Japanese Psychological Research 1999, Volume 41, No. 3, 179 185 Short Report Physical and conceptual priming effects on picture and word identification JUNKO MATSUKAWA Department of Psychology, Faculty of Law and Literature, Shimane University, Matsue 690-8504, Japan Abstract: In the study phase of these experiments, subjects were asked to think of an item suggested by the omission in an incomplete sentence, and then look at a picture or word describing an item and say whether it was the same as theirs. In the test phase, they were asked to identify studied and nonstudied items presented briefly in either picture or word form. Subjects were then required to recall the words or pictures shown in the study phase. Experiment 1, with a within-subjects design, revealed that the studied pictures were identified more readily than studied words and nonstudied pictures. This indicates a physical priming effect. In word identification, studied words were identified more readily than nonstudied words; however, there was no difference between studied words and studied pictures, and the performance for studied pictures and nonstudied items were largely the same. The physical priming effect on picture identification was also shown in Experiment 2, with a between-subjects design. Different processing mechanisms in picture and word identification are discussed. Key words: physical priming, conceptual priming, picture identification, word identification. In a typical implicit memory task, subjects read a list of words in the study phase and perform a word fragment completion task in the test phase. Completion rates for words which subjects read in the study phase are superior to rates for words not read in the study phase, indicating the physical modality effect. This effect is mainly explained by the transferappropriate processing hypothesis (e.g., Roediger, 1990) or the perceptual representation hypothesis (e.g., Tulving & Schacter, 1990). According to these hypotheses, the priming effect would not be predicted if the stimulus modality differs between study and test phases. However, Jacoby (1983), Roediger and Blaxton (1987), and Weldon and Roediger (1987) observed such a conceptual (crossmodality) effect, although it was smaller than the physical modality effect. Hirshman, Snodgrass, Mindes, and Feenan (1990) found direct evidence of the conceptual priming effect in a picture fragment completion task. They found that sentence generation with the words presented in the study phase could affect picture fragment completion in the test phase. However, Srinivas (1993) asked subjects to identify briefly presented recoverable and nonrecoverable picture fragments after a word-generation task in the study phase but failed to find a conceptual priming effect. The different types of memory tests used, fragment completion tests and identification tests, might explain this discrepancy. Witherspoon and Moscovitch (1989) suggested that perceptual identification depends more on visual information than word completion tests, because a word completion is a more conceptual problem-solving task that requires lexical 1999 Japanese Psychological Association. Published by Blackwell Publishers Ltd, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA.
180 J. Matsukawa search (p. 23). They also suggested that the two types of tests are stochastically independent. The first purpose of this paper is to compare conceptual priming and physical priming on picture identification. According to Witherspoon and Moscovitch (1989), perceptual identification should be more sensitive to the physical features of stimuli when they are presented briefly. If so, we would not expect to see a conceptual priming effect in picture identification, although Witherspoon and Moscovitch were primarily discussing word identification. The conceptual priming effect will not be found in picture identification if Srinivas (1993) and Witherspoon and Moscovitch (1989) are correct. However, we would expect to find the effect of physical priming on picture identification because many researchers have found this effect in perceptual identification (e.g., Jacoby, 1983; Jacoby & Dallas, 1981; Roediger, 1990; Tulving & Schacter, 1990). Second, we will examine the relative roles of conceptual and physical priming on picture and word identification. Nelson, Reed, and Walling (1976) proposed the sensory-semantic model on the basis of their paired-associate learning results for pictures and words. This model is related to the superiority of the sensory code for pictures that contributes to their identification. Their proposal suggests that picture identification depends on the sensory or physical features of pictures rather than on their conceptual meaning. If so, the conceptual priming effect, if present at all, will be smaller than the physical priming effect on picture identification. On the other hand, this will not be the case for word identification. Experiment 1 Method The subjects were 36 undergraduate students at Shimane