Our objective was to develop machine-smoking protocols (i.e. puff volume, puff duration and puff frequency) reflecting the distribution of human smoking behaviour as a complement to the International Standard (ISO) protocol. For this purpose, a modelling approach based on smoking topography and excretion data of urinary nicotine metabolites from clinical studies in the UK and Japan was used. Three smoking protocols (LOW, MEDIUM and HIGH) were based on the 10th percentile, the mean and the 90th percentile of the puff volumes obtained from topography data. The corresponding puff durations were estimated by linear regression analysis of the puff duration/puff volume relationship. Finally, puff frequency was estimated using a predictive model assuming that the nicotine yield is a linear function (i.e. proportional) to the amount of air actually drawn through the burning part of the cigarette. This approach was termed "nicotine proportion method" where the puff frequency is changed iteratively until it equals the corresponding nicotine uptake measured in human smokers during a clinical study. The assumption was verified by measuring the nicotine yield in the laboratory on three modelled protocols for four conventional cigarettes. In 10 out of 12 cases, a very good agreement was observed between the estimated nicotine yields obtained with our modelling approach and the measured nicotine yields obtained from smoking machines in the laboratory. Although smoking virtually identical cigarettes, smoking protocols based on topography and excretion data of urinary nicotine metabolites from the UK clinical study were different than those based on data from the study performed in Japan. This may be explained by different smoking behaviour between these two populations of smokers.