A series of experiments carried out in a climatic chamber under various conditions of confinement, ventilation or of filtration have revealed various difficulties inherent in the characterisation of environmental tobacco smoke. The results obtained show that in measurements of the particulate phase the most appropriate method is counting by particle-size category; the amount of environmental smoke can be evaluated either from the increase in the number of micellae of between 0.4 and 0.6microm, or, paradoxically, from the decrease in the number of micellae less than 0.2microm in diameter. Nicotine, the most specific tracer, poses problems particularly in relation to sampling, because it is present in both partic and vapour form. The ratio of the two forms depends on time and on the ventilation or filtration conditions. Carbon monoxide is the most stable marker and is quite easy to measure under the usual experimental conditions; however, other sources are likely to p roduce major increases in the CO level. The concentrations of the volatile organic constituents vary widely with time and are subject to transfer phenomena between the gas and particulate phases. The phenomena detected show that the relationships between the concentrations of the various constituents of smoke are highly dependent on the ventilation and/or filtration conditions and of the time between the emission of the smoke and the determination. In view of these reservations, it would appear to be hazardous to deduce the level of a compound on the basis of simple knowing the level of a tracer of tobacco smoke. Similarly, these facts must be taken into account when the exposure of non-smokers to environmental smoking is assessed.