This paper re-examines the effects of breastfeeding on long-term educational outcomes using longitudinal data on siblings. While family-fixed effects allow controls for all shared family factors, these estimators are sensitive to compensating or reinforcing behaviors by parents. These biases may be particularly important for estimating the effects of parental investment such as breast feeding, where sibling discordance may be difficult to treat as a random outcome and may result in persistence in differential investments between siblings. This paper uses a unique question asked to adolescent siblings about parental favoritism to adjust for potential reinforcing behavior by parents. Standard fixed effects estimates suggest important long-term educational effects of breastfeeding; however, these effects are uniformly eliminated after focusing on families who treat siblings equally. These findings shed light on the mechanisms linking associations between breastfeeding and longer term outcomes. Copyright © 2010 John Wiley & Sons, Ltd.