During the fall of 1984, we conducted a survey of umbilical cord blood lead levels of 723 live births that occurred at 5 hospitals located in 5 cities in California. Historical ambient air lead levels were used as a qualitative surrogate of air and dust exposure. The area-specific cord blood means (all means approximately 5 micrograms/dl), medians, deciles, and distributions did not vary among locations. The California distributions included means that were lower than the 6.6 micrograms/dl reported in Needleman et al.'s Boston study in 1979. Indeed, the entire California distribution was shifted to the left of the Boston study distribution, even though 3% of the California cord lead levels exceeded 10 micrograms/dl--the level above which Needleman et al. have documented psychoneurological effects in children during the first few years of life. Fourteen percent of premature babies had cord blood lead levels above 10 micrograms/dl. The association between prematurity (i.e., less than 260 d gestation) and elevated (greater than 5 micrograms/dl) cord blood lead was observed in all hospitals and yielded a relative risk of 2.9 (95% CI: .9, 9.2) and a population attributable risk of 47%. Further research is needed to confirm this association and to explore the roles of endogenous and exogenous sources of lead exposure to the mothers who give birth to premature infants.