The variation of snow albedo with wavelength across the solar spectrum from 0.3 μm in the ultraviolet (UV) to 2.5 μm in the near infrared (IR) was measured at Amundsen-Scott South Pole Station during the Antarctic summers of 1985–1986 and 1990–1991. Similar results were obtained at Vostok Station in summer 1990–1991. The albedo has a uniformly high value of 0.96–0.98 across the UV and visible spectrum, nearly independent of snow grain size and solar zenith angle, and this value probably applies throughout the interior of Antarctica. The albedo in the near IR is lower, dropping below 0.15 in the strong absorption bands at 1.5 and 2.0 μm; and it is quite sensitive to grain size and somewhat sensitive to zenith angle. Near-IR albedos were slightly lower at Vostok than at South Pole, but day-to-day variations in the measured grain size due to precipitation, drifting, and metamorphism were found to cause temporal variations in near-IR albedo larger than those due to any systematic geographical change from South Pole to Vostok. The spectrally averaged albedos ranged from 0.80 to 0.85 for both overcast and clear skies, in agreement with measurements by others at South Pole and elsewhere in Antarctica. Using a two-layer radiative transfer model, the albedo can be explained over the full wavelength range. Tests were made to correct for systematic errors in determining spectral albedo. Under clear skies at about 3000-m elevation the diffuse fraction of downward irradiance varied from 0.4 in the near UV to less than 0.01 in the near IR; knowledge of this fraction is required to correct the measured irradiance for the instrument's deviation from a perfect cosine-response. Furthermore, the deviation from cosine response is itself a function of wavelength. Under clear skies a significant error in apparent albedo can result if the instrument's cosine collector is not parallel to the surface; e.g., if the instrument is leveled parallel to the horizon, but the local snow surface is not horizontal. The soot content of the snow upwind of South Pole Station was only 0.3 ng/g. It was somewhat greater at Vostok Station but was still too small to affect the albedo at any wavelength. Bidirectional reflectance at 0.9-μm wavelength, measured from a 23-m tower at the end of summer after the sastrugi (snow dunes) had diminished, showed a pattern remarkably similar to the spectrally averaged pattern obtained from the Nimbus 7 satellite.