FREE RECALL OF VERBAL AND NON-VERBAL STIMULI

Q. JI exp. Psychol. (1970) 22, 215-221 FREE RECALL OF VERBAL AND NON-VERBAL STIMULI JEFFREY R. SAMPSON Department of Computing Science, The University of Alberta In two experiments, 40 and 72 male subjects viewed 24 items, half as words and half as drawings. In Experiment I, all subjects were given a learning set; free recall was tested immediately after stimulus presentation and 30 min. later. In Experiment 11, half the subjects were set; recall was tested immediately and/or I day later. Picture recall was superior to word recall in all conditions (P < 0.005), in agreement with the findings of other investigators. Major extensions to previous procedures involved incidental learning, mixed stimulus lists, non-verbal reporting of pictures and delayed recall. A double-encoding explanation for the superior non-verbal performance is discussed. Introduction Several investigators (Ducharme and Fraisse, 1965 ; Lieberman and Culpepper, 1965; Paivio and Csapo, 1969; Paivio, Rogers and Smythe, 1968; Scott, 1967) have compared free recall performance on objects or pictures of objects with performance on the words that name them. In all these reports, memory for the non-verbal stimuli was better than for the verbal ones. Most explanations of this finding appear to involve a version of the following hypothesis: a familiar object usually evokes a verbal encoding in addition to the figural one, while names of such objects are often encoded only verbally; when recall is tested, the availability of two encodmgs facilitates memory for the objects. The best empirical support for this hypothesis comes first from the finding of Ducharme and Fraisse (1965) that memory for pictures and names presented together was virtually identical to that for the pictures presented alone, and second from the finding of Paivio and Csapo (1969) that better memory for pictures disappeared when the stimuli were presented rapidly enough to prevent the presumed implicit naming response. Several aspects of this hypothetical double-encoding process deserve further attention. First, the studies cited above provide little information about the spontaneity of the mechanism, since they all involved subjects who were biased toward labelling behaviour by the nature of the experimental procedures. In all five studies it would appear that the subjects knew before stimulus presentation that they would be expected to report only the names of the objects they saw. Ducharme and Fraisse (1965) analysed data from only a second task, after the subjects had experienced the verbal-report situation, while Lieberman and Culpepper (1965) and Scott (1967) gave the subjects lined response sheets before stimulus presentation. Paivio et al. (1968) and Paivio and Csapo (1969) used multiple tasks and appear to have provided response sheets in advance in some 215

216 J. R. SAMPSON cases (the latter study used oral report in the first experiment). If the doubleencoding mechanism is a natural means for processing figural information, it should be possible to find better memory for pictures in a setting where verbal responses are not implied beforehand. To this end, the present experiments (a) stressed only memory for the items, in pre-stimulus instructions to the subjects given a learning set, and (b) employed a number of subjects who were unaware of the memory test to follow. Another question left open by the procedures of the earlier experiments is whether subjects can demonstrate comparable performance differences when shown lists of items containing both verbal and non-verbal stimuli. In all five studies mentioned above, a given subject saw only one type of item and could thus conceivably have used a particular, fixed memory system (as Atkinson and ShXrin, 1968, employ the term) best suited to the type of data being processed. The present experiments used mixed lists, where superior performance on figural stimuli would be more suggestive of a manipulable control process (again in the Atkinson-Shiffrin sense) contributing to improved storage and/or retrieval of the pictures. Third, the double-encoding hypothesis would imply that subjects can usually recall the mode of presentation of the stimulus (since words could usually be distinguished by their single, verbal code). Both the unmixed list and purely verbal report aspects of the earlier studies precluded consideration of this question. In the present experiments, on the other hand, subjects were encouraged to draw the stimuli they remembered as pictures. Finally, the previous studies tested recall immediately after stimulus presentation. The present experiments include a delayed as well as an immediate test, to see if the superior memory for non-verbal stimuli persists over time. It should be emphasized that the experiments to be reported here do not test the validity of the double-encoding hypothesis; nor were they intended to. The plausibility of that mechanism is based upon the evidence cited above. What these experiments do test is the validity of the finding of superior memory for non-verbal stimuli in situations which differ critically from those in previous research. If the double-encoding mechanism is accepted as the basis for that superior memory, then these experiments may also be viewed as revealing something about the generality of that mechanism. This distinction will not always be made explicit in discussions which follow. Experiment I This experiment provided preliminary tests of the mixed list format and of the non-verbal report of pictures technique. It also served to suggest a number of refinements for the design of Experiment 11. Methodological detail concerning procedures not relevant to the present discussion is available in Kaplan, Kaplan and Sampson (1968). Method The subjects were 40 undergraduate males, whose participation partially fulfilled an introductory course requirement.

