THE NATURE OF THE COOPERATION ACQUIRED FOR AN EVALUATION PROJECT IS A MAJOR DETERMINANT OF THE DESIGN AND OUTCOME OF THE RESEARCH. THE CONSENT OF AGENCY AUTHORITIES IS NEEDED, BUT THE SOCIAL SCIENTIST MUST CONSIDER THE MOTIVATION OF THE AUTHORITIES IN GRANTING PERMISSION; THIS FACTOR MAY LEAD TO DIFFERENT LEVELS OF COOPERATION AFFECTING THE RELIABILITY, VALIDITY, PRECISION, EFFICIENCY, AND RELEVANCE OF THE RESEARCH PROJECT. IF THE AGENCY DIRECTOR AGREES TO A RESEARCH PROJECT EXPECTING IT TO BRING STATUS, PRESTIGE, OR FUTURE FUNDING TO THE AGENCY, THEN COOPERATION WILL OCCUR ONLY WHEN THE RESEARCH IS PROVIDING SYSTEMATIC EVIDENCE THAT THE AGENCY, ADMINISTRATION, AND STAFF ARE OPERATING EFFICIENTLY AND PRODUCTIVELY. COOPERATION OF THE STAFF IS IMPORTANT, AND RESEARCHERS SHOULD REALIZE THAT A STATEMENT FROM THE AGENCY DIRECTOR THAT THE STAFF WILL COOPERATE IN A PROJECT DOES NOT GUARANTEE THEIR COOPERATION. CONTROL OVER CLIENTS IN EVALUATION RESEARCH IS ANOTHER PROBLEM. WHEN THE CLIENT IS A RECIPIENT OF SERVICES, SUCH AS IN AN EVALUATION OF TREATMENT PROGRAMS, PARTICIPATION IN THE RESEARCH IS VOLUNTARY. THE SAMPLE, THEREFORE, IS NOT NECESSARILY RANDOM AND THE DROPOUT RATE IS LIKELY TO BE HIGH. THE RESEARCH DESIGN MUST BE ALTERED TO ACCOMMODATE THESE FACTORS. IN A SURVEY WITH CLIENTS AS RESPONDENTS, THE TYPE AND LEVEL OF CONTROL OVER CLIENTS WILL AFFECT THE RESPONSE RATE; E.G., INSTITUTIONALIZED RESPONDENTS MIGHT ANSWER IN A WAY PERCEIVED TO IMPROVE THEIR CHANCES FOR RELEASE. IN INTERVIEWS, THE SITUATION AND THE RELATIONSHIP BETWEEN AN INTERVIEWER AND INTERVIEWEE ARE REFLECTED IN VERBAL RESPONSES. BEHAVIOR SAMPLES OFTEN ARE SUGGESTED AS BETTER PREDICTORS OF BEHAVIOR, BUT CRITERION VARIABLES MUST BE DECIDED ON PRIOR TO RESEARCH. WHEN THE CRITERION VARIABLE IS AN UNSPECIFIED SET OF FUTURE ACTIONS, AN ATTITUDE MEASURE MAY BE THE BEST PREDICTOR. (DAG)