* Despite time-lags and nonlinearity in ecological processes, the majority of our knowledge about ecosystem responses to long-term changes in climate originates from relatively short-term experiments. * We utilized the longest ongoing snow removal experiment in the world and an additional set of new plots at the same location in northern Sweden to simultaneously measure the effects of long-term (11 winters) and short-term (1 winter) absence of snow cover on boreal forest understorey plants, including the effects on root growth and phenology. * Short-term absence of snow reduced vascular plant cover in the understorey by 42%, reduced fine root biomass by 16%, reduced shoot growth by up to 53% and induced tissue damage on two common dwarf shrubs. In the long-term manipulation, more substantial effects on understorey plant cover (92% reduced) and standing fine root biomass (39% reduced) were observed, whereas other response parameters, such as tissue damage, were observed less. Fine root growth was generally reduced, and its initiation delayed by c. 3 (short-term) to 6 weeks (long-term manipulation). * Synthesis. We show that one extreme winter with a reduced snow cover can already induce ecologically significant alterations. We also show that long-term changes were smaller than suggested by an extrapolation of short-term manipulation results (using a constant proportional decline). In addition, some of those negative responses, such as frost damage and shoot growth, were even absolutely stronger in the short-term compared to the long-term manipulation. This suggests adaptation or survival of only those individuals that are able to cope with these extreme winter conditions, and that the short-term manipulation alone would overpredict long-term impacts. These results highlight both the ecological importance of snow cover in this boreal forest, and the value of combining short- and long-term experiments side by side in climate change research.