A classic experiment by Henri Tajfel provides evidence for the conclusion that the division of a group into subgroups is enough to trigger discriminatory behavior, even if there is no reason for such behavior in terms of the individual’s own interest. I don’t challenge that conclusion; but I question an implicit assumption which is suggested by the experimental setup and by the language used by Tajfel in describing the experiment. The assumption is that an initially coherent group will typically experience (...) division as a result of outside influence. A totally different picture of the situation and totally different social policy recommendations will follow if we believe instead that groups are evolving structures, and specifically structures that constantly and autonomously come into conflict with themselves. (shrink)