We tested the hypothesis that habitat temperature is the driving variable determining both the short-term and seasonal patterns of abundance (biomass) of Eying adult non-biting midges (Chironomidae) in the high Arctic. Malaise traps were used to record the Eight activity patterns of adult chironomids during the 1992 and 1993 seasons at two contrasting sites, a polar semi-desert and a tundra heath adjacent to Ny Angstrom lesund, Spitsbergen, Svalbard. The relationship between flight activity, measured as chironomid biomass captured, and macro and microclimatic parameters was examined. Mean and minimum screen temperatures were the best single predictors of chironomid activity. Best subsets regression was used to establish predictive relationships between biomass captured and climatic parameters for each site. Again, macroclimatic temperature variables emerged as the best predictors, together with much smaller contributions from relative humidity and precipitation. The species composition of the chironomid community changed as the summer progressed and the 'average' individual became smaller. However, when cumulative biomass captured per trap was plotted against cumulative day degrees above zero, starting at snow melt, the relationship on both sites in 1992, an average year, was approximately linear or at most slightly curved. This suggested that biomass capture was mainly a function of temperature and time. By contrast, in 1993, the warmest year on record (1961-93), the initial response was again almost linear but later became asymptotic, suggesting that the supply of chironomids available for capture was depleted, despite suitably high temperatures. This contrast between years (100 day degrees) provided an opportune analogue for the likely effects of climate warming on chironomid activity. The implications for the food supply of tundra-breeding birds are discussed in the context of the known emergence patterns of Arctic chironomids