The Deese/Roediger/McDermott (DRM) illusion is one of the most widely used methods for studying false memory. Early studies provided normed false recall and false recognition data for DRM lists, where recognition is preceded by prior recall tests, and reported regression analyses that revealed backward associative strength (BAS) as one of the strongest predictors of false memory. As an extension of that line of research, we collected new recognition norms that are not confounded by prior recall tests and included gist strength (GS) as a theory-driven predictor of false memory. In Study 1, we normed true and false recognition for 55 DRM lists without prior recall tests, and in Study 2, we normed these lists for their perceived levels of gist strength. In Study 3, we fit a series of multiple linear regression models to the recognition data from Study 1 as well as recall and recognition data from prior false memory norms to disentangle the effects of BAS and GS on false recognition (with and without prior recall) and immediate false recall. Our results revealed that levels of true recognition and the recall–recognition correlation were inflated by prior recall tests. More importantly, GS was the strongest predictor of false recognition, whereas BAS was the strongest predictor of immediate false recall. A GS × BAS interaction was consistently observed for false recall and false recognition, in which the effects of BAS declined as GS increased. This suggests that the two variables compete with each other rather than reinforce each other’s effects.
