Confidence,Correctness,and Difficulty with Non-Psychophysical Comparative Judgments

Abstract

Three 4-alternative forced-choice tests were constructed, representing different levels of difficulty. Individual items on each test consisted of the names of four states and a 5-point confidence rating scale. S's task was to choose the largest state in each item and indicate his degree of confidence in his choice. Although confidence assignments were somewhat higher for relatively easy judgments than for more difficult ones, the difference was not sufficient to keep the relationship between confidence and correctness invariant with respect to the difficulty of the judgmental task. In short, the results indicate that the value of confidence expressions as indicators of the probable correctness of “intellectual” judgments varies considerably with the difficulty of the judgments involved.