Correlations between imagery and memory across stimuli and across subjects

Bulletin of the Psychonomic Society 1979. Vol. 14 (5), 368-370 Correlations between imagery and memory across stimuli and across subjects JOHN T. E. RICHARDSON Brunei University, Uxbridge, Middlesex UB83PH, England Stimulus imageability has been reliably established as an excellent predictor of memory performance. Variation among subjects in evoked mental imagery shows no such correlation with recall. Two experiments showed these generalizations to hold even when correlations across stimuli and across subjects were computed from the same data. It was concluded that ratings of the subjective vividness of evoked mental imagery are not valid indexes for research on individual differences in memory performance. Under a wide variety of experimental conditions, it has been reliably established that the image-arousing potential of the stimulus material is highly correlated with its memorability. Indeed, Paivio (1971, Chapter 7) concluded that imageability was the best predictor of performance of all the stimulus attributes investigated by experimental psychologists. This is in marked contrast to the failure of studies of the variation among individual subjects to demonstrate a reliable correlation between evoked mental imagery and memory performance. In this paper, I shall attempt to resolve this fundamental discrepancy in the experimental evidence for the importance of mental imagery as a psychological process. Stimulus imageability is measured by the use of questionnaires administered to large groups of subjects, in which they are asked to evaluate items along 7-point rating scales. Such questionnaires customarily employ instructions similar to those used by Paivio, Yuille, and Madigan (1968), who defined an item's imagery value as the ease with which it aroused a mental image. This value is computed by taking the average rating across the subjects employed in the rating task. The imagery ability of an individual subject may be measured either by the use of questionnaires on the subjective vividness of experienced imagery or by tests of spatial manipulation ability (Ernest, 1977). When careful experimental procedures are employed to avoid experimenter bias and other sources of variation in demand characteristics, neither method shows any correlation between imagery ability and memory performance (Berger & Gaunitz, 1977 ; Richardson, 1978a). This is puzzling, since the questionnaires on the vividness of experienced imagery are at least superficially quite similar to the rating tasks used to measure stimulus imageability . The subjects are once again asked to evaluate the ease with which various items arouse vivid mental imagery, and a subject's imagery ability is measured by taking the average rating across the stimuli employed (e.g., Sheehan, 1967) . One possible explanation is that the apparent similarity between the two tasks hides crucial procedural differences that prevent variation among subjects in evoked mental imagery from having predictive capacity in investigations of human memory . This can be easily tested by obtaining measures of stimulus imageability and subject imagery ability within the same experimental situation and considering the correlations between these measures and subsequent memory performance . If both indexes are valid measures of variation in evoked mental imagery, they should be equally successful in predicting recall. This idea is tested in Experiment 1. On the other hand, a more radical suggestion may be made. It could be argued that mental imagery is essentially private and solipsistic (e.g., Doob, 1972). In particular, it is likely that subjects lack a definite origin and a defmite unit of measurement for making these subjective judgments, and that, in consequence, their appreciation of the vividness of mental imagery can be taken to define at best an ordinal scale. It is possible that such judgments are adequate to ensure that relatively gross comparisons among different stimulus items predict variations in the memorability of those items. Nevertheless, since there is no unique mapping of this ordinal subjective scale of vividness onto the 7-point rating scale (which may be variously interpreted as representing either an interval scale or a ratio scale), the absolute values assigned by a subject to a given stimulus will have no meaning , and comparisons among different subjects in terms of their average ratings will not predict variations in memory performance. There fore, even if measures of stimulus imageability and subject imagery ability are obtained within the same experi mental situation, the former may predict recall perfor mance, but the latter should not. Requests for reprints should be sent to Dr. J . T. E Richardson, Department of Psychology, Brunei University, Uxbridge, Middlesex UB8 3PH, England. Method Copyright 1979 Psychonomic Society, Inc. EXPERIMENT 1 Twenty-seven students participating in an introductory course in laboratory methods at Brunei University were asked to 368 0090-5054/79/110368-03$00.55/0 CORRELATIONS BETWEEN IMAGERY AND MEMORY evaluate 40 English nouns with frequencies of AA or A in the count of Thorndike and Lorge (944). The items were presented for rating on imageability following the procedure of Paivio, Yuille, and Madigan (1968) . The subjects received a response booklet, the first page of which contained instructions for giving imagery ratings, with four additional nouns as examples. The second page contained 40 7-point rating scales. The ends of each scale were labeled "low imagery" and "high imagery," and each scale was numbered from I to 40. When all of the subjects had read the instructions, they turned to the second page of the response booklet. The words were read aloud by the experimenter at a rate of one item every 10 sec, and the subjects responded by rating each item on the 7-point scale. After they had rated the last item, the subjects received an unanticipated recall test on the 40 items. They received standard free recall instructions, their responses were written on the reverse of the response booklet, and they were allowed 3 min for recall. Results Data from the first five items and the last five items were ignored to avoid primacy and recency effects in recall. Thus, rating and recall data were employed from 30 items and from 27 subjects. The average recall per subject was 9.81 items, and the average imagery rating was 4.55. The first analysis considered the effect of stimulus image ability upon recall. The correlation between the mean imagery rating assigned to each item and the proportion of subjects who recalled that item was +.28. This approached, but did not attain , statistical significance by a one-tailed t test [t(28) = 1.56, .1 > P > .0 5] . The second analysis considered the effect of the subjects' imagery ability upon recall. The correlation between the mean imagery rating given by each subject and the proportion of items recalled by that subject was - .07. A on (...truncated)