The present review of the literature suggests that it is useful to distinguish between two types of consensus information-normative expectancies (e.g., Jones & McGillis' prior probability concept) and explicit base rates (e.g., Kelley's conception of observed covariation across actors). Normative expectancies, which may be derived from a knowledge of one's own behavior (i.e., the false-consensus effect) or the behavior of others, provide one basis for prediction and causal inference. Explicit, sample-based consensus may also be employed, but under somewhat restrictive conditions: (a) when prior expectations are neutralized and/or (b) when the consensus manipulation is particularly strong, salient, easily translatable, representative of the criterial population, and causally relevant. A number of additional issues are reviewed (e.g., the cognitive strategies by which observers reject base rates), and recommendations for the direction of future research are made.I would like to thank the following people for their helpful comments on earlier drafts of the manuscript: