median plane. This fact is not consistent with a purely cognitive interpretation which would only take into account apparent position. According to Kinsbourne’s model, directing attention to a target at 90” to the right should increase the activity of the left hemisphere, whereas directing attention to a target at 90” to the left should decrease it to the same extent. The hypothesis that orientation of attention towards one side may determine differences in level of activity between the hemispheres and in that way give rise to lateral asymmetry in performance is not supported by the present results. In fact, the condition with both fictitious and real sources at 90” was the only one to give a clear right-srde advantage. Moreover, when real loudspeakers were at 45”, relative performance for a particular side estimated either on the basis of percent of errors or on the basis of intrusions was manifestly independent of position of the fictitious loudspeakers. It seems, therefore, that the false belief of the listener about the spatial origin of the messages and the corresponding orientation of his voluntary attention may annul but not create a lateral asymmetry in performance. Some auditory correlate of 90” sources as well as belief in this direction must be present in order. to obtain a right-side advantage effect.
The effects o,f ventriloquism
131
One aspect of the auditory correlates of lateralized signals is probably a stronger projection on the left-dominant hemisphere of signals from 90” to the right than of signals from 90” to the left. This neural feature could be potentially able to raise lateral differences in performance. However, even if the processing of lateralized signals is constrained by the relative strength of their left-hemisphere projection, particularly when they are delivered simultaneously, the present work shows that the effects of neural constraints are not transferred directly in performance but undergo the action of cognitive control processes. On the other hand, the auditory correlate of signals which is necessary for the occurrence of a laterality effect does not need to be a feature privileging by itself one particular side. This feature might be anything else but the presence or absence of a neural representation for sounds with a particular spatial origin. Let us consider one of Kinsbourne’s hypotheses, i.e., that relative degree of activation of the two hemispheres determines shifts of covert attention in the lateral plane. We may adapt it to the present results by supposing that, following left-hemisphere activation, more attention capacity is automatically distributed to a limited region in the right side of the subject; this attention imbalance would be a potential one, and before determining its effect, capacity should be spatially allocated by an effort of voluntary attention which depends also on information derived from the other senses and from expectancy. By directing attention away from the preferred region as a consequence of a false belief, the capacity supplement would be lost or shared out in some way between the two signals. Conversely, if by a false belief attention is allocated to the preferred region without a cortical correlate of stimulation in that region, no supplement of capacity would be available for the processing of the signal actually presented. At this point in the discussion, it is interesting to reconsider the classical structural model of laterality effects. We have undertaken two stages of its reductionist approach. The first was to deduce a psychological phenomenon from the functional organization of a neurological system. The second was to delete the real auditory world in which this functional organization evolved. Disposing of electrophysiological data on the microstructure of the auditory system, it was tempting to account for lateral asymmetries in audition in terms of numbers of neural units responding to stimulation in each ear. As Putnam (1973) pointed out, the way by which psychology - a ‘higher-level’ science - is reduced to the laws of its ‘lower-level’ science is the theory of the brain and nervous system. The process of reductive analysis has been rewound in the case of the ‘ear differences’ phenomenon by showing, first, that we privilege positions in space, not the laboratory right ear, and second, that structural factors are an element of the explanation, not the explanation.
138
Jo.@ Morais
REFERENCES Broadbent, D. E. (1974) Division of function and integration of behavior. In F. 0. Schmitt and F. G. Worden (Ed%), The neurosciences. ‘Third Study Program. Massachusetts, M.I.T. Press. Pp. 3 l-41. Bryden, M. P. (1963) Ear preference in auditory perception. J. exp. Psychol., 65, 103-105. Darwin, C. J. (1971) Ear differences in the recall of fricatives and vowels. Q. J. exp. Psychol., 23, 46-62. Gerber, S. E., and Goldman, P. (1971) Ear preference for dichotically presented verbal stimuli as a function of report strategies. J. acoust. Sot. Amer., 49, 1163m 1168.
Haggard, M. P. (1971) Encoding and the REA for speech signals. Q. J. exp. Psychol., 23, 34-45. Jackson, C. V. (1953) Visual factors in auditory localization. Q. J. exp. Psycho/., 5, 52-65. Kimura, D. (196 1) Cerebral dominance and the perception of verbal stimuli. Can. J. Psychol., 15, 166-171. (1964) Left-right differences in the perception of melodies. Q. J. exp. Psychol., 16, 355-358. (1967) Functional asymmetry of the brain in dichotic listening. Cortex, 3, 163-178. Kinsbourne, M. (1970) The cerebral basis of lateral asymmetries in perception. In A. F. Sanders (Ed.), Attention and performance, Vol. 3. Amsterdam, North Holland. Pp. 193-201. (in press) The mechanism of hemispheric control of the lateral gradient of attention. In Attention and performance, Vol. 5. Krashen, S. (1972) Language and the left hemisphere, Working papers in phonetics 24, October. Los Angeles, University of California. Mimer, B., Taylor, L., and Sperry, R. W.
(1968) Lateralized dichotically-presented commissural section 161,
suppression of digits after in man. Science,
184-185.
Morais, J., and Bertelson, P. (1973) Laterality effects in diotic listening. Pert., 2, 107~-111. and Bertelson, P. (1975) Spatial position versus ear-of-entry as determinant of the auditory laterality effects: A stereophonic test. Unpublished manuscript. Morton, J., Crowder, R. G., and Prussin, H. A. (1971) Experiments with the stimulus suffix effect. J. exp. Psychol. Mono., 91, 169-190. Myers, T. F. (1970) Asymmetry and attention in phonic decoding. In A. F. Sanders (Ed.), Attention and performance, Vol. 3. Amsterdam, North Holland. Pp. 158- 177. Putnam, H. (1973) Reductionism and the nature of psychology. Cog., 2, 131-146. Rosenzweig, M. R. ( 19 5 1) Representations of the two ears at the auditory cortex. Amer. J. Psychol., 167, 147158. (1954) Cortical correlates of auditory localization and of related perceptual phenomena. J. camp. physiol. Psychol.,
47, 269-276.
Shankweiler, D., and Studdert-Kennedy, M. (1967) Identification of consonants and vowels presented to the left and right ears. Q. J. exp. Ps_Ychol., 19, 59-63. Simon, J. R., Small, A. M., Ziglar, R. A., and Craft, J. L. (1970) Response interference in an information processing task: Sensory versus perceptual factors. J. exp. Psychol., 85, 311-314.
Sparks, R., and Geschwind, N. (1968) Dichotic listening in man after section of neocortical commissures. Cor-
The effects
tex, 4, 3-16. Studdert-Kennedy, M., and Shankweiler, D. (1970) Hemispheric specialization for speech perception. J. acoust. Sot. Amer., 48, 579-594. Thompson, C. L., Samson, D. S., Cullen, J. K., and Hughes, L. F. (1973) The effect of varied bandwidth, signal-tonoise ratio, and intensity on the perception of consonant-vowels in a
of ventriloquism
139
dichotic context: Additivity of central processing. Paper delivered at 86th Meeting, Acoustical Society of America, Los Angeles, OctoberNovember. Witkin, H. A., Wapner, S., and Leventhal, T. (1952) Sound localization with conflicting visual and auditory cues. J. exp. Psychol.,
43, 58-67.
RCsumt Les sujets devaient rappeler un de deux messages provenant de haut-parleurs caches qui Btaient situ&s a gauche et a droite du plan median, soit a 90 degres, soit a 45 degres. I1 leur etait dit que les messages provenaient de deux faux haut-parleurs, visibles et sit& aussi soit a 90 degres, soit a 45 degres. L’indication du cot& a rappeler Btait donnee avant la presentation. Une superiorit& du cot8 droit significative a et& obtenue dans la condition avec haut-par-
leurs reels et fictifs a 90 degres, mais pas dam les autres conditions. Ces resultats montrent qu’on peut obtenir une superiorite du coti: droit avec presentation P travers des haut-parleurs et rappel unilateral, et permettent de refuter une interpretation purement structurelle ou purement cognitive des asymmetries laterales dans l’audition. Le role des facteurs structuraux et cognitifs est disc&e.
3
Dual processes
in reasoning?*
P. C. WASON University
College London
J. ST. 6. T. EVANS Plymouth
Polytechnic
Abstract Previous results have shown that the introduction of negatives into the sentences used in a deductive problem affected behavior in a systematic way which was independant of the logical structure of the problem. In the present investigation, the subjects were asked to justify their responses when reasoning about such sentences. In accordance with previous results, the responses were dominated by the terms in the sentences regardless of whether they were negated. However, the justifications did vary when negatives were introduced in accordance with the logical consequences of the responses. The interpretation of these justifications as causes of behavior seemed implausible. It was suggested that they were rationalizations, or that there was at least some form of dual processing between behavior and conscious thought.
The aim of this investigation is to elucidate the character of the reasons given for attempted solutions to the selection task, or four-card problem (Wason, 1966, 1968a). The problem is now fairly well known. In essence it consists in establishing the truth value of a conditional sentence, e.g., ‘If a card has a vowel on one side, then it has an even number on the other side’, by selecting for inspection the necessary and sufficient cards from a set consisting of a vowel, a consonant, an even number and an odd number (under the restriction that each has a letter on one side and a number on the other side). The solution is to select the vowel and the odd number because only these two
* We thank I. Keddie for conducting the experiment, and Dr. Neal Smalley for invaluable criticism. First author’s address:
Psycholinguistics Research Unit, Woltson House, 4, Stephenson Way, London N.W. 1 2HE. Cognition
3(2). pp. 141 ~- 1.54
P. C. Wason and J. St. B. T Evans
values on the same card could falsify the test sentence; the even number could merely verify it vacuously. But previous studies (with abstract material) have demonstrated that the most frequent erroneous solutions conSist in selecting the values mentioned in the test sentence. The problem is recalcitrant to correction (Wason 1969a), and its difficulty is not restricted to test sentences with the surface structure of a conditional; in a recent version (Wason and Golding, 1974) the sentence is an assertion e.g., ‘a letter is above each number’. The reader might well expect some justification for yet another paper on the problem. The issues suggested by previous research still remain puzzling and have only recently been investigated systematically. For instance, Bracewell (1974) has argued that the subjects’ remarks, when confronted by subsequent contradictions to their initial choice, (e.g., ‘That is doing it the other way round’) are ‘correct’ inferences based on inadequate premises: The subject is supposed to lack access to all the implicit features of the problem. Our view is that such utterances are rationalizations determined by erroneous selections. The issue is of some generality, and the present study shifts the argument to the character of elicited reasons which accompany the choice of cards. The problem has also been turned to other purposes, e.g., a critique of formal operations (Wason, 1975), the logic of natural language (Van Duyne, 1974) and interhemispheric differences in reasoning (Gelding, Reich and Wason, 1974). It has, in fact, become a tool for investigating a variety of theoretical issues rather than an object of study in its own right. An information-processing model (Johnson-Laird and Wason, 1970) has been postulated as an explanation of performance. A subject’s degree of insight into the problem is defined as a function of his appreciation that falsification of the test sentence is more relevant to the solution than its verification. Goodwin and Wason (1972) found a correlation between the postulated degrees of insight (defined by choice of cards) and reasons given for the choice. Evans and Lynch (1973), however, adduced evidence that error is due not to a tendency to verify but to a tendency to ‘match’ the mentioned cards with the actual cards. When the consequent of the.conditional was negated, they observed a highly significant tendency for a subject to be correct for apparently spurious reasons. For example, given the sentence, ‘if there is a W on one side of a card, then there is not a 9 on the other side’, the W and the 9 tended to be (correctly) selected. A number which is not 9 available, as (say) 8, is (correctly) rejected. This result might seem to indicate that negating the consequent allows the subjects to be ‘right’ because they simply ignore the ‘not’. The phenomenon, however, is more general. When the antecedent was negated, they observed a highly significant tendency for the value mentioned
ha1
processes
in reasoning?
143
to be (wrongly) selected. Given the sentence, ‘If there is not a T on one side of a card, then there is a 5 on the other side’, the T and the 5 tended to be selected although the correct solution would be to select the letter other than T and the number other than 5. Evans (1972b) argues that matching bias is a hypothesis based on a statistical regularity and that it il,teracts with the logical processes demanded by the problem. Its lack of theoretical status has been challenged by Van Duyne (1973). But whatever its interpretation, the results obtained by Evans and Lynch (1973) cast some doubt on the Johnson-Laird and Wason (1970) model. Verification and matching coincide (under affirmative sentences), but matching is evidently more fundamental because it appears indifferent to the presence of a negative. The thought processes elicited by this problem seem to be more primitive than had been supposed. Evans and Lynch did not, however, ask their subjects to give reasons for their solutions. The present study aims to see how solutions are related to reasons when the consequent of the conditional is negated. On this particular rule matching bias should induce the logically correct solution, but obviously for the wrong reasons. It is a matter of considerable interest to see whether the subjects’ justifications still reflect the ‘complete insight’ observed by Goodwin and Wason (1972) to accompany the correct solutions.
