This study was designed to investigate the effects of retrieval on memory for prose in second-graders and adults. Specifically, this study assessed the effects of retrieving an interleaved story in canonical form on the subsequent retrieval of that story in its originally presented interleaved form. Subjects were randomly assigned to one of five conditions. The critical comparisons were between the invalidation condition and its control condition. In the invalidation condition, the experimenter said that the initial instructions to recall the story were incorrect and that the subjects should now recall the story exactly as presented. Subjects in the control condition were asked to recall the story exactly as presented after recalling the story canonically. The results indicated that the nature of recall was greatly influenced by the demands of the retrieval task. Second-graders and adults made more theme-irrelevant elaborations in the invalidation condition than in the control condition. Second-graders' recall protocol was more congruent with the most recent recall instructions when their former recall instructions were invalidated than in the control condition. In contrast, adults' recall protocol was congruent with the most recent recall instructions whether or not their initial recall instructions were invalidated. Congruence was measured by episode clustering (ARC) and by input-output Spearman Rank Order Correlation. These findings suggest that conditions at retrieval and not at encoding appear to determine what form recall will take. Further, the findings of this study suggest that second-graders are more likely to internalize experimenter provided retrieval plans than are adults and are more likely to abandon these retrieval plans when provided with negative evidence for their internalized plan. Schmidt and Schmidt (in preparation) have recently argued that recall is influenced by a retrieval plan and that one of the sources through which a retrieval plan can be generated is through the retrieval context. They also argue that the retrieval context should have its most powerful impact when limited or inappropriate retrieval plans are available to the subject. The differences between second-graders and adults reported here serve to support this view. / Ph. D.