for their helpful comments on the dissertation work. We thank C. MacLeod, H. Roediger, E. Tulving, J. Gardiner, and two anonymous reviewers for helpful comments on earlier versions of this article.

In the reported experiments, the spacing of repetitions improved performance on cued-memory tests (a frequency judgment test and graphemic cued-recall test) when items were studied in an intentional or an incidental-semantic condition but not in an incidental-graphemic study condition. The results imply that spacing effects on cued tests depend on level of processing carried out on the stimuli rather than on intentionality of learning per se, as suggested by R. L. Greene (1989Greene ( , 1990. The findings undermine a voluntary rehearsal account of spacing effects in cued tests. Alternative accounts are discussed, including the view that involuntary processes akin to semantic (or lexical) priming play a critical role in spacing effects on cued-memory tests.

In three experiments, we examined why pictures are remembered better than words on explicit memory tests like recall and recognition, whereas words produce more priming than pictures on some implicit tests, such as word-fragment and word-stem completion (e.g., completing -l_plLJit or ele as elephant). One possibility is that pictures are always more accessible than words if subjects are given explicit retrieval instructions. An alternative possibility is that the properties of the retrieval cues themselves constrain the retrieval processes engaged; word fragments might induce data-driven (perceptually based) retrieval, which favors words regardless of the retrieval instructions. Experiment 1 demonstrated that words were remembered better than pictures on both the word-fragment and word-stem completion tasks under both implicit and explicit retrieval conditions. In Experiment 2, pictures were recalled better than words with semantically related extralist cues. In Experiment 3, when semantic cues were combined with word fragments, pictures and words were recalled equally well under explicit retrieval conditions, but words were superior to pictures under implicit instructions. Thus, the inherently datalimited properties of fragmented words limit their use in accessing conceptual codes. Overall, the results indicate that retrieval operations are largely determined by properties of the retrieval cues under both implicit and explicit retrieval conditions. Generally, people remember pictures better than words (e.g., Paivio, 1971;Paivio, Rogers, & Smythe, 1968;Shepard, 1967). Although the basis of the picture superiority effect is not entirely clear, several aspects of encoding have been identified as causal. For example, the dual coding hypothesis (Paivio, 1971(Paivio, , 1983(Paivio, , 1986 posits that under ordinary circumstances, pictures are likely to become redundantly encoded because they evoke both imaginal and verbal codes, whereas words are likely to evoke only verbal codes. The sensory-semantic model (Nelson, 1979) implicates the role of stimulus surface features: pictures provide more distinctive visual representations than do words, and hence they are more memorable. Other researchers (lntraub & Nicklos, 1985;Nelson, Reed, & McEvoy, 1977;Smith & Magee, 1980) have suggested that pictures naturally engage deeper' 'levels of processing" (Craik & Lockhart, 1972) than do words. Experimental evidence includes the findings that pictures can be categorized faster than words (Potter & Faulconer, 1975;Smith & Magee, 1980), and that words processed semantically or elaboratively are remembered as well as pictures (Durso & Johnson, 1980). Despite the differences in the specific mechanisms postulated in the various theories of the picture superiority effect, all the theories share Preparation of this article was supported in part by Grant HD15054-07 from the National Institute of Child Health and Human Development. We thank Douglas Nelson for helpful comments on an earlier draft. The second author is now at Rice Universit...

Experiments are reported in which effects of repeating words exactly (e.g., elephant, elephant) or repeating some meaningful aspect-a synonym (pachyderm), an associate (tusk), or a category coordinate (hippopotamus)-were examined on free recall and word-fragment completion. In free recall, large effects of both exact repetition and conceptual repetition were found; the magnitude of the latter was about half that of the former. In contrast, in primed word-fragment completion, repetition effects were rather small and there was no evidence of indirect (or conceptual) priming. Also, presentation of synonyms, associates, and coordinates in isolation failed to prime word-fragment completion. The results provide further evidence that the basis of primed word-fragment completion is different from that of free recall; the former seems to have a perceptual (or perhaps lexical) basis, whereas the latter relies more on meaningful processing.

Subjects saw or heard words presented once, or repeated 4 or 16 times in massed fashion, and then received an implicit or explicit memory test. Massed repetition did not increase priming on word fragment completion beyond that obtained from a single presentation but did enhance performance on various explicit tests (free recall, recognition, question cued recall, and word fragment cued recall) and an implicit general knowledge test. Modality of presentation affected implicit and explicit word fragment cued tests but did not affect performance on any of the other tests. Levels of processing affected performance on implicit and explicit question cued tests. These results are consistent with a transfer appropriate processing account of dissociations among memory measures and imply that massed repetition promotes conceptual processing but does not entail a repetition of perceptual-based processes responsible for priming on word fragment completion.

No abstract

Subjects saw or heard words in a list (e.g. limerick) and then took two successive tests. The first was a yes/no recognition test in which auditory/visual modality of test words was manipulated orthogonally to the study modality. The second test varied with experimental conditions: subjects produced words to either perceptual (fragment) cues (l- -e-ick) or conceptual cues (What name is given to a lighthearted five-line poem?), under either explicit or implicit retrieval instructions. The major findings were: (a) that regardless of the type of retrieval cue (perceptual or conceptual) the degree of dependency between recognition and cued recall was greater than that between recognition and implicit retrieval; and (b) that modality shifts adversely affected perceptually cued explicit and implicit retrieval, whereas they had no effect either on conceptually cued retrieval or on recognition. These results suggest that the memory system subserving, and the processes involved in, conceptual priming differ from those underlying recognition and perceptual priming.

A number of predictions derived from Bower's (1981) associative network model of mood and memory were investigated by inducing elated, neutral, or depressed moods in different groups of subjects. A mood-evaluation checklist verified the success of the mood-induction procedure. While in induced moods, subjects performed a lexical decision task in which the target stimuli were positive or negative trait adjectives, neutral nontrait adjectives, or nonwords. As predicted, subjects in elated moods showed facilitated processing of positive target words relative to processing of negative and neutral target words. Contrary to prediction, subjects in depressed moods did not respond faster to negative target words than they did to positive target words. Following the lexical decision task, subjects were administered an incidental free recall test for the words presented during the lexical decision task. Contrary to prediction, the three mood groups recalled similar numbers of words for each type of word target category. The asymmetric effects of elation and depression on memory that were observed in this study and a number of other studies are discussed.Fundamental to the predictive validity of Bower's (1981) associative network model of memory and emotion are two assumptions. The first is that specific emotions are represented as separate nodes in the same associative network as are words, concepts, and propositions of a verbal symbolic nature. The second assumption is that activation of an emotion node, by whatever means, will serve to lower the threshold of excitation of an associatively linked node via the process of spreading activation. This activation should increase the likelihood that related information would become more accessible to consciousness than it otherwise would, or at least become more accessible than information not associatively linked to the emotion node .