This study was undertaken to determine the relation between self-reported number of cigarettes smoked per day and urine cotinine concentration during pregnancy and to examine the relations between these two measures of tobacco exposure and birth weight. Data were obtained from the Smoking Cessation in Pregnancy project, conducted between 1987 and 1991. Cigarette smoking information and urine cotinine concentration were collected for 3,395 self-reported smokers who were receiving prenatal care at public clinics in three US states (Colorado, Maryland, and Missouri) and who delivered term infants. General linear models were used to quantify urine cotinine variability explained by the number of cigarettes smoked per day and to generate mean adjusted birth weights for women with different levels of tobacco exposure. Self-reported number of cigarettes smoked per day explained only 13.9% of the variability in urine cotinine concentration. Birth weight declined as tobacco exposure increased; however, the relation was not linear. The sharpest declines in birth weight occurred at low levels of exposure. Furthermore, urine cotinine concentration did not explain more variability in birth weight than did number of cigarettes smoked. These findings should be considered by researchers studying the effects of smoking reduction on birth outcomes.