We examined well-sintered, mid-winter snow from the Sierra Nevada of California with an environmental scanning electron microscope equipped with a cold stage. Filaments were observed at all grain boundaries in a coated sample, while filaments were much harder to find in uncoated samples. X-ray microanalysis failed to detect soluble impurities in grain boundary grooves or on ice grain surfaces. It seems reasonable to conclude the filaments are composed of soluble impurities since, were they made of ice, they would have sublimated in vacuum and during specimen preparation. Observations of the grain boundaries themselves are not consistent with a sintering mechanism dominated by vapor diffusion.