VERBAL AND NON-VERBAL STIMULI 2x7 The stimuli consisted of 24 items: ARROW, BELL, BOAT, BOTTLE, CAT, CIRCLE, FUNNEL, LAMP, PIPE, STAR, SUN, WINDOW, BIRD, BOOK, CANE, CUP, FISH, FLOWER, FORK, MOON, PENCIL, SQUARE, WATCH and WHEEL. Half the subjects saw the first 12 of these as words (typed, capital letters) and the remainder as pictures (simple line drawings), while the situation was reversed for the other half. This division, combined with two orders of presentation, yielded four stimulus lists, which were each seen by 10 subjects. The stimuli were displayed individually on every third slide of a group of 73 slides, each of which was presented for 4 sec. No more than three consecutive stimulus slides were of the same type (word or picture). Except for an initial slide containing the word READY, the non-stimulus slides contained colours, which the subjects were instructed to name. They were also told not to respond to the words and pictures, but to study them carefully for a later recall task. Immediately after presentation of the slides, the subjects were given paper and pencil and asked to reproduce as accurately as possible as many of the stimulus items as they could recall, in any order. Subjects who inquired if they were to draw the pictures were told to do so. After 30 min. of tasks independent of the recall situation, all subjects were again asked to reproduce the items they could still remember. Results In the immediate recall condition, subjects remembered 32 per cent of the words and 47 per cent of the pictures; in the 30-min. test, the corresponding figures were 29 and 46 per cent. An analysis of variance revealed significantly better memory for pictures (see Table I below); there was no significant difference between the immediate and 30-min. performances, and no interaction. Reversals-reproductions of pictures as words or vice versa-were rare, constituting only about 4 per cent of items recalled in the immediate condition (virtually the same items were again reversed at 30 min.). Eleven of the 14 (immediate) reversals were in the direction of words represented as pictures. (Two subjects, who, despite the instructions, reported all stimuli verbally were excluded from the reversal results.) Discussion It would appear that better memory for non-verbal stimuli can be demonstrated (a) when subjects do not use (and do not necessarily anticipate using) verbal reports of figural items, and (b) when each subject views both types of stimulus items. Thus the proposed verbal-encoding mechanism would not seem limited to situations where its use is suggested by the nature of the recall task. Furthermore, subjects appear to be capable of using such a mechanism even when they are not observing a single type of material, suggesting that a fairly flexible process is involved. The low incidence of reversals implies that subjects are able to distinguish the double-encoded pictures from the words at the time of recall. The simplest hypothesis is that words are usually encoded only verbally. Experiment 11 This experiment was designed to test the validity of the results of Experiment I in a wider context. Since the 30-min. delay in the first experiment did not produce any major performance differences, the delayed recall task was advanced to I day

