A probabilistic model (INDAIR) has been developed to predict air pollutant concentrations in home microenvironments in the UK. The model has been parameterised using probability functions for four pollutants simultaneously (NO2, CO, PM10 and PM2.5), under three emission scenarios (no source, cooking, smoking). Model predictions are broadly consistent with data on indoor concentrations in UK homes. Modelled mean concentrations were most sensitive to variation in outdoor concentrations, air exchange rate and deposition velocity in no-source scenarios, while modelled peak concentrations in source rooms were most sensitive to variation in emission rate and room size. Under model assumptions, smoking and cooking made a significant contribution to annual mean indoor concentrations of PM10 and PM2.5, gas cooking made a significant contribution to annual mean indoor NO2 concentrations, while annual mean CO concentrations were dominated by infiltration of outdoor air. The modelled frequency distributions of 24 h mean values showed 95 percentile concentrations that were typically twice the mean concentrations in no-source scenarios, and 3-4 times the mean concentration during emission peaks. The higher exposure of residents in homes at the upper ends of the frequency distributions may be associated with adverse health outcomes, and probabilistic modelling approaches can contribute to identification of the characteristics of homes with high indoor concentrations. (c) 2006 Elsevier Ltd. All rights reserved.