Concreteness and relational effects on recall of adjective-noun pairs

Concreteness and relational effects on recall of adjective-noun pairs

Allan Paivio

Abstract Extending previous research on the problem, we studied the effects of concreteness and relatedness of adjective-noun pairs on free recall, cued recall, and memory integration. Two experiments varied the attributes in paired associates lists or sentences. Consistent with predictions from dual coding theory and prior results with noun-noun pairs, both experiments showed that the effects of concreteness were strong and independent of relatedness in free recall and cued recall. The generally positive effects of relatedness were absent in the case of free recall of sentences. The two attributes also had independent (additive) effects on integrative memory as measured by conditionalized free recall of pairs. Integration as measured by the increment from free to cued recall occurred consistently only when pairs were high in both concreteness and relatedness. Explanations focused on dual coding and relational-distinctiveness processing theories as well as task variables that affect integration measures.

This study further tested alternative hypotheses concerning the effects of concreteness and relational variables on free recall, cued recall, and measures of integrative recall. The theoretical and empirical issues were reviewed in detail by Paivio, Walsh, and Bons (1994). We summarize the pertinent aspects of that background and then present the rationale for the present research.

The alternative hypotheses and predictions were based on dual coding theory (e.g., Paivio, 1971, 1991) and Marschark and Hunt’s (1989) relational/distinctiveness processing theory. Dual coding theory explains positive effects of word concreteness in target tasks primarily in terms of the following empirically supported assumptions: (a) nonverbal images are more likely to be aroused by concrete than abstract words; (b) the memory traces of the activated images are “stronger” than the verbal traces of the words themselves; (c) the image and verbal traces are mnemonically independent and additive; (d) concrete word pairs promote activation of compound images that function as integrated memory traces; and (e) the integrated image can be redintegrated by presentation of one pair member as a retrieval cue, thereby mediating response recall. The independence-additivity assumption accounts for most of the concreteness effect in free recall and some of the effect in cued recall. The imagery integration-redintegration hypothesis accounts for the findings that the concreteness effect is larger in cued recall than in free recall, and that concrete items are especially effective as retrieval cues. The integrative and retrieval functions of compound images define the conceptual peg hypothesis of imagery effects in paired associate learning, which we discuss further after describing Marschark and Hunt’s alternative approach.

Marschark and Hunt (1989) proposed that the effects of concreteness on memory arise from relational and distinctive processing of items rather than imagery or dual coding mechanisms. Relational processing entails responding to word pairs or sentences on the basis of inter-item relational information inherent in the items (associative or semantic relations) or activated by experimental procedures (e.g., instructions to relate the items in some way). Distinctive processing entails responding to items on the basis of any information that distinguishes items from each other.

Marschark and Hunt reasoned that memory for response words from a list of pairs depends on the activation of both relational and distinctive information at encoding. Relational information that is reactivated at retrieval delineates a search set of word pairs, and distinctive information then permits discrimination of each target pair and response word from the set. Concreteness-induced imagery, though a possible source of both kinds of information, especially enhances distinctive processing. Therefore, given activation of the appropriate relational information, concrete items should be recalled better than abstract items because the former are distinctive. If relational processing is absent during encoding or retrieval, concreteness effects should be attenuated or eliminated (Marschark & Hunt, 1989, p. 711).

The above hypothesis implies that concreteness will interact with conditions that maximize or minimize relational processing during encoding or retrieval. The results of a series of experiments by Marschark and Hunt (1989) were generally consistent with the predictions. They were also interpreted to be inconsistent with dual coding theory on the grounds that the latter attributes concreteness effects to imagery as an additional memory code rather than to imagery-based distinctiveness in the context of relational processing.

Paivio et al. (1994) responded that relational and distinctiveness processing mechanisms do not in themselves distinguish the two theories because dual coding theory also assumes that imagery and verbal processes can affect memory by enhancing relational processing (e.g., in paired associate learning) or distinctiveness (e.g., in recognition memory) of items. The theories differ, however, in regard to two hypotheses that are relevant to concreteness effects on memory. One is the dual coding hypothesis that imaginal and verbal codes are mnemonically independent and, therefore, additive in their effects on item memory. This explanation was the focus of Marschark and Hunt’s (1989) opposition to dual coding. The other important difference is that, whereas Marschark and Hunt proposed that concreteness effects depend on relational processing, Paivio et al. reasoned from an empirical and theoretical basis that concreteness and relational variables are independent in their effects on memory.

