Association false recognition occurs without strategic criterion shifts

In the DRM (Deese/Roediger and McDermott) false memory paradigm, subjects s tudied lists of words associated with nonpresented critical words. They wer e tested in one of four instructional conditions. In a standard condition, subjects were not warned about the DRM Effect. In three other conditions, t hey were told to avoid false recognition of critical words. One group was w arned before study of the lists (affecting encoding and retrieval processes ), and two groups were warned after study (affecting only retrieval process es). Replicating prior work, the warning before study considerably reduced false recognition. The warning after study also reduced false recognition, but only when critical items had never been studied; when critical items we re studied in half the lists so that subjects had to monitor memory for the ir presence or absence, the warning after study had little effect on false recognition. Because warned subjects were trying to avoid false recognition , the high levels of false recognition in the latter condition cannot be du e to strategically guessing that critical test items were studied. False me mories in the DRM paradigm are not caused by such liberal criterion shifts.