University. They participated in this experiment as part of a course requirement or as volunteers. If any students could identify all stimuli or no stimuli in the identification tasks, they were rejected as subjects. Sixty common pictures were selected from the standardized set of Snodgrass and Vanderwart (1980). The items depicted belonged to various categories. The verbal counterpart of each picture was selected from Matsukawa (1983), the Japanese edition of Snodgrass and Vanderwart (1980). The mean name agreement for them was 93.0% and mean image agreement was 3.92 (5-point rating scale with endpoints labeled very low image agreement and very high image agreement). These items were divided into three sets of 20. Sentences for concept generation were constructed so that subjects would invariably think of the item when the word was deleted from the sentence and replaced with circles for letter markers (e.g., in English, There are lots of cigarette butts in the OOOOOOO ). The experiment used a two-factor within-subjects design. The first factor was study type with three levels (pictures, words, nonstudied), while the second factor was test type with two levels (pictures vs. words). Three sets of materials were counterbalanced so that for the subjects each item appeared in each experimental condition equally often. A TKK three-channel slide tachistoscope and an NEC microcomputer were used for this experiment. Subjects were tested individually. They were told that the task in the study phase was to think of an item corresponding to the missing word in the sentences presented on the screen. They were asked to read the sentence for 7 s. After that the correct answer (item), which was either a picture or a word, was presented for 2 s on the screen. Subjects were asked to consider whether the item they thought of corresponded to the correct answer and pressed the yes key if so or the no key if not. After the subjects completed the first task, they were given a 2-min backward counting distracter task before proceeding to the second task. The second experiment (the test phase) was picture and word identification. Subjects were presented with a series of 30 pictures and a series of 30 words. The presentation times were 43 ms for the pictures and 53 ms for the words. These presentation times were determined in a
Physical and conceptual priming effects on picture and word identification 181 preliminary experiment as ones which avoided subjects identifying all the stimuli or none at all. The picture identification task comprised pictures of 10 items that had been presented as pictures as answers in the study phase and 10 that had been presented as words. Another 10 items were nonstudied. The word identification task similarly included 10 items that had been presented as words, 10 that had been presented as pictures, and 10 nonstudied items. Each picture and word was counterbalanced so as to be equally presented to the subjects. Half the subjects were presented the picture series first and the word series second, and the other half were presented the word series first and the picture series second. After the picture and word identification tasks, subjects were again given a 2-min backward-counting distracter task. After that, they were asked to recall the correct items which they had generated during the study phase of the experiment. Results A 3 2 analysis of variance of the percentage of items identified was conducted with the factors of study type (picture, word, nonstudied) and test type (picture vs. word). The mean percentage correct identification rates for the six experimental conditions are shown in Table 1. There was a main effect of study type, F(2, 70) = 9.61, MSe = 2.17, p.001, picture = 52%, word = 51%, nonstudied = 43%, in which the studied items were more Table 1. Mean percentage picture and word identification, and correct recall for each condition in Experiment 1 Study type Picture Word Nonstudied Identification (%) Picture 55.3 46.1 41.9 Word 50.0 56.4 43.6 Correct recall (%) Picture 41.7 32.2 Word 39.2 31.4 easily identified than nonstudied items. The multiple comparison analyses indicated that the percentages for the studied items in both the picture and word test conditions were significantly higher than the percentage of nonstudied items identified, p.001. There was no main effect of test type (F 1), picture = 48%, word = 50%. The interaction between study type and test type was significant, F(2, 70) = 3.37, MSe = 3.59, p.05. The simple main effects and the multiple comparison indicated that the percentage of studied pictures identified was significantly higher than the percentage for either the studied words or the nonstudied pictures, F(2, 140) = 6.35, MSe = 2.88, p.05, when picture identification was asked of the subjects. When word identification was asked of the subjects, the percentage of studied words identified was significantly higher than the nonstudied words, F(2, 140) = 5.51, MSe = 2.88, p.05. Although the percentage identification for