Confidence Ratings and Level of Performance on a Judgmental Task

Abstract

A 4-alternative forced-choice test was administered to 96 Ss. S's task was to attempt to select the correct alternative from each test item and to indicate his degree of confidence in his choice on a 5-point rating scale. The objective was to compare the confidence assignments of Ss who did relatively well on the primary judgmental task with those of Ss who did poorly. It was found that Ss who performed poorly on the primary task (LP Ss) tended on the average to use lower confidence ratings than Ss who did relatively well (HP Ss). Although few used either high or low ratings exclusively, all Ss tended to use one end of the confidence scale much more frequently than the other. However, whereas HP Ss were fairly consistent in using the high end of the scale, LP Ss were about evenly divided between those using the high end and those using the low. For both groups, performance tended to be monotonically related to expressed confidence. In terms of measures developed by Adams and Adams, HP Ss made more “realistic” confidence judgments than did LP Ss; however, there was no striking difference between groups in terms of differences in performance associated with step increases in expressed confidence.