THE TRADITIONAL EXPERIMENTAL-CONTROL GROUP PROCEDURES USED IN EVALUATING PROGRAMS ARE RARELY SUCCESSFUL WITH DELINQUENCY PREVENTION PROGRAMS SINCE SELECTION OF RANDOM SAMPLES IS NEARLY IMPOSSIBLE AND IS USUALLY REPLACED WITH RECRUITMENT OF A COMPARISON GROUP. FURTHERMORE, THE PROCEDURES CREATE STRESS WHICH ENDANGERS THE OVERALL SUCCESS OF THE PROGRAMS AND PRODUCE HOSTILITY ON THE PART OF PROGRAM STAFF AGAINST RESEARCHERS ATTEMPTING TO GAIN KNOWLEDGE ABOUT THE IMPACTS OF A PROGRAM. COMPLEX EVALUATION PROGRAMS ONLY INCREASE THE LIKELIHOOD OF STRESS. AN ILLUSTRATION OF AN INAPPROPRIATELY COMPLEX RESEARCH DESIGN IS PRESENTED TO SHOW THAT EACH ATTEMPT TO INTRODUCE AND CONTROL FOR A NEW EXPERIMENTAL VARIABLE INCREASES THE RISK OF COMING INTO CONFLICT WITH THE COMMUNITY, ANOTHER AGENCY, OR MEMBERS OF THE PROJECT STAFF. THESE TYPES OF CONFLICTS ENLARGE THE POSSIBILITY OF ERROR JUDGMENTS ABOUT THE RELATIONSHIPS AND INTERACTIONS AMONG VARIABLES AND ABOUT THE INTEGRITY OF THEIR USE. IF SOPHISTICATED RESEARCH SKILLS WERE DIRECTED TOWARD THE ANALYSIS OF DATA GATHERED FROM LARGER NUMBERS OF CASES AND FROM MORE READILY AVAILABLE DATA SOURCES, MANY OF THE NEGATIVE CONSEQUENCES OF EVALUATIONS COULD BE MINIMIZED. HOW FINDINGS ARE TREATED IS AN IMPORTANT FACTOR SINCE THE EVALUATION RESULTS ARE PRESENTED IN A POLITICAL ARENA AND CAN HAVE GREAT IMPLICATIONS FOR THE PROGRAMS. OBJECTIVE ASSESSMENT MAY DAMAGE FLEDGLING PROJECTS WITHOUT ALTERING INEFFECTIVE BUT ENTRENCHED PROGRAMS. RESEARCHERS SHOULD BE OBSERVERS RATHER THAN DOERS AND SHOULD FOCUS ON THE TASK OF OBSERVING, GATHERING, AND MARITAL DISPUTES, INDIVIDUAL PROBLEMS, NEIGHBORHOOD EVALUATION. REFERENCES ARE PROVIDED. (DAG)