Word frequency and the conceptually-driven/data-driven processing distinction of memory retrieval.

Abstract

The effects of word frequency and two encoding contexts were directly compared using semantic or graphemic cued recall. Subjects generated and read both common and rare words, but were given only one type of cued recall task. semantic cued recall (i.e., a conceptually-driven task) required the recall of previously studied words when given cues that have the same meaning, whereas graphemic cued recall (i.e., a data-driven task) required the use of cues that have a similar physical appearance as previously studied words to aid recall. While generate and read study contexts dissociate conceptually-driven from data-driven tasks (Blaxton, 1989), the current results add that the frequency of stimulus items contributes to this effect. Generated common relative to rare words were recalled better in semantic cued recall. In contrast, graphemic cued recall was better for rare than common words when read without context during encoding. The results further indicate that the semantic similarities typically found between common words is likely to benefit more by conceptual processes. Likewise, when semantic information is not available to guide the recall of rare items in graphemic cued recall, the physical of perceptual features will. The transfer-appropriate processing framework accounts for these results best in that the specific encoding operations that occurred with common and rare words overlapped greatest when matched with the retrieval demands for each cued recall task.