Mixed-effects models were used to estimate within-person and between-person variance components, and some determinants of environmental exposure to particulate matter (PM 2.5 ), black smoke (BS) and trace elements (Cl, K, Ca, Ti, Fe, Ni, Cu, Zn, and Pb) for personal measurements from 30 adult subjects in Gothenburg, Sweden. The within-person variance component dominated the total variability for all investigated compounds except for PM 2.5 and Zn (in which the variance components were about equal). Expressed as fold ranges containing 95% of the underlying distributions, the within-person variance component ranged between 5-fold and 39-fold (median: sixfold), whereas the between-person variance component was always osixfold (median: threefold). The relatively large within-person variance components can lead to attenuation bias in exposure-response relationships and point to the importance of obtaining repeated samples of PM exposure from study subjects in epidemiological investigations of urban air pollution. On the basis of the variance components estimated for the various particulate species, between 3 and 39 repeated measurements per subject would be required to limit attenuation bias to 20%. Significant determinants for personal exposure levels were urban background air concentrations (PM 2.5 , BS, Cl, Zn, and Pb), cigarette smoking (PM 2.5 , BS, K, and Ti), season (PM 2.5 , Fe, and Pb), and the time spent outdoors or in traffic (Fe).