Experiment Performance under a test sentence with an unnegated consequent was compared with that under a negated consequent. The sentences had the following general form: (1) ‘If there is a [specified letter] on one side of a card, then there will be a [specified number] on the other side.’ (2) ‘If there is a [specified letter] on one side of a card, then there will not be a [specified number] on the other side.’ The four cards which accompany each test sentence have the following general form: The specified letter (L+); the specified number (N+); an unspecified letter (L-); and an unspecified number (N-). Under (1) the solution is (L+) and (N-); under (2) it is (L+) and (N+). It was predicted that values of (L+) and (N+) would be selected under both (1) and (2), the solutions being nominally wrong under (1) and nominally right under (2). Reasons for either selecting or rejecting each card were written down by the subjects, but no predictions were made about their content. A related group design was used. Twenty-four unpaid volunteer students of
144
P. C. Wason and J. St. B. T Evans
the City of London Polytechnic, who had no previous experience with the problem, were assigned alternately to one of two groups and tested individually. One group first performed the ‘affirmative task’ and then the ‘negative task’. The other group performed the tasks in the opposite order. The subjects were first acquainted with the fact that all the cards had a letter on one side and a number on the other side by inspection of an ‘example set’ of eight cards which were not actually used in the experiment. Each test sentence was presented with four cards displaying (respectively) the two specified and two unspecified symbols. For each subject all the symbols were different on the two tests, and the order of presenting the cards from left to right was random. After the subjects had examined the ‘example sets’ of cards, the following instructions were read by them, and they were then Finally, the salient points were asked whether they had any questions. repeated by the experimenter to clarify the procedure. ‘You will be given two sheets of paper one at a time. At the top of the sheet is stated a simple rule connecting the combinations of letters and numbers which are written on the cards. The rule only applies to the cards in front of you and it may be true or false. You must decide which card (or cards) need definitely to be turned over in order to establish whether the rule is true or false. ‘Below the cards, on the left-hand side of the sheet, the identification numbers of the cards are arranged in columns. When you have decided which cards must be turned over, place a “Yes” or “No” in the column next to the appropriate identification number: “Yes” - for turning the card over “No” - for leaving the card alone I also want you to write down your reasons for choosing to examine or to ignore each card in the third column. The problem is not so easy as it looks, so please think carefully before giving your answer.’ No time limits were imposed, and the subjects were not allowed to turn over any of the cards.
Quantitative
results
The frequency of different selections is presented in Table 1. It will be noted that there is such strong support for the prediction that statistical evaluation is unnecessary. Fifteen out of the 24 subjects were correct in the negative task compared with none in the affirmative task. Furthermore, exactly half the subjects made the matching response (L+) and (N+) in the affirmative task,
Dual processes in reasoning?
145
out of a possible 15 responses. It will also be noted that there is no apparent order effect. However, there were 8 out of 24 atypical responses (‘others’) defined as selections which either omit (L+) or include (L-), in the affirmative task, compared with three in the negative task (of these 11 cases, four consisted in the selection of all four cards). Previous research (Johnson-Laird and Wason, 1970) suggests that atypical responses are probably indicative of either misunderstanding or guessing. No reason is apparent for the relatively greater number of such responses in the affirmative task. Table
1.
Frequency
of responses in affirmative and negative tasks Affirmative
Order of permormance/ values selcted (L+) (N+) @+I &+I W-1 others N
Negative
1st
2nd
Total
1st
2nd
Total
12 4 o* 8
9 2 0 1
6 4 0 2
15* 6 0 3
12
12
6
6
2 0 4
2 0 4
12
12
24
24
* = Correct response.
Qualitative
results
A small minority of protocols were so idiosyncratic that no consistent thought process could be inferred from them. An attempt was first made to classify the remainder with respect to selected cards into the following categories: (a) Verification, (b) falsification and (c) matching. These exemplars illustrate the criteria for assigning a protocol to each category. (a) Verification. ‘Should be J on the other side if rule holds.’ ‘If there is not a 3 on the other side the rule is proved. ’ ‘To verify the rule as to there being a is B on the opposite side. ’ ‘If there is an A on the other side the statement true. ’ (b) Falsification. ‘If there is a 6 on the other side, then the statement is false.’ ‘To see if there is a Y on the other side to disprove the statement.’ ‘If rule is true, then there will not be a 2. If rule is false there will be.’ ‘If this card is overturned and there is an F the statement is false.’ (c) Matching. ‘Uppermost symbols may correspond with those on the back.’ ‘It would be possible for there to be C on the other side.’ ‘If it has a 4
146
P. C. Wason and J. St. B. T. Evans
overleaf then Q and 4 are associated. ’ ‘The rule only says that 7 is related to J.’ Some difficulties were experienced in attempting to distinguish between verification and matching. For example, ‘To find out if there is a 4 on the other side’ could be interpreted as a weak statement of verification. Similarly, ‘To check the truth or falsity of the statement’ seems like a simple compliance with the instruction. The source of the difficulty is that terms indicative of verification are linguistically unmarked. This difficulty does not arise with the falsification category because its terms are linguistically marked, e.g., ‘untrue’, ‘disproof’ or cognate expressions. Hence a more stringent decision was adopted: Each protocol was classified as either ‘falsification’ or ‘not falsification’. In the affirmative task four out of the 24 protocols were classified as ‘falsification’. In the negative task 1 1 out of 24 were classified as falsification, nine of which were associated with correct solutions. Of these nine, eight occurred in the group who performed the negative task first. This significant order effect (p < 0.05, Fisher exact test) may have been due to the fact that the subjects were not allowed to turn over the cards. This may have induced frustration and hence diminished interest for the second task although there was no evidence for this supposition. The remaining two subjects, who yielded falsification protocols but incorrect solutions in the negative task, selected in each case just (L+), one in the first performance and the other in the second performance. These cases reveal only an ‘indirect falsification effect’, i.e., a reason in terms of falsification to justify the selection of the (unnegated) antecedent when the consequent is negated, e.g., ‘Easiest thing to do to prove the rule false is to turn T over to find out whether it is in fact also 5.’ This is distinguished from the a ‘direct falsification effect’, i.e., a similar type of reason to justify selection of the (negated) consequent. Of the four subjects who yielded falsification protocols in the affirmative task, three did so when that task was performed second, and all four yielded fasification protocols in the negative task. The protocols of three subjects show how the reasons given for the choice of the mentioned cards are quite different in the affirmative and negative tasks when both make exactly the same matching responses. In order to ease the readers’ comprehension, the affirmative and negative sentences have been corrected so that they are lexically equivalent. It will be noted that all three cases reveal both a ‘direct’ and an ‘indirect’ falsification effect. In addition, the reasons given for not selecting the unmentioned cards reveal a process of ‘mismatch’.
Dual processes in reasoning?
Cards and responses
Negative
Affirmative
task
If there is a B .
G2.
S3
B
yes
task
there will not be a 3
If there is a B there will be a 3.
Reasons
Reasons
‘If the rule was false, would be a 3 on the side. If true there not be a 3. B and 3 should be of the as part assumption.’
then
there other would taken same
147
then
‘The rule only says that B is related to 3. It does not say anything about there being a logical sequence of letters to numbers, so no assumptions about letters and numbers other than B and 3 can be made.’
3
yes
‘If the rule was false, there would be a B on the other side.’
‘As above.’
U
no
‘The rule only states that there is no relation between 3 and B. It does not state whether there is a relation between other numbers and letters.’
‘Logical argument
6
no
‘As above.’
‘As above.’
(32,
s2
B
yes
‘To see that
3
yes
‘To ensure B.’
U
no
‘It need thing.’
extension above.’
of
it is not a 3.’
‘To ensure that the reverse is 3.’
that it is not a
‘To ensure that the reverse is B.’
not
prove
any-
‘The result inconclusive.’
might
be
148
P. C. Wason and J. St. B. T. Evans
6
ll0
G2,
S6
B
yes
3
‘It need thing.’
not
prove
any-
‘The result inconclusive.’
might
be
‘If there is a 3 on the other side, then the statement is false.’
‘If there is a 3 on the other side then the statement is true.’
‘If there is a B on the other side then the statement is false.’
‘If there is a B on the other side, then the statement is true: otherwise it is false’.
U
n0
‘Whatever number is on the other side will not show if statement is true or false.’
‘Any number may be on the other side.’
6
no
‘Any letter may be on the other side, therefore no way of knowing if statement is true.’
‘If numbers are fairly ranom, then there may be any letter on the other side, thereby giving no indication unless the letter is B.’
It will be noted that the protocols differentiate between the tasks, but at a more basic level they are similar. They reveal what might be called a ‘secondary matching bias’: The subjects not only select the values mentioned (‘matching bias’), but justify their selections in terms of the named values on the other side of the card.
Discussion In the negative task (at least when it constitutes the first test) the justifications for choice strongly suggest logically appropriate thought processes. If such reasons had been associated with the unnegated sentences used in previous studies, they would have been interpreted as indicating ‘complete insight’. There are three hypotheses which attempt to reconcile matching responses with qualitative observations incompatible with such matching: The behaviorist hypothesis, the rational hypothesis and the dual process hypothe-
Dual processes in reasoning?
149
sis. (These terms are intended merely as mnemonic labels which have only an approximate reference to the hypotheses enumerated.) The behaviorist hypothesis is that the ‘reasons’ are verbal responses which cannot be interpreted as revealing anything about the cause of behavior. This hypothesis is dubious because reasons and tasks are correlated. On the first task 8.3 percent (1 out of 12) of the affirmative protocols are classified as ‘falsification’, compared with 75 percent (9 out of 12) of the negative protocols. The rational hypothesis is that the reasons express the cause of the choices. It follows that in the negative task the appropriate reasons indicate a genuine logical insight into the structure of the problem. But it is rather implausible to suppose that a genuine insight would suddenly vanish in the subsequent affirmative task. Furthermore, when the antecedent rather than the consequent of the conditional is negated (Evans and Lynch, 1973), the subjects tended to choose the negated value which, in this case, is logically inappropriate. If the effect of a negative is assumed to be the same on both the antecedent and consequent of a conditional sentence, it follows that the negated consequent in the present study does not seem to confer genuine insight into the logic of the conditional. And it follows from this that the rational hypothesis is untenable. It is, however, to be regretted that we did not include test sentences with a negated antecedent in the present study. The dual process hypothesis postulates that performance and introspection reflect different underlying processes. It makes two fundamental assumptions: ( 1) The processes underlying the reasoning performance, e.g., matching bias, are not generally available for introspective report. (2) Introspective accounts of performance reflect a tendency for the subject to construct a justification for his own behavior consistent with his knowledge of the situation. The process specified in (2) might be termed ‘rationalization’ although we do not wish to emphasize the motivational aspects of the Freudian concept. More directly relevant is the work of social psychologists who have shown a need for individuals to maintain consistency between their beliefs (Zajonc, 1968). There are several consequences of the two assumptions. It follows from (1) that introspective protocols will be of little assistance in the construction and validation of theories concerned with predicting behavior. It follows from (1) and (2) that identification of processes underlying behavior will not be much help in predicting the nature of associated introspective protocols. The latter will depend on what the subject perceives to be the requirements of the situation, in the light of which he will interpret his own behavior. Thus the predominance of verification and falsification among the explicitly stated
150
P. C. Wason and J. St. B. T. Evans
protocols of the present experiment arises from the fact that the subject was instructed to test the truth value of an affirmative and negative sentence (respectively). He tends to interpret his choice in terms appropriate to these contexts. It also follows from the two assumptions that, instead of introspection revealing the causes of behavior, behavior is seen as one of the principal causes of the introspection. This reversal of the common sense direction of causality is reminiscent of the James-Lange theory of emotion. But in what way does the presence of a negative elicit a justification expressed predominantly in terms of falsification? One answer would be that a negative induces the idea of falsification with respect to the terms which it negates. It does appear that the matching response is susceptible to an interaction between a specified truth value and the syntactical form of the conditional. Evans (1972a) instructed subjects to construct verifying and falsifying instances of conditional sentences in which the components were systematically negated. On the first trial, under a verifying instruction, the matching response was made to the negated antecedent in only two cases out of 48; but under a falsifying instruction, it was made in 27 cases out of 48. In other words, given a sentence in the form, ‘if not (L+) then (N+)‘, a verifying instance tended to be constructed (correctly) as (L-) and (N+), but a falsifying instance constructed (incorrectly) as (L+) and (N+) rather than the correct (L-) and (N-). This account illuminates the ‘direct falsification effect’, but it leaves unexplained the ‘indirect falsification effect’ ~ the selection of the unnegated antecedent terms with appropriate falsifying reasons. But perhaps the explanation is not so deep: Falsification arises because it fits both the responses and the context of the task. In its strongest form the dual process hypothesis, that response determines conscious thought, may be an oversimplification. A weaker (but more plausible) assumption is that there is a dialectical relation between them: A process of rapid continuous feedback between tendencies to respond and consciousness rather than two temporally distinct phases. This ‘interpretive’ view of reasoning in the selection task has been well expressed by Smalley (1974): ‘ . . . reasoning is not the orderly, linear process we have sometimes imagined it to be. The kind of reasoning typified by orderly syllogistic deduction from premises to conclusion [should be] replaced by a more disorderly process in which extracted features are organized into an interpretation . . . the process is disorderly in the sense that a shifting back and forth between various pieces of information may be necessary to come to an interpretation.’ The hypothesis is formulated in its strong form so that it can be tested. It is well known that the processing of a negative normally imposes more cognitive load than the processing of an affirmative and consumes more time. Hence the dual process hypothesis and the rational hypothesis generate different predic-
Dual processes in reasoning?