the studied pictures in the word identification test was higher than for nonstudied words, this did not reach statistical significance. There was no significant difference between the percentage of studied words and studied pictures identified. Priming scores were computed by subtracting each subject s percentage identification rates in the nonstudied condition from the percentage in the studied picture and word conditions. When the studied item had been a picture, the priming scores were 13.3% for pictures in the test phase and 4.2% for words in the test phase. When the studied item had been a word, the priming scores were 12.8% for words in the test phase and 6.4% for pictures in the test phase. The mean percentage correct recall for each condition is shown in Table 1. A 2 2 factor analysis of variances of the percentage recall by study type (word vs. picture) and test type (word vs. picture) was conducted. The main effect of study type was significant, F(1, 35) = 18.76, MSe = 1.42, p.001, indicating that studied pictures were recalled at a higher rate than studied words. There was neither a main effect of test type nor an interaction between
182 J. Matsukawa study type and test type. These results concur with previous reports of better recall with pictures (Nickerson, 1965; Weldon & Roediger, 1987; Weldon, Roediger, & Challis, 1989). Discussion The picture identification results from Experiment 1 agree with Srinivas (1993) but not Hirshman et al. (1990). No conceptual priming effect was found in this experiment. We therefore agree with the proposal of Witherspoon and Moscovitch (1989) that picture identification is more sensitive to physical features, and any conceptual priming effect would be much smaller than the physical priming effect as a consequence. The word identification results also indicated the physical (modality) priming effect. This finding largely concurs with previous findings (e.g., Jacoby, 1983; Jacoby & Dallas, 1981). The conceptual (cross-modality) effect was not conclusively demonstrated, because identification rates for the studied pictures did not differ significantly from those for either the studied words or the nonstudied words. However, it seems likely that conceptual priming influences subsequent word identification more than picture identification. Weldon (1991) examined perceptual, lexical, and conceptual processing effects in priming and found that printed words produced more priming than auditory or generated words. She suggested that lexical access is more important than physical similarity because only those subjects instructed to think of the original words by mentally interchanging the vowels exhibited significant priming. In our experiment, it is likely that the subjects had to access to their lexicon to search for the appropriate word in order to complete the sentences. This suggests that both lexical and semantic systems relate to conceptual processing in this experiment. We used a within-subjects design in Experiment 1. According to Brown and Mitchell (1994) and Challis and Brodbeck (1992), there are differences between the effect in a withinsubjects design and the one in a betweensubjects design. For example, they could manipulate both picture and word conditions during the study phase because subjects had been asked to look at either pictures or words to determine whether they corresponded to the concept that subjects generated. If that were the case, a conceptual priming effect may be found in perceptual identification. Experiment 2 examines this possibility with a between-subjects design. Experiment 2 Method The subjects were 32 undergraduate students at Shimane University. The materials used in Experiment 2 were the same as in Experiment 1. The experiment used a three-factor design. Prime type was the first factor, with two levels (pictures vs. words) arranged between subjects. The other two factors were study type, with two levels (studied vs. nonstudied), and test type, with two levels (picture vs. words), arranged within subjects. Except for the following points, the procedure was generally the same as in Experiment 1. During the study phase, half the subjects were presented pictures as answers to the missing word in the sentence. The other half were presented words as answers. During the test phase, subjects were presented with a series of 30 pictures and a series of 30 words. The presentation times were 43 ms for the picture series and 60 ms for the word series. Fifteen pictures represented items that had been presented as answers in the study phase, and the other 15 represented nonstudied items. Fifteen words represented items that had been presented as answers in the study phase, and the other 15 represented nonstudied items. Each picture and word was counterbalanced so as to be equally presented to subjects. Half the subjects were presented the picture series first and the word series second, while half of the subjects were presented the word series first and the picture series second in the test phase. Results A 2 2 2 analysis of variance of the percentage of items identified was conducted,
