The "productivity puzzle," or gender gap in research productivity, is often thought to be due, in part, to gendered child care responsibilities. The time, energy, and money devoted to child-rearing can reduce the research productivity of scholars, especially women. However, there is little systematic research on this issue. The present study contributes to the literature by exploring the productivity puzzle in light of the influence of eight age sets of children. Data are from the National Research Council's 1995 Survey of Doctoral Recipients. They refer to 11,231 PhDs, in the sciences and engineering, who are employed in academia. Controlling for children, structural factors and personal characteristics of the PhDs, the present study finds that women still publish significantly less than men. Second, productivity is higher for PhDs with children less than 11. Third, productivity is relatively low for women with young children. Fourth, for the social sciences, the area with the highest proportion of women PhDs, gender is unrelated to productivity, but women in the social sciences who have young children have relatively low productivity. The leading predictors of productivity included location in a research university and effort (hours worked). The full model explains 39% of the variance in productivity. Children are not a strong predictor of productivity, but the influence that they do have followed a gendered pattern.