The distinction between implicit and explicit tests of memory has given rise to much recent experimental research and theoretical speculation (Moscovitch, Goshen-Gottstein, & Vriezen, 1994; Moscovitch, Vriezen, & Goshen-Gottstein, 1993; Richardson-Klavehn & Bjork, 1988; Roediger & McDermott, 1993; Schacter, 1987). In explicit tasks such as free recall, cued recall, and recognition memory, the respondent consciously attempts to remember the wanted information, whereas in implicit tasks such as perceptual identification, word-fragment completion, and word-stem completion, previously presented information affects performance on the task although the respondent may be unaware that memory is being assessed. One major source of evidence for the distinction comes from experiments showing that different types of information affect performance on implicit and explicit memory tasks. Whereas explicit tests are affected primarily by the degree of semantic elaboration induced at the time of encoding (e.g., Craik & Tulving, 1975), implicit tests are little affected by level-of-processing or other semantic manipulations (Graf & Mandler, 1984; Graf, Mandler, & Haden, 1982; Jacoby & Dallas, 1981). On the other hand, implicit tests appear to be heavily dependent on the similarity of surface information between study and test conditions. For example, Jacoby and Dallas (1981) found less priming on a visual word identification test if the studied words were presented auditorily as opposed to visually in the learning phase. Similarly, Graf, Shimamura, and Squire (1985) showed that word-stem completion was negatively affected by presenting words for study auditorily rather than visually; the same result was reported for wordfragment completion by Roediger and Blaxton (1987a). In a review of several such experiments, Kirsner and Dunn (1985) delineated the circumstances under which the perceptual record of past experience persists to affect current memory performance.