Designs for Discrimination
So far we have discussed only situations in which one stimulus at a time is evaluated. In psychophysics, even these designs are sometimes termed discrimination because they permit estimates of the ability to distinguish two stimuli or stimulus classes, and in common psychological usage discrimination means telling two things apart. In this chapter, we introduce paradigms in which the process of discrimination is more salient because two or more stimuli are explicitly compared on each trial.
There are two types of such paradigms, which we term comparison designs and classification designs. Comparison designs (considered in this chapter) resemble one-interval designs in that the observer makes a binary decision based on an underlying representation containing only two distributions. These paradigms require the observer to make a direct comparison between two stimulus presentations. In classification designs (chap. 9), the observer again compares (two or more) stimuli, but the world of possible stimuli and their representation contains more than two distributions. Because multiple distributions are mapped onto each response, the task facing the observer is more complex than a simple comparison.
There are only two comparison designs: two-alternative forced-choice and the reminder design. It is possible to analyze both with only the one-dimensional, flatland tools of Part I. The advantage of considering two-dimensional representations is that each design offers multiple alternative decision rules, and the perspective of a multidimensional spaceland view renders the relations among these rules more visible. A representation containing a pair of bivariate distributions in a two-dimensional perceptual space was illuminating in analyzing the compound detection problem in chapter 6, and we construct similar decision spaces to describe two-alterna-