Past published studies establishing a significant, positive link between education and agricultural productivity have used Asian data almost exclusively. With one exception, these studies have not checked to see if the positive impact of education comes about only by screening for ability. Using household level data from coffee regions of Kenya and Tanzania, this study finds that households in which the agricultural decision-maker is numerate and literate produce 30 percent more agricultural output, holding other variables constant. This estimate of the impact of education is markedly higher than those made using years of schooling as an independent variable. Increasing cognitive skills beyond basic numeracy and literacy has no additional impact on production. Evidence from Kenya (but not from Tanzania) suggests that reasoning ability and the presence of at least one educated person in the household also increase productivity substantially.