Memory and the theory of signal detection.

TLDR

Recognition memory research has long struggled to combine hits and false alarms into a single performance index, raising concerns that correct response rates may be contaminated by factors beyond memory state and underscoring the need for a theory of the decision system that maps memory states to overt responses. The study seeks to develop quantitative theories of recognition memory that explicitly model how memory and decision processes interact to produce observed responses. To evaluate such theories against experimental data, the authors propose deriving performance measures that are independent of decision parameters, allowing memory models to be tested directly.

Abstract

A recurrent problem in the study of recognition memory has been that of combining hits and false alarms (or correct and incorrect responses) into a single index of performance. That the proportion of correct responses in a yes/no or multiple-choice recognition test may in some way be contaminated by factors other than the state of the memory system has never been seriously disputed, but only recently has the question received detailed and systematic attention. Traditionally the problem has been viewed in terms of correcting for chance success; in current terminology, and stated more generally, it is a problem of providing an adequate theory of the decision system which maps a given state of the memory system into an overt response. The reasons for this increased attention to decision processes in memory are not difficult to trace. In the first place, the development of quantitative theories of recognition memory has necessitated a precise and explicit account of how the memory and decision systems interact to produce a given response. If a theory of memory is to be tested against data from a recognition-me mory experiment, it is necessary to obtain performance measures that are independent of those parameters