Paivio et al. (1994) tested the two theories using free and cued recall of response items from lists of concrete and abstract paired associates in which the pair members were either associatively related or unrelated. The alternative hypotheses led to different predictions. According to dual coding theory, concreteness and relatedness should have independent and additive effects in both free and cued recall. According to Marschark and Hunt’s (1989) theory, however, the two variables should interact so that a concreteness advantage is greater with related than with unrelated pairs. Moreover, because suing reinstates relational processing, a triple interaction should emerge in which the concreterelated pair advantage would be further augmented by suing.

The results of two experiments were consistent with dual coding predictions in that concreteness and pair relatedness were independent and additive in cued recall and free recall. In one experiment, free recall showed a suggestive interaction (significant over participants), the form of which was inconsistent with relational/distinctiveness processing theory. The clear independence of the effects of the two variables in cued recall and the absence of a triple interaction provided the strongest support for dual coding theory.

The Paivio et al. (1994) experiments also investigated concreteness effects on integrative memory. Begg (1972) found that integrative memory as measured by the increment from free recall to cued recall was higher for concrete than for abstract adjective-noun phrases. Begg’s interpretation was that concrete phrases are stored as integrated images whereas abstract phrases are stored as individual words. In contrast, using cued recall of sentences, Marschark and Paivio (1977) found that, although overall recall was higher for concrete than abstract sentences, recall was equally integrated for the two types of materials. This equivalence seemed inconsistent with Begg’s results and interpretations, as well as with dual coding theory, which also attributes special integrative properties to imagery. Paivio et al. suggested that the apparently anomalous results obtained by Marschark and Paivio could be explained in terms of strong intrasentence verbal associations, which provided a basis for integrative recall of abstract sentences even in the absence of imagery. Accordingly, Paivio et al. predicted that pair concreteness and relatedness would have independent and additive effects on integrative recall as measured by the increment from free to cued recall and by free recall of pairs conditionalized on recall of response members, a measure highlighted by Horowitz and Prytulak (1969; for a discussion of both measures, see Begg, 1972). These predictions were confirmed, justifying the conclusion that strong verbal associations were necessary for integrative recall of abstract pairs whereas high imagery value was sufficient for integration of concrete pairs.

The main aims of the present experiments were to (a) test the generality of the independent effects of pair concreteness and pair relatedness on free and cued recall using adjectivenoun phrases and sentences rather than noun-noun pairs, and (b) reconcile the different patterns of integration effects observed in the studies reviewed earlier. The change in materials was motivated by Marschark and Hunt’s (1989) suggestion that Begg’s (1972) results could be attributed to encouragement of relational processing by his use of meaningful adjective-noun pairs, and because dual coding and relational/distinctiveness processing theories have not been compared using sentences.

The integration outcomes in particular hinge on the extent to which the patterns of results for adjective-noun pairs are comparable to those obtained with noun-noun pairs, which in turn could depend on how integration is measured. The free-to-cued recall increment measure in particular would be affected by any factor that differentially affects overall recall levels for the two tasks. Relevant evidence on cued recall can be found in early experiments concerned with the conceptual peg hypothesis, involving imagery effects in paired associate learning (see Paivio, 1971, pp. 272-276). Results uniformly showed that nouns are superior to adjectives as retrieval cues for response terms and that the difference is largely attributable to the generally higher imagery value of nouns. No single experiment directly compared adjective-noun and noun-noun pairs, but different experiments using comparable items suggested that cued recall of noun responses was generally lower for adjective-noun pairs (e.g., Yuille, Paivio, & Lambert, 1969) than for noun-noun pairs (e.g., Paivio, Smythe, & Yuille, 1968).

Free recall data show a different pattern in the only relevant comparison known to us. Specifically, free recall was as high_for response nouns from adjective-noun pairs (Begg, 1972, Experiment 2) as from noun-noun pairs (Paivio et al., 1994, Experiment 1), suggesting that free recall may be less affected than cued recall by the stimulus change from noun to adjective.

The implications of the above comparisons for the present experiments are clear: To the extent that adjectivenoun cued recall here turns out to be lower than noun-noun cued recall in Paivio et al. (1994) with no corresponding change in free recall levels, integration as measured by the increment from free to cued recall would be lower in the present experiments than in Paivio et al. (1994). Moreover, if sufficiently depressed across all pair types, the expected additive effects of concreteness and relatedness on increment-defined integration might not appear with adjective-noun pairs.

