Abstract

SUMMARY This article has two purposes. The first is to describe four theoretical models of yesno recognition memory and present their associated measures of discrimination and response bias. These models are then applied to a set of data from normal subjects to determine which pairs of discrimination and bias indices show independence between discrimination and bias. The following models demonstrated independence: a two-highthreshold model, a signal detection model with normal distributions using d' and C (rather than beta), and a signal detection model with logistic distributions and a bias measure analogous to C. Cis defined as the distance of criterion from the intersection of the two underlying distributions. The second purpose is to use the indices from the acceptable models to characterize recognition memory deficits in dementia and amnesia, \bung normal subjects, Alzheimer's disease patients, and parkinsonian dementia patients were tested with picture recognition tasks with repeated study-test trials. Huntington's disease patients, mixed etiology amnesics, and age-matched normals were tested by Butters, Wolfe, Martone, Three major points are emphasized. First, any index of recognition memory performance assumes an underlying model. Second, even acceptable models can lead to different conclusions about patterns of learning and forgetting. Third, efforts to characterize and ameliorate abnormal memory should address both discrimination and bias deficits. A common method for measuring memory performance is by use of a recognition memory test. Here we consider the yesno form of recognition memory test in which studied (old) items are presented mixed with distractor (new) items and the subject's task is to decide whether each item is old (by responding "yes") or new (by responding "no"). Performance on a yesno recognition test is summarized by two measures: the hit rate, the probability that the subject classifies an old item as old, and the false alarm rate, the probability that the subject classifies a new item as old. A given model of the recognition memory process specifies how these two measures are to be combined to characterize memory performance. Since