Detection Theory: A User's Guide

We return again to designs for studying discrimination. The tasks described to this point—yes-no with or without a rating response or a reminder, and 2AFC—provided the experimental cornerstone for detection theory in psychology. They are natural paradigms for studying the detection of weak signals and, as we have seen, are simply related to each other on theoretical grounds.

Each design, however, has shortcomings. The failure of the predicted relation between yes-no and 2AFC (Eq. 7.2) led us to the suspicion that participants are limited in the one-interval task by imperfect memory. Twoalternative forced choice, which survives this criticism, is subject to anothe: In some applications, the task is difficult to describe to participants. Observers in 2AFC are instructed to “choose the picture you think you have seen before” or “choose the interval that contained a tone added to the noise background. ” The dimension of judgment—recency and “tone-ness, ” in these examples—is made explicit. But observers may not share the experimenter's definition of the dimension being judged, and may even be able to distinguish the stimuli without having names for them at all. Many listeners in simple auditory tone-detection experiments, for example, discover that “tone-ness” is not, introspectively, the basis for judgment: The experience of a very weak stimulus is not a small version of a more intense one, but participants usually learn to respond appropriately with training.

Frequently, the problem of describing the dimension on which the stimuli differ is not so readily solved. Sometimes the physical dimension is difficult to characterize for participants; the experimental design precludes training; or participants are unsophisticated, and forced-choice instructions are difficult to convey. We now discuss three participant-friendly designs that seem

-213-