This paper compares winter soil temperatures at five high arctic sites (Ny Alesund, West Spitsbergen) and one subarctic site (Slattatjakka, Abisko) during 1992/93 and 1993/94. At the high arctic sites snow cover afforded slight insulation where minimum air temperatures were as low as -32 degrees C (March 1993). However, snow did not accumulate significantly until late winter, by which time the ground had cooled to approximately -20 degrees C. The polar night aided soil cooling by minimizing solar heat gain. Soil temperatures at 3 cm depth during the autumn freeze were initially higher than surface temperatures, but once frozen, the zone inhabited by soil microarthropods (approximately 10 cm depth) remained isothermal and closely tracked air temperature. By contrast, throughout the spring thaw, the soil at 3 cm depth was cooler than the surface. Hence, snow cover reduced absolute minimum temperatures in late winter but prolonged the effective winter period. Hence soil organisms may be inactive for up to 79% (289 d) of the year, owing to the extended period that the ground is frozen. The incidence of daily ground freeze/thaw events was reduced at high arctic sites compared with a subarctic location. Similarly, there were differences in temperature means and minima at the adjacent high arctic sites dependent on location and topography; for example, on opposite coasts of the Broggerhaloya, West Spitsbergen the minimum temperatures in 1993/94 were -15.7 degrees C (Stuphallet) and -8.2 degrees C (Kjaerstranda). Terrestrial microarthropods inhabiting sites with late snow accumulation and cold air temperatures experience extreme low soil temperatures and hence require effective cold-hardiness strategies.