Comparable depression would not be expected here for the integration measure based on conditionalized free recall of pairs because the relevant experiments cited above suggest that free recall is generally unaffected by the difference in pair type. If true here, overall integration scores should be comparable for our adjective-noun pairs and the noun-noun pairs used by Paivio et al. (1994); even so, interactive rather than additive effects of concreteness and relatedness could still occur here.

We report two experiments in which participants read phrases or sentences containing adjective-noun pairs that vary in rated concreteness and intrapair relatedness.’ Presentation was followed by unannounced free recall or cued recall tests. The predictions from dual coding theory were that concreteness and relatedness would have independent and additive effects on both types of recall. Additivity of the two variables is also predicted for integrative memory, with the possible qualification suggested above for the increment measure of integration. Word pairs were used in Experiment 1, and sentences in Experiment 2.

Experiment 1

METHOD

We first constructed related and unrelated adjective-noun pairs that were either concrete (high in imagery value) or abstract (low in imagery). We used rating data to obtain pairs that differed as much as possible on each variable while controlling word familiarity. Lists that were mixed on the two variables were prepared from these materials and then used in free recall and cued recall tests.

Participants. The volunteer participants were either psychology graduate students or undergraduate students (the latter volunteering for course credit) from three Ontario universities: McMaster, Wilfrid Laurier, and the University of Western Ontario. Normative ratings on imagery and relatedness were provided by 23 graduate and 20 undergraduate students. Sixty-four introductory psychology students participated in the memory experiment.

Materials. We aimed to construct adjective-noun pair lists comparable in length to those used in prior related experiments (e.g., list length ranged from 24 to 40 pairs in experiments by Begg, 1972; Marschark & Hunt, 1989; and Paivio et al., 1994), while maximizing the range of variation on imagery value and relatedness of familiar words. This required an iterative procedure in which we (a) constructed tentative pair candidates from a large pool of words, (b) had the pairs rated on the relevant attributes by groups of participants, and (c) used these ratings to construct shorter lists that were again rated by different groups. We ended up with 28 related and 28 unrelated pairs, 14 of each relatedness level being concrete and 14 abstract. Appendix A lists the target pairs and their rated attributes. The following is a detailed description of the construction procedure.

In stage 1, we selected a set of 975 nouns and 273 adjectives from the Francis and Kucera (1982) word frequency norms, where they were identified as either nouns or adjectives. Imagery ratings for the entire set of 273 adjectives were obtained from unpublished norms (Paivio, 1974). Imagery values for the nouns were obtained from two sources: imagery ratings for 907 of the nouns were available in Paivio, Yuille, and Madigan (1968), and ratings for the remaining 68 nouns were obtained from 16 subjects using instructions from Paivio et al. (1968). Five concrete and five abstract nouns from the published norms were also included as fillers to provide a reliability check. The correlation between the Paivio et al. imagery values and the present imagery ratings for the 10 words was .99.

In stage 2, we formed 24 concrete related pairs by combining nouns and adjectives that are relatively high in imagery (mean item rating >= 4.5 out of 7), and 24 abstract related pairs by combining adjectives and nouns that were relatively low in imagery (

The two relatedness-counterbalanced mixed lists constituted the experimental materials. Each list contained seven related and seven unrelated concrete pairs and an equal number of abstract related and unrelated pairs. The lists were taped on a videocassette and presented in uppercase on an Apple IIe computer at a rate of 8 s per pair (cf. Marschark & Hunt, 1989; Paivio et al., 1994). For the cued recall test, the 28 adjectives were typed one per line on a sheet of paper. A blank sheet was used for the free recall test. A page containing instructions for the appropriate memory test was attached to each sheet.

Procedure. Each of the two basic lists was presented to 32 participants, who were tested in groups of 4 to 8. The participants were told simply to read silently the word pairs that would appear on the TV screen. Immediately following the presentation of pairs, half of the participants in each group received a free recall booklet and were instructed to write down as many of the pairs or individual words as they could remember. The other half received a cued recall test booklet and were told to write down, beside each adjective, the noun that had been paired with it. They were given a maximum of 8 min for the memory test.

