False recognition of associates: How robust is the effect?

False recognition of nonpresented words that were strong associates of 12 words in a study list was examined. Six lists were read to subjects; each list contained the 12 strongest associates to a critical nonpresented word. False-alarm rates to the 6 critical nonpresented words were obtained under several different conditions. The manipulations included varying the level of processing done to the study lists, varying the recognition-test procedure, repeating each of the study lists three times, and mixing the words from the six study lists together. A reliable false-recognition effect for critical nonpresented words was obtained in all conditions. However, the effect was not impervious to all of the manipulations. Significantly lower false recognition was obtained when learning was incidental as well as when the words on the six lists were mixed together. Neither level of processing nor repetition significantly influenced false recognition. This last result is inconsistent with Hintzman’s (1988) MINERVA 2 global memory model, but agrees with predictions from Shiffrin, Ratcliff, and Clark’s (1990) SAM model.