THE ARTICLE IS DIRECTED TO THE PROGRAM ADMINISTRATOR OR RESEARCHER WHO IS UNFAMILIAR WITH THE TECHNIQUES OF RESEARCH DESIGN. AFTER AN INTRODUCTION WHICH POINTS OUT THE IMPORTANCE OF COMPARING CLIENTS WITH SIMILAR CHARACTERISTICS WHEN EVALUATING PRETRIAL DIVERSION PROGRAMS, TWO RESEARCH DESIGNS FOR OUTCOME ANALYSIS ARE EXPLAINED. IT IS POINTED OUT THAT PERSONS SELECTED FOR DIVERSION PROGRAMS OFTEN ARE LOW-RISK CLIENTS WHO WOULD ORDINARILY HAVE A LOWER RECIDIVISM RATE THAN THOSE INCARCERATED. A GOOD OUTCOME ANALYSIS RESEARCH DESIGN, THEREFORE, WOULD NOT LOOK ONLY AT RECIDIVISM RATES; IT WOULD ALSO LOOK AT CLIENT CHARACTERISTICS AND COMPARE RECIDIVISM RATES FOR SIMILAR PERSONS IN THE PRETRIAL DIVERSION PROGRAM BEING EVALUATED AND IN A TRADITIONAL CRIMINAL JUSTICE PROGRAM (INCARCERATION OR PROBATION). BOTH EXPERIMENTAL DESIGN AND QUASI-EXPERIMENTAL DESIGN ARE EXPLAINED. EXPERIMENTAL DESIGN RANDOMLY ASSIGNS PERSONS TO ONE OF TWO PROGRAMS. EXPERIMENTS ARE DIFFICULT TO IMPLEMENT BECAUSE PROGRAM ADMINISTRATORS AND CRIMINAL JUSTICE OFFICIALS PERCEIVE LEGAL AND ETHICAL PROBLEMS. A QUASI-EXPERIMENTAL DESIGN, WHICH RANDOMLY SELECTS PERSONS WITH LIKE CHARACTERISTICS FROM COURT RECORDS, IS THEREFORE GENERALLY USED. THREE METHODS OF SELECTING A CONTROL GROUP ARE EXPLAINED, AND EXAMPLES OF ERRORS WHICH SHOULD BE AVOIDED ARE GIVEN. SOURCES OF ADDITIONAL INFORMATION ARE GIVEN IN THE TEXT.