A proper understanding of infant health requires the integration of sooloeconomic, behavioral, and biomedical models. A methodology is presented for assessing the effects of “underlying” social factors and proximate behavioral and biomedical factors on infant morbidity, growth, and mortality. The method is applied to data collected from over 3,000 children in Cebu, Philippines, over the first 2 years of life. Data were collected between 1983 and 1985. A central theme is that mothers recognize certain observable and nonobservable threats to the health of their infants, and that the mothers take measures to reduce the risk from such threats. It is shown that if conventional statistical techniques (which do not take such behaviors into account) are used, the estimates of the effect of the risk factors on health are incorrect. Procedures for obtaining correct estimates are described. The application of the methodology is illustrated by modeling childhood diarrhea, and by showing how maternal education Induces behavioral changes, and how these changes, in turn, induce changes in the prevalence of childhood diarrhea. Am J Epidemiol 1991; 133:185-201.