1.51
tions. According to the former, the cognitive load would be imposed primarily in finding the ‘reasons’ (justifications); according to the latter it would be imposed primarily in determining the response. Hence the dual process hypothesis would predict that the ratio of ‘justification time’ to ‘selection time’ would be less for affirmative than negative sentences. The rational hypothesis would predict the converse relation. However, Smalley (personal communication) has pointed out that the semantic representation of negatives may be difficult to express as sentences regardless of which hypothesis is correct. He proposes the following simpler test. The rational hypothesis would predict that affirmatives will be faster than negatives in both ‘selection time’ and ‘justification time’. If the prediction is not confirmed, then the alternative dual process hypothesis is corroborated. This test can be made more economical and decisive by just considering ‘selection time’. Corroborative evidence for the dual process hypothesis is to be found in previous results on the four-card problem (Wason and Johnson-Laird, 1972). These experiments suggested that subjects tended to engage in highly distinctive rationalizations to preserve an initial erroneous solution in the face of contradictory evidence. For example, they may insist that the cards are ‘irreversible’ or acknowledge that a card falsifies the rule but then deny its relevance. Such phenomena seem to’reveal distinct thought processes which fail to interact so that conflict may remain unresolved. One experiment (Wason, 1969b) is particularly vulnerable to reinterpretation in terms of dual processing. The solution to the problem was presented, and the task was to give reasons why it was correct. All the subjects accomplished this, and the inference was that they had been prevented from imposing their own structure on the task. But an equally plausible interpretation is that the reasons did not reflect insight but were simply constructed to fit the solution. The rival interpretations could be tested by presenting erroneous solutions as ‘correct’. The dual process hypothesis would predict that ‘reasons’ would be found to satisfy the purported correctness of any common wrong solution. It also follows that the verbal protocols (Goodwin and Wason, 1972) collected to corroborate the Johnson-Laird and Wason (1970) model would not be interpreted as independent evidence for the stages postulated in the model. The protocols reflect the choices which define the stages. Circumstantial evidence for the hypothesis may be found in an inductive reasoning problem (Wason, 1960, 1968b). The task was to discover a rule by generating triads of numbers (with feedback about whether they conformed to the rule). It was quite frequently found that the same hypothesis about the rule would be reformulated without awareness after the first formulation had been pronounced wrong. For example, one subject announced this rule on the
152
P. C. Wason and J. St. B. T. Evans
basis of confirming evidence: ‘The rule is to start with a basic number, then double it and thirdly multiply it by three.’ On being told it was wrong, five more confirming triads were generated, and this rule was announced: ‘The rule is that the second number is double the first and two-thirds of the third.’ The hypothesis, x 2x 3x, (where x = any whole number) continues to exert itself unconsciously but allows a conscious displacement to fulfill the requirements of the task. The dual process hypothesis, however, assumes that an individual’s rationalizations may be wholly appropriate when the problem lies within his competence or experience. For example, when the problem is presented in a realistic guise (e.g., Johnson-Laird, Legrenzi and Sonino Legrenzi, 1972), it is not at all difficult. The familiar content obviates the matching response because the structure of the problem is readily apparent. Indeed such a response seems to be elicited by the perplexity aroused by an abstract conditional because when a simpler logical connective (e.g., the disjunctive) is substituted, there is little evidence of matching (Wason and Johnson-Laird, 1969; Van Duyne, 1973). The matching response may represent in miniature form a feature of intuitive thought which is so characteristic of judgments in unclear situations. In this sense it may be analogous to the ‘common-element fallacy’ in disjunctive concept attainment (Bruner, Goodnow and Austin, 1956, p. 168). In its abstract form the four-card problem is almost certainly not a satisfactory technique for investigating how the conditional is construed in a natural language, but it may be a potential technique for investigating intuitive and ill-defined thought processes. Such processes have been neglected by experimental psychologists, perhaps because they are the antithesis of rational thought. And yet they are, of course, a commonplace in scientific discovery. For example, mathematicians (e.g., PoincarC) sometimes report that the solutions to their problems occur ‘intuitively’ and that the conscious construction of the proof is worked out after the insight. Indeed, since deduction cannot generate new knowledge ana induction is philosophically discredited, some kind of intuitive process is probably the source of hypotheses. We have argued that the results of the present experiment represent a (banal) example of such processes. The fact that a trivial alteration in the task predisposes the subjects to be nominally right or wrong is beside the point, and the use of conditional sentence structure to elicit the phenomena is wholly fortuitous.
Dual processes
in reasoning?
153
REFERENCES Bracewell, R. J. (1974) Interpretation factors in the four card selection task. Paper read at the Trento conference on the selection task. Bruner, J-S., Goodnow, J. J., and Austin, G. A. (1956) A study of thinking. London, Wiley. Evans, J. ST. B. T. (1972a) Interpretation and ‘matching bias’ in a reasoning task. Q. J. exp. Psychol., 24, 193-199. (1972b) On the problem of interpreting reasoning data. Cog., 1, 373384. and Lynch, J. S. (1973) Matching bias in the selection task. Brit. J. Psychol., 64, 391-397. Golding E., Reich, S. S., and Wason P. C. (1974) Inter-hemispheric differences in problem solving. Perception., 3, 231-235. Goodwin, R. Q., and Wason, P. C. (1972) Degrees of insight. Brit. J. Psychol., 63,205-212. Johnson-Laird, P. N., and Wason, P. C. (1970) A theoretical analysis of insight into a reasoning task. Cog. Psychol., 1, 134-138. Legrenzi, P., and Sonino Legrenzi, M. (1972) Reasoning and a sense of reality. Brit. J. Psychol., 63, 395-400. Smalley, N. S. (1974) Interpreting sentences and relating them to possible instances: An information processing approach. Paper read at the Trento conference on the selection task. Van Duyne, P. C. (1973) A note on Evans’ matching response hypothesis. Cog., 2(2), 239-242.
Ptsumt Des resultats prealables ont montre que l’introduction de negatives dans les regles utilisees dans un probleme de deduction
(Ed.), New horizons in psychology I. Harmondsworth, Penguin. (1968a) Reasoning about a rule. Q. J. exp. Psychol., 20, 273-218. (1968b) ‘On the failure to eliminate hypotheses’ . . . a second look. In P. C. Wason and P. N. Johnson-Laird, (Eds.), Thinking and reasoning. Harmondsworth, Penguin. (1969a) Regression in reasoning? Brit. J. Psychol., 60, 47 l-480. (1969b) Structural simplicity and Some complexity: psychological thoughts on a novel problem. Bull. Brit. psychol. Sot., 22, 281-284. (in press) The theory of formal operations - a critique. In B. Geber (Ed.), Structure and development in general and social psychology: The contribution of Piaget’s theory. London, Routledge. and Johnson-Land, P. N. (1972) Psychology of reasoning: Structure and content. London, Batsford. and Golding E. (1974) The lanBrit. J. guage of inconsistency. Psychol., 65. 537-546. Zajonc, R. B. (1968) Cognitive theories in social psychology.. In G. Lindzey and E. Aronson (Eds.), The handbook of social psychology, Vol. 1. London Addison-Wesley. (1974) Realism and linguistic complexity in reasoning. Brit. J. Psychol., 65,59-67. Wason, P. C. (1960) On the failure to eliminate hypotheses in a conceptual task. Q. J. exp. Psychol., 12, 129-140. (1966) Reasoning. In B. M. Foss
affectait le comportement de facon systematique independante de la structure logique du probleme. Dans l’enqubte pre-
154
P. C. Wason and J. St. B. T. Evans
sente on demandait aux sujets de justifier leurs reponses en raisonnant sur ces rbgles. ConformCment aux rksultats prkalables, les rkponses Btaient dominkes par les termes dans les rkgles sans se saucier de la nkgation. Les justifications variaient nkanmoins quand on introduisait des negatives conformes aux conskquences logiques des
rkponses. L’interpretation de ces justifications comme cause du comportement paraissait peu plausible. M est suggCrC qu’elles Btaient des rationalisations oti qu’il y avait au moins quelque forme de traitement double entre comportement et pens&e consciente.
Discussion
A Reply to Brainerd SIDNEY STRAUSS Tel-Aviv
University
Having written the first of a three-part review (Strauss, 1972) of the training literature, Brainerd’s review [Cognition 2(3), 349-3701 fell on my desk and forced me to re-read many of the cited studies - a thoroughly thankless task. I reached the decision to answer his claims because I thought they might be representative of other’s objections to what I had written in my review. I must confess to having had second thoughts about the conclusions reached in my review article after a first reading of Brainerd’s critique; but after having re-read many of the articles over which we disputed and having re-examined my arguments and his, I am now unconvinced that his conclusions are preferable to mine. Before analyzing Brainerd’s comments, three short, preliminary remarks are necessary. First, Brainerd pointed out that my review was highly selective. This was because there are two other parts to my three-part review and what is missing in the first part appears in the other two. Second, Brainerd mentioned several times that studies or their conclusions were not discussed in enough detail. He is surely aware of the editor’s ubiquitous plea to cut one’s manuscript’s length. In my case, it was cut considerably. For the same reason, I will be brief in the discussion of many articles in this piece, as well. Third, for some peculiar reason, Brainerd held me responsible on several occasions for not citing or taking into account the findings of studies which were published at the same time as or after the publication of my review article. Telepathy has never been one of my strong points. We shall now begin to discuss Brainerd’s arguments in the order in which they appeared in this piece. His first claim, put in the form of a query, was ‘ . . . does the conclusion (about the organismic-developmental stage hypothesis) itself provide nontrivial evidence for the Piagetian stage hypothesis?’ The answer offered was that the evidence I provided was of the trivial variety. Brainerd’s plea for nontrivial evidence for the organismic-developmental stage hypothesis (or any other hypothesis, for that matter) sounds perfectly Cognition
3(2), pp. I55
- 185
156
Sidney Strauss
reasonable. However, an analysis of what Brainerd understands as nontrivial evidence and its uses belies both an overestimate of the contemporary state of psychology and a basic misunderstanding of scientific methodology. His distinction between trivial and nontrivial evidence centers around the uniqueness of findings for a theory. Brainerd (p. 350) argued that ‘Trivial evidence takes the form of findings (logical or empirical) that are consistent with both the hypothesis of interest and with its chief competitors. .. Nontrivial evidence, on the other hand, consists of logical or empirical findings which are more or less unique to this hypothesis’. By this account then, Chomsky’s revolutionary approach to the description of language in Syntactic structures was trivial since his original proposed model of generative grammar did not improve upon the phase structure grammar description of language. That is, as originally proposed, it could not solve problems which remained unsolvable with the phrase structure grammar description. Yet Chomsky and’othkrs understood his model to be nontrivially deeper because it provided simpler solutions to problems. Brainerd’s understanding of scientific methodology seems to be unnecessarily simplistic, limiting and, consequently, inadequate. Also, since Brainerd is not interested in the accumulation of ‘nontrivial’ evidence simply for the sake of its accumulation, one gets the very distinct impression that he is advocating that crucial experiments be conducted. On the basis of the ‘nontrivial’ findings from these experiments, one should be able to choose between theories. Otherwise, why conduct crucial experiments? There are two reasons why this is difficult to countenance. First, theories in the social sciences in general and psychology in particular are not sufficiently developed for one to produce hypotheses specific and precise enough that empirical findings could refute the hypotheses and call the theories into question. And even if we were capable of the theoretical and empirical precision of, say, the physical sciences, it is not clear that we would be better off than the physical sciences with respect to crucial experiments. Lakatos (1970) has explained in detail how crucial experiments in the physical sciences were understood to be crucial only decades and sometimes even generations after the experiment. Similarly, Schagrin (1973) observed that what appeari to be a defeat for a theory in a crucial experiment may be only a temporary setback due to lack of imagination on the part of the experimenters. This mitigates against the ‘instant rationality’ which Brainerd apparently advocates and which Lakatos reasonably rejects. It is of interest that Lakatos (1974) developed his thesis further and concluded that crucial experiments are impossible to conduct. Even if one does not accept this radical notion, Lakatos’ lucid and cogent arguments as to the alternative
A reply to Brainerd
approaches to the rejection of crucial experiments ingly cautious about advancing such approaches.
