CRITICAL INVESTIGATIONS OF THE Z-ROC CURVE: EVIDENCES FOR A SINGLE-PROCESS MODEL OF RECOGNITION MEMORY
A dominant approach to recognition memory is dual-process theory, comprising both signal- detection (familiarity) and high-threshold (recollection) mechanisms. The finding that z-ROC slopes are invariably smaller than unityâ€”establishing larger variance for the hypothetical target distribution than that of the lure distributionâ€”is taken as support for the involvement at retrieval of recollection, in addition to familiarity (thereby increasing the variance of the target distribution). Critically, in a seminal study, Yonelinas et al. (2001) obtained confidence ratings (from which z-ROC slopes could be calculated) and, within each level of confidence, a subjective Remember-Know (R-K) judgment was made. In an attempt to simulate recognition unaffected by recollection, 'Remember' (R) trials were then selectively excluded. Consistent with dual-process theory, the z-ROC slope increased to unity, demonstrating the involvement of only a single processâ€”Familiarity. Here we argue that Yonelinas et al. inadvertently biased participants to respond R only when they were highly confident that the item had been studied and that, in addition, the authors did not consider the possible existence of Recollection-based false alarms. When running the experiment with unbiased instructions, R responses spread across all confidence levels. We next excluded both R and K responses for both FAs and hits. Similar to Yonelinas, we too found an increase in the slope of the z-ROC curve. Our analysis suggests a strength-based interpretation to the data which include not only recollection-based hits, but also recollection-based FAs.