We examined the detection of smoking by 50 randomly-selected general practitioners. The practitioners said that they should detect smokers among their patients with a high degree of accuracy: a median of 100% of smokers under ideal conditions, and a median of 80% of smokers given the realities of practice. Practitioners actually identified 56.2% of their patients who were smokers--360 of 641 smokers among 2044 patients in the study. Some 65% of patients with a smoking-caused or smoking-exacerbated disease were detected, and this degree of detection was not related to age. For smokers of greater than 50 years of age and without any obvious smoking-related condition, the detection rates were similar to those of patients with smoking-related disease; however, in those patients who were less than 30 years of age, only 41% of smokers were detected--although the prevalence of smoking was higher in the younger than it was in the older patients. For all age-groups, the prevalence of smoking in men was about 15% higher than it was in women, but there was no significant difference between the detection rates in the sexes. In spite of their higher risks from smoking, the detection rates for pregnant women, or women who were taking oral contraceptive agents, were no higher than those for other women of less than 35 years of age. The doctors said that they had treated for smoking 78% of the detected smokers who had smoking-caused disease, compared with 35% of detected smokers with no smoking-related condition. They rated the value of such treatment higher in the former group than in those patients who had not yet developed smoking-related problems. Doctors appear to respond to the evidence of disease that is caused by smoking more than to the habit itself, which is a handicap in the prevention of smoking-related disease.