151
should make one exceed-
1. The criterion problem Brainerd pursued at some length the question of the criteria one uses to establish whether a subject has acquired concrete operational reasoning for a particular concept. I have argued that one should use a criterion which includes a correct judgement (e.g., in the case of length conservation, the subject would judge the objects to be equivalent in length), a logical justification (reversibility, addition/subtraction, transitivity, compensation, identity), and resistance to a verbal counter-suggestion. Brainerd argued that the problem can be reduced to whether one accepts a judgments-only or a judgments-plus-explanation criterion. Without going into the details of his arguments, we can isolate one of his main conclusions. That is, ‘the principal findings of this literature are not seriously affected by the choice between judgments-only and judgments-plusexplanations’. Space considerations do not allow a full discussion of his claim, and by and large I would agree with his conclusion. However, we shall see that occasionally one finds reports of studies in which children produced correct judgments yet could not provide logical justifications. Brainerd would opt for the judgments-only criterion while I would choose the judgments-plusexplanation criterion. This will be slightly elaborated upon in section 2.2.1.
2. Training experiments I had reviewed the literature in such a way that three general classes of training procedures were described: Disequilibrium, mental operations and regression. 2.1 Disequilibrium In my original article, I had noted that Piaget describes disequilibrium as a mechanism for structural transformation and elaboration. Briefly, I had distinguished between two types of equilibrium: Adaptational and organizational. The former pertains to the equilibration between environmental information and the cognitive structure. The latter refers to the equilibrated cognitive organization which comprises more than one structure. If potential contradictions from this structural mixture do not become actualized, the
158
Sidney Strauss
remains
interactions adaptational emphasis
in equilibrium.
Disequilibrium
similar in both cases is that there are between cognitive structures and environmental information. For disequilibrium the emphasis existing cognitive structure. of organization disequilibrium is on the cognitive internal contradictions. internal contradictions
category
here
was what
I referred suggests
in feedback actions. This lack of fit inducing a state of disequilibrium’ Brainerd correctly pointed
could
inherent outcome
was hypothesized
conflict
training
that a
should the
mental
negative shortcomings
result
structure, findings
procedure. conflict with Ss’ predictions distort environmental outcomes so that it would fit their cognitive structure. reasons why, as Brainerd whether or not concrete-operational concepts can be via adaptational disequilibrium treatments’. pointed out that these two shortcomings problems which are of interest methodological interest since we do not have any of conflict post-hoc of cognitive change. measure of surprise varying (Miller, Watson, 1970), and research measures seems to me to be quite important. valuable since we should be interested perhaps more important, when are they not distorted. another way of asking how information perturbs it. All of the above was not even mentioned by Brainerd conflict is not the only adaptational disequilibrium training category. Other kinds of training category are: (a) Verbal feedback, (b) conformity
A reply to Brainerd
159
(c) dimensional discrimination training. These three categories of training are described in an unpublished manuscript (Strauss, in preparation), and, for the sake of completeness, 1 shall comment on some of Brainerd’s points. The verbal feedback method is one in which Ss are told if their judgments were correct or incorrect. Studies which fall under this heading were conducted by Overbeck and Schwartz (1970), Figurelli and Keller ( 1972), Siegler and Liebert (1972), Ahr and Youniss (1970) and Brainerd (1972b). Let us begin our analysis with Overbeck and Schwartz (1970), who attempted to induce weight conservation through a verbal correction procedure. This study had two general problems. The first problem is that the training method may have been confounded on two counts. The first confounding was that in one-third of the training tasks, Ss were presented a deformation of one of two equal weight clay balls plus the addition or subtraction of clay from the non-deformed ball. This training may have made functional those mental operations which Piaget has argued are necessary for conservation. Consequently, it could be claimed that mental operation exercise may have been partially responsible for the reported cognitive advance. The second confounding is that this type of training meets the requirements of Smedslund (196 1b) as conflict-producing. That is, a potential conflict for an intuitive S exists when he judges a deformed clay object to have less clay and also judges the other to have less because some clay has been removed. Here we have two alternative explanations of the effects of the type of training which Brainerd has labeled verbal feedback. The objections to the confounded training may be partially dismissed since it was suggested (Strauss, 1972) that: (1) Mental operations training which exercises these operations does not necessarily coordinate them - a prerequisite for concrete operational reasoning, and (2) conflict training, as described above, has not been very successful in inducing structural transformation of elaboration. The second problem in this study arises when we analyze the Ss’ structural profiles. Unfortunately, Overbeck, and Schwartz assessed Ss’ stage usage for only the weight concept. We have no evidence that Ss were not transitional between weight and, say, length conservation. Despite this lack of direct evidence for such transitionality, the results suggest that this may have been the case. For example, there was an all or none tendency to conserve. This is similar tc the findings reported by Langer and Strauss (1972) that strictly intuitive Ss rarely constructed the concrete operations structure after training. In contrast, those who were intuitive for a more complex problem (length conservation) and transitional within or concrete operational for a less complex problem (discontinuous quantity conservation) were very likely to change to concrete operational reasoning for length conservation. In the Overbeck and Schwartz study it was found that older Ss were more likely to
160
be influenced by were also more
to be concrete
it is operational
to argue that these for some other an extension of Beilin’s
Ss
I had an example of multiple operations training reversibility, addition/subtraction be classified in A of the above or both in either of As such, Brainerd’s claim A study in which were provided the above test was conducted by Siegler and Liebert (1972). In this experiment, Ss were administered training in which they were shown deformations of objects and either were or were not provided corrective feedback and rules of reversibility, etc. The findings were that while more of those who were provided both rules and feedback conserved on the posttest than those who were provided only rules or only feedback, the differences were not significant. Thus, the test of the above question remains inconclusive. We shall have more to say about this later. Ahr and Youniss (1970) conducted a training study in which Ss were simply told if they were correct or incorrect when solving class inclusion problems. They wrote that the ‘training was predicated upon the possibility that a S who understood the inclusion relation would perform appropriately if he were given the opportunity to correct himself (Ahr and Younniss, 1970, p. 138). Note that their claim was not that a preoperational S would become operational due to corrective feedback training. Yet this is what Brainerd purported in his article, and he made no effort to explain this discrepancy. Finally, Brainerd (1972b) also attempted to induce concrete operational reasoning through corrective feedback, this time among Ss who were intuitive for number conservation. The S was to judge if two rows of an equal number of chips had the same number after a deformation, The two possible judgments were ‘same’ and ‘not the same’. Eventually, through training, most Ss learned to say ‘same’. One objection to this study and others of its kind is that there was no conservation of unequal number on the posttest; thus, the training might have simply provided the S with the information that he should say ‘same: in all cases without his having learned the principle underlying the tasks. In addition, there is evidence that Brainerd (1972b) was providing what he had subsequently argued was ‘the trivial and uninteresting fact that subjects who partially possess a to-be-trained concept tend to derive more benefit from training than subjects who do not possess the concept at all’ (p. 356). Some of his Ss were transitional (partially possessed the to-be-trained concept), and categorized the latter study Ss were provided rules As consequence, test of whether not
A reply to Brainerd
161
they were almost surely the stage II Ss who were more likely to conserve after training than stage I Ss. In a later section we shall present an argument against Brainerd’s point about transitional Ss; nevertheless, the point of these remarks should be clear here. The main question to be asked at this point is: Did Brainerd intend studies in the verbal feedback category to provide evidence of de nova construction of concrete operational reasoning among intuitive Ss? Given the above quote, one must conclude that the reply would be affirmative. Direct, reported evidence or inferences from Ss’ ages in the above studies do not confirm this. What can be concluded from the data is that some transitional Ss are influenced by information that their judgments are incorrect. This information could be understood as an indication to the S that of the two structures which he has applied (intuitive and concrete operational), only those judgments produced by the latter are correct. Apparently in cases where Ss were strictly intuitive, training was not effective. The second type of training which Brainerd argued could fall under the heading of adaptational disequilibrium is what he called conformity and observational learning studies. What Brainerd called conformity training is exemplified in cases where an intuitive S was placed with a conserving S and the two were presented conservation problems with the instructions that they must arrive at an agreed-upon decision about the task (Murray, 1972; Silverman and Geiringer, 1973). This training condition is one of the few in which apparently strictly intuitive Ss produced posttest conserving judgments and logical justifications. This was explicitly the case in Murray (1972) but was impossible to tell from Silverman and Geiringer (1973). The condition of a ‘here and now’ conflict of structural products in a social interaction situation was argued by Smedslund (1966) to be one which could result in structural transformation and elaboration. Strauss and Cohen (1974) found that this type of training was somewhat successful in inducing the mental operations necessary for the solution of perspective tasks (cJ: Huttenlocker and Presson, 1973). In observational learning or imitation studies, intuitive Ss observed a model during training who produced concrete operational judgments or judgments and justifications (Rosenthal and Zimmerman, 1972; Sullivan, 1967, 1969; Waghorn and Sullivan, 1970; Murray, 1974). This aspect of the literature is quite unilluminating, and Brainerd’s claim that ‘temporally durable and conceptually generalizable conservation concepts have been induced’ by this type of training is a result of an incomplete survey on his part. An example of the obfuscation in imitation studies can be illustrated by the three Sullivan studies. In his first study, Sullivan (1967) attempted to induce operational thought through a filming technique. There were four groups in
162
Sidney Strauss
the experiment: (a) Film modeling-verbal principal (VP); (b) film modeling-no verbal principal (NVP); (c) natural conservers; and (d) a control group. The difference between the VP and NVP conditions was that in the VP conditions, the intuitive observer heard a model produce a conservation judgment and justification while in the NVP condition, the model produced a conservation judgment only. The results were that significantly more VP (84%) and NVP (76%) than control Ss (10%) conserved on the posttest. In a follow-up study, Sullivan (1969) conducted two experiments. The first was designed to measure the effect of amounts of training upon inducement, generalization and extinction posttest tasks. A modification in the training was that one VP group viewed the modeling film once and a second VP group viewed it three times. Similarly, one NVP group viewed a modeling film once and a second group viewed it three times. It was reported that there were no significant differences between Ss viewing the one or three film sessions. Furthermore, contrary to the Sullivan (1967) study findings, the VP S’s in each group performed significantly better than their NVP counterparts on all three posttest phases. Sullivan’s (1969) second experiment was designed to measure the effects of: (a) The order of the three-phase posttest presentation and (b) the time interval between the training and posttests. The same procedure as in the 1967 study was followed except that the posttest presentation order was varied. The results were that corztrary to the findirlgs of the 1967 study (where there were significant differences between the NVP, VP and the control Ss) and cotltrary to the fi’ndirzgs of the first experiment of the 1969 study (where VP Ss performed better than NVP and controls), there were no significant differences between the NVP, VP and control Ss. In another study, Waghorn and Sullivan (1970) found results similar to Sullivan (1967) and Waghorn (1970) studies, one certainly gets the picture of controls (36%) conserved on the posttest. Since Brainerd quoted only the Sullivan (1967) and Waghorn (1970) studies, on certainly gets the picture of considerable consistency when, as 1 have attempted to show, a more comprehensive account of the imitation litarature on this topic indicates that more caution is needed. Space limitations prevent a comprehensive account of the literature on imitation here. But it should be mentioned that Brainerd’s label of social learning theory did not distinguish between various theories such as Bandura’s (1969), Aronfreed’s (1968), and Miller and Dollard’s (1941). In addition, no mention was made of an alternative understanding of imitation studies which radically break from the Bandura and Aronfreed traditions. The cognitive developmental approach suggests that these studies provide information to Ss about what performances are being requested of them by the experimenter
A reply to Brainerd
163
and that Ss comply with these requests within the constraints of their structural limits. For reviews of this explanation, the reader is referred to Kuhn ( 1973), Kuhn and Langer ( 1968) and Langer and Turiel(1969). Finally, Brainerd created what is probably a theoretical obfuscation by combining two types of training procedures (conformity and observational learning) which should remain separate. The first type involves active give and take in a social interaction setting. This is more in line with claims of structuralist theory adherents that S must be an active, interacting participant in the construction of novel cognitive structures. In contrast, observational learning places S in a relatively passive role which is more characteristic of the environmentalist approach of behaviorist theory. The third type of training which Brainerd thinks falls within the adaptational model is dimensional discrimination training. Peters ( 1967) provided an excellent rationale for this type of training when he noted that young children do not discriminate between quantitative and qualitative changes in conservation problems because irrelevant, qualitative cues are often more salient than the relevant, quantitative ones. For example, in a discontinuous quantity conservation task, an increase in a row’s length increases the salience of irrelevant cues (length and density) and correspondingly decreases the salience of the relevant cue (number). This suggests that S will learn appropiate conservation responses if one trains him to: (a) Attend to the relevant cues of a deformation and (b) discriminate these cues from the irrelevant but more salient ones. As Brainerd indicated, Gelman (1969) conducted a study of this sort where Ss were provided corrective feedback when they chose among cues inherent in a conservation task. It was found that with this type of training, Ss started learning immediately and reached an asymptotic learning level of approximately 95% correct responses. On the posttest, 95% and 96% of Ss in this condition were assessed to be at the concrete operations level for number and length conservation, respectively. In this study, Gelman noted that the rate of learning during the training sessions indicated that Ss were able to work with notions of quantity with considerable ease. This suggests that Ss had this ability prior to the training which, in effect, precipitated it. This quantification hypothesis has received indirect support from a study conducted by Christie and Smothergill (1970). They attempted to partially replicate Gelman’s study by administering her length training procedure to Ss who were pretested as not having conserving responses for length tasks. This study was different from Gelman’s in two significant ways: (a) No number pretests or training were administered, and (b) the children in their study were much younger (3,6-5,O). They reported that during training the mean number of correct responses was 61.5% which
164
Sidney Strauss
