The screening hypothesis suggests that intereducational earnings differentials, even when standardized for differences due to noneducational factors, reflect no direct productivity-enhancing effects of education but only its effects as a device for signaling preexisting ability differences. But three predictions in the spirit of the hypothesis are not in fact borne out. First, rates of return to uncompleted courses (dropouts) are as high as to completed courses. Second, standardized educational differentials rise with age, although employers have better information about older employees' abilities. Third, if screening is the main function of education, it could be done more cheaply by simpler testing procedures. Yet these are not used widely despite the profits that could, according to the hypothesis, be made by those developing them.