RESULTS AND DISCUSSION

The results differ from the Paivio et al. (1994) study in that integrative memory as defined by the increment from free to cued recall was reliably augmented by concreteness and (independently) by pair relatedness in the case of nounnoun pairs, whereas here the concreteness effect fell short of significance in the case of adjective-noun pairs. Moreover,the integration scores for all four pair-attribute combinations are much lower for adjective-noun than noun-noun pairs. In contrast to these differences, strong integration effects were confirmed here for both concreteness and relatedness when integration was measured by the conditional probability of pair recall in the free recall task.

The variations in integration results across experiments could be descriptively explained if only cued recall were selectively depressed with adjective-noun pairs in the manner suggested by the early studies cited in the introduction. The expected pattern is confirmed by a comparison of the data of the present experiment in Table 1; and in Table 1 in Paivio et al. (1994), which also are based on lists in which the pair attributes were mixed. Cued recall proportions for all combinations of concreteness and relatedness are consistently lower for adjective-noun than noun-noun pairs, with the overall proportions being .40 and .53, respectively. Free recall, however, is at least as high for adjective-noun as noun-noun pairs, the overall proportions being .28 and .27, respectively. Consequently, the free-tocued increment scores are necessarily lower for adjectivenoun than noun-noun pairs. Experiment 2 provides further evidence on this descriptive interpretation.

Experiment 2

Experiment 2 extended Experiment 1 by using complete sentences rather than adjective-noun word pairs. Thus, it tested the generalizability of previous results and interpretations with materials that are grammatically more complex, though still simple enough to permit us to use the same memory tests as before.

METHOD

General Discussion

We discuss the results in relation to the issues that distinguish dual coding theory and Marschark and Hunt’s (1989) theory, Begg’s (1972) results, and the variations in the pattern of effects across experiments.

Marschark and Hunt (1989) argued that concreteness effects are dependent on relational processing during encoding and retrieval, so that a concreteness advantage will be absent or attenuated when materials or experimental conditions minimize the opportunity for relational processing. The implications for the present experiments are that concreteness effects should be greater for related than unrelated pairs, especially in cued recall because it requires relational processing at retrieval. In contrast, Paivio et al. (1994) reasoned from dual coding theory that imagery processes primarily account for concreteness effects and that these are independent of relational processes, so that concreteness and relational variables will have independent and additive effects in both recall tasks but with the effect of each variable being stronger in cued recall than in free recall. Our results – like the earlier results with noun-noun pairs – generally supported dual coding theory, with the qualification that relatedness did not add significantly to the strong effect of concreteness on free recall in Experiment 2 of the present study.

It is especially notable that the recall advantage of concreteness remained highly reliable and substantial for unrelated as well as related pairs, in both recall tasks, and for both phrases and sentences. For example, in free recall, the ratio advantage of concrete over abstract nouns in the two experiments averaged 3:1 for unrelated pairs and 2:1 for related pairs. Similarly, in cued recall, the average ratio advantage of concrete over abstract nouns was 4:1 for unrelated pairs and slightly more than 2:1 for related pairs. Thus, the concreteness advantage was at least as great with unrelated as with related pairs in both tasks, which is inconsistent with Marschark and Hunt’s dependency hypothesis. The free recall results are specifically at odds with their suggestion (Marschark & Hunt, 1989, p. 712) that the concreteness advantage obtained by Begg (1972) in free recall could have been due to his use of meaningful adjectivenoun phrases, which prompted relational processing even in the absence of relational suing. That suggestion is refuted by our finding of strong positive effects of concreteness on free recall with unrelated as well as related adjective-noun pairs in both experiments.

The integration effects are more problematic theoretically because they depended on how integration is defined. We first address the conditionalized free recall results because their interpretation is relatively straightforward and the analysis also bears on the free recall results already discussed. Free recall integration benefited strongly from concreteness and relatedness in both experiments. Except for a slight (probably artifactual) interaction in Experiment 1 and an uninterpretable interaction involving list in Experiment 2, the effects of the two variables were generally independent and additive, just as they were in Paivio et al. (1994).

The free recall integration results can be explained as follows: The conditionalized measure highlights recall of intact pairs relative to isolated words. In the absence of explicit retrieval cues, pair recall will be facilitated by factors that promote unitization of each pair. Theoretically, both imagery (cf. Begg, 1972) and relatedness (Paivio et al., 1988) of pair members are such unitizing factors. This interpretation of conditionalized pair recall is reinforced and clarified by separate analyses of the components of the conditionalized scores, namely recall of intact pairs and of unpaired words. Pair recall was independently facilitated by both concreteness and relatedness (p

We turn finally to the problematic results for integration as measured by the increment from free recall to cued recall. In contrast with the pair recall measures, neither of the two experiments showed the independent and additive effects of concreteness and relatedness that were expected from dual coding theory and were obtained by Paivio et al. (1994) with noun pairs. Our experiments yielded significant increment (integration) scores only for concrete related pairs, but even in that case the increment scores (averaging .21) were much lower than for the corresponding noun pairs (averaging .42) in the earlier experiments.

