Smoking behaviour and exposure to carbon monoxide, nicotine, and tar were studied in 19 middle-tar smokers. All smoked their own brands for three weeks and then switched to either a conventional low-nicotine, low-tar brand (control) or a medium-nicotine, low-tar cigarette for a further three weeks, the order then being reversed. The medium-nicotine, low-tar brand also had a low delivery of carbon monoxide. With the medium-nicotine, low-tar cigarette mouth-level delivery and intake of nicotine was similar to that with the smokers' usual brands, and significantly greater than with the control low-tar cigarette. Intake of carbon monoxide from the medium-nicotine, low-tar cigarette was significantly less than with either own or control brands. With both low-tar brands mouth-level exposure to tar was reduced relative to smokers' usual cigarettes. There was no evidence, however, that the reduction in tar exposure was greater with the medium-nicotine brand than with the control low-tar cigarette. Both low tar brands were "'oversmoked" relative to subjects' usual middle-tar cigarettes. The medium-nicotine, low-tar cigarette was marginally more acceptable than the control brand, and the particular design used in the study resulted in a lower intake of carbon monoxide. In terms of reducing mouth-level exposure to tar, however, the medium-nicotine, low-tar cigarette had no advantage over the control low-tar product. In part this was because of the ratio of tar to nicotine delivery obtained by human smokers was not the same as that obtained by smoking machine.