Signal Detection and Matching: Analyzing Choice on Concurrent Schedules
A. W. Logue State University of New York at Stony Brook
Monica L. Rodriguez Columbia University
The matching law ( Herrnstein, 1961, 1970) and signal-detection theory ( Green & Swets, 1966) both provide quantitative methods for analyzing choice. Comparisons of the similarities and differences between these two methods can indicate how each may best be used and can also improve our understanding of both methods and of choice in general. The two experiments reported here compare these two approaches empirically. First, however, it will be useful to outline briefly some of the purposes and procedures of each approach.
The usual signal-detection procedure employs discrete trials. One of two explicit stimuli (signal plus noise or noise alone) is presented at the start of each trial. The subject's task is to make one of two responses indicating which stimulus was presented. A payoff, a punisher, and/or feedback is ordinarily then presented, the trial ends, and after an intertrial interval, a new trial begins. This procedure results in four possible types of response (see Fig. 8.1). Substituting payoffs for each of the four classes of response in Fig. 8.1 yields a payoff matrix.
The purpose of signal-detection theory has been to provide independent measures of the effects on responding of (a) outcomes, and (b) the discrimination of the explicit stimuli. According to traditional signal- detection theory, "bias" will ideally be affected only by motivational factors (e.g., the relative values in the cells of the payoff matrix), whereas "sensitivity" will ideally be affected only by manipulations of the physical character of the signal and of the noise ( Green & Swets, 1966; Nevin, 1981; Swets, Tanner, & Birdsall, 1961). When the values of the payoff matrix have been varied, bias, not sensitivity, has indeed usually been affected