is quite a bit lower than in Gelman’s study. Even more significant, Christie and Smothergill reported that no Ss conserved on any of the posttest length conservation tasks. The indirect support this provides for the quantification hypothesis comes from the age differences of Ss in Gelman’s (x = 5,6) and Christie’s and Smothergill’s (x = 4,3) studies. The younger children in the latter study might not have had as advanced quantification abilities prior to the training and, consequently, were less able to discriminate these cues from the others. More telling data from Christie and Smothergill’s study would have been the percentage of correct training responses coming from Ss in the upper age range, which came close to the ages of Gelman’sSs. Also, it would have been revealing if individual Ss had scored 95% training success, as in Gelman’s study, and failed to conserve on the posttest. The reason that this evidence is only indirect is that we have no measure of quantification other than conservation or non-conservation responses. A study of a descriptive, normative nature whose purpose was to study the development of young children’s spontaneous levels of quantification might be able to provide us with the measure necessary to test this quantification hypothesis (Strauss, Hess, Evron, Feldman and Kruglanski, 1974). In sum, it appears that we have a bit of a disappointment due to the unfulfilled expectations produced by Brainerd’s enthusiastic endorsement of additional types of adaptational disequilibrium training procedures. The three additions (verbal feedback, conformity/imitation and dimensional discrimination training) did not produce the ‘temporally durable and conceptually generalizable conservation concepts’ that he claimed. The one exception to this might be conformity training where apparently intuitive Ss produced concrete operational reasoning in one of two studies in which such training procedures were administered. The dearth of cases of this type of research make for precarious conclusions; however, it is a line of investigation which should be continued. The results of the other types of training were contradictory and not unitary as we were led to believe by Brainerd’s account. 2.1.2 Organizational disequilibrium This category of training touches on the areas of oblique reciprocal assimilations and intrastructural oscillations. As I noted in my original article, few studies have addressed themselves to this type of training, yet, despite Brainerd’s criticisms, I believe that it can be a very revealing and productive area of research. Evidence of its potential productivity has been described by CellCrier (1972) who pointed out that recent studies of this type conducted by Inhelder, Sinclair and Bovet (1974) had been partly responsible for changes in Piaget’s formulation of the equilibration process. Cellerier indi-
A reply to Brainerd
165
cated that Inhelder et al.3 studies showed that children apply different schemes and concepts in the solution of a problem and that different cognitive categories evolve at different rates. Thus, there are interactions between different categories (e.g., classification, conservation, time) and different operative levels. CellCrier (1972, p. 121) argued, therefore, that ‘Piaget’s picture of development now incorporates vertical relations (intracategory filiations), horizontal ones (intercategory lateral interactions), and obEique ones (interactions between elements of different operatory levels)‘. In my article, what I referred to as transitional reasoning or structural mix is what CellCrier referred to as potential oblique interactions. The point of organization disequilibrium training procedures is to exploit this natural developmental phenomenon by creating a problem situation in which a child is using elements from different operatory levels in the solution of a problem. As a point of clarification, I shall present two examples of training strategy between different categories. The first example comes from a study conducted by Strauss and Rimalt (1974). The categories were spatial, as measured by area conservation, and logico-mathematical, as measured by discontinuous quanity conservation. Among other things, we investigate.d the effects of training for area conservation (for which all Ss were at the intuitive level) on Ss who were either intuitive or concrete operational for discontinuous quantity conservation. The training obligated the experimental condition Ss to successively alternate judgments about the number of cubes (farms) on two rectangles (pastures) and the area remaining on these pastures. That is, Ss who were at the concrete operations operatory level for discontinuous quantity conservation would have the potential for oblique interactions - or what I called organizational disequilibrium - while Ss who were intuitive for discontinuous quantity conservation would not have this potential. The results were that Ss at different pretest operatory levels who were administered the training condition were more likely to conserve on area posttest problems than (a) controls at the same operatory levels who were not administered training and (b) Ss who were assessed as intuitive for both discontinuous quantity and area conservation. The second example of a training study where oblique interactions were part of the training procedure is that conducted by Strauss and Han (in press). The Ss were intuitive or operational for the conceptual categories of space, as represented by length, and for time, as represented by speed. We investigated the interactions between nonequivalent length conservation (for which some Ss were intuitive) and speed (for which all Ss were intuitive). The main result of interest here was that for Ss who received training there was little posttest change (10%) in reasoning about the speed concept for those with no structural
166
Sidney Strauss
mix, while 55% of Ss who had structural mix changed. In addition, there was transfer to the child’s conception of movement (Piaget, 1970). In both of these studies, the attempt was to ask questions such that in one problem, S successively alternated the different operatory levels. This alternation or oscillation and the attendant contradictions which may accompany it has been argued by Inhelder (1968) and Langer (1969) to be one of the sources of progressive equilibration. Thus, these training studies were an attempt to induce in an experimental setting what has been hypothesized to occur in the natural development of a child. Now let us look at the objections Brainerd raised. Apparently he does not object to the category itself but to the studies I included within it. He argued that the behavioral definitions of transitional and preoperational are dependent upon the to-be-trained concept, and, therefore, the definitions are circular. He further argued that independent measures of Piagetian thinking structures would be tests of Piaget’s eight groupment structures. Empirical evidence which supported Piaget’s theory (Inhelder and Piaget, 1958) indicated that either a child has none of the eight or all of them. If he has none of the eight he is intuitive while if he has all of them he is concrete operational. This places us back where we were before, but instead of claiming that a child is concrete operational on the basis of his judgments and reasoning on, say, seriation problems, we now say that he has constructed a coordinated grouping structure. Brainerd’s claim notwithstanding, these statements do not differ in any important sense. The next argument offered by Brainerd was that transitional Ss are those who have a concept in competence but not in performance, while preoperational Ss do not have a concept in either competence or performance. As a consequence, ‘only performance is trained with transitional subjects but the concept itself is trained with preoperational subjects’ (p. 356). This distinction is similar to the one I drew between what 1 had called learning and development (CJ! Piaget, 1964). Learning was defined as learning to apply rules of an already constructed structure to tasks to which these rules were not applied. Development was defined as the construction of new structures. The main difference between our distinctions pertains to what is thought to be acquired by the preoperational child. Brainerd apparently believes that ‘the concept itself is trained with preoperational subjects’. If this is so, then what is the competence? I confess to be at a loss here since Brainerd did not provide us a clear picture of what he understands a concept to be, nor does he indicate its relations to competence and performance. I have written to the effect that a theory of competence is an idealized system of rules hypothesized to constitute an individual’s mental organization. As such, it also constitutes a theory of mind. A theory of performance pertains to the
A reply to Brainerd
161
operations which an individual must perform in solving tasks. It is not a direct product of competence since it has other factors which are peculiar to it and which constrain it, e.g., the subject’s memory limits and the complexity of the task. An example of the relations between a theory of competence and performance can be found in a study recently completed by Strauss, Hess, Evron, Feldman and Kruglanski (1974). The task which was studied was that of discontinuous quantity conservation. Piaget has argued that the mental organization required for task solution is the grouping structure. Strauss et al. and Strauss, Hess, Hoffman, Levy, and Ze’eri (1974) have shown that a child must go through a two-step process in producing judgments about discontinuous quantity conservation. First he must compare each row with itself from the pre-transformation to the post-transformation states. The second step requires S to compare his within-row judgments which were produced in the first step. It was found that both concrete operational and preoperational children go through the same process and that it was possible to order the difficulty of tasks accordingly on the basis of this process. Note that there is no obvious direct relation between Piaget’s competence theory and the performance theory just outlined. If this characterization is accepted, then one is forced to reject Brainerd’s conclusion that the findings of studies I cited as exploiting (not inducing, as Brainerd wrote) structural mixture are ‘just about as predictable and uninteresting as the finding that transitional subjects are more susceptible to training than preoperational subjects’ (p. 356). The reason for this is that if our investigations of cognitive development were limited to changes in underlying competence alone, as Brainerd would have it, we could examine only those moments when a child constructs a new structure. All other acquisitions are ‘uninteresting’ changes. Investigations of structural consolidation, for example, would hold little interest for us according to my understanding of Brainerd’s argument. Neither would the relations between different conceptual categories (cf. Inhelder, 1972) be an interesting topic for empirical test if underlying them were different structural performances and competencies. If I were to borrow an example from another discipline which would illustrate what I perceive to be the argument, it would come from the Kuhn-Popper debate over the nature of change in theories of science. Popper’s (1970, 1972) argument was that the methodology which produces revolutionary change in the structure of scientific theories should be the guiding methodology for scientists. On the other hand, Kuhn (1962, 1970) argued that scientific revolutions are rare events and that most of the time scientists are engaged in what he called normal science. ‘l‘his is essentially the research
168
Sidney Strauss
of ‘puzzle-solving’ that is conducted within a theoretical structure which serves to elaborate and test it. Much of the Kuhn-Popper debate revolves around the question of whether or not Kuhn’s characterization of normal science is descriptive or prescriptive. It appears to me that Brainerd is assuming an extreme form of the Popperian notion that the only important events worth describing are those in which changes in the underlying structure take place. Since, as in the history of science, these are uncommon occurrences, I have argued that we should also examine what seems to be happening during practically all of the child’s life, e.g., structural elaboration and consolidation. This was what 1 intended when writing about virtuosity.
2.2 Mental
operations
There have been two research strategies with respect to mental operations. The first has been to train for individual mental operations. It has been argued (Strauss, 1972) that preoperational children possess the action rules which are external forms of the operations of addition, subtraction and reversibility. If this is the case, we should not expect that training for what Ss already possess should result in significant changes in their forms of reasoning. The second research strategy has been to train for multiple operations. I had argued that operational coordination might ensue from this type of training. 2.2.1 Training individual operations As just mentioned, there is a basic research strategy related to mental operations which is ‘to induce mental operations that are characteristic of one stage more advanced than the child’s current level of competence. To date, all the research on this idea has been limited to training children at the intuitive stage to progress to the concrete operations stage. Consequently, the procedure has been to train Ss for one of the logical operations that normally develops at the concrete operational stage (either the operations of addition, subtraction, or reversibility)’ (Strauss, 1972, p. 340). For addition/subtraction training, Brainerd agreed with my conclusions that the studies cited do not indicate that this is a powerful training procedure. He then cited several articles from Japanese journals whose findings, he contended, can be interpreted as supporting the addition/subtraction training procedure in terms of its effectivity (Hatano and Ito, 1966; Hatano and Suga, 1969, 1972). Of these empirical 1969; Inagaki, 1970; Inagaki and Hatano, studies, only the research by Hatano and Suga (1969) is in English. The other studies have English abstracts which are insufficiently detailed to allow an independent assessment of their training procedures. In the Hatano and Suga
A reply to Brainerd
169
(1969) study, variants of Smedslund’s (196 1a) conflict training procedure were employed. The training procedures were confounded and, as such, do not constitute a good test of the addition/subtraction training strategy. Similarly, Brainerd’s claim that ‘conservation concepts have been induced via the revised (above mentioned) procedure’ is unfounded in the case of Hatano and Suga’s (1969) research. They found that most of the Ss who appeared to conserve on the posttest responded incorrectly on a nonequivalence task. That is, these Ss claimed that two sets of objects were equivalent when they were not. This finding suggests that the Ss learned no more than a short-cut rule during training. In the same vein, it was found that 86% of the Ss who were apparently posttest conservers gave up their conservation judgments after the experimenter provided a verbal counter-suggestion. A dual interpretation can be made from Brainerd’s subsequent caution that ‘there are insufficient grounds at this time for the conclusion that addition/ subtraction effects are “partial or negligible” ‘. The first interpretation is that it is wise to withhold judgment until more information is available. This is an approach which is not necessarily peculiar to the case of addition/subtraction training and appears more pious then judicious. The second and more likely interpretation is that the evidence he has provided necessitates a second look at the conclusions I reached (and with which he agreed) from experiments conducted in Western societies. The experiment described above does not confirm his doubt. If Brainerd reads Japanese, he would perform a great service in translating the other studies he cited as providing evidence for the effectiveness of this training strategy. If he neither reads Japanese nor received a translation of the above articles, I can only comment that perusing abstracts makes for easy reading. For reversibility training, Brainerd pitted my conclusions that this type of training has not yet been shown to be productive against the conclusions of others (Beilin, 197 1; Brainerd and Allen, 197 1; Glaser and Resnick, 1972) that it has been a quite successful training procedure. Since 1 did not fall into line with ‘what others regard as a reasonably safe conclusion’, Brainerd sought to analyze the reasons for this. First, wrote Brainerd, ‘Strauss invokes a variation on the necessary-but-notsufficient objection . . .‘, i.e., reversibility is a necessary but not sufficient condition for conservation. Brainerd then attempted to show that Piagetian theory ‘amounts to a conditional of the form: If conservation, then reversibility’ and that my objection contains a converse conditional fallacy. Brainerd is mistaken about Piagetian theory, and the conditional should be changed into a bi-conditional. That is, if conservation, then reversibility and if reversibility, then conservation. The reason for this mutual implication is that Piaget argues that operations appear together (not singly) and that their appearance is
170
Sidney Strauss
signaled by conceptual products such as conservation. Thus, conservation without reversibility and reversibility without the other operations which lead to conservation are contrary to Piaget’s theory. What I had attempted to argue in my article was that part of the problem of reversibility and other operational training is the ambiguity in Piaget’s theory about the link between the physical action of renversibilitk and its mental operation equivalent - reversibilitt. Children who cannot conserve argue that if the deformation action is reversed (e.g., tl e spread-out row is returned to its original configuration) the number of objects will be equivalent again. Similarly, these same children produce judgments that if one of two equivalent rows has an object added to it, it has more and if an object is removed, it has less. Hence, these children have rules which allow them to produce correct judgments about individual, isolated renversible, addition and subtraction actions. How these become internalized as coordinated operations is still an outstanding problem. The second reason for my ‘negative conclusions about reversibility training is that most of the successful reversibility experiments cited by previous reviewers are not cited by Strauss’ (p. 359). As Brainerd pointed out in the same paragraph, ‘the only rationale for not considering these studies as reversibility training experiments is that their treatments focus on at least one other operation and, therefore, their findings do not establish anything about reversibility training per se’ (p. 360). This seems to me to be perfectly legitimate since renversibility was confounded and, therefore, not tested in the studies mentioned (Bearison, 1969; Beihn, 1965; Gruen, 1965; Rothenberg and Orost, 1969; Smith, 1968; Winer, 1968). The third reason which Brainerd attributed to my conclusions about reversibility training was that I did not include four studies in my review, two of which were published after my review article (Brison, 1966; Hamel and Riksen, 1973; Schnall, Alter, Swanlund and Schweitzer, 1972; Roll, 1970). We shall now see how these studies alter the conclusions which I reached from reviewing other studies. The reader should keep in mind that Brainerd’s first two arguments were about theory and methodology and that the above four studies’ findings were cited in an attempt to show that empirical data from unconfounded studies contradict my conclusions that renversibility training was not effective. Hamel and Riksen (1973) conducted a renversibility training study in which it was shown that training Ss performed significantly better on the trained-for conservation tasks (continuous and discontinuous quantity) than control Ss who were not administered training. Contrary to Brainerd’s claim, there was MO ‘far transfer’ effect since there were no significant differences between renversibility training Ss and control Ss on concepts not trained for.
