Standard theories in economics generate predictions of market behavior by invoking two fundamental assumptions. First, agents are self-interested in that their utility function depends only on their own material payoffs. Second, market behavior is at equilibrium so that no individual agent can achieve a higher payoff by unilaterally deviating from the equilibrium. Recent advances in behavioral economics relax both assumptions by allowing agents, for example, to care about others' payoffs and to make mistakes (see Matthew Rabin, 1998, Colin F. Camerer, George Loewenstein, and Rabin, 2003, and Teck-Hua Ho, Noah Lim, and Camerer, 2006a, for comprehensive reviews). This paper focuses on the self-interested assumption and investigates how social comparison may lead to fairness concerns between peers.A simple and powerful way to demonstrate that people are not purely self-interested is to study the so-called ultimatum game. In this game, a leader and a follower divide a fixed pie. The leader moves first and offers a division of the pie to the follower. The follower can accept or reject. If the follower accepts, the pie is distributed according to the proposal. If the follower rejects, both players earn nothing. When players care only about their own material payoffs, the subgame perfect equilibrium predicts that the leader should offer a small amount (e.g., a dime) to the follower and the follower would accept (since a dime is strictly preferred to nothing). However, data from many experiments (where subjects are motivated by substantial financial incentives) cast doubt on this sharp prediction. Typically, there are almost no offers below 20 percent of the pie. A majority of offers are between 30 percent to 40 percent. Low offers are frequently rejected and the frequency of rejection increases as the offer decreases. These findings are robust to stake size (Robert Slonim and Alvin E. Roth, 1998), persist with repeated trials (Roth et. al, 1991), and prevail across diverse cultures (Joseph Henrich, 2000;Camerer et. al, 2001;Miguel Costa-Gomes and Klaus G. Zauner, 2001). * Ho and Su: Haas School of Business, University of California, Berkeley, CA 94720, Email: Ho: hoteck@haas.berkeley.edu; Su: xuanming@haas.berkeley.edu. Authors are listed in alphabetical order. Direct correspondence to any of the authors. We thank George Akerlof, Eduardo Andrade, Colin Camerer, Vince Crawford, Laura Gardner, Juanjuan Zhang, and three anonymous reviewers for helpful comments. Esther Hwang provided superb research assistance. Taizan Chan designed and developed the software system for running our experiments.