Unit analysis of implicit and explicit memory tests
Schacherer, Christopher William
Roediger, Henry L., III
Doctor of Philosophy
The present study compares the cognitive processes underlying two perceptual implicit memory tests (word stem completion and word fragment completion) and four explicit memory tests (word stem cued recall, word fragment cued recall, free recall, and recognition). Like many previous studies (and as is predicted by most current memory theories) manipulation of the level, or depth of cognitive processing engaged during the study phase dissociated the explicit tests from the implicit tests. That is, for the explicit tests, processing the study items under a deep level of processing resulted in a greater number of words being recalled or recognized (compared to performance under a shallow level of processing at study). On the implicit tests, this manipulation had very little effect. These differential effects are often interpreted as evidence that qualitatively different processes underlie performance on implicit and explicit tests. However, in looking at which (instead of how many) items are produced on the tests, the conclusions are somewhat different. In the present study this "unit analysis" approach described by Rubin (1985) showed that: (1) implicit and explicit tests correlated more strongly within stimulus type (stem/fragment) than they did within test type (implicit/explicit), (2) both part-word cued recall tests (word stem and word fragment cued recall) correlated strongly with recognition even though they correlated only modestly with each other, and (3) free recall did not correlate positively with any of the other tests (implicit or explicit). These results are explained in terms of a generate/recognize model that incorporates transfer appropriate processing assumptions. Briefly, it is suggested that the implicit tests and their explicit counterparts involve the same data-limited process, and that recognition is not similarly limited--relying almost exclusively on conceptually driven processes. However, this generate/recognize explanation fails to explain why free recall does not correlate positively with any of the other tests. The failure of free recall to correlate positively with any of the other tests is interpreted as suggesting that free recall may rely on qualitatively different processes.
Experimental psychology; Psychology; Psychometrics; Cognitive psychology