A behavioral model for performance on signal-detection tasks is presented, It is based on a relation between response and reinforcement ratios which has been derived from both animal and human research on the distribution of behavior between concurrently available schedules of reinforcement, This model establishes the ratio of obtained reinforcements for the choice responses, and not the probability of stimulus presentation, as the effective biaser in signal-detection research, Furthermore, experimental procedures which do not control the obtained reinforcement ratio are shown to give rise to unstable bias contours. Isobias contours, on the other hand, arise only from controlled reinforcement-ratio procedures,
371The theory of signal detection (Peterson, Birdsall, & Fox, 1954;Tanner & Swets, 1954;van Meter & Middleton, 1954) holds the promise of extracting two independent measures to describe behavior in a detection task. The two measures are stimulus discriminability, a measure of the subject's ability to tell two stimulus conditions apart, and bias (or criterion), a measure of how performance can be changed by nonsensory motivational or payoff variables. Most research in contemporary psychophysics has placed primary emphasis upon the sensory performance of human subjects, and attempts to relate stimulus parameters to the physical properties of the stimuli, independently of bias, are well documented. As a result, rather less effort has been expended in the search for a bias parameter which remains invariant with changes in discriminability (Dusoir, 1975;Luce, 1963).The term "bias" (or "criterion") has frequently been used in both an explanatory and a descriptive sense-often with serious confusion (Treisman, 1976). In addition, as Dusoir (1975) pointed out, there is no generally accepted way of measuring bias and, hence, there is little agreement on the true shape of empirical isobias contours. Dusoir suggested the need for a measure of bias which was unaffected by changes in variables which, on a priori grounds, might be expected to change only discriminability (e.g., stimulus values). The measure must, however, be affected by operations which should manipulate The research reported here was supported entirely by the New Zealand University Grants Committee, to which organization we continue to be most grateful. We thank the Associate Editor, Dr. A. Kristofferson, and an anonymous reviewer for helpful comments, and also Michael Corballis for constructive discussion. Requests for reprints may be sent to Dianne McCarthy, Department of Psychology, University of Auckland, Private Bag, Auckland; New Zealand.Copyright 1981 Psychonomic Society, Inc.bias (e.g., stimulus-presentation probability and payoff). Dusoir, reviewing the then-current theories of bias (e.g., Broadbent, 1971; Hardy & Legge, 1968;Healy & Jones, 1973;Luce, 1963;Parks, 1966;Thomas & Legge, 1970;Treisman, 1964), found no measure of bias satisfying the above requirements.Here, we review a behavioral approach to bias which: (1) unlike signal-detection...