Alternative models of responsiveness are evaluated for goodness of fit to moves made by negotiators in six international negotiations: five arms-control talks and one base-rights negotiation. Three models, developed originally by Stoll and McAndrew (1986), are referred to as directional (tit-for-tat), trend, and comparative reciprocity. Seven models are variants on trend and comparative reciprocity including various lags and weightings for earlier and recent moves, as well as a combined trend and comparative model. The results indicate clear support for the comparative model in all of the negotiations analyzed: Negotiators responded to the difference between moves made by themselves and others in the previous round, adjusting their next move in the direction of the other's previous move. They did not respond to a monitored trend in the other's moves. Nor did the combined trend and comparative model improve predictions over the comparative model alone. Implications are discussed for the importance of fairness in the process of negotiating, for theories of information processing in bargaining, and for the “threshold-adjustment” hypothesis.