Problems in conducting an outcome evaluation under uncertain knowledge of effects are addressed. This evaluation was designed to obtain outcome assessments of a National Science Foundation Workshop, which was working with 20 high school chemistry teachers and 20 high school biology teachers to stimulate effective teaching, assist teachers in developing instructional modules, and provide a broader network of scientists to relate to high school science activities. The initial evaluation design called for self-reports of program impact from participants, but hard data, in terms of test scores, became necessary to document impact more directly. Three measures were selected to gather data on workshop impact: (1) a questionnaire determining the reactions of participants; (2) a questionnaire examining the utility for teachers of activities in the workshop as translated into their teaching; and (3) an instrument, Our Class and Its Work (OCIW), developed by M. J. Eash and H. Waxman (1983), which was used to gather information from students of workshop teachers (N=442) and students of control teachers (N=384). Students' perceptions of classroom approach paralleled teacher self-reports; when teachers reported a change in classroom behavior, OCIW student data suppored a change. The problems in outcome evaluation illustrated by this study indicate the need to rely on more than just test scores in evaluating program impact. (SLD)