One of the classic debates in psychology concerns the stability of personality. With rare exception, studies that have correlated objective behavior on two occasions have obtained coefficients below .30. Not only has the direct measurement of objective behavior failed to provide evidence of stability, but selfreport scales in attitude and personality inventories, as well as ratings of behavioral samples by judges (although themselves stable), have produced low correlations with objective behavior. Does this indicate, as some have suggested, that stability of behavior lies primarily in the eye of the beholder? The issue can be resolved by recognizing that most single items of behavior have a high component of error of measurement and a narrow range of generality. In four separate studies it was demonstrated that when measures of behavior are averaged over an increasing number of events, stability coefficients increase to high levels for all kinds of data, including objective behavior, selfratings, and ratings by others, and that objective behavior can then be reliably related to self-report measures, including standard personality inventories. The observation that it is normally not possible to predict single instances of behavior, but that it is possible to predict behavior averaged over a sample of situations and/or occasions, has important implications not only for the study of personality but for psychological research in general.