The negative results to an assessment of a Bank Street-sponsored Follow Through program raised questions about conventional ways of assessing the effects of educational programs. Children in a Follow Through program using the developmental-interaction approach and children in conventional classrooms were observed and tested for comparison. The classroom observation data were quite different for the two types of classrooms, but there were no significant testing differences attributable to the program. Test responses are affected by unique factors such as examiner variables. Certain aspects of the test situation receive little attention but are especially relevant to the assessment of educational programs. Test responses also assess an individual's ability to generalize and transfer from one situation to another. Children in a conventional classroom, with its emphasis on the teacher's dominant role, are better attuned to the testing situation, in which the examiner questions and the child responds. There is, as well, greater uniformity of experience in the conventional classroom. In the child's passive role in the testing situation, demonstration of cognitive ability is quite dependent on language usage. The approach that may be appropriate for assessing some kinds of functions and groups may lead to misevaluation of others. What is needed is a flexible, responsive, diversified use of formative evaluation and a delay in summative evaluation until a program has had a chance to take effect. (KM)