We use multilateral bargaining experiments to examine how the order of bargaining (simultaneous or sequential) and the nature of contracts (contingent or non-contingent) affect the duration of bargaining, the efficiency of exchange, and the distribution of the surplus in a laboratory land-assembly game with one buyer and two sellers. While theory predicts an earnings advantage for the first seller when contracts are sequential and contingent, and for the second seller when contracts are sequential and non-contingent, we find that when a seller has an earnings advantage in the laboratory, it is the first seller to bargain in the non-contingent contract treatments. This result contradicts conventional wisdom and a common result from the land-assembly literature that it is advantageous to be the last seller to bargain, a so-called "holdout". We also find evidence that sequential bargaining leads to more aggressive seller bargaining and greater bargaining delay than simultaneous bargaining, ceteris paribus, and that non-contingent contracts increase bargaining delay and the likelihood of failed agreements. The majority of sellers indicated a preference for being the first seller to bargain in all sequential bargaining treatments.