In a training situation, the standard procedures for predicting performance entail long delays between the request for a predictive instrument and its delivery. Methods were developed for constructing psychological prediction instruments at the time of the request, rather than waiting for the maturation of the criterion data. The two synthetic predictive methods developed were about as effective as instruments developed by classical experimental validation methods, with the advantage that they are immediately available and can be used even before a course has been given for the first time. The synthetic methods use estimates of performance made by expert judges as substitutes for actual performance. One technique is based on the experts' ratings of the individual trainees, and the second is a factorial evaluation based on a rating of the importance of various factors to the success of the training. The type of training was found to influence the effectiveness of the predictors, but for all the military training situations tested, the criteria were adequate. Evaluations by several experts were found to be more effective than those by an individual expert. The amount of professional experience did not appear to be a factor in a judge's ability to apply the techniques successfully. (Author/MT)