Evidence that broader processing facilitates delayed retention

Bulletin of the Psychonomic Society 1977. Vol. 10 (1).28-30 Evidence that broader processing facilitates delayed retention WILLIAM F. BATTIG and GILLES O. EINSTEIN University oj Colorado, Boulder, Colorado 80309 After rating words on one to three of either three bigbly-correlated dimensions jconcreteness, imagery, and categorizability) or three relatively uncorrelated dimensions jconcreteness, pleasantness, and number of word features), 48 college students were tested without warning for 48-h delayed recall and recognition of the 36 words. The latter broad-processing group showed significantly better recognition than the former narrow-processing group. Group differences in recall, however, were nonsignificant, and the narrow group was slightly superior on words rated for all three dimensions. The results indicate delayed recognition, if not recall, to be better for words which have been processed more broadly or elaborately. Memory is further suggested to be more closely dependent upon breadth or elaboration than depth of semantic processing. That memory for verbal materials should be improved if these have undergone broader or more elaborate processing is clearly implied by the recent expansion of the levels-of-processing approach to incorporate elaboration as well as depth of processing as critical determinants of memory (e.g., Craik & Tulving, 1975). With the notable exception of Klein and Saltz (l976), however, there is little evidence for such facilitative effects of more elaborate processing. More specifically, Jenkins and his associates have found no evidence that words processed either by two semantic tasks (Hyde. 1973) or a semantic and nonsemantic task (Walsh & Jenkins, 1973) are recalled any better than words processed with a single semantic task. The semantic-nonsemantic result was also obtained for 24-h delayed recognition by Schwarz (1976). All of these experiments, however, required that both types of processing be performed during a single relatively short (4- to 9-sec) presentation and (except for Schwarz) used very short retention intervals. The present paper reports an exploratory use of quite different and hopefully more appropriate techniques for investigating the effects of breadth of processing upon delayed retention . The major distinguishing features (as described in more detail below) include (l) multiple presentations of the same word, with a different type of processing during each presentation; (2) within-subject variations from one to three types of processing on a given word; (3) a longer (48-h) retention interval followed by both recall and recognition tests; (4) variations in breadth of processing through the use of highly intercorrelated (concreteness, imagery, and categorizability) and relatively uncorrelated rating dimensions (concreteness, pleasantness, and number of word features). After the present data had been collected, Klein and Saltz (l976) reported poorer immediate recall for words processed on two highly correlated than two less correlated semantic dimensions selected from happy-sad, biglittle, pleasant-unpleasant, and fast-slow. Their procedure, however, was quite different in requiring subjects to shift across four to six different types of ratings with only 24 words, with two ratings durfng a single presentation and somewhat variable short retention intervals. Thus, the present experiment represents a significant extension of the paradigm and the results of Klein and Saltz (1976). METHOD Each of 48 students in an introductory research methods course provided data during two classroom sessions 48 h apart. During the fIrst session, all students were given three rating sheets each, with different rating instructions and 24 words to be rated accordingly, so that all 24 words on each sheet were rated for the same one of the three rating dimensions. These 72 ratings on 36 different words (which were identical for all subjects) provided that 12 words each appeared on only 1,2 , or all 3 of these sheets and thus were rated for 1, 2, or 3 different dimensions. During the next class 2 days later, students rust were given (without prior warning) recall sheets for the 36 words they had rated. After completing and turning in the recall sheet , each subject was given a sheet containing all 36 rated words intermixed randomly with 36 distractors (each of which was either semantically or structurally similar to at least one of the 36 rated words). For each of these 72 words, subjects were instructed to circle one of four symbols: Y (quite sure you rated the word), YG (unsure, but best guess is that you rated the word), NG (unsure, but best guess is that you did not rate the word), or N (quite sure you did not rate the word). Number of times (one to three) the word had been rated was also estimated for words eliciting Y or YG responses. Both the recall and recognition tests were subject paced, as were the previous word ratings. This is Publication 68 of the Institute for the Study of Intellectual Behavior, University of Colorado, and was supported by Grant BNS 72-02084 from the National Science Foundation. We thank Lewis O. Harvey, Jr., for making possible the present signal detection (d') analysis of the present data. 28 BROADER PROCESSING F ACILIT ATES DELAYED RETENTION The groups differed only in whether their three rating dimensions were highly interrelated or relatively uncorrelated. The correlations were based upon a large word-scaling project currently nearing completion at Colorado, wherein 2,854 words were rated for each of seven dimensions. Three dimensions (concreteness, imagery , and categorizability) all yielded pairwise 'correlations of at least .88, and therefore were selected as the rating dimensions for the narrow-processing group. Another set of three dimensions (concreteness, pleasantness, and number of features or attributes of the word) were all intercorrelated .39 or less, and thus were selected as rating dimensions for the broad group. Elimination of those subjects who failed to complete all tasks in both sessions left 25 in the broad group and 23 in the narrow group (one of the latter failing to complete the recall task). Within each group, each rating dimension was used for each of the three sets of 12 words with approximately equal frequency and the order in which these dimensions were rated was also counterbalanced across subjects. The 24 words on each rating sheet always included all 12 words being rated for all 3 dimensions, 8 words rated for 2 dimensions (meaning that 4 of these words were rated for each possible combination of 2 dimensions), and 4 words rated only for that single dimension. Although the same 12-word set was always rated 3 times (as well as twice and once), these 3 sets of 12 words were carefully chosen so that each was closely comparable to the other 2 sets in ratings on all 7 dimensions from the aforementioned word norms (including also meaningfulness and familiarity). Moreover, the (...truncated)