Daily PM2.5 samples were repeatedly collected (1-8 times) in the homes of elderly nonsmoking individuals with coronary heart disease in Amsterdam, The Netherlands (33 individuals) and Helsinki, Finland (44 individuals). Sources of indoor PM2.5 were evaluated using a two-way multilinear engine model. Because the indoor elemental data lacked a traffic marker, separation of traffic related PM was attempted by combining the indoor data with fixed site outdoor data that also contained NO. Six outdoor sources, including long-range transport (LRT), urban mixture, oil combustion, traffic, sea-salt, and soil were identified, and three indoor sources were resolved: resuspension, potassium-rich and copper-rich sources. The average contribution of the indoor factors was 6% (1.1 microg m(-3)) and 22% (2.4 microg m(-3)) in Amsterdam and Helsinki, respectively. The highest longitudinal correlations between source-specific outdoor and indoor PM2.5 concentrations were found for LRT and urban mixture; the median R was above 0.6 for most sources. The longitudinal correlations were lower in Helsinki than in Amsterdam. Indoor-generated PM2.5 was not related to ambient concentrations. We conclude that using outdoor and indoor data together improved the source apportionment of indoor PM2.5. The results support the use of fixed site outdoor measurements in epidemiological time-series studies on outdoor air pollution.