This essay focuses on a key change that the author has experienced in the practice of program evaluation research over the past 20 years: the accommodation of the logic of science and the psychology of politics. This change has substantially altered the role of the evaluator, the evaluator's relationship with clients, the design of an evaluation study, and the training of evaluators at the Ph.D. level. The role of evaluator used to be that of a scientist highly reliant on technical skills who was called upon to do brief studies, draft and implement the design, and deliver the final report, all for a fixed price contract. During the last 10 years, this role has changed considerably, primarily because of the longer duration of the evaluation projects and the higher level of engagement between the evaluator and the client. As a consequence, evaluation designs have become more evolutionary in design; an initial "evaluation research matrix" is modified over time in consultation with the client, and the final report is replaced by a series of interim reports on specific questions which in turn are revised as data come in. This shift in design and practice of evaluation arose in response to the need to resolve conflict among contending parties at an institution by delivering useful information in a timely manner. Preparation of Ph.D. students in evaluation research accordingly has come to rely much more on their early first-hand engagement with field problems as members of evaluation teams. The author concludes that evaluation research is likely to continue to be involved with contextual politics as well as technical demands. (TE)