We report analysis of data on outdoor air pollution and respiratory symptoms in children collected in the Czech part of the international Small Area Variations in Air pollution and Health (SAVIAH) Project, a methodological study designed to test the use of geographical information systems (GIS) in studies of environmental exposures and health at small area level. We collected the following data in two districts of Prague: (1) individual data on 3,680 children (response rate 88%) by questionnaires; (2) census-based socio-demographic data for small geographical units; (3) concentrations of nitrogen dioxide (NO2) and sulfur dioxide (SO2) measured by passive samplers in three 2-week surveys at 80 and 50 locations, respectively. We integrated all data into a geographical information system. Modeling of NO2 and SO2 allowed estimation of exposure to outdoor NO2 and SO2 at school and at home for each child. We examined the associations between air pollution and prevalence of wheezing or whistling in the chest in the last 12 months by logistic regression at individual level, weighted least squares regression at small area (ecological) level and multilevel modeling. The results varied by the level of analysis and method of exposure estimation. In multilevel analyses using individual data, odds ratios per 10 microg/m3 increase in concentrations were 1.16 (95% CI = 0.95-1.42) for NO2, and 1.08 (95% CI = 0.97-1.21) for SO2. While mapping of spatial distribution of NO2 and SO2 in the study area appeared valid, the interpolation from outdoor to personal exposures requires consideration.