A bivariate analysis of infant mortality in the 1986/87 Ondo State Demographic and Health Survey data indicates that children of secondary school graduates experienced a higher rate of infant mortality than children of less educated mothers. Although this pattern has also been shown in a few other Demographic and Health Surveys, this paper explores the Ondo State data to explain why infant mortality showed such a counterintuitive pattern in relation to maternal education. This search for an explanation of the pattern started with an examination of the data for errors and then proceeded to examine the importance of some intermediate mechanisms that had been suggested for the education-child survival relationship. The results suggested that data errors, use of health services and quality of maternal care were not enough to explain the relationship. Rather, results of a logistic regression analysis showed that breastfeeding duration and maternal age at childbirth were statistically the most significant variables for predicting infant survival in Ondo State. The inverse relationship between mother's education and infant mortality rates that was not shown by bivariate analysis came out clearly only after controlling for the effect of breastfeeding duration. The linkage between these findings and broader social and economic realities of Nigeria was provided through reviews of available information. The conclusion from the study was that, although breastfeeding and maternal age showed up as the most statistically significant variables, they apparently are just the variables that effectively captured the effects of the harsh economic conditions, especially among secondary school graduates, that prevent most young mothers from translating their child-rearing ideals into reality.