A previous study of pregnant women whose water supply was potentially contaminated revealed a significant increase in spontaneous abortions among those drinking tap as opposed to bottled water, regardless of exposure to the contamination. The relation between reported prenatal water consumption and risk of spontaneous abortion was therefore analyzed in an independent cohort of pregnancies seen in 1981-1982 at a health maintenance organization in three counties in northern California. The study used a nested case-control design. Since early miscarriages are left-truncated, gestational age at which a pregnancy comes to medical attention may confound results of pregnancy outcome studies. Risk set analyses were therefore conducted in two stages: 1) the life of table-adjusted risk for those drinking mainly bottled water was 8.4%; the risk for those drinking mainly tap water was 12.5%, and 2) the Cox proportional hazards model was used to control for multiple confounders, yielding a hazard ratio for spontaneous abortion of 1.5 (95% confidence interval 1.1-2.0) for consumers of tap water compared with bottled water. Tap water drinkers whose home source of water included ground water had the greatest risk (13.8) and, after controlling for confounders, their hazard ratio was 1.7. Based on external data from comparable studies, bottled water drinkers appeared to have had unusually low risks, and tap water drinkers who received ground water may have had slightly high risks. Inconsistencies in the reporting of tap water consumption suggest recall bias. Causal factors could not be ruled out, however, although no reproductive toxins, either biological or chemical, have yet been identified in the tap water in this area.