This paper investigates the relationship between local air pollution and urban structure with an emphasis on urban fragmentation. Using a unique dataset of 249 Large Urban Zones (LUZ) across Europe, a Bayesian Model Averaging selection method is employed to empirically identify the determinants of within-LUZ concentration of three air pollutants: NO2, PM10 and SO2 for the year 2006. Several indices of land use are considered among possible determinants. These are supplemented by a dataset on various economic, demographic and meteorological variables that can explain the variation of air pollution. The results of this econometric analysis support the hypothesis that urban structure has significant effects on pollution concentration. In particular, they suggest that fragmented urban areas experience higher concentrations of NO2 and PM10 and that densely populated urban areas suffer from higher SO2 concentration. The findings suggest that policies favouring continuous urban areas may result in environmental improvements.