Although studies have succeeded in devising statistically sophisticated prediction schemes for use in screening or placing clients in rehabilitation programs, they typically lack comparison with simpler methods, e.g., the single best predictor method. In this study of 296 disabled clients referred to the St. Louis Jewish Employment Service for vocational evaluation, such a comparison was made. Variables included race, sex, age, education, 13 Wechsler Adult Intelligence Scale subtest and total scores, and 10 ratings of workshop performance. Employment status, assigned to the evaluated clients, was a further seven-categorized variable. Three prediction techniques were employed: (1) the multiple linear regression technique, (2) the multiple linear regression of factor scores, and (3) the single best predictor method. A cross-validation sample was utilized. The methodology was described and the results discussed. Two points were demonstrated: (1) the most useful prediction model may be the least statistically sophisticated model; and (2) shrinkage in predictive power upon cross-validation may be considerable, especially when regression models are used. Possible uses of prediction schemes were considered. (TL)