A reply to Brainerd
171
A problem which arises in interpreting this study’s results is that the reader is not provided information about pretest and posttest raw scores. Instead, we are shown mean changes from the pretests to the posttests. The Ss’ scores were arrived at by giving one point for a correct judgment on a conservation task. The first pretest was on continuous and discontinuous quantity which means that Ss could obtain a maximum score of 2.0. We were shown that the mean change score for the renversibility training Ss was 1.lO while for the control group it was -.05. A change of 1.10 means that on the average, Ss acquired conservation for one of the tasks on which they were nonconservers prior to the training. What we do not know is who changed to what. For example, if one-half of the Ss had a 0 score on the pretest and all of them changed to a score of 2.0 and none of the other Ss changed, the mean change score would be 1.0. Similarly, if the one-half of the S who had pretest scores of 1.O changed to 2.0 and none of the pretest Ss with a score of 0 changed, the mean change score would also be 1.0. We are not given this kind of information in the results section, nor do we know what percentage of Ss have pretest scores of 0 and 1, nor can we calculate this from the data we were given. This information seems critical since the training was for both. concepts which were on the pre- and posttests. As for the Schnall et al. (1972) study, we find that groups of Ss were administered four renversibility training conditions without proper control groups. The point could be made, however, that the almost totally negative results for three of the training condition’s Ss (2/60 pretest nonconservers conserved on the posttest) could serve as a control for the one condition in which there was some change (6/20 conserved on the posttest). In a replication and extension study, Strauss, Hess and Gilad (1974) administered the original, successful Schnall et al. (1972) training procedure and a second, analogous procedure. Control groups were used in a variant of the Solomon four-group design. Our results indicated no training effects whatsoever. Consequently. Brainerd’s conclusions about the Schnall et al. study should be understood as equivocal. As mentioned, Brainerd resolutely failed to note to the reader that there were three unsuccessful reversibility training conditions in the Schnall et aE. study. These conditions were significant since they were analogous to the training procedures in the Brison (1966) and Roll (1970) studies, the remaining two that he claimed indicate successful renversibility training conditions. This apparent contradiction will be shown to have a resolution which lies in an analysis of these two studies showing that they are not what Brainerd made them out to be. In the former study, Brison (1966, p. 320) cautioned that ‘the main
172
Sidney
Strum
limitation of this study is the failure to isolate the variables in the experimental training responsible for the induction of conservation’. In my opinion, a close inspection of his training condition sustains Brison’s warnings, and I cannot account for the reasons that Brainerd did not heed them. As such, Brison’s analysis of his own study does not particularly underline Brainerd’s contention that this is an example of a study which is ‘not subject to any methodological challenges [and] . . . focuses narrowly on inversion reversibility’. In the latter study, Roll (I 970) described a procedure in which, during one month’s time, 16 S’s were administered 44 trials of renversible deformations. Of these 16 Ss, eleven produced conservation judgments on the posttest, but only four were able to justify their judgments. At this point, the criterion problem looms large since if we accept the judgments-only criterion, 69% of the Ss conserved as opposed to only 25% posttest conservers if we accept the judgments-plus-explanations criterion. As was mentioned earlier, it is not common to find judgments and justifications at different levels, yet in the study 44% of the subjects produced this discrepancy. A second finding, which might also call the assessment procedure into question, was that the age range for Ss who were nonconservers for discontinuous quantity was 5,7 to 7,11. This concept is usually acquired naturally around age 5. Roll described a second training procedure which was conducted with children who were assessed as intuitive for the same concept and whose age range was from 5,9 to 8,l. The findings from the Roll study when taken together seem to show that if relatively older Ss are administered a large number (44) of renversible deformation trials (where Ss see that after being spread out, a row of objects can be returned to their original positions and will be the same number) over a long period of time, approximately 43% will be able to argue on a posttest that, when deformed, the objects in the rows are equivalent in number but that only l/4 can explain why this should be the case. This is hardly the ‘substantial evidence of the effectiveness of reversibility training’ which Brainerd claimed. In sum it appears that Brainerd’s claim that ‘there appears to be strong support for the effectiveness of treatments which focus narrowly on this [inversion] operation’ is exaggerated, at best. A more considered account of the data from the inversion studies which Brainerd cited and from the studies which I originally included in my review does not convince me that my original assessment of renversibility training was incorrect. Indeed, Brainerd’s choice of experiments and their findings seem to strengthen my claim that ‘research related to this mental operation has not been productive in inducing progressive structural reorganization’ (Strauss, 1972, p. 343). Finally, Brainerd included a series of studies whose purposes were to
A reply to Brainerd
173
investigate the influence of training for the identity operation on conservation acquisition (Hamel, 197 1; Hamel and Riksen, 1973; Hamel, Van der Veer and Westerhof, 1972). As we shall see once again, the gap between Brainerd’s statements that a form of training has been effective (in this case it is identity training) and the experimental evidence supporting this claim is enormous. For the first experiment (Hamel, 1971), I shall simply quote from his abstract. ‘The effect that conservation responses are furthered by reminding the child beforehand of the sameness of the water in the containers which changes in appearance, was not established by us. The striking improvement in conservation on the posttest which Bruner mentions, was also not confirmed’ (my italics added, p. 39). How Brainerd construed the content of this experiment as supporting his claim is not at all clear. For the Hamel and Riksen (1973) study, I have the same difficulty in interpreting their findings for identity training as I had for their renversibility training procedure (see above comments). Finally, the Hamel et al. (1972) study showed that some children offer identity justifications for conservation judgments. Contrary to Brainerd’s claim, the children in this study were administered relational terms training and not identity training. The results were that children who were transitional were more susceptible to the training than those who were strictly intuitive. As more of these children conserved on the posttest and since conservation judgments are sometimes accompanied by identity justifications, the frequency of posttest identity justifications increased. This somewhat unstartling finding when combined with the other studies’ findings leads me to question the soundness of Brainerd’s contention that identity training has induced conservation. In sum, the review of findings for studies whose research strategy was to train for individual operations did not sustain Brainerd’s claims. For the addition/subtraction operation training, the Japanese literature could not be evaluated, thus leaving us with Brainerd’s confirmation of my original assessment of this training procedure as being ineffective. The renversibility and identity training strategies were also shown to be unproductive. These findings are important not only for the sake of these operations alone, but also for their implications for multiple operations training. 2.2.2 Training multiple operations This heading is what Brainerd interpreted to be my intention for what I had labeled ‘operational coordination’. My argument ran as follows: (I) Intuitive thought is characterized by the reasoning that addition results in more of something and subtraction results in less, that one can empirically return to the former state of equivalence, etc. (2) These action rules remain segregated in the sense that intuitive children do not use them to come to conservation
174
Sidney Strauss
judgments. (3) Underlying concrete operational thought is an organized structure whose elements are the above rules which are coordinated as internalized operations. (4) As a consequence, an important feature that differentiates the intuitive from the concrete operations structure is the integration of what already appears in the intuitive stage. (5) Thus, training for operational integration among intuitive children could induce concrete operational thought products such as conservation. Experimental evidence for this hypothesis was sought in studies in which more than one operation was included in the training. The rationale for this was that the use of more than one operation could serve as an external analogue of coordinated operations. For example, one could add clay to one of two equivalent-amount balls of clay and could then remove the same clay. Here we have the physical action rules which are analogues to’ the addition operation and its inverse, the subtraction operation, which produce a final state of invariance. I showed that the literature indicated that training of individual operations was unsuccessful and training which included more than one operation was successful in inducing concepts such as conservation. From this evidence it was concluded that this successful training, in which there is an external analogue of coordinate operations, can be interpreted to confirm the hypothesis that the coordination of functionally segregated operations could induce structural transformation and elaboration. I further pointed out (p. 346) that the operational coordination training strategy ‘does not explain how a subject internalizes that information and coordinates his own mental operations’. Brainerd’s first argument against the above was that in none of the studies which I cited as providing evidence for the above hypothesis were subjects tested for having possessed the operations. As a consequence, he argues, ‘it is equally reasonable to conclude that the effectiveness of these treatments stems from the inculcation of the operations themselves’ (p. 36 1). I agree with Brainerd’s caution. On the other hand, I believe that the assumption that these rules were already possessed is not rash, especially since the children in these studies were at least five years old - an age at which children generally have the above-mentioned rules. Brainerd ( 1972a; p. 116) apparently thought the same when he wrote elsewhere about ‘the well-known fact that some nonconservers possess reversibility’. Why he now takes the position that these children probably do not possess these rules and that training which includes more than one rule might inculcate them remains unexplained. After the above suggestion, Brainerd then took a harder line in the follow‘If the coordinative interpretation of multiple operations ing paragraph: treatments cannot be accepted, then we are left with only the aforementioned conclusion that this type of training has been “markedly successful” ’ (my
A reply to Brainerd
175
italics, p. 361). He took issue with my conclusions that: (a) Operational coordination training has been successful in inducing conservation acquisition, and (b) this training technique has been more successful than training for individual operations. The objection to my interpretation of the success of operational coordination training was that these studies incorporated reversibility in their training treatments. Brainerd reasoned that since he had provided evidence of the success of reversibility operation training, the success of multiple operations can be attributed to the reversibility operations within these studies. However, it was shown above (cJ: section 2.2.1) that Brainerd’s argument about the efficacy of renversibility training was attractively unconvincing. Thus, the point remains that training which includes one operation has not been successful while training which has an analogue of coordinated actions has been successful in inducing conservation acquisition. 2.3 Extinction
studies
In my review, I noted that there were apparently two contradictory sets of findings of methodologically similar investigations about the ‘extinction’ of concepts which Piagetians argue are structural products, e.g., conservations. The training procedure in these studies was to present conserving Ss empirical evidence of nonconservation and to assess whether these Ss accepted or rejected this information. The first set of findings comes from the original extinction studies conducted by Smedslund (1961 b), Hall and Kingsley (1968) and others. Miller (1971) noted that these studies were not based on a solid theoretical base, and I had suggested that their similar findings provided some commonality. I then wrote that Miller concluded that the results of these studies were that ‘extinction of conservation is ubiquitous, rather easy to obtain, and a majority of Ss seem susceptible to such training’ (Strauss, 1972, p. 347). A second set of findings can be found in studies which I had interpreted to be formally similar to the above-mentioned studies. The second type of study was conducted by Turiel (1966), Turiel and Rothman (1972), Rest, Turiel and Kohlberg (1969) and Kuhn ( 1972). In these studies, Ss were presented a form of reasoning one below (-1) their predominant stage. The results were that children understood the ~ 1 reasoning yet rejected it as inadequate. I then interpreted the extinction studies as being formally similar to the -1 studies since conserving children, who are concrete operational, were shown the empirical phenomenon of nonconservation for which these very same children had argued in the -1 intuitive stage. As a result of the above, the argument was developed that there can be
176
Sidney Strauss
several explanations of the differences in these sets ot fmdmgs. The first explanation pertained to diagnosis since some experimenters probed Ss’ reasoning and others did not. That probing can show a much different picture of Ss’ understanding of the -1 problem than nonprobing was shown by Strauss and Liberman (1974). A second explanation pertained to the medium of presentation of a --I structure of reasoning. In the case of --1 studies, Turiel (1966) presented the -1 structure in argument form, Kuhn (1972) presented arguments and classification of materials, while Smedslund (196 1b), Hall and Kingsley (1968), etc., presented -1 empirical outcomes of physical deformations. Thus, I wrote that ‘one would imagine that the violation of an expected empirical law is of a different order of complexity than an argument or a classification of material which one had previously used predominantly’ (Strauss, 1972, p. 35 1). A third explanation of the discrepant results pertained to the content of the -1 information: Turiel’s work was about reasoning concepts, Kuhn’s dealt with logical concepts and the other studies’ content was physical concepts. This factor could also influence the results of the above studies and could alone or in combination with the others produce the discrepant results. This was the major thrust of my argument, and, true to form, Brainerd’s comments did not even tangentially touch upon this interpretation. His first argument was that I was on shaky empirical ground when suggesting that a stage theory interpretation is preferable to Miller’s (1971) conclusion that extinction of conservation is ubiquitous and rather easy to obtain. Brainerd understood that I was using Smedslund’s (196 la) findings to support my suggestion. On the contrary, I had argued that Smedslund’s findings were tzot consistent with organismic-developmental theory. Instead, the empirical data I used to support the organismic-developmental claim were drawn from Turiel’s and Kuhn’s -1 studies. As a consequence, one does not have to make the choice between Miller’s and my interpretation of extinction that Brainerd had characterized. Instead, one can disagree or agree with my interpretation of extinction studies as being equivalent in form to Turiel’s and Kuhn’s studies. After inadequately characterizing my position and comparing it to Miller’s, he then advanced his own interpretation. The first point made was that resistance to extinction is largely a function of ‘the ordinal position that concept occupies in the natural emergence hierarchy for conservation concepts’ (p. 363). As a result, number conservation, for example, should be more difficult to extinguish than weight conservation. Two reasons were offered for this, and we shall now examine them. For reasons of elucidation, the second reason will be examined first. Here Brainerd claimed that concepts such as quantity conservation which are logical prerequisites for concepts such as weight should be more difficult to extinguish.
