A battery of four psychomotor tests was developed and evaluated as a measure of the potential of mental patients of varying diagnoses for rehabilitation training. The tests were to be suitable for administration and interpretation by non-professional level technicians. The relationship between performance on the 5 days of tests and patient outcome 1 year later was studied in 996 patients of varied diagnoses. Results indicated that the major share of significant differences on psychomotor performance occurred between two clusters of diagnoses: schizophrenic, manic depressive, and personality disorder on one hand; and chronic brain syndrome plus mentally deficient on the other. Through multiple correlation and regression analyses, accuracy of test prediction was found to be 70 to 78% correct. However, variations in discharge rate and level of performance among diagnostic groups limited the use of a cut-off score suitable for all diagnoses, and the small number of cases per diagnosis prevented development of separate norms. It was concluded that an inverse relationship exists between quality of psychomotor performance and severity of mental disease (coefficient of multiple correlation = .50 to .65). The method developed has applications in the selection of patients for train ng. An 86-item bibliography, 46 tables, and 17 figures are provided. (Author/JD)