The potential effects of indoor NO2 and tobacco smoke on the respiratory system were studied using pulmonary function data from a longitudinal study. In 1982, NO2 was measured in 163 homes of non-smoking adult women and detailed information on smoking inside the house was gathered over the period 1965-1982. Personal exposure to NO2 was calculated from the measured levels and the activity pattern of the study population (range 11-139 micrograms NO2/m3). Historical exposure was estimated using regression models of NO2 on house characteristics. It is shown that estimation of historical exposure to indoor NO2 on the basis of house characteristics only is, at present, too inaccurate for use in epidemiological studies. Actual measurement of NO2 is unavoidable for exposure assessment in health effect studies of indoor exposure to NO2.