A review of the literature fails to reveal any definitive study of the validity of item interaction variances for predicting criteria. The tools for such a study are developed here, in the form of programming for an electronic computer, and are applied to two criteria of sales success in Company X, using the items of the Strong Vocational Interest Blank. Using a 20-pair key built with the new tools, a criterion of quota performance is predicted with a cross-validity of .34; making allowance for effects of length and reliability, it is estimated that a validity in excess of .50 can be achieved, accounting for over twice as much criterion variance as the conventional linear criterion key. A criterion of continuation with Company X is found to be more complicated. Several paradigms portraying some possible mechanisms of these relations are given in the final section, and illustrated.