These discrepancies reflect the fact that cued recall of responses from adjective-noun pairs was consistently lower and by comparable magnitudes in both experiments than cued recall of responses from noun-noun pairs in Paivio et al. (1994). By contrast, this was not the case with free recall. Moreover, these contrasting patterns for the two types of materials and recall tasks are completely consistent with the patterns that emerged from comparisons of different experiments in the 1960s and early 1970s as described in our introduction. Thus, the absent or reduced increment-defined memory integration for adjective-noun pairs as compared to noun-noun pairs is entirely an artifact of reduced noun recall when cued by adjectives as compared to nouns.

The conceptual peg hypothesis provides a theoretical explanation of the difference. Among item attributes, imagery value is the best predictor of word recall, especially when varied among items that serve as retrieval cues in associative memory tasks. When varied over the same range, noun imagery and adjective imagery have comparable effects on response recall but imagery is generally higher for nouns than adjectives. This difference accounts largely for the consistent superiority of nouns as retrieval cues for response items (Paivio, 1971, pp. 272-276).

The analysis is applicable here as well. Averaging over concrete and abstract words, the nouns that served as stimuli in the experiments by Paivio et al. (1994, Appendix) have median imagery values of 4.6 as compared to 3.6 for the adjectives in the present experiments (Appendix A). The difference could account for the cued recall difference and, accordingly, it could also account for the increment-defined integration difference given that noun-noun and adjectivenoun pairs did not differ in overall free recall of the response nouns.

Our interpretations of the different ways in which imagery and relatedness affected the two measures of integration does not mean that the free recall of pairs is superior to the increment score as a measure of integration. The differences simply show that. integration is measuredependent, as are other inferential psychological concepts.

We have presented new data showing once more that dual coding theory provides a better explanation of concreteness effects on memory than does relational/distinctiveness processing theory. This should not be taken as a general comment on the latter theory, as it is applicable to other memory tasks in which it has been shown to have predictive and explanatory value (e.g., Begg & Nicholson, 1994; Einstein & Hunt, 1980). Our results and interpretations speak instead to the specific problem addressed by Marschark and Hunt (1989), namely the role of imagery in associative and item memory. We conclude that imagery enhances both distinctiveness (e.g., itemspecific recognition) and relational processing. Both types of processing can be accomplished in nonimaginal ways, but the distinction cannot explain imagery effects. Apparently, we need both dual-relational and dual-distinctiveness information for a full capture of imaginal versus other bases for the two types of information.

This research was supported by Operating Grants OGP0004866 to Allan Paivio and OGP0008122 to Ian Begg from the Natural Sciences and Engineering Research Council of Canada. We thank Ann Anas and Neil Khanna for help in testing participants.

Correspondence concerning this article should be addressed to Allan Paivio, Department of Psychology, University of Western Ontario, London, Ontario N6A 5C2 (E-mail: apaivio@julian.uwo.ca).

In a third experiment using the same adjective-noun pairs with conditions varied between lists, the basic pattern of results was the same as is the two reported ones.

References

Anderson, R. C., & Ortony, A. (1975). On putting apples into bottles – A problem in polysemy. Cognitive Psychology, 7, 167-180.

Begg, I. (1972). Recall of meaningful phrases. Journal of Verbal Learning and Verbal Behavior, 11, 431-439.

Begg, L, & Nicholson, S. (1994). Semantic and episodic relations are experimentally dissociable and stochastically independent. Canadian Journal of Experimental Psychology, 48, 399-417.

Einstein, G. O., & Hunt, R. R. (1980). Additive effects of individual-item and relational processing. Journal of Experimental Psychology: Human Learning and Memory, 6, 588-598.

Foss, D. J., & Harwood, D. A. (1975). Memory for sentences: Implications for human associative memory. Journal of herbal Learning and Verbal Behavior, 14, 1-16.

Francis, W. N., & Kucera, H. (1982). Frequency analysis of English usage: Lexicon and grammar. Boston, MA: Houghton Mifflin Company.

