The theme of the proposed research is that everyday inferential behavior is much more accurate than previously thought. People are not general-purpose inference machines, but are instead adapted to the structure of the environment in which they typically operate. This allows them to exploit strategies that are both simple and accurate. The proposed experiments examine the role of an important environmental feature: The extent to which a datum, or observation, is rare or unexpected. Rare observations are more informative than common ones, and people appear to make strong assumptions about what kinds of events are rare. These assumptions make sense outside of the laboratory and help explain many previous findings and lead to new predictions. The proposed experiments examine in detail people's sensitivity to rarity and the implications for several areas of research on thinking and reasoning.