THE FLORIDA INTENSIVE SUPERVISION PROJECT (ISP), WHICH INVOLVED OVER 9,000 SUBJECTS, IS DESCRIBED, WITH EMPHASIS ON METHODS FOR THE SELECTION, ASSIGNMENT, AND DIFFERENTIAL TREATMENT OF SUBJECTS. PROBLEMS ENCOUNTERED IN SAMPLING, POSSIBILITIES FOR MATCHING, AND RESTRICTIONS ON MATCHING TECHNIQUES ARE DISCUSSED, TOGETHER WITH STATISTICAL TACTICS AND PROCEDURES. THE DISCUSSION TOUCHES ON COMPARABILITY OF EXPERIMENTAL AND COMPARISON GROUPS, SPECIFIC PROBLEMS IN SAMPLING AND MATCHING, THE EXISTENCE OF UNCONTROLLED DIFFERENCES AND THEIR STATISTICAL TREATMENT, AND PROBLEMS OF CLIENT MOBILITY AND EXPERIMENTAL CONTROL. SUGGESTIONS FOR EXPERIMENTAL DESIGNS AND STATISTICAL ANALYSIS TECHNIQUES OF USE IN AVOIDING UNINTERPRETABLE AND UNGENERALIZABLE FINDINGS IN FIELD EVALUATION STUDIES ARE OFFERED. THE FOLLOWING MAJOR RECOMMENDATIONS ARE PRESENTED: EMPHASIS OF QUALITY OF SAMPLING OVER QUANTITY, CONSIDERATION OF SUBJECT FACTORS IN THE INITIAL DESIGN OF RESEARCH, MEASUREMENT OF PRESUMED CHARACTERISTICS OF SUBJECTS ON INDEPENDENT AND DEPENDENT VARIABLES OF INTEREST, PERFORMANCE OF SUBTLE STATISTICAL ANALYSES AS NEEDED TO ILLUMINATE POSSIBLY CONFOUNDING SUBJECT DIFFERENCES, AND PREPARATION FOR THE PRACTICAL PROBLEMS TO BE EXPECTED IN LARGE-SCALE FIELD RESEARCH. A LIST OF REFERENCES IS INCLUDED. (LKM)