THIS CASE STUDY CONCERNS TWO PLANNED VARIATION EVALUATIONS--ONE FOR OPERATION HEAD START AND ONE FOR A PROGRAM CALLED FOLLOW THROUGH. BECAUSE IT IS IMPOSSIBLE TO ISOLATE CHILDREN AND ASSUME THEY WILL HAVE NO EDUCATION AT ALL, CLASSICAL EXPERIMENTAL DESIGN IS INAPPROPRIATE FOR EVALUATION OF SUCH SOCIAL PROGRAMS. PLANNED VARIATION ASSUMES THAT A MULTIPLICITY OF TECHNIQUES WILL BE USED AND A VARIETY OF OUTCOMES WILL BE OBTAINED. THE OBJECT IS TO IDENTIFY THE TECHNIQUES ASSOCIATED WITH THE MOST POSITIVE OUTCOMES. THE KNOWLEDGE ABOUT LEARNING PROCESSES, PARENT-CHILD INTERACTIONS, PARENT-SCHOOL INTERACTIONS, AND TEACHER TRAINING WHICH RESULTED FROM THESE EVALUATIONS IS BRIEFLY SURVEYED. THE EVALUATOR IS WARNED NOT TO EXPECT CONCRETE FINDINGS. RATHER, THE KNOWLEDGE GENERATED WILL BE BROAD. HOWEVER, AS THESE EXPERIMENTS HAVE SHOWN, IT CAN BE EXTREMELY USEFUL. VARIOUS EXPERIMENTAL AND ETHNOGRAPHIC APPROACHES WHICH CAN HELP GENERATE KNOWLEDGE AND ALTERNATIVE STRATEGIES ARE EXPLORED. THE EVALUATOR IS URGED TO HAVE A DEFINITE PLAN, TO BE FLEXIBLE IN RECONCILING CONFLICTING DATA, AND TO CARRY OUT ALL EVALUATIONS IN CONJUNCTION WITH SOMEONE WHO IS AN EXPERT IN THE FIELD. REFERENCES ARE PROVIDED. (GLR)