Conventional approaches to program evaluation have followed the experimental and psychometric traditions dominant in educational research. Their aim (unfulfilled) of achieving fully objective methods had led to studies that are artificial and restricted in scope. Illuminative evaluation is introduced as a total re-appraisal of the rationale and techniques of program evaluation. It follows a contrasting 'anthropological' research paradigm. Illuminative evaluation recognizes that an innovative instructional program cannot be separated from its learning milieu--a network of cultural, social, institutional, and psychological forces. The evaluator concentrates on processes within the classroom rather than on outcomes (goals, objectives) derived from a specification of the instructional system. Data is collected from observation, interview, questionnaires, and documentary information. Illuminative evaluation thus concentrates on the information gathering rather than on the decision making component of evaluation. The paper concludes with a discussion of the problems and potentialities of the new approach: its range of applicability; objectivity; validity; obligations of the research worker; how explorations of the learning milieu can meet the need for theoretical advance; and how the illuminative approach can clarify decision-making. (Author/CP)