Aerosol particles serve as cloud condensation nuclei worldwide, and they affect the Earth's radiation budget both directly and indirectly. These particles consist mostly of sulfate compounds. Ice core measurements can be used to infer past variations of atmospheric sulfate concentration, but to do so requires knowledge of the deposition mechanisms. Significant ''dry'' deposition may occur by filtering when air moves through the snow due to changes in pressure caused by wind blowing over a rough surface (wind pumping). The filtering efficiency of snow was measured at South Pole Station, using an optical particle counter and a condensation nucleus counter. The number size-distribution of ambient aerosol peaks at a dry particle diameter of 0.13 μm, the volume size-distribution at 0.17 μm. Less than 5% of the particles have diameters >0.3 μm. Diffusion from interstitial air to snow grains appears to be the primary mechanism of dry deposition for particles <0.6 μm in diameter, but another mechanism, probably impaction, becomes significant for larger particles. Aerosol deposition by filtering occurs with an e-folding time of 1--3 s depending on particle size, corresponding to an e-folding depth of 0.5--1 cm for an estimated air velocity of 0.4 cm s-1 within the surface snow. Even for long residence times, a small number of particles (<0.1%) are found in filtered air, suggesting a small degree of new particle formation or reentrainment. However, both the e-folding depth and the reentrainment rate are small enough that smoothing of the sulfate records in ice cores should be negligible. Three mathematical models for filters agree in describing filtering by snow as dominated by diffusion, but all underpredict the filter efficiency. Capture of aerosol particles is found to be 2--3 times as rapid as that assumed by Cunningham and Waddington <1993>, supporting their conclusion of nearly total removal of particles from air entering the snow. Blowing snow might also be expected to collect aerosol particles; however, a calculation suggests that deposition to blowing snow on the Antarctic Plateau is insignificant. If wind pumping and diffusion contribute significantly to total deposition, the flux of air into the snow and the residence time of the air within the snow control the deposition rate. Both air flux and residence time are functions of wind speed and surface roughness, so that the aerosol flux to the snow depends on these factors as well as atmospheric concentration, complicating the interpretation of paleoclimate records for aerosol-bound substances. ¿ American Geophysical Union 1996