As epidemiological studies report associations between ambient air pollution and adverse birth outcomes, it is important to understand determinants of exposures among pregnant women. We measured (48-h, personal exposure) and modeled (using outdoor ambient monitors and a traffic-based land-use regression model) NO, NO(2), fine particle mass and absorbance in 62 non-smoking pregnant women in Vancouver, Canada on 1-3 occasions during pregnancy (total N=127). We developed predictive models for personal measurements using modeled ambient concentrations and individual determinants of exposure. Geometric mean exposures of personal samples were relatively low (GM (GSD) NO=37 ppb (2.0); NO(2)=17 ppb (1.6); 'soot', as filter absorbance=0.8 10(-5) m(-1) (1.5); PM(2.2)=10 microg m(-3) (1.6)). Having a gas stove (vs. electric stove) in the home was associated with exposure increases of 89% (NO), 44% (NO(2)), 20% (absorbance) and 35% (fine PM). Interpolated concentrations from outdoor fixed-site monitors were associated with all personal exposures except NO(2). Land-use regression model estimates of outdoor air pollution were associated with personal NO and NO(2) only. The effects of outdoor air pollution on personal samples were consistent, with and without adjustment for other individual determinants (e.g. gas stove). These findings improve our understanding of sources of exposure to air pollutants among pregnant women and support the use of outdoor concentration estimates as proxies for exposure in epidemiologic studies.