The use of ability testing for job selection has become widespread in the Federal Government and in the U.S. Employment Service, which assists private sector employers. The justification for the practice is based largely on research findings claiming a high level of validity for such tests in predicting job performance. More recently, such claims have been translated into the dollar increases in productivity that would result if optimal testing strategies were used for selecting employees for jobs. However, a careful review of the claims indicates that they are not supported by research evidence. The utility of any selection procedure depends on (1) its ability to predict worker performance better than alternatives; (2) the selection ratio of employer openings to applicants; and (3) the economic value of the better employee selection relative to the costs of the selection. On the first point, the evidence that general ability tests are superior to other selection criteria in predicting the various indicators of worker performance is not convincing. Furthermore, much of the research on ability testing for job selection ignores the second point, and much contains many unsubstantiated conclusions and overstatements with regard to the third point. (MN)