Objective. The contact hypothesis is difficult to test because of selectivity bias: the direction of causation between contact and attitudes cannot be definitively determined. Selectivity bias can be avoided given a sample for which, for reasons unrelated to group member attitudes, whites divide into groups that differ systematically in the amount of contact they have with blacks. This paper uses a natural experiment that provides such a test: All-Star voting by baseball players versus fans. The average white player has more substantial and persistent contact with African Americans (as teammates) than does the average white fan, and the contact hypothesis would thus predict that players will discriminate less than fans, all else equal. Methods. Logistic, ordinary-least-squares (OLS), and tobit regressions are used to analyze the effect of candidate race on votes received, controlling for other factors. Results. No evidence is found of a differential in discrimination related to the amount of contact, although discrimination by both groups is found to decline between 1970 and 1980. Conclusion. The contact hypothesis is not supported; exogenous variations in the amount of interracial contact are not associated with difference in levels of discrimination.