We investigated the consequences for exchange outcomes of the violation of an assumption underlying most social psychological research on exchange. This assumption is that the negotiated direct exchange of commodities between two actors (pure exchange) can be validly represented as two actors splitting a fixed pool of resources (split pool exchange). We designed five experimental conditions to determine differences in bargaining behavior in pure exchange and split pool exchange. We conclude that the validity of research using the split pool exchange approach is questionable, since we observed much less variance in outcomes and more efficient agreements in split pool exchange than pure exchange. Moreover, although theories accurately predicted outcomes of split pool exchange, they could not accurately predict outcomes of pure exchange. We discuss possible implications of our findings for exchange and research on exchange. (Contains 1 figure, 8 tables, and 10 footnotes.)