218 J. R. SAMPSON after stimulus presentation; furthermore, not all subjects participated in both recall sessions, in order to determine if a substantial delay before the first test of recall would affect relative memory for words and pictures. Finally, as a further check on the generality and spontaneity of the superior non-verbal memory phenomenon, some of the subjects were not given a learning set; this seemed advisable because the possibility could not be ruled out in Experiment I that some subjects believed (without evidence) that recall would be purely verbal. Further methodological detail for Experiment 11 (particularly with respect to skin-resistance recording, which was actually used in both experiments but will be mentioned only incidentally here) is available in Sampson (1969a, &the latter reference also contains the raw data for this experiment). Method The subjects were 72 undergraduate males, whose participation partially fulfilled an introductory course requirement. The stimulus lists and their manner of presentation were the same as in Experiment I. The four lists were balanced across each of the treatment groups to be described. Two conditions were systematically varied, producing 6 12-subject treatment groups in a 2 x 3 design. The three recall conditions involved whether subjects had a single task, either immediately after stimulus presentation or I day later, or were doubly tested both immediately and at a day. The two set conditions involved whether or not subjects were told, as part of the pre-stimulus instructions, that they would later be asked to remember the items. Non-set subjects were told only that it was important for them to pay careful attention to the items so that a meaningful measure of their skin resistance (which was monitored throughout stimulus presentation) could be obtained. The recall task was conducted in the same manner as in Experiment I. For the I -day test, the subjects involved returned during the same hour on the day after stimulus presentation; subjects participating in only the immediate test did not return. Each non-set subject was asked if he had anticipated his recall session($. Results The instructions manipulating learning set were reasonably effective, as can be seen from the finding that only 6 of the 48 non-set recall tests were reported as other than a complete surprise. The percentage recall figures for the various conditions are presented in Table I (along with the data from Experiment I). Three separate analyses of variance revealed significantly better memory for pictures in all conditions. (The.maineffect differences related to set and delayed red1 were also significant at the 0.05 level or beyond ; no significant interactions were found between the word-picture factor and these other main effects.) Reversals were rare, comprising less than 5 per cent of all items reported. Of the 24 items reversed in all recall tests, 14 were in the direction of words represented as pictures. There were variations in reversal frequency among the groups, with immediate recall by set subjects showing the lowest incidence (less than I per cent) and 1-day first-session recall by non-set subjects showing the highest (about 26 per cent). ~cus~on The finding that subjects who admitted to being surprised by the recall test did relatively as well on the pictures as subjects with a learning set would seem to

VERBAL AND NON-VERBAL STIMULI 219 TABLE I Mean percentage recall of wor& and pictures Experiment Condition N Words Pictures (per cent) (per cent) I Immediate recall 30 min. recall 40 40 I1 Immediate: set non-set Day, Test 2: set non-set Day, Test I: set non-set 24 24 12 I2 12 12 imply that the hypothesized attachment of verbal labels to figural stimuli (and the consequent facilitation of memory) is not strictly a mnemonic technique; strong cognitive habits appear to be involved. In addition, non-verbal recall is better at least as long as I day later, suggesting that the effect of such labels is not especially transient. The last observation may need to be tempered in light of the reversal findings in Experiment 11. While still a rare feature of the performance of most subjects, reversals made up about a quarter of the responses in the most demanding memory task (first recall at I day by non-set subjects). It would seem that memory for the mode of the stimulus decreases with overall recall performance. While it is possible to interpret this finding as evidence for deterioration of the double code for figures it should be noted that the figural part would have to fade to explain the reproduction of a picture as a word. Furthermore, this explanation would not account for reversals in the other direction, which were actually more common in both experiments. General Discussion The present research suggests that the previously established phenomenon of better free recall performance on pictures than on the words that name them is just as apparent when (a) the learning is incidental, (b) the stimulus lists contain both kinds of items, (c) the pictures are reported nonverbally, and (d) the recall test is delayed up to I day. The persistence of this phenomenon suggests an underlying mechanism which is both stable and pervasive. As indicated at the outset of this report, certain results of earlier investigators strongly suggest a double-encoding of familiar figures as the basis for the superior memory. It might be of interest to speculate in a little more detail about the nature and origins of such a mechanism.

