The systematic discrepancy between A ' for overall recognition and remembening: A dual-process account

Signal detection accounts of recognition assume that all item endorsements arise from the assessment of a single continuous indication of memory stren gth, even when subjects claim to categorically separate items accompanied b y contextual recollection from those that are not (viz., remembering vs. kn owing). Dissociations of these response types are held to occur because the former require a higher response criterion for item strength than does the latter. Meta-analytic and individual subject data suggest that when the A' metric is used, accuracy for remembering can systematically deviate from t hat of overall responding for individual subjects. This occurs because, unl ike the symmetric and rigid receiver-operating characteristic (ROC) implied under A', empirical ROCs are asymmetric and plastic. A dual-process model predicted that the magnitude of the deviation would vary as a systematic fu nction of the proportion of overall recognition accompanied by subjective r emember reports for individual subjects. The predictions were confirmed usi ng multiple regression on Monte Carlo and experimental data sets and were a lso shown to generalize to the double equal-threshold, single high-threshol d [i.e., H - FA; (H - FA)/(1 - FA)], and the equal variance signal detectio n d' corrections. The unequal variance signal detection model was also show n to mirror the data, but only under the post hoe assumption that every sub ject adopts a very similar remember criterion placement rule. The results d emonstrate that the systematic failure of tightly constrained models of rec ognition constitutes valuable regression data for more complex models and s imultaneously highlights why single-point measures of accuracy are unsuitab le as summaries across conditions or groups. Furthermore, the results show that remember rates carry unique information regarding the underlying proce sses governing individual subject performance that cannot be gleaned from t he overall hit and false alarm rates in isolation.