We compare social preference and social norm based explanations for peer effects in a three-person gift-exchange game experiment. In the experiment a principal pays a wage to each of two agents, who then make effort choices sequentially. In our baseline treatment we observe that the second agent's effort is influenced by the effort choice of the first agent, even though there are no material spillovers between agents. This peer effect is predicted by a model of distributional social preferences (Fehr-Schmidt, 1999). As we show from a norms-elicitation experiment, it is also consistent with social norms compliance. A conditional logit investigation of the explanatory power of payoff inequality and elicited norms finds that the second agent's effort can be best explained by the social preferences model. In further treatments with modified games we find that the presence/strength of peer effects changes as predicted by the social preferences model. As with the baseline treatment, a conditional logit analysis favors an explanation based on social preferences, rather than social norms following for these treatments. Our results suggest that, in our context, the social preferences model provides a parsimonious explanation for the observed peer effect.