The effects of temporal instability in attitudes on the attitude-behavior relationship were examined in a study of volunteering to tutor blind children. A mailed appeal was received by 286 Israeli undergraduates who had completed a questionnaire either 3 months, 6 months, or both 3 and 6 months earlier, or not at all. Embedded in the questionnaire were attitude items on altruistic acts (including tutoring blind children) and on various controversial issues. The attitude-behavior correlation was higher over the shorter time interval (.47 vs. .13), and data from the group whose attitudes were measured twice indicated this was due to real change in individuals' attitudes. A specific attitude and its corresponding behavior correlated more strongly among those whose general set of altruistic attitudes showed high rather than low temporal stability (.47 vs.-.03), but stability of the specific attitude did not moderate this correlation. Characteristics of attitudes that might influence their stability are discussed, and it is shown that attitude stability is not a general trait.In 1931, Thurstone suggested that temporal instability in attitudes weakens the relation of attitudes to behavior. Although numerous theorists and researchers have reiterated this point over the years (e.g., Alwin, 1973;Fishbein, 1967;Schuman & Johnson, 1976;Schwartz & Tessler, 1972;Wicker, 1969), not a single published research study has directly examined whether real change in individuals' attitudes contributes to attitude-behavior discrepancies. The present study addresses this question from two vantage points. First, it compares the predictive validity of specific attitudes measured at different temporal removes from behavior. Second, it examines the relative predictability of behavior among groups that differ in the stability of their underlying general attitudes.Two studies concerned with other questions have mentioned data bearing on whether the The research reported here was supported by National Science Foundation Grant SOC72-05147. I am grateful to Esther Geva and Avi Gottlieb for their aid in conducting the study and to Thomas Heberlein, Gerald Harwell, Jane Piliavin, G. William Walster, and J. Stanley Black for their comments on an earlier draft.Requests for reprints should be sent to Shalom H. Schwartz, Department of Sociology, University of Wisconsin, Madison, Wisconsin 53706. predictive validity of attitudes diminishes with the passage of time prior to behavior. Kelley and Mirer (1974) found that attitudes correctly predicted voting for an average of 85% of their sample in the presidential elections from 1952 to 1964. For respondents whose attitudes reflected conflict or indifference over the election, 28% of the variance in errors of prediction was explained by the number of days that intervened between the interview and the election. With each doubling of time, the error rate rose by 4 percentage points.This evidence foi the importance of change in attitudes cannot be generalized easily to other attitude-behavior settings, howev...