This conference report identifies research needs in evaluating and documenting large scale school improvement efforts to serve disadvantaged populations. Summaries of the conference presentations are provided, grouped into three sections. The first section examines the basis of conducting evaluations. Several basic assumptions are challenged, among them that well-planned innovative programs have an appreciable effect, that research findings influence educational decisions, and that the use of scientific inquiry is a valid basis for social change. It is suggested that a "judicial method" can involve school staff in constant controversy and keep them learning as they weigh "trial" evidence. The second section discusses methods of knowing whether programs are being implemented. Discussions of a method to calibrate degrees of implementation by teachers, the "banking" of a retrievable group of generalized outcome measures (which would serve Follow Through programs that are ill-served by standardized achievement tests), and the diverse requirements for federal administrators are also included. Uses and misuses of tests and the inadequacies of the testing system for language-minority students are discussed in the third section. The report concludes with discussion, conclusions, and recommendations by conference members; a list of participants; and a list of papers presented. (AS)