The Will Rogers Illusion in Judgments About Social Groups

Abstract

Subjects' judgments of the mean of 12 scores were influenced by the way in which the scores were dichotomized. The estimated mean was higher when the three highest scores formed one group (e.g., payments for women) and the nine lowest formed the other (e.g., payments for men) than when the nine highest were one group and the three lowest the other. We call this phenomenon the Will Rogers Illusion (WRI). The WRI occurred only when estimates of the subgroup means were made prior to the estimates of the mean of the whole group. When the latter mean was judged first, the WRI was reversed. These and other data indicate that the means of subgroups can influence judgments of group means, a finding that is relevant to research on social stereotypes.