Traditional gift-exchange experiments were designed with corner equilibria so that evidence of positive reciprocity could not be disentangled from one-sided errors. Our first treatment replicates the traditional design and finds that effort is an increasing function of wage for mid-range wage offers, but this relationship is not significant for high and low offers. The second treatment has an interior equilibrium, asymmetric marginal costs of reciprocity (positive and negative) and lower efficiency gains. There is evidence of a decrease in the deviations from the subgame perfect Nash equilibrium. However, there is still significant reciprocal behavior (positive and negative).