Discriminant analysis of Holland's Self Directed Search and Vocational Preference Inventory significantly improved the predictive validity of his classification scheme, over the traditional method of assigning high point codes. These results were obtained with both an initial and a replication sample of undergraduates pursuing various degrees. The average increase in classification accuracy from discriminant analysis over that obtained from high point codes of one letter was about 16%. The criterion for forming discriminant groups in this study was the subject's own choice of anticipated occupation. The full theoretical sophistication of the six-letter high point code can be achieved without the use of the 720-cell design as would be required in the high point code method. Using discriminant analysis, one can make inferentially stable predictions with as few as 60 subjects and have better than reasonable confidence that the results would replicate on an independent sample. (Author/CP)