There is currently much interest in understanding how biodiversity loss affects the functioning of ecosystems, but few studies have evaluated how ecosystem processes change in response to one another following biodiversity loss. We focused on a well-described gradient of 30 forested lake islands in northern Sweden, where island size determines the occurrence of lightning-ignited wildfire, which in turn determines successional stage, plant species composition, and productivity. We investigated the effect of biodiversity loss on biological nitrogen fixation by feathermosses through an experiment consisting of factorial removals of three understory shrub species (Vaccinium myrtillis, Vaccinium vitis-idaea, and Empetrum hermaphroditum) and two plant functional groups (shrubs and tree roots). We tested the hypothesis that, following vascular plant species loss, N fixation rates would be impaired by changes in pools or processes that increase extractable soil N, because changes in the supply rate of N to feathermosses should influence their demand for newly fixed N. Further, we hypothesized that the effects of removals on N fixation would depend on environmental context (i.e., island size), because it has been previously demonstrated that the effect of vascular plant species removal on N recycling pools and processes was strongest on productive islands. The data demonstrated that removal of two shrub species (V. vitis-idaea and E. hermaphroditum) negatively aflected the N fixation of Hylocomium splendens, but positively affected Pleurozium schreberi, resulting in unchanged areal N fixation rates. In the functional removal experiment, tree root removal resulted in a significant negative effect on N fixation. The effects of shrub and root removals on N fixation occurred only on small islands and thus were context dependent. This pattern did not correspond to the effect of shrub and root removal treatments on N-recycling pools or processes, which only occurred in response to specific vascular plant removals on large or medium islands. The data thus did not support our hypothesis that N fixation was directly responsive to changes in N-recycling pools or processes caused by vascular plant species removals, but instead highlighted the importance of species-specific interactions and environmental context in determining the manner in which biodiversity loss alters ecosystem processes.