AN EXPERIENTIALLY DERIVED BASE-RATE ERROR IN HUMANS

Abstract— Two experiments tested college students in a delayed matching-to-sample procedure designed to explore how multiple competitive sources of control interact in determining choice behavior. The primary sources of potential control were the frequency with which each of two colors was correct (base rates) and the accuracy of a sample in predicting the correct choice (accuracy rate). In one experiment, subjects ignored the base rates, and therefore made nonoptimal choices. These results mimicked the well-known base-rate error found in verbal responses to questions. In a second experiment, manipulations to increase subjects' motivation did not alleviate the error significantly. Together, these results suggest that the base-rate error may represent a broad behavioral pattern not restricted to verbal arenas. The results are compared with previous attempts to reconcile behavioral and cognitive lines of experimentation.