The conclusion of this study is that the most valid figure-classification test for predicting grades at the Naval Academy is a moderately speeded test that about 70% of the candidates can finish. The score on speeded tests is a function of two orthogonal factors, the factor of ability being more valid than the speed factor. Ordinarily one assumes that those who are able to solve difficult problems in a test will work more quickly than those who cannot. This study suggests that this is not always the case. The hypothesis is advanced that not every mark on the answer sheet represents a subject's reasoned conception of an adequate solution. Many of the answers may be guesses, although not necessarily random answers. The group of subjects who reach the end of a speeded test may include some who can solve the problems quickly and also some who answer the problems before they complete the solutions. These subjects may not understand what they are expected to do or they may prefer to guess in order to work quickly. Thus, while it may not always be possible to describe the full effects of speededness by stating the proportion of answers marked, nevertheless the evaluation of a test cannot be complete without some knowledge of the speed at which the subjects were required to work.