Signal-Detection Analysis of Illusions and Heuristics
John A. Nevin University of New Hampshire
In the fall of 1959, Bill McGill introduced me to the theory of signal detection in a first-year graduate course at Columbia. I was hooked at once, and the hook is still holding firm. In my current version of McGill's course, I teach the classical theory, its fundamental challenge to the concept of threshold, and some of its elegant applications to vision and psychoacoustics, as well as recent behavioral versions of detection theory (e.g., Davison & Tustin, 1978). But what I really try to convey is something more general, which I will call here the signal-detection approach to quantitative behavior analysis. In its broadest terms, the signal-detection approach is concerned with the analysis of choice between discriminated operants -- responses under the joint control of stimuli and consequences ( Skinner, 1969; for a discussion in relation to signal detection, see Nevin, Jenkins, Whittaker, & Yarensky, 1982). The analysis involves separating the discriminative effects of the stimuli from the biasing effects of the consequences, each of which can be shown to be invariant with respect to variations in the other (see, e.g., Nevin, 1984). As such, it is a model of scientific analysis, and it has the further advantage of unifying two areas of behavioral research -- stimulus control and reinforcement -- that, although related, have long differed in their methods of analysis and styles of theory.
A coherent account of behavior that treats the effects of stimuli and reinforcers in common terms would be a major advance. Some steps in that direction appear in Vol. 1 of this series ( McCarthy & Davison, 1981; Nevin, 1981); some more recent advances are explored by Davison and Jenkins ( 1985) and by Alsop and Davison (this volume). But that is not my topic