Toward a Generate/Recognize Model of Performance on Direct and Indirect Tests of Memory

In two experiments, we compared performance on different tests of memory. Word stems were presented to be completed as an indirect test of memory. For a direct test of memory, the same stems were presented as cues for recall of earlier-presented words. Subjects in a generate/recognize test condition generated a completion for each word stem and then judged whether the generated word was presented earlier. Interactions between test conditions and a prior processing, and a materials variable were successfully predicted by a generate/recognize model of recall that postulates two bases for memory decisions. Cuedrecall differs from stem-completion performance in that recognition processes are involved in the former but not the latter task. However, subjects do sometimes output words as recalled without doing a recognition-memory check. This failure to reliably do a recognition check results in recall of unrecognized words. o 1990 Academic RUSS, 1~.