Detection theory and psychophysics: a review.

Abstract

I am particularly pleased to discuss the application in psychophysics of the general theory of signal detection at a symposium commemorating Fechner's founding work. Although this effort is in part a theoretical and experimental critique of Fechner's principal concepts and methods, which indicates that they should be replaced, I suspect that he would have welcomed it warmly---in fact, that he would have been among the first to recognize the value of these new tools had they become available in his time. I suspect this on the basis of his interest in Bernoulli's early ideas about statistical decision ([3], p. 284), and in the notion of subthreshold, or "negative," sensations ([3], p. 293)--two central concepts in the psychophysical application of the theory of signal detectability. The theory of signal detectability (henceforth called TSD) was developed most fully in the years 1952-1954 by Peterson, Birdsall, and Fox [35] at the University of Michigan, and by Van Meter and Middleton [55] at the Massachusetts Institute of Technology. At the same time, although working apart from TSD, Smith and Wilson [39] at the Massachusetts Institute of Technology and Munson and Karlin [33] at the Bell Telephone Laboratories were conducting psychoacoustic experiments that demonstrated the relevance of the theory to human observers; their experimental results led them to suggest a similar theory of the human detection process. Meanwhile, Tanner, and Swets [52, 53] were making a formal application of TSD in the field of vision. Since then, other general discussions and reviews similar to this one have appeared; I should mention the 1955 paper by Swets, Tanner, and Birdsall [47] that includes a complete review of the data collected in vision; the 1956 paper by Tanner, Swets, and Green [54] that includes several studies in audition; Green's [22] exposition in the current series of tutorial articles in the Journal o] the Acoustical Society; and Licklider's chapter [30] in the series edited by Koch for the American Psychological Association. The present discussion is distinguished from the first three of these in that it is not tutorial, detailed, nor documentary. The statistical and psychophysical bases of the work are not considered here, and no data are presented. In this sense, the present discussion is most like Licklider's. Its only ad