A reply to Brainerd
177
Why this should be the case is not very clear. According to this reasoning, a logical derivative of a concept is easier to extinguish than the concept from which it was derived. In the philosophy and methodology of science, it is taken as axiomatic that if a logically derived concept of phenomenon is refuted by empirical evidence, one must give up the concept from which the phenomenon was derived - the primary concept. Thus, in its present form, Brainerd’s argument appears weak. Of course, one could argue that young children do not behave as scientists (should). If this is thought to be the case, one must provide more evidence for it than a simple declarative sentence. The other argument offered by Brainerd as to why the resistance to extinction decreases as we follow the natural emergence hierarchy for conservation concepts is more complex. It runs as follows: Since number conservation is acquired earlier, it requires less experiential confirmation; therefore, it must be experienced as more logically necessary and should be more difficult to extinguish via disconfirming experiential data. This argument is peculiar, especially given the prior argument. The prior argument proposed a development of concepts which have a prerequisite relationship, i.e., number is a prerequisite for weight, hence weight cannot emerge before number. Now Brainerd argues that number is a more logically necessary concept since it emerges before weight. This suggests that both have an equal probability of emerging simultaneously and that, because number is more logically necessary, it emerges first. 1 personally do not see a resolution of these two seemingly incompatible arguments, one of which is that some concepts are prerequisites for others and thus cannot emerge simultaneously, the other being that concepts can emerge simultaneously but do not because of differences in their logical necessity. Let us put this objection aside for the moment and proceed to an examination of the supporting evidence which Brainerd brings to bear on his hypothesis. As evidence for high resistance to number conservation extinction, Wallach and Sprott (1964) are cited as having attempted to extinguish this concept in both natural and trained conservers. A more critical reading of that study indicates that there was y1o attempt made to extinguish number conservation nor were natural conservers given training. In that study, Ss were provided a verbal counter-suggestion about nonconservation during training. That is, when objects (dolls and beds) were deformed and S judged them to be equivalent or nonequivalent, E noted that due to the deformation, a bed did not have a doll in front of it. This is certainly not an extinction study like the others which Brainerd cited (Brison, 1966; Hall and Kingsley, 1968; etc.). With respect to the inclusion of natural conservers, Wallach and Sprott wrote that ‘the experiment ended with the pretest for Ss in the Unequal When Paired, Conservation, and Partial Conservation categories’ (p. 1060, italics
178
Sidney Strauss
added). As a consequence of the above, it can be fairly argued that Brainerd’s ‘supporting evidence’ for resistance to number nonconservation is simply mistaken. On the other hand, there is some empirical evidence which disconfirms Brainerd’s hypothesis. In a recent study, Strauss and Liberman (1974) investigated the effects on conservers of observing the empirical violation of conservation in discontinuous quantity and weight. Briefly, the findings were that apparent spontaneous acceptance of -1 experiences was found to be a form of rejection of these experiences after probing. By the end of the probing procedure, almost all Ss had rejected both the -1 discontinuous quantity and weight experiences. Thus, these two concepts were equally resistant to -1 procedures, yet Brainerd would have predicted the latter to be less resistant. A second set of data disconfirm an extension of Brainerd’s argument. The extension is that two concepts which emerge at approximately the same time in the ontogenetic course should be similarly resistant to extinction procedures. Miller and Lipps (1973) tested Ss’ resistance to weight conservation and -1 weight transitivity, both of which emerge at about the same time. Their findings were that Ss were considerably more resistant to -1 weight transitivity than to -1 weight conservation. Brainerd could argue that his hypothesis was narrowly stated, pertaining to conservation concepts alone, and that the cited study is not a fair test of his hypothesis. I see these findings as being consistent with my (Strauss, 1972, p. 351) suggestion that one might find differences in resistance to -1 experiences for physical concepts (such as weight) and logico-mathematical concepts (such as transitivity). As mentioned earlier, Brainerd argued that we can determine which concepts are more or less logically necessary by their appearance in a child’s cognitive history. That is, ontogenetically prior concepts require less experiential confirmation and are, therefore, more logically necessary than ontogenetitally later concepts. The word ‘necessity’ is a modal, and the term ‘logical necessity’ falls within the domain of modal logic (cJ: Hughes and Cresswell, 1968). Logical necessity does not refer to the truth of statements about the physical world, as Brainerd would have it. Instead, it refers to the tautological nature of logically true sentences such as ‘It is either raining or not raining’. But sentences such as these, which include the disjunction of an atomic sentence and its denial, are empty in that they do not carry any semantic information (cf Hintikka, 1973). Wittgenstein in his Tractatus logicophih’sophiczts made this point quite clearly in stating that ‘I know nothing about the weather when I know that it is either raining or not raining’. As conservation laws are basic laws of the physical world, logical necessity does not explain them.
A reply to Brainerd
179
I understand Brainerd to be writing about what is sometimes called nomic necessity (Nagel, 1961) or causal necessity (Carnap, 1966). Nomic necessity refers to the necessity of the occurrence of basic, universal physical laws such as gravitation, conservation and those laws pertaining to the operating of a lever. The combination of basic laws in the ‘extinction’ task reads that if two objects which contain the same amount of the same matter are placed equidistantly from the center of a properly functioning balance scale, the force of gravity, interacting with both objects will result in the objects balancing, and we would claim that they are equivalent in weight. In sum there are four basic laws here: (a) Conservation of weight, (b) conservation of matter, (c) the operating of a lever and (d) gravity. The surreptitious removal of clay from one of the objects in the ‘extinction’ studies has been understood to pertain to the conservation of weight alone. In fact, though, Ss do not see nonconservation of weight; rather, they observe the nonbalancing of a scale. This phenomenon violates one of the four basic laws and needs not be restricted to weight conservation. Strauss and Liberman (1974) found that Ss most readily give up lever laws which suggests that they are the least understood or necessary. This interpretation, of course, remains to be tested with young children since none of the investigators who conducted ‘extinction’ studies determined if Ss understood these basic laws. This type of study was, however, recently conducted (Strauss, Danziger and Ramati, 1974) with college students, and it was found that none gave up laws of weight and matter conservation or gravity. Of the 70% of those who accepted the nonbalancing phenomenon, all explained it in terms of the laws of the lever. This data calls into question a number of ‘extinction’ studies since all of the experimenters understood implicitly and explicitly that they were tapping Ss’ notions of weight conservation alone. Further research of this type is necessary, and, if the above conclusions are sustained, we would then have an answer to Brainerd’s query about how it could be that trained conservers of weight are no less resistant to ‘extinction’ procedures than natural weight conservers. In sum, we can see once again that Brainerd’s points were off the mark. His analysis was also characterized by his steadfast refusal to come to grips with my argument that studies of this sort might offer insights into the organismicdevelopmental claim that cognitive functioning is hierarchically integrated (Bearison, 1974; Goldstein, 1939; Turiel, 1974; Werner, 1948).
180
Sidney Strauss
3. Summary Brainerd concluded with a set of seven summary statements which he believed were consistent with training study data. Due to space considerations, I will not always quote him directly and will paraphrase some of his points and briefly present a summarizing rejoinder to them. 1. Training treatments which introduce disequilibrium into the cognitive system from external sources have been successful. This, of course, refers to what had been termed adaptational disequilibrium. Of the types of such disequilibrium (prediction-outcome conflict, verbal feedback, conformity, observational learning and dimensional discrimination learning), only conformity training met the criterion of success offered by Brainerd. There have been so few studies of this nature that strong conclusions of the type offered by Brainerd should be tempered. 2. Subjects who evidence a given concrete operational concept on the pretests tend to be more susceptible to training experiences designed to induce that concept than subjects who do not evidence the concept on the pretests. This fact, in and of itself, permits no further conclusions about ‘structural mixture’ and training susceptibility. This category pertains to organizational disequilibrium. The reader will be reminded that Brainerd had claimed that the issue here was one of competence versus performance and that Brainerd’s definitions of these notions and their relations to concepts was quite unclear. In the final analysis, the question seemed to be whether one should accept only de now structural development (as Brainerd would have it) or both structural development and elaboration (as I had claimed). 3. If we have two concrete operational concepts A and B such that A invariably precedes B during the course of normal cognitive growth, A probably will prove easier to train than B in subjects who possess neither concept. The point of my arguments was missed as is captured nicely here by Brainerd’s comment. For reasons of space I could not comment on Brainerd’s thoughts in the main text and for the same reasons I elect to refrain from commenting here. Suffice it to say that the notion of structural mix has little to do with this (cf: Strauss and Ilan, in press). 4. Training treatments for a single operation (rule) have successfully induced conservation concepts, especially in the case of inversion reversibility and to a lesser extent for addition-subtraction, identity and reciprocity reversibility. As was shown
in the present
article,
more
careful
readings
of those
articles
A reply to Brainerd
18 1
claimed by Brainerd to support the above position lead one to exactly the opposite conclusions. 5. There is no definitive evidence that multiple operations training will induce conservation, in part because it is confounded with inversion training. In view of statement 4, however, it seems reasonable to suppose that multiple operations training eventually will prove successful also. Since it had been shown that statement 4 was incorrect, multiple operations training may very well be an interesting avenue of research to follow, but not for the reasons offered by Brainerd. One way to test this idea was suggested by Brainerd, namely, to factorially compare multiple operations with individual operations training. 6. The resistance of conservation concepts to extinction treatments is neither universally low nor universally high. Instead, resistance appears to vary as a function of the order emergence of conservation concepts. Kesistance to ‘extinction’ was shown to have little to do with the order ot emergence of conservation concepts. Earlier developing concepts such as discontinuous quantity conservation were no more resistant to ‘extinction’ than weight conservation. Similarly, concepts emerging at approximately the same time were not equally resistant. 7. To date, appreciable differences in the extinction resistance of trained and natural conservers have not been observed. It was argued in the present article that this may be a function of the testing procedures. Strauss, Danziger and Ramati (1974) found that college students understood what experimenters have thought was weight conservation extinction to be related to laws of the operating of the lever. If these findings could be replicated, Brainerd’s statement would then be incorrect. After the seven statements, Brainerd suggested that each one contradicts one or more of the conclusions I had reached (Strauss, 1972). Since I had stated these as partial evidence for the organismic-developmental view, Brainerd concluded that there was no support for my claim. I have presented further evidence in the text of the present article that the seven summary statements presented above were simply incorrect. Thus, my original assessment of the training literature seems to me to be sustained.