220 J. R. SAMPSON At least two arguments can be advanced in support of the hypothesis that it is usually more natural for most adults to think of the name of a simple familiar object when they see it than it is for them to form a clear mental image in response to the name. First, there is usually less ambiguity about what to call a common object than there is about the exact form of the image a word names. A second reason why the naming habit may be more frequent and natural rests in the nature of childhood experiences with language. The infant is constantly bombarded with the names of the objects he sees; and the older child asks the name of each new feature of his environment. Thus, if any order is likely when both object and name are present during early learning, it is object first, name second. So it is not unreasonable that the association between object and name is often biased so that the object will elicit the name more easily than vice versa. One feature of the Ducharme and Fraisse (1965) study lends further credence to this analysis of the role of early learning. In contrast to the adult subjects, children (of mean age around 8 years) performed no better on the pictures alone than on the words alone, but somewhat better on the combined stimuli. It would appear that the naming process is not so spontaneous in 8-year olds that it cannot be assisted by overt cues. Further evidence comes from the f ink by Dale (1969) that 4-yearold children perform better in a delayed colour-matching task if they tend spontaneously to say aloud the names of the test colours. Dale s hypothesis is, in part, that both a representation and a verbal code are stored during stimulus presentation, but that the verbal code is quickly lost by the non-verbalizing (silent) children. Further research could contribute to an even better understanding of the ontogeny of spontaneous verbal encoding. For the adult, however, the function of verbal encoding in the experimental context is rather easily understood. Both the figural and verbal traces of the nonverbal stimulus would normally be retrieved (because of their close association) ; and the subject could report the item as a picture. In the rare cases where only the verbal trace is retrieved, the subject would reverse the mode and report the item as a word. Individual instances of sharp image-formation in response to words, perhaps more frequent in a subject with exceptional capacities for visual thinking, could explain the somewhat more common occurrence of reversals in the word-topicture direction. The experimental portion of this investigation was supported in part by Research Grant MH-xx599 from the National Institute of Mental Health, United States Public Health Service, to Drs. S. and R. Kaplan, to whom the author is indebted for advice and assistance received while working in their laboratory at the University of Michigan, Ann Arbor, Michigan. References ATKINSON, R. C. and SHIFFRIN, R. M. (1968). Human memory:a proposed system and its control processes. In SPENCE, K. W. and SPBNCE, J. T. (Eds.), The PsychoZogy of Learning and Motivation, Vol. 2. New York: Academic Press. DALE, P. S. (1969). Color naming, matching, and recognition by pre-schoolers. Child Dev. 40, I 135-44. DUCHARME, R. and FRAISR, P. (1965). Etude genetique de la memorisation de mots et d images. Can. J. Psychol. 19, 253-61.

VERBAL AND NON-VERBAL STIMULI 221 KAPLAN, S., KAPLAN, R. and SAMPSON, J. R. (1968). Encoding and arousal factors in free recall of verbal and visual material. Psych. Sn'. 12, 73-4. LIEBERMAN, L. R. and CULPEPPER, J. T. (1965). Words vs. objects: Comparison of free verbal recall. Psychol. Rep. 17, 983-8. PAIVIO, A. and Csm, K. (1969). Concrete image and verbal memory codes. J. exp. Psychol. 80,279-85. PAIVIO, A., ROGERS, T. B. and SMYTHE, P. C. (1968). Why are pictures easier to recall than words? Psychon. Sn'. XI, 137-8. SAMPSON, J. R. (1969~). Further study of encoding and arousal factors in free recall of verbal and visual material. Psych. Sci. 16, 221-z. SAMPSON, J. R. (1969b). A neural subassembly model of human learning and memory. Technical Report No. 08226-12-T, Department of Computer and Communication Sciences, University of Michigan, Ann Arbor, June. Scm, K. G. (1967). Clustering with perceptual and symbolic stimuli in free recall. J. verb. Leamakg verb. Behmr. 6, 864-6. Received 24 November 1969