We examined the relationship of machine-estimated nicotine yield by cigarette brand with the level of cigarette consumption and two biochemical measures of smoke exposure (expired-air carbon monoxide and plasma thiocyanate) in a large, population-based sample of smokers (N = 713). The lower the nicotine yield of the cigarette, the greater the number of cigarettes smoked per day. Prior to adjusting for number of cigarettes smoked per day, nicotine yield was not related to the actual measures of smoke exposure. Smokers of ultralow-yield cigarettes had laboratory tests of smoke exposure which were not significantly different from those of smokers of higher-yield brands. Only after adjustment for number of cigarettes smoked per day did nicotine yield become significantly related to expired-air carbon monoxide and to plasma thiocyanate. In multivariate analysis, the number of cigarettes smoked per day accounted for 28 per cent and 22 per cent of the variance in observed expired-air carbon monoxide and plasma thiocyanate levels, respectively, whereas nicotine yield accounted for only 1 per cent and 2 per cent of the variance, respectively. The relative lack of an effect of nicotine yield on the biochemical measure appears to be due to the fact that smokers of lower nicotine brands smoked more cigarettes per day, thereby compensating for reduced delivery of smoke products. Our data do not support the concept that ultralow-yield cigarettes are less hazardous than others. Machine estimates suggesting low nicotine yield underrepresent actual human consumption of harmful cigarette constituents.