Hypermnesia for Words in Serial Learning

Subjects were exposed to 16 words under serial learning instructions and then received a serial recall test. Following this initial test, subjects (a) attempted recognition, (b) were prompted with extralist associative stimuli, or (c) again attempted serial recall. Subjects in the latter group then attempted recognition or were prompted for recall. Analyses of recall probabilities conditional on prior failure to recall showed that (a) subjects were more likely to recall prior failures of recall on a second serial test than on a second recognition test, and (b) recognition yielded more recalls of prior failures than did prompting. Secondary analyses showed that neither response criterion shifts, item selection artifacts, nor extra recall time accounted for the superiority of the second serial test. The results are interpreted to be inconsistent with a critical assumption underlying encoding specificity theory, and to be consistent with a recent stimulus-sampling version of generation-recognition theory.