Two different approaches for treating response bias in the process-dissociation procedure were assessed: a multinomial approach proposed by A. Buchner, E. Erdfelder, and B. Vaterrodt-Pliinnecke (1995) and a dual-process, signal-detection approach proposed by A. P. Yonelinas, G. Regehr, and L. L. Jacoby (1995). The authors examined data presented by Buchner et al. and found that, although the signal-detection-based model worked slightly better than the multinomial model, the data did not provide a strong test of either model. However, an examination of other recognition data showed that the multinomial model produced distorted estimates of recollection and familiarity, and it was unable to account for observed receiver operating characteristics (ROCs). In contrast, the dual-process, signaldetection model produced unbiased estimates and was able to account for the observed ROCs. The authors also provide an overview of the general controversy surrounding the processdissociation approach. Buchner, Erdfelder, and Vaterrodt-Pliinnecke (1995) proposed a modification of the process-dissociation procedure that incorporates response bias. We begin this article by describing the process-dissociation procedure and then describe the multinomial model that Buchner et al. proposed, contrasting their model with a dual-process, signaldetection method that we (e.g., Yonelinas, 1994; Yonelinas, Regehr, & Jacoby, 1995) have developed to take differences in response bias into account. Next, we examine the data that Buchner et al. used as evidence to support their model and show that our model deals with those data slightly better than does their model. We argue, however, that the manipulations used in their experiments may have influenced both response bias and recollection, and thus their data may not provide a strong test of models. We assess the models further by examining changes in response criterion associated with differences in response confidence and show that the multinomial model, but not the dual-process, signaldetection model, produces distorted estimates of recollection and familiarity. Finally, we examine receiver operating characteristics (ROCs) and find that the model that underlies the multinomial approach is in conflict with the ROC data. In contrast, we show that our signal-detection model accounts well for observed ROCs. While considering models of response bias, we provide a road map for contro