Using the Deese-Roediger-McDermott paradigm, recollection of distinctive information about cognitive operations was shown to improve recognition accuracy. Requiring participants to generate (vs. read) each list word by solving an anagram substantially decreased false recognition of non-presented critical lure words. In addition, correct recognition increased when generation occurred at both study and test--even if different anagram cues were presented at test-suggesting that reinstating cognitive operations at test can facilitate correct recognition. When participants generated half the lists and read the other half at study, false recognition decreased for both generated and read lists, relative to a group who read all list words at study. This finding suggests that generation induced participants to adopt a distinctiveness heuristic at test, whereby they took the absence of memory for generation as evidence for a word's nonoccurrence at study, and is less consistent with an impoverished relational encoding account that attributes false recognition to impaired encoding of relations between list words.
Bibliography: p. 51-59