Physical and conceptual priming effects on picture and word identification 183 Table 2. Mean percentage picture and word identification, and correct recall for each condition in Experiment 2 Picture Non- Word Nonstudied studied studied studied Identification (%) Picture 65.0 50.8 46.6 47.1 Word 63.8 65.8 68.3 62.1 Correct recall (%) Picture 42.1 31.3 Word 41.7 32.1 with the factors of prime type, test type, and study type. The mean identification percentages for each condition are shown in Table 2. There was a main effect of test type, F(1, 30) = 14.94, MSe = 7.66, p.001, picture = 53% vs. word = 65%, which indicated that words were identified more easily than pictures. The interaction of prime type test type study type was also significant, F(1, 30) = 4.35, MSe = 5.43, p.05. The interaction of prime type test type was not significant, F(1, 30) = 3.09, MSe = 8.32, p.10. The results of the simple interaction analysis indicated that the interaction between test type and study type for picture prime type was significant, F(1, 30) = 4.37, p.05, which indicated that studied pictures were identified more easily than nonstudied pictures, F(1, 30) = 6.65, MSe = 5.43, p.05, but studied words were identified no more easily than nonstudied words statistically (F 1). For word prime type, there was no interaction (F 1). When the prime type was a picture, the priming scores were 14.2% for pictures and 2.0% for words in the test phase. When the prime type was a word, the priming scores were 6.25% for words and 0.5% for pictures in the test phase. A 2 2 analysis of variances on the percentage recall with the factors prime type and test type was conducted. The main effect of prime type was significant, F(1, 30) = 6.58, MSe = 5.70, p.001, indicating that pictures were recalled at a higher rate than words. There was no main effect of test type nor an interaction between prime type and test type. These results coincide with the ones in Experiment 1 and previous results (Nickerson, 1965, 1968; Weldon & Roediger, 1987; Weldon et al., 1989). Discussion With picture identification, Experiment 2 again showed the physical modality effect, while the results of word identification did not indicate significant physical or conceptual priming effects. The clear lack of a conceptual priming effect suggests that there is a difference for processing words between a within-subjects design and a between-subjects design, as noted above. Nevertheless, the experimental design cannot explain why there was no physical priming effect for word identification. This result differs from our expectation and previous findings (e.g., Jacoby, 1983; Jacoby & Dallas, 1981). The possibility is that the presentation time of the word condition in Experiment 2 might have been too long, because even performances on nonstudied word items indicate an identification rate of over 60%. This possibility was therefore examined in Experiment 3. Experiment 3 Method The subjects were 24 undergraduate students of Shimane University. Materials used in the experiment were the same as those in Experiments 1 and 2. A two-factor experimental design was used. One factor was test type with two levels (picture vs. word) and the other was study type with two levels (studied vs. nonstudied), and they were arranged within subjects. Except for the following, the procedure was the same as in Experiment 2. During the study phase, all subjects were presented words as answers for the incomplete sentences. The presentation time was 53 ms for the word series in the test phase, which was the same as the presentation time for the word series in Experiment 1.
184 J. Matsukawa Table 3. Mean percentage word identification, and correct recall for each condition in Experiment 3 Studied Nonstudied Identification (%) Picture 59.7 58.1 Word 56.4 51.4 Correct recall (%) Picture 35.2 Word 29.4 Results and discussion The means for the four experimental conditions are shown in Table 3. A 2 2 analysis of variance on the percentage of word and picture identification was conducted with the factors test type and study type. There were no main effects (F = 1.03 for test type; F 1 for study type) or interaction (F 1). Priming scores were 2.2% for the picture condition and 5.0% for the word condition. The studied words were significantly more readily identified than nonstudied words when a sign test was conducted on the priming scores of word and picture identification tasks (p =.048). This result concurs with that of Jacoby and Dallas (1981), although there were no significant effects discovered by the analysis of variance. The correct recall results concurred with those in Experiments 1 and 2. sensory-semantic model that is related to the superiority of the sensory code for pictures that contributes to their identification. Thus, picture processing depends more on the sensory or physical features of pictures than on conception or meaning. In addition to a physical