Spatially disaggregated surveys of smoking behaviour are rare and hence estimating the geography of the incidence of smoking is difficult. The main aim of this study is to develop a technique for estimating smoking probability for different age/sex groups in small areas across the whole of Scotland using information on smoking behaviour from the Scottish Household Survey. This is useful not only in its own right, but as an aid to studies of geographical variations in diseases such as lung cancer that, as a first step, need to control for smoking behaviour. The method developed uses individual-level characteristics from the Scottish Household Survey combined with a set of output area and pseudo-postcode sector measures from the 1991 census to model the probability of smoking. The parameters from this model are then used to make smoking predictions by age and sex for output areas across Scotland. This is the first time that such geographically detailed estimates of smoking have been made available.