We conducted a retrospective cohort study to evaluate the relation between the level of total trihalomethanes in drinking water and adverse birth outcomes. The study population comprised women residing in an area with municipal surface water who had a singleton birth in Nova Scotia between January 1, 1988, and December 31, 1995, or a pregnancy termination for a major fetal anomaly. We found little association between trihalomethane level and the outcomes related to fetal weight or gestational age, but we found an elevated relative risk for stillbirths for average trihalomethane levels during pregnancy of 100 microg/liter or greater (adjusted relative risk = 1.66; 95% confidence interval = 1.09-2.52) relative to women exposed to trihalomethane levels of 0-49 microg/liter. We saw little evidence of an elevated prevalence or dose-response pattern for congenital anomalies, with the possible exception of chromosomal abnormalities (adjusted prevalence ratio = 1.38 and 95% confidence interval = 0.73-2.59 for women exposed to trihalomethane levels of 100 microg/liter or greater).