We investigated the relation between caffeine beverage consumption and spontaneous abortion in 2,967 pregnant women planning to deliver at Yale-New Haven Hospital in 1988-1992. We evaluated coffee, tea, and soda drinking in the first month of pregnancy in interviews before the end of the sixteenth week of gestation. We obtained information on 98.2% of the pregnancies (including 2,714 singleton livebirths and 135 spontaneous abortions). As compared with abstention from caffeine beverages (coffee, tea, and soda), the adjusted odds ratios for spontaneous abortion associated with consumption of 1-150, 151-300, and > 300 mg caffeine daily were 0.81 [95% confidence interval (CI)) = 0.54-1.20], 0.89 (95% CI = 0.48-1.64), and 1.75 (95% CI = 0.88-3.47), respectively. Drinking > or = 3 cups of tea or coffee was associated with elevated risks of spontaneous abortion (adjusted odds ratio = 2.33, 95% CI = 0.92-5.85; and adjusted odds ratio = 2.63, 95% CI = 1.29-5.34, respectively). These results, if replicated, suggest that some ingredient (or correlate) of tea or coffee may account for some of the observed association of caffeine with spontaneous abortion. In this study, caffeine consumption is more strongly related to spontaneous abortion than alcohol or cigarette use in early pregnancy.