When Do Base Rates Affect Predictions

Recent studies have shown that when people make predictions, they often neglect base-rate considerations. Instead of considering what typically happens in situations like the one being judged, they rely on the extent to which the judged case is representativ e of the possible prediction categories—although if representativeness fails to provide a clear guide to prediction, people will resort to base-rate considerations. Manis, Dovalina, Avis, and Cardoze recently argued against this conclusion, presenting as evidence a series of experiments in which subjects predicted the category membership of individuals depicted in each of a set of photos. They found that base rates had a clear effect on discrete predictions (i.e., a majority of the photos were predicted to belong to the larger category) and a smaller effect on the confidence subjects attached to those predictions. However, only a minority of the photos could be readily classified by representativeness. As a result, Manis et al.'s findings can be reinterpreted in a way that makes them compatible with previous findings. In this light, their study emerges as a constructive replication of earlier results demonstrating judgment by representativ eness. A burgeoning area of research is the study of the role of background (or base-rate) information in people's predictions about specific target individuals (Borgida & Brekke, in press; Kassin, 1979). The impetus for much of this work was a study by Kahneman and Tversky (1973) demonstrating that base rates may be largely ignored in the presence of even flimsy specific evidence. They argued that predictions are often governed by judgments of "representati veness." The user of this heuristic predicts that people belong to groups whose prototypes they most resemble. Such predictions are made even when the target individuals are characterized by only a brief, unreliable description and even when they resemble low base-rate groups, that is, groups with few members in the population from which the individuals are drawn. In the course of this base-rate research, Kahneman and Tversky's original design