Slope of the receiver-operating characteristic in recognition memory

Two proposals concerning the slope of the receiver-operating characteristic (ROC) in recognition memory were tested-the constancy-of-slopes generalization (R. Ratcliff, C.-E Sheu, & S. D. Gronlund, 1992) and the dual-process hypothesis (A. P. Yonelinas, 1994). The constancy-of-slopes generalization states that changes in accuracy leave ROC slopes unchanged. The dual-process hypothesis states that ROC slopes are less than 1.00 because they are produced by two independent memory factors: recollection and familiarity. Four experiments tested the constancy-of-slopes generalization. The results contradicted the generalization. Slopes changed when accuracy was changed by varying study time, word frequency, and encoding task. A survey of the Literature showed slope changes produced by those variables and also by other variables. Analyses of the shapes of ROCs were used to test the dual-process hypothesis. The results contradicted that hypothesis.