Seven hundred sixty-six black and 462 white non-Hispanic mothers who had given birth in Alameda County (Calif.) in 1987 were interviewed, the number of controls (birth weights of 3,000 or more grams) in each ethnic group equal to the number of cases (birth weights of 500 to 2,499 grams). Women who used heroin and/or methadone during pregnancy were excluded. After adjusting for confounders, the relative risk of low birth weight for women who use cocaine throughout pregnancy was found to be 4.0 (95 percent CI = 1.8, 8.9). Maternal cocaine use increased the relative risk of low birth weight, pre-term low birth weight, and intrauterine growth retardation, but not if ceased during the first trimester of pregnancy. In assessing the total problem, it is speculated that eligible women who could not be located for interview were more likely than interviewees to use cocaine, and that, among interviewees, underreporting of cocaine use was more prevalent than overreporting. 3 tables, 17 references.