EVALUATORS NEED TO BE AWARE OF THE FACT THAT ADMINISTRATORS OFTEN TAKE CONSIDERABLE LICENSE IN DEFINING AN INTERVENTION AS DIVERSION. FOR INSTANCE, WHEN DIVERSION IS THOUGHT OF AS ANY EFFORT TO MINIMIZE PENETRATION INTO THE CRIMINAL JUSTICE SYSTEM, A POSTCONVICTION RESIDENTIAL ALTERNATIVE TO JAIL COULD BE REGARDED AS DIVERSION. OFTEN DIFFERENCES IN PROCEDURE AMONG PROGRAMS CALLED DIVERSION MAKE COST-EFFECTIVENESS COMPARISONS IMPOSSIBLE. EVALUATORS SHOULD KEEP IN MIND THAT DIVERSION CAN BE A NEW MECHANISM WITHIN A CONVENTIONAL PROCESSING SYSTEM, OR IT CAN BE MERELY NEW TERMINOLOGY FOR EXISTING PROCEDURES. A SECOND PROBLEM CONFRONTING EVALUATORS IS THAT PROGRAM ADMINISTRATORS OFTEN ARE SKEPTICAL ABOUT EVALUATIONS, USUALLY BECAUSE THEY BELIEVE THAT RESEARCHERS PRODUCE NEGATIVE, BIASED INFORMATION THAT ENDANGERS PROGRAM FUNDING, BECAUSE THEY QUESTION THE VALIDITY OF EVALUATIVE RESEARCH. EVALUATORS NEED TO COME TO AN AGREEMENT WITH PROGRAM ADMINISTRATORS ABOUT RESEARCH GOALS. EVALUATORS ALSO NEED TO ATTEND TO THE SUBTLETIES OF THE RESEARCH SETTING AND INTERVENTION STRATEGY, OBSERVING PROGRAM OPERATIONS AND CLIENT-STAFF INTERVENTIONS AND NOT JUST RELYING ON STATISTICS FROM PROGRAM RECORDS. A THIRD SOURCE OF PROBLEMS IN DIVERSION PROGRAM EVALUATION IS INADEQUATE PLANNING FOR DATA ACQUISITION. RESEARCHERS OFTEN USE INAPPROPRIATE DESIGNS THAT FAIL TO ANTICIPATE PROGRAM CHARACTERISTICS. EVALUATORS SHOULD BE PREPARED TO SACRIFICE A DEGREE OF SCIENTIFIC QUALITY FOR FEASIBILITY, BUT MUST DECIDE IN ADVANCE ON THE PROPER BALANCE BETWEEN THE TWO. EVALUATORS SHOULD ALSO BE CERTAIN TO INCLUDE ADMINISTRATIVE AND SERVICE STAFF IN THE PLANNING STAGES OF THEIR RESEARCH. EXAMPLES OF DIVERSION EVALUATION IN CALIFORNIA ILLUSTRATE THE PROBLEMS IDENTIFIED. A LIST OF REFERENCES IS INCLUDED. (LKM)