The smoking habits of 300 consecutive pregnant women and of other members of their household were obtained by interview at 16 and 32 weeks' gestation and after delivery. Urine samples for estimation of cotinine were collected at the same times during pregnancy and on admission in labour. Both self-reports of active maternal smoking and urinary cotinine levels were significantly associated with lighter babies but urinary cotinine correlated the better. Active maternal smoking was associated with a decrease in birth-weight of 12 g for every cigarette smoked in a day while there was a decrease of 25 g in birth-weight for every microgram of cotinine/mg of urinary creatinine. Passive maternal smoking was associated with a decrease of 66 g in mean birth-weight but this decrease was not statistically significant. Self-reports of active and passive smoking accounted for 43% of the variation in urinary cotinine levels suggesting that urinary cotinine assays can be used to assess and monitor exposure to tobacco smoke in pregnancy.