We studied the sensitivity of testing the newborn infant's hair, meconium, and urine in detecting gestational cocaine exposure. The infants were born to 59 women who were interviewed to determine their use of cocaine during pregnancy and whose hair was analyzed for the presence of cocaine. Regression analysis was used to evaluate the relationship between cocaine in newborn hair and in maternal hair. Radioimmunoassay of infants' hair and gas chromatography-mass spectrometry of meconium were more sensitive than immunoassay of urine (p less than 0.02), which failed to identify 60% of cocaine-exposed infants. The quantity of benzoylecgonine in the newborn infant's hair correlated best with the proximal-segment maternal hair, representing the last 12 weeks of antepartum hair growth (R = less than R less than 0.83). Approximately half (52%) of the variation in infants' hair was explained by variation in the proximal maternal hair segment. Correlation (R = 0.77) and explained variation (59%) improved slightly when premature infants (n = 9) were excluded. We conclude that analysis of the newborn infant's hair by radioimmunoassay or of meconium by gas chromatography-mass spectrometry is more sensitive than analysis by immunoassay of urine, and can detect fetal cocaine exposure that occurred during the last two trimesters of pregnancy.