Memory scanning for words in visual images

This study attempts to replicate an experiment reported by Seamon (1972). In the present investigation, as in the study by Seamon, the scanning of short-term memory was compared when its contents were rehearsed words vs. a mental image. Memory sets were composed of either one, two, or three words. In the relational imagery group, subjects were required to form a single interactive mental scene of the entities which the memory set words represent. Nonimagery subjects were given instructions to covertly rehearse the memory set. In both groups, the usual memory set size (m) effect was obtained, i.e., reaction time (RT) increased linearly with m. Moreover, the set size effect was the same for both groups. This latter finding stands in marked contrast to the result obtained by Seamon; he found no effect of set size when subjects were given interactive imagery instructions. Because of the failure to replicate Seamon, an additional group of subjects were given imagery instructions. For this latter group, some of the procedural discrepancies between the relational imagery group of the present study and the corresponding group in Seamon’s study were resolved. Also, in this additional group, the set size effect was examined as a function of the subjects’ ratings of the quality of the images which they had formed. The same set size effect was found for this additional group, and the effect was independent of image quality.