We evaluate four theories that predict the distribution of power in exchange networks. All four theories-core theory, equidependence theory, exchange-resistance theory, and expected value theory-assume actors rationally pursue self-interests. Three of the theories add social psychological assumptions that place the pursuit of self interest in an interactive context. Predictions of exchange earnings by the four theories are evaluated against data from eight experimental networks, including types of networks not previously studied. These networks vary conditions that affect the chances that a position can be excluded from exchange. We find that when the theories base predictions on a network position's structural potential for exclusion, exchange-resistance theory provides the bestfit, but when predictions are based on actual experiences of exclusion, expected value theory fits best. Our discussion focuses on the distinction between the a priori potential for exclusion versus experienced exclusion as factors in the genesis of power T he problem of power distribution in exchange networks has captured the attention of a variety of theorists. The appeal of the problem derives from the combination of the formal representation of social structure as network (Wellman and Berkowitz 1992) and sociology's perennial concern with power. A growing body of experimental studies now permits researchers to test various theoretical formulations. How does location in a network confer advantages on a person or a corporate body in their dealings with others? Consider the promotion prospects of two senior accountants, Andy and Bob. Because Andy's work involves accounts at various regional offices, his coworkers typically do not know each other. Bob, on the other hand, deals with corporate accounts, so his coworkers typically associate with each other. Thus, Andy and Bob are surrounded by two very different networks and it is not obvious that Andy's network favors him for promotion (Burt 1992). In a second ex