We experimentally test competing theories of three-player majoritarian bargaining models with fixed, known disagreement values. Subjects are randomly assigned to three roles: a proposer and two types of voters. Each role is randomly assigned a disagreement value, i.e. a given amount of money he/she will receive if the proposal is rejected. These values are known to all players before any decision is made. Proposers then make a take-it-or-leave-it offer on how to split a fixed, known amount of money among the players. If a majority of players accepts the proposal, the players' payoffs are determined by the proposal; if the proposal is rejected, each player receives his or her reservation value. We assess the ability of three behavioral hypotheses – self-interest, egalitarianism, and inequality-aversion – to account for our results. Our primary design variable is the proposer's reservation value, which allows us to obtain different implications from each hypothesis. We find that each hypothesis is inconsistent with our data in important respects. However, subjects strongly respond to changes in reservation values as if they were interpreted as a basic form of entitlement.