Perceived Similarity and Accuracy of Peer Ratings

The hypothesis that raters will be more accurate in rating peers perceived to be similar to themselves, suggested by Mumford (1983) and derived from social comparison theory, was examined. Subjects were 681 Israeli entrants to a military training program. Shortly after course inception, subjects were asked to review the performance of squad members and to forecast their final grade. Subjects also judged peers' similarity to self overall (general similarity), in course achievement (foreground similarity), and in military experience (background similarity). Analysis revealed that accuracy was markedly lower in the evaluation of dissimilar others. Although the same pattern of results was observed for all forms of similarity, stronger effects resulted when similarity was measured in terms of general and foreground characteristics. Implications for future theory and research as well as for the practical application of peer assessment are discussed.