priming effect, it seemed likely that conceptual priming could have affected subsequent word identification in Experiment 1, although we could not show this conclusively. In contrast, the result from Experiment 2, with a between-subjects design, failed to show any conceptual priming effect. This result cannot be considered as a ceiling effect because the results for conceptual priming were repeated in Experiment 3, in which the presentation time was shorter than in Experiment 2. This means that the experimental design could have affected the priming effect on word identification. A dissociative effect due to the experimental design (e.g., a between-subjects vs. a within-subjects design) has also been reported by other investigators (Challis & Brodbeck, 1992; Challis & Sidhu, 1993). Implicit memory could be affected by various experimental conditions (Brown & Mitchell, 1994; Nelson, Schreiber, & McEvoy, 1992). All of these researchers suggest that implicit memory is sensitive to the method of processing the stimuli in the study phase. The results of the present series of experiments further suggest that the experimental design could affect perceptual identification. General discussion Only the physical priming effect was observed on picture identification in Experiment 1, which had a within-subjects design. Furthermore, subjects recalled the items presented as pictures in the study phase more often than the items presented as words, indicating the picture superiority effect. These results were also replicated in Experiment 2, with a betweensubjects design. These results suggest that picture processing is more sensitive to physical than to conceptual features. Nelson et al. (1976) proposed a References Brown, A. S., & Mitchell, D. B. (1994). A reevaluation of semantic versus nonsemantic processing in implicit memory. Memory and Cognition, 22, 533 541. Challis, B. H., & Brodbeck, D. R. (1992). Level of processing affects priming in word fragment completion. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 595 607. Challis, B. H., & Sidhu, R. (1993). Dissociative effect of massed repetition on implicit and explicit measures of memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19, 115 127.
Physical and conceptual priming effects on picture and word identification 185 Hirshman, E., Snodgrass, J. G., Mindes, J., & Feenan, K. (1990). Conceptual priming in fragment completion. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16, 634 647. Jacoby, L. L. (1983). Remembering the data: Analyzing interactive processes in reading. Journal of Verbal Learning and Verbal Behavior, 22, 485 508. Jacoby, L. L., & Dallas, M. (1981). On the relationship between autobiographical memory and perceptual learning. Journal of Experimental Psychology: General, 110, 306 340. Matsukawa, J. (1983). A study of characteristics of pictorial materials (1): With Snodgrass and Vanderwart s pictures. Memoirs of the Faculty of Law and Literature of Shimane University, 6, 97 139. Nelson, D. L., Reed, V. S., & Walling, J. R. (1976). The pictorial superiority effect. Journal of Experimental Psychology: Human Learning and Memory, 2, 523 528. Nelson, D. L., Schreiber, T. A., & McEvoy, C. L. (1992). Processing implicit and explicit representations. Psychological Review, 99, 322 348. Nickerson, R. S. (1965). Short-term memory for complex meaningful visual configurations: A demonstration of capacity. Canadian Journal of Psychology: Review of Canadian Psychology, 19, 155 160. Nickerson, R. S. (1968). A note on long-term recognition memory for pictorial material. Psychonomic Science, 11, 58. Roediger, H. L. (1990). Implicit memory: Retention without remembering. American Psychologist, 45, 1043 1056. Roediger, H. L., & Blaxton, T. A. (1987). Effects of varying modality, surface features, and retention interval on priming in word-fragment completion. Memory and Cognition, 15, 379 388. Snodgrass, J. A., & Vanderwart, M. (1980). A standard set of 260 pictures: Norms for name agreement, image agreement, familiarity, and visual complexity. Journal of Experimental Psychology: Human Learning and Memory, 6, 174 215. Srinivas, K. (1993). Perceptual specificity in nonverbal priming. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19, 582 602. Tulving, E., & Schacter, D. L. (1990). Priming and human memory systems. Science, 247, 301 306. Weldon, M. S. (1991). Mechanisms underlying priming on perceptual tests. Journal of Experimental Psychology: Learning, Memory, and Cognition, 17, 526 541. Weldon, M. S., & Roediger, H. L. (1987). Altering retrieval demands reverses the picture superiority effect. Memory and Cognition, 15, 269 280. Weldon, M. S., Roediger, H. L., & Challis, B. H. (1989). The properties of retrieval cues constrain the picture superiority effect. Memory and Cognition, 17, 95 105. Witherspoon, D., & Moscovitch, M. (1989). Stochastic independence between two implicit memory tasks. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15, 22 30. (Received March 15, 1997; accepted May 9, 1998)