Marschark, M., & Hunt, R. R. (1989). A reexamination of the role of imagery in learning and memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, IS, 710-720.

Marschark, M., & Paivio, A. (1977). Integrative processing of concrete and abstract sentences. Journal of Verbal Learning and Verbal Behavior, 16, 217-231.

Paivio, A. (1971). Imagery and verbal processes. New York: Holt, Rinehart & Winston. Reprinted 1979, Hillsdale, NJ: Erlbaum.

Paivio, A. (1974). Imagery and familiarity ratings for 2311 words. Unpublished manuscript, University of Western Ontario, London, Ontario, Canada.

Paivio, A. (1991). Dual coding theory: Retrospect and current status. Canadian Journal of Psychology, 45, 255-287.

Paivio, A., Clark, J. M., & Khan, M. (1988). Effects of concreteness and semantic relatedness on composite imagery ratings and cued recall. Memory & Cognition, 16, 422-430.

Paivio, A., Smythe, P. C., & Yuille, J. C. (1968). Imagery versus meaningfulness of nouns in paired associate learning. Canadian Journal of Psychology, 22, 427-441.

Paivio, A., Walsh, M., & Bons, T. (1994). Concreteness effects on memory; When and why? Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 1196-1204.

Paivio, A., Yuille, J. C., & Madigan, S. A. (1968). Concreteness, imagery, and meaningfulness values for 925 nouns. Journal of Experimental Psychology Monograph Supplement, 76 (1, Pt. 2).

Philipchalk, R. P., & Begg, I. (1971). Context concreteness and form class in the retention of nonsense syllables. Journal of Verbal Learning and Verbal Behavior, 10, 499-505.

Yuille, J. C., Paivio, A., & Lambert, W. E. (1969). Noun and adjective imagery and order in paired-associate learning by French and English subjects. Canadian Journal of Psychology, 23, 459-466.

Manuscript accepted September 7, 1999

Sommaire

En elargissant des travaux anterieurs sur la question, nous aeons mis a l’epreuve d’autres hypotheses quant aux effets que la concretude et le degre de connexion daps des couples adjectif nom exercent sur le rappel fibre, le rappel indice et des mesures de rappel integratif. Selon la theorie du double encodage, routes ces mesures du rappel des noms devraient reveler l’independance (l’additivite) de refs effets. Les predictions decoulant toutefois de la theorie du traitement relationnel/distinctif soar que l’interaction des deux variables entrainerait une plus grande efficacite de la concretude dans les couples avec connexion que dans ceux sans connexion, surtout en rappel indice. Dans deux experiences, les caracteristiques des couples oat varie a l’interieur de listes ou de phrases. En accord avec a la fois les predictions fondees sur la theorie du double encodage et de precedentes donnees issues de couples nom-nom, les deux experiences montrent que le rappel fibre et le rappel indice soar fortement influences par la concretude, independamment du degre de connexion. Tandis que les effets de ce degre soar en general positifs, ceux sur le rappel libre ne soar pas significatifs.

Les deux variables ont aussi des effets independants sur la memoire integrative telle que mesuree par une difference positive entre rappel fibre et rappel indice. D’apres cette mesure, l’integration ne se produit systematiquement que pour les couples a niveau eleve de concretude et de connexion; ce patron etait absent des donnees obtenues a (aide de couples nom nom puisque les deux variables avaient alors des effets positifs independants sur la meme mesure d’integration. Pour rendre compte de maniere descriptive des differents profils des effets d’integration degages par l’ensemble des mesures a travers les deux experiences, if est possible d’invoquer le fait que les adjectifs sont generalement moins efficaces que les noms lorsqu’ils servent de stimuli pour les noms en rappel indice, mais pas en rappel fibre une difference qui n’a selectivement affecte que la mesure d’integration. L’efficacite differentielle des adjectifs et des noms ‘a titre d’indice s’accorde avec l’hypothese d’un support conceptuel assure par l’imagerie, dans le cadre de la theorie du double encodage. Les presents resultats sont donc juges davantage conformer a cette theorie qu’a celle centree sur un traitement relationnel/distinctif.

ALLAN PAIVIO and MUSTAQ KHAN, University of Western Ontario IAN BEGG, McMaster University

Copyright Canadian Psychological Association Sep 2000

Provided by ProQuest Information and Learning Company. All rights Reserved