Litter decomposition is an important driver of terrestrial systems, and factors that determine decomposition rate for individual litter species have been widely studied. Fewer studies have explored the factors that regulate how mixing litters of multiple species affects litter decomposition and nutrient dynamics, and only a handful of studies have investigated how litter-mixing effects may differ among different habitats or ecosystems, or how they respond to environmental gradients. We used a well-established retrogressive chronosequence involving thirty lake islands in northern Sweden in which time since fire disturbance increases with decreasing island size; smaller islands therefore have reduced rates of aboveground and belowground ecosystem processes. On each of these islands we utilized plots with and without the long-term experimental removal of shrubs. Litters from the six most common plant species on the islands were prepared in single-, three- and six-species litterbags, and placed on both the shrub-removal and non-removal plots on each island to decompose for one year. We found significant non-additive effects of litter mixing on litter decomposition rates, on final litter N and P concentrations, and on litter N loss, but these non-additive effects varied both in direction and magnitude with changed number of species, and even among litter mixtures with the same number of species. Further, the magnitude of non-additive effects of litter mixing on both litter decomposition and nutrient dynamics was significantly influenced by both island size and the interaction between island size and shrub-removal treatment. When shrubs were present, there was a U-shaped relationship between these non-additive effects and island size, while the relationship was positive when shrubs were removed. Hence, our results support previous findings that litter mixing may produce non-additive effects on litter decomposition and nutrient dynamics, and that these effