As the study of negotiation and mediation has grown rapidly over the last three decades, so have the number of approaches to it. Behavioral scientists of all persuasions bring their ideas and methods to bear on the study of mediation. This paper identifies some of the more significant of these approaches, and argues that many of them are predicated on erroneous, even unrealistic, assumptions. It argues that the best way to conduct research on mediation is to study such behavior in the real, not the simulated, world, and choose data that is directly generated by parties or mediators in conflict. The paper argues for a systematic archival research, and presents the broad outlines of such an approach can be organized and conducted. The findings presented suggest the importance and relevance of this method of research into mediation.