Predictions from the bounded and unbounded reciprocity hypotheses and from social identity theory (SIT) were examined in a minimal group experiment in which ingroup outcome dependence, outgroup outcome dependence, and the strength of social identity were orthogonally manipulated. Both ingroup and outgroup outcome dependence affected reward allocations. Participants made more ingroup-favoring reward allocations across all conditions. The identification manipulation produced hypothesized effects on social identification measures and marginal effects of identification on reward allocations in the no-dependence condition. Support was found for both an unbounded and bounded version of the reciprocity hypothesis and marginal support for a SIT approach to intergroup discrimination. The study highlights insufficiencies of both theoretical approaches and suggests possibilities for integration and elaboration.