A TSI Model 3800 aerosol time-of-flight mass spectrometer (ATOFMS) was deployed for single-particle analysis in Shanghai during the World Exposition (EXPO), 2010. Measurements on two extreme cases: polluted day (1st May) and clean day (25th September) were compared to show how meteorological conditions affected the concentration and composition of ambient aerosols. Mass spectra of 90496 and 50407 particles were analyzed respectively during the two sampling periods. The ART-2a neural network algorithm was applied to sort the collected particles. Seven major classes of particles were obtained: dust, sea salt, industrial, biomass burning, organic carbon (OC), elementary carbon (EC), and NH -rich particles. Number concentration of ambient aerosols showed a strong anti-correlation with the boundary layer height variation. The external mixing states of aerosols were quite different during two sampling periods because of different air parcel trajectories. Number fraction of biomass burning particles (43.3%) during polluted episode was much higher than that (21.6%) of clean time. Air parcels from the East China Sea on clean day diluted local pollutant concentration and increased the portion of sea salt particle dramatically (13.3%). The large contribution of biomass burning particles in both cases might be an indication of a constant regional background of biomass burning emission. Mass spectrum analysis showed that chemical compositions and internal mixing states of almost all the particle types were more complicate during polluted episode compared with those observed in clean time. Strong nitrate signals in the mass spectra suggested that most of the particles collected on polluted day had gone through some aging processes before reaching the sampling site.