Valid assessment of professional competence has proven to be an elusive goal. Objective tests, direct observation of performance, overall ratings of competence, and simulations have been tried and found wanting in one way or another. Objective test items are criticized as being unrealistic and therefore invalid. Direct observation tends to be very unreliable and thus invalid. Simulations and overall ratings of competence share both of these flaws to some extent. The difficulties inherent in evaluating professional competence are outlined, and some ways to minimize the impact of these difficulties are suggested. A general framework is proposed for evaluating the validity of measures of competence, and this framework is used to examine the strengths and weaknesses of three approaches to the assessment of professional competence: (1) direct observation; (2) simulation; and (3) objective tasks. In evaluating the validity of such an assessment, it is important to give special attention to the weakest links in the argument. For performance testing, evaluation and generalization are the weakest links. For objective tests, extrapolation is the weakest link. For simulations, any of the links can be weak or strong depending on the simulation. There is a 28-item list of references. (Author/SLD)