Objective: We explored the relationships among measures of psychosocial well-being, maternal size, and smoking in predicting infant size at birth. Methods: Participants in this population-based cohort study were drawn from public health prenatal clinics in Jefferson County, Alabama during 1985-1988. Para 1 and 2 women were screened for 11 risk factors for low birth weight, including small stature, a previous low birth weight infant, and smoking. Results: Poor scores on five of six psychosocial scales, as well as on a combined profile, were associated with a significantly higher relative risk of fetal growth retardation (FGR) only in thinner women, defined as having a body mass index less than the median (relative risk [RR] 2.11, 95% confidence interval [CI] 1.47, 3.04). A significant association between the psychosocial profile and birth weight was demonstrated for thin women in a multivariate analysis adjusting for gestational age, race, infant sex, and smoking (P = .0003). The relationship remained significant when hypertension, alcohol and drug use, and weight gain were added to the model (P = .003). In women with a body mass index above the median, a poor psychosocial profile showed little association with FGR (RR 1.20, 95% CI 0.73, 1.98) and did not have a significant association with birth weight. A poor profile had a greater association with FGR in nonsmokers (RR 2.04, 95% CI 1.29, 3.22) than in smokers (RR 1.4, 95% CI 0.95, 2.06). Conclusions: Greater pre-pregnancy weight for height appears to protect against the adverse effects of a poor psychosocial profile in a population of poor, primarily black women. In thinner women, both smoking and a poor psychosocial profile were associated with a substantially increased rate of FGR, indicating a subgroup of women who may receive greater benefits from intervention programs.