BACKGROUND: This study investigates the prevalence of under-reported use of tobacco among Bangladeshi women and the characteristics of this group. METHODS: The 1999 and 2004 Health Survey for England included 996 Bangladeshi women aged 16 years and above, 302 with a valid saliva sample and 694 without. The main outcome measure was the prevalence of under-reported tobacco use. RESULTS: Fifteen per cent of Bangladeshi women with a saliva sample under-reported their personal tobacco use. Under-reporters were very similar to self-reported users except for being much more likely to report chewing paan without tobacco (47% versus 9%, P < 0.001). Under-reporters differed significantly from cotinine-validated non-users in most respects. Regression analyses confirmed that under-reporters and self-reported users were similar in age, education level and exposure to passive smoking. Under-reporters were older and less educated than cotinine-validated non-users. Both self-reported users [odds ratio (OR): 0.11, 95% confidence interval (CI): 0.04-0.30] and cotinine-validated non-users (OR: 0.42, 95% CI: 0.20-0.89) were far less likely to report chewing paan without tobacco compared with under-reporters. CONCLUSIONS: Contrary to our a priori hypothesis, under-reporters were not young, British-born, English-speaking women likely to be concealing smoking but resembled self-reported tobacco users except for being much more likely to report chewing paan without tobacco.