The available objective evidence suggests that the accuracy of predicting which students will succeed in a particular graduate school is often no better than modest, especially if such predictions are based only upon a test or a grade record. Taken together these two types of predictors do a reasonably good job, considering the restricted range of ability involved. The best way to improve selection of graduate students will be to develop improved criteria of success. This is no small job for graduate faculties, but it carries the promise of more effective utilization of talent and greater assurance of equity in admitting students to advanced levels of training and the privilege associated with such programs. (Author)