The purpose of this study was to discover what amount of speededness of a nonverbal-reasoning test would result in the highest validity as a predictor of grades at the U.S. Naval Academy at Annapolis. In order to do this, a group of figure-classification tests were given in 1948 to all the midshipmen then in their first summer at the Academy. All of the tests were administered with twelve-minute time limits but different tests included ten, twenty, or thirty items. Each of the men took five different tests including at least one at each speed level. Correlations were found between scores on the different tests and also between the tests and the grades in the first year's work at the Academy. These intercorrelations were analyzed by factorial methods. The following conclusions were drawn from this study: that the moderately speeded tests were the most valid, and that the ability to answer the problems correctly had a slight negative correlation with the tendency to answer the problems quickly.