Little is known about the accuracy of usual-care providers' detection of pregnant smokers. This study explored the proportion of pregnant women misclassified as nonsmokers by midwives in a public antenatal clinic. A sample of 204 women whom midwives classified as nonsmokers were asked to complete a self-administered questionnaire and to provide a urine specimen for cotinine analysis. Results indicate that midwives failed to detect a significant proportion of smokers. The conservative estimate of the proportion of midwife-identified nonsmokers who could be reclassified as smokers on the basis of the questionnaire and urinalysis procedures was 7.4% (95% CI 3.8-10.9%), the medium estimate was 8.8% (95% CI 4.9-12.7%) and the worst-case estimate was 15.2% (95% CI 10.3-20.1%). To increase the coverage achieved by smoking-cessation programs, antenatal clinics should consider incorporating biochemical measurements into routine screening procedures. Future studies examining smoking status in pregnancy should detail the methods used to classify subjects and document response rates in relation to each self-report and biochemical measurement.