2.16 Retrieval Processes in Memory

words (only verbal code) (Paivio et al., 1968; Paivio, 1969). Yet, in a study from our lab (Hamilton and Rajaram, 2001) we found that on implicit tests such as word fragment completion (_ t r _ _ b _ r _ _) and implicit general knowledge test (what fruit wears its seeds on its skin?), there was an equivalent 0 0.1 0.2 0.3 0.4 0.5 0.6 Free recall Explicit GK Implicit GK Type of task P ro po rt io n co rr ec t Implicit WFC Explicit WFC Shallow New Deep Figure 1 Mean proportion of correct responses for each item type for each of the five tasks used (GK, general knowledge task; WFC, word fragment completion task). Adapted from Experiment 2 in Hamilton M and Rajaram S (2001) The concreteness effect in implicit and explicit memory tests. J. Mem. Lang. 44(1): 96–117. Retrieval Processes in Memory 269 advantage for producing studied concrete words and studied abstract words over their nonstudied counterparts (Figure 1). In other words, on an implicit memory task there was no memorial benefit of concrete words over abstract words. Therefore, the concrete/abstract distinction in memory cannot be discussed only in the context of differential encoding or storage. This distinction demands a more complex explanation because this effect is not ubiquitous – it can be specific to a particular mode of retrieval. A converse pattern can also emerge by changing the retrieval mode such that some study differences do not affect explicit memory but produce changes in implicit memory. For example, changing the modality 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24 0.26 Fragment completion Stem completion Test R e la tiv e p ri m in g Visual encoding Audito Figure 2 Relative priming for the four implicit memory tasks ac and Roediger HL (1993) Direct comparison of four implicit memo of presentation at study – presenting words either in the auditory or the visual modality – does not change the level of free recall (Blaxton, 1989; Srinivas and Roediger, 1990; Rajaram and Roediger, 1993), and this null finding suggests that modality of presentation at study does not matter. However, this conclusion is only partly correct as we (Rajaram and Roediger, 1993) found in our study with four different implicit tasks involving perceptual cues (see Figure 2). When subjects were presented with impoverished cues such as word fragments, word stems, rapidly presented words in the word identification task, or anagrams (brtaserwyr) to solve in the anagram solution task (strawberry) and were asked to perform these tasks with the first solution that comes to mind, performance improved on these implicit tests if the study and test materials were presented in the same modality compared to different modalities (see also Jacoby and Dallas, 1981; Kirsner et al., 1983; Graf and Mandler, 1984; Roediger and Blaxton, 1987; Blaxton, 1989; Srinivas and Roediger, 1990; Weldon, 1991). In other words, the impact of a study variable is sometimes detectable only when subjects used the implicit retrieval mode. (As an aside, but consistent with the general argument about the impact of retrieval cues in modulating memory performance, we also found that studied pictures produced the worst performance on these implicit tasks that provided word-based cues. This, of course, is contrary to the pattern that is typically observed in free recall and recognition, where memory for pictures is better than that for words (Paivio et al., 1968; Madigan, 1983; Weldon and Roediger, 1987; Rajaram, 1993).) Anagram solution Word identification conditions ry encoding Pictorial encoding ross different encoding conditions. Adapted from Rajaram S ry tests. J. Exp. Psychol. Learn. Mem. Cogn. 19(4): 765–776. 270 Retrieval Processes in Memory The impact of retrieval mode on memory performance can be seen even when the retrieval cues themselves are held constant and only the mode of retrieval is varied through instructions, or when the retrieval intentionality criterion is met by the experimental conditions (Schacter et al., 1989). For example, when performance on free recall is contrasted with performance on a task such as implicit word fragment completion, the retrieval mode changes, but so do the test cues (no test cues in free recall and perceptually degraded cues in fragment completion, e.g., _ t r _ _ b _ r _ _). But the role of retrieval mode in detecting memory would be more persuasive if dissociations between explicit and implicit memory could be observed even when the same test cues were used in both conditions. Graf and Mandler (1984) reported such a dissociation between an explicit word stem cued-recall task and an explicit word stem completion task where the same test cues (e.g., ele_______) were used, and the levels of processing effect occurred on the explicit but not the implicit version of the task. In a largescale study, Roediger et al. (1992) reported similar patterns of performance when they contrasted explicit and implicit versions of the test using identical stem cues (e.g., ele_______) as well as explicit and implicit versions of the test using identical fragment cues (e.g., _ l _ p _ a n _). Along the same lines, in our study with concrete and abstract words we just described, we found that the presence of the concreteness effect in free recall also extended to the explicit retrieval version of the general knowledge test (complete the following question with a studied word: ‘‘What fruit wears its seeds on its skin?’’) but, as noted earlier, not to the implicit version of the general knowledge test (Hamilton and Rajaram, 2001). In other words, we demonstrated dissociative effects on explicit and implicit versions of a different type of task, namely general knowledge, while holding the test cues constant (see again Figure 1). 2.16.4.2 Differentiating between Conceptual and Perceptual Retrieval Cues While explicit and implicit retrieval modes can produce various dissociations of theoretical significance such as the ones just described, dissociations can also occur within one retrieval mode – for example, with differences in implicit retrieval cues provided to the participants. In previous sections, we described how differences in retrieval cues can bring about changes in memory performance within the context of explicit memory tasks such as free recall, cued recall, and recognition. The tenets of the transfer-appropriate processing framework (Roediger et al., 1989; Roediger, 1990) predict systematic differences even within implicit memory tasks, depending on the type of process demanded by the tasks. If the implicit memory task largely depends on perceptual processes for its successful completion – as in the cases of word fragment completion, word stem completion, perceptual identification tasks – then encoding orientation produces one type of effect. But if the implicit task mainly relies on conceptual processes – as in the cases of the general knowledge test described earlier and also tests such as implicit category association test (given a category name, participants are asked to produce all the exemplars that come to mind within 30 s) – then the same encoding orientation produces the opposite effect. So, for example, the modality effect observed across different perceptual implicit tasks described in the study by Rajaram and Roediger (1993) disappears on conceptual implicit tasks because changes in modality of presentation are not important for accessing meaning (e.g., Blaxton, 1989; Srinivas and Roediger, 1990). Encoding variables that differentiate the extent to which meaning (rather than surface information as in the modality manipulation) is varied during study produce the opposite effect: As we described earlier in this section, the levels-ofprocessing effect that reflects changes in conceptual processing of information disappears on perceptual implicit tasks. But this effect is reliably observed on conceptual implicit tasks such as the implicit general knowledge test (Hamilton and Rajaram, 2001) and the implicit category association test (Hamann, 1990; Srinivas and Roediger, 1990). Together, these patterns show that, while retrieval modes of explicit and implicit retrieval change accessibility in dramatic ways, the nature of the process demanded by the retrieval task (conceptual versus perceptual) can also change accessibility in a manner that is powerful and that can be orthogonal to the retrieval