EQUIVALENCE INTERVAL AS A MEASURE OF UNCERTAINTY

Subjects made a series of 15 psychophysical judgments, one about each of five stimuli in three different stimulus classes. Then, supposing that their answers were not exactly correct, half of the subjects indicated the range of values within which they were reasonably certain that the correct answer did in fact lie. This range is termed an equivalence interval. The remaining subjects simply rated their confidence in the accuracy of each judgment. It was found that the width of the equivalence intervals correlated quite highly with the confidence ratings, and that they became wider the further the judged stimulus was from a clearly defined standard of comparison. Interestingly, however, although the equivalence intervals increased in size as the subjects' judgments became less accurate, they still failed to enclose the correct value nearly 42% of the time. The results were discussed in terms of the usefulness of the equivalence interval as a measure of uncertainty. © 1979.