OBJECTIVE To assess whether there is an association between the level of in utero cocaine exposure and findings on neonatal cranial ultrasound, controlling for potentially confounding variables. METHODS In a prospective longitudinal study, three cocaine exposure groups were defined by maternal report and infant meconium assay: unexposed, heavier cocaine exposure (>75th percentile self-reported days of use or of meconium benzoylecogonine concentration) or lighter cocaine exposure (all others). Neonatal ultrasounds from 241 well, term infants were read by a single radiologist who was masked to the exposure group. RESULTS Infants with lighter cocaine exposure did not differ from the unexposed infants on any ultrasound findings. After controlling for infant gender, gestational age, and birth weight z scores and for maternal parity, blood pressure in labor, ethnicity, and use of cigarettes, alcohol, and marijuana during pregnancy, the more heavily cocaine-exposed infants were more likely than the unexposed infants to show subependymal hemorrhage in the caudothalamic groove (covariate adjusted odds ratio: 3.88; 95% confidence interval: 1.45, 10.35). CONCLUSIONS This is the first study to demonstrate that ultrasound findings suggestive of vascular injury to the neonatal central nervous system are related to the level of prenatal cocaine exposure. Inconsistency in previous research in identifying an association between prenatal cocaine exposure and neonatal cranial ultrasound findings may reflect failure to consider dose effects.