Data on behavior may not be directly observable or may be distorted by the officer for personal interests. Multiple sources of information must therefore be used. These should include a structured psychiatric interview, observation during the interview, psychological tests, departmental records, medical records, and interviews with collateral informants. Inferences from behavioral data should be based as much as possible on research results. Research in substance abuse and its consequences may be particularly relevant. The recommendation as to whether or not the officer should continue on active duty should be based on departmental police regarding the level of evidence required when an officer is judged to be fit or unfit for duty. The evaluator should not be left to set policy in this area. 9 references.