To investigate whether drinking tap or bottled water during pregnancy affects the risk of spontaneous abortion, we asked questions about water consumption in a large case-control study (626 cases, 1,300 controls). The study ascertained cases from hospital pathology laboratory reports of pregnancies that began in 1986 and obtained controls from birth certificates. The crude odds ratio for consumption of any vs no cold tapwater at home during the first trimester was 1.2 (95% confidence interval = 1.0-1.5), with no dose-response effect. The crude odds ratio for any bottled water consumption was 0.79 (95% confidence interval = 0.65-0.96), with a downward trend by amount consumed. Adjusting for many potential confounders did not alter these associations appreciably, although some variables appeared to be effect modifiers. The point estimates were stronger among women who were more difficult to contact, suggesting the possibility of bias.