REFERENCES Ahr,
P. R., and Youniss, J. (1970) Reasons for failure on the class inclusion problem. Child Devel., 41, 131-143. Aronfreed, J. (1968) Conduct and con-
science. New York, Academic Press. Bandura, A. (1969) Principles of behavior modification. New York, Holt, Rinehart & Winston.
182
Sidney
Strauss
Bearison, D. J. (I 969) The role of measurement operations in the acquisition of consveration. Devel. Psychol., 1, -
653-660. (1974) The construct
of regression:
A Piagetian
Merrill-Palmer
approach.
Q., 20, 21-30.
Beilin, H. (1965) Learning and operational convergence in logical thought. J. Exp. -
child Psychol.,
2, 3 17-339.
(1971) The training and acquisition of logical operations. In M. F. Rosskopf, L. P., Steffe and S. Taback (Eds.), Piagetian cognitive developmental research and mathematical education. Washington, National
Council of Teachers of Mathematics. Brison, D. W. (1966) Acceleration of conservation of substance. J. genet. Psychol., 109, 31 I-322. Brainerd, C. J. (1972a) Reinforcement and reversibility in quantity conservation acquisition. Psychon. Sci., 27, 114116.’ (1972b) The age-stage issue in conservation acquisition. Psychon. Sci., 29, 115-117. -
-
(1973) Neo-Piagetian training experiments revisited: Is there any support for the cognitive-developmental stage hypothesis. Cog., 2, 349-370. and Allen, T. W. (1971) Experimental inductions of the conservation of ‘first order’ quantitative invariants. Psychol.
Carnap, R. tions
Bull., 75, 128-144. (1966) Philosophical foundaof physics. New York, Basic
Books. Cellerier, G. (1972) Information processing tendencies in recent experiments in cognitive learning - theoretical implications. In S. Franham-Diggory (Ed.), Information processing in children. New York, Academic Press. Christie, J. F., and Smothergill, D. W. (1970) Discrimination and conservation of length. Psychon. Sci., 21, 3366337.
Figure& J. C., and Keller, H. R. (1972) The effects of training and socioeconomic class upon the acquisition of conservation Child concepts. Devcl.,
43, 293-298.
Gelman, R. (1969) Conservation acquisition: A problem of learning to attend to relevant attributes. J. exp. child Psychol., 7, 167-l 87. Glaser, R., and Resnick, L. B. (1972) lnstructional psychology. In P. H. Mussen and M. R. Rosenzweig (Eds.), Annual review of psychology. Palo Alto, Annual Reviews, Inc. Goldstein, K. (1939) The organism. New York, American Book Company. Gruen, Cl. E. (1965) Experience affecting the development of number conservation in children. Child Devel., 36, 963-979.
Hall, V. C., and Kingsley, R. (1968) Conservation and equilibration theory. J. genet. Psychol., 113, 195-213. and Simpson, G. 1. (1968) Factors influencing extinction of weight conservation. Merrill-Palmer Q., 14, 197-210. Hamel, B. R. (1971) On conservation of liquids. Hum. Devel., 14, 39-46. and Riksen, B. 0. M. (1973) Identity, reversibility, verbal rule instruction, and conservation. Devel. Psychol., 9, 66-72. ~~
Van der Veer, M. A. A., and Westerhof, R. (1972) Identity, languageactivation training and conservation. Brit. J. educ. Psychol.,
42, 186-19
1.
Hatano, C. A., and lto, Y. (1966) The acquisition of conservation of area in second grades. Jap. J. Psychol., 37, 1855194. and Suga, Y. (1969) Equilibration and external reinforcement in the acquisition of number conservation. Jap. psychol. Res., 11, 17-3 1. Hintikka, J. (1973) Logic, language-games, and information: the philosophy
Kantian themes in of logic. Oxford,
A reply to Brainerd
Clarendon Press. Hughes, G. E., and Cresswell, An introduction
M. J. (1968) to modal logic. Lon-
don, Methuen. Huttenlocher, J., and Presson, C. C. (1973) Mental rotation and the perspective problem. Cog. Psychol., 4, 277-299. Inagaki, K. (1970) The effect of cognitive motivation on receiving and gathering of information. Jap. J. educ. Psychol., -
-
18, 14-25.
and Hatano, G. (1969) Motivational influence on epistemic observation. Jap. J. educ. Psychol., 16, 191-202. and Hatano, G. (1972) The effect of cognitive motivation aroused by positive infirming instances. Jap, J. educ. Psychol.,
20.
The diagnosis of reasoning in the mentally retarded. New York, John Day.
Inhelder,
-~
-
B.
(1972) Falsification and the methodology of scientific research programmes. In 1. Lakatos and A. Musgrave (Eds.), Criticism and the growth of knowledge. Cambridge, Cambridge University Press. and Piaget, J. (1958) The growth of logical
-
(1968)
thinking
from
childhood
to
adolescence.
New York, Basic Books.
Sinclair, H.,
and
Bovet, M. (1974)
Apprentisage et structures de la connaissance. Paris, Presses Universitaires
de France. Kuhn, D. (1972) Mechanisms of change in the development of cognitive structures. Child. Devel., 43, 833-844. -(1973) lmitation theory and research from a cognitive perspective. Hum. Devel., 16, 157-180. and Langer, J. (1968) Cognitive developmental determinants of imitation. University of California, Berkeley. Unpublished manuscript. Kuhn, T. S. (1962) The structure of scientific revolutions. Chicago, University of Chicago Press. (1970) Logic of discovery on psy-
183
chology of research. In 1. Lakatos and A. Musgrave (Eds.), Criticism and the growth of knowledge. Cambridge, Cambridge University Press. Lakatos, L. (1970) Falsification and the methodology of scientific research In I. Lakatos and programmes. A. Musgrave (Eds.), Criticism and the growth of knowledge. Cambridge, Cambridge University Press. (1974) The role of crucial experiments in science. Stud. Hist. Phil. Sci., 4, 309-325.
Langer, J. (1969) Disequilibrium as a source of development. In P. H. Mussen, J. Langer and M. Covington (Eds.), Trends and issues in developmental psychology. New York, Holt, Rinehart & Winston. and Strauss, S. (1972) Appearance, reality and identity. Cog., 1, 1055128. and Turiel, E. (1969) Role of imitation and conflict in cognitive acceleration. University of California, Berkeley. Unpublished manuscript. Miller, N., and Dollard, J. (1941) Social learning and imitation. New Haven, Yale University Press. Miller, S. A. (1971) Extinction of conservation: A methodological and theoretical analysis. Merrill-Palmer Q., 11, 319-334. (1973) Contradiction, surprise, and cognitive change: The effects of disconfirmation of belief on conservers and nonconservers. J. exp. child Psychol., 15, 47-62. and Lipps, L. (1973) Extinction of conservation and transitivity of weight. J. exp. child Psychol., 16, 388-402.
Murray, F. B. (1972) The acquisition of conservation through social interaction. Devel. Psychol., 6, I-6. Murray, J. P. (1974) Social learning and cognitive development: Modeling effects on children’s understanding
184
Sidney
Strauss
Brit. J. Psychol.,
of conservation.
65,
1.5~160. Nagel, E. (1961) The structure of science. New York, Harcourt, Brace & World. Overbeck, C., and Schwartz, M. (1970) Training in conservation of weight. J.
child
exp. Peters,
Psychol.,
D. L. (1967) Task individual differences servation of number.
able. Phil. Sci., 40, 1-9. Schnall,
in a conservation task: A closer analogue of reversibility than empirical return. ChildDevel., 43, 101221023.
9, 253-264. variations in Piaget’s
and con-
Merrill-Palmer
Shantz,C., and sessment through
Q., 13, 295-308. Piaget,
J. (1964) ment. In Rockcastle
-
Rest,
Siegler,
Sci., 18, 93-94. R. S., and Liebert, R. M. (1972) Effects of presenting relevant rules and complete feedback on the conservation of liquid quantity task.
The child’s conception of and speed. London, Rout-
Devel. Psychol., 7, 133-l 38. Silverman, I. W., and Geiringer, E. (1973) Dyadic interaction and conservation induction: A test of Piaget’sequilimodel. Child. Devel., 44, bration
ledge & Kegan Paul. K.,R. (1970) Normal science and I. Lakatos and its dangers. In A. Musgrave (Eds.), Criticism and the growth of knowledge. Cambridge,
Smedslund, J, (1961a) The acquisition of conservation of substance and weight in children. 1: Practice in conflict without reinforcement. situations
School (1970)
of education.
movement Popper,
Cambridge University Press. (lY72) Objective knowledge: An approach. Oxford, evolutionary
815-820.
Scan. J. Psychol., -.
Clarendon Press. Kohlberg, L. Turiel, E., and J., (1969) Level of moral development as a determinant of preference and comprehension of moral judgments made
by
others.
J.
Pers.,
37,
2, 1566160. (1961 b) The acquisition of conservation of substance and weight in children. 111: Extinction of conservation of weight acquired ‘normally’ and by means of empirical controls on a balance
-
225-252. Roll,
Watson, J, S. (1970) ASegocentrism spatial of violation. expentancy
Psychon.
Learning and developR. E. Ripple and V. N. (Eds.), Piaget redis-
covered: A report of the conference on cognitive studies and curriculum Cornell University, development. -
M., Alter, E., Swanlund, T., and Schweitzer, T. (1972) A sensorymotor context affecting performance
S. (1970) Reversibility training and stimulus desirability as factors in Child number. conservation of
Devel., 40, 707-726. Rosenthal, T. L., and (J Y 72) Modeling instruction and
Devel. Psychol.,
Rothenberg, B. B., and Orost, J. H. (1969) The training of conservation of number in young children. Child Devel.,
40,10?-126. Schagrin,
M. L.
(I 973) On being unreason-
Scan. J. Psychol.,
origines sociales de la In F. Bresson and decentration. M. de Monmollin (Eds.), Psychologie
et epistemologie genetiyue: Piagetians. Paris, Dunod. Smith,
Zimmerman, B. J. by exemplification conservation. in
6, 302-401.
scale.
2, 85-87. (1966) Les
Themes
I. D. (1968) The effects of training procedures upon the acquisition of conservation of weight. Child Devel.,
39, 5155526. Strauss,
S. (1 Y 72) Inducing cognitive development and learning: A review of short-term training experiments. 1: organismic-developmental apThe proach. Cog., 1, 3299357. (in preparation) Inducing cognitive
A reply to Brairzerd
-
-
-
development and learning: A review of short-term training experiments. II: The behaviorist approach. and Cohen, J. (1974) A study of the relations between mental operations underlying rotation and perspective tasks. Tel-Aviv University. Unpublished manuscript. Danziger, Y., and Ramati, T. (1974) College students’ understanding of possible violations of physical laws of conservation of mass, weight, operating of the lever, and gravity. TelAviv University. Unpublished manuscript. Hess, P., and Gilad, T. (1974) Training for reversibility: Implications for reversibility structure inducement. Tel-Aviv University. _Unpublished manuscript. Hess, P., Evron, R., Feldman, E., and Kruglanski, F. (1974) A two-step model of reasoning about discontinuous quantity. Tel-Aviv University. Unpublished manuscript. and Han, J. (in press) Length conservation and speed: Organizational disequilibrium training between concepts. J. ed. psychol. and Langer, J. (1970) Operational thought inducement. Child. Devel., 41, 163-175. D. (1974) The and Liberman, empirical violation of conservation laws and its relation to structural change. J. exp. child psychol., 18, 464-479. and Rimalt, I. (1974) The effects of organization disequilibrium training
185
on structural elaboration. Devel. Psychol., 10, 526-533. Sullivan, E. V. (1967) Acquisition of conservation of substance through film In D. W. techniques. modelling Brison and E. V. Sullivan (Eds.), Recent research on the acquisition of conservation of substance. Toronto, Ontario Institute for Studies in Education. (1969) Transition problems in conservation research. J. genet. Psychol., 115,41-54. Turiel, E. (1966) An experimental test of the sequentiality of developmental stages in the child’s moral judgments. J. Pers. sot. psychol., 3, 61 l--61 8. (1974) Conflict and transition in adolescent moral development. Child Devel., 45, 14L29. and Rothman, G. R. (1972) The influence of reasoning on behavioral choices at different stages of moral development. Child Devel., 43, 741756. Waghorn, L., and Sullivan, E. V. (1970) The exploration of transition rules in conservation of quantity (substance) using film mediated modeling. Acta Psychol., 32, 65-80. Wallach, L., and Sprott, R. L. (1964) Inducing number conservation. Child Devel., 35, 1057-1971. Werner, H. (1948) Comparative psychology of mental development. New York, International Universities Press. Winner, S. A. (1968) Induced set and acquisition of number conservation. Child Devel., 39, 195- 205.