The risk of delivering a low-birth-weight infant as the result of exposing a nonsmoking pregnant woman to environmental tobacco smoke (ETS) is not well defined. The method of ascertaining ETS exposure during pregnancy may explain the lack of consistent study findings. In a large sample of pregnant women, we compared distributions between two methods of ETS exposure: self-report and cotinine, a nicotine metabolite, from serum. At livery, subjects were asked about duration and location of exposure to ETS during their second trimester. A single cotinine measurement was assayed from serum collected at 15-19 weeks gestation (limit of detection=0.05 ng/mL). Self-reported (hours per day) ETS exposure was correlated (r=0.38) with cotinine concentration. Regression analysis revealed that while self-reported ETS was significantly associated with (log) cotinine, it did not explain a large amount of total variation. While 72% of subjects reported no exposure to ETS, almost all had measurable levels of cotinine. Studies of pregnant women based upon an hours per day ETS question have likely misclassified a sizable portion of ETS-exposed women as "unexposed." Since there is recent evidence that low levels of ETS exposure result in unfavorable pregnancy outcomes, these studies have underestimated the effect of ETS.