Effects of Cue Consistency and Value on Base-Rate Utilization

In two experiments using "Bayesian" probability judgment tasks, we examined the effects of numerical values of base rates and case cues, the degree of consistency in these values, and the narrowness of the populations to which these cues pertained. Both experiments showed that the "base-rate fallacy" is observed only when (a) one combines base and case cues that lead to dissimilar judgments when each is considered alone, and (b) the case cue is high in numerical value, reflecting high diagnosticity of the case source. We obtained similar results in within- and between-subjects versions of the first experiment and in a second between-subjects experiment using a different problem. We conclude that in tasks such as those studied, the base-rate fallacy is far less general than has been suggested. The appearance of such a fallacy is produced when the numerical value of the case cue is held constant at high levels, as is typical in most published research on this topic. However, the observed insensitivity to base rates disappears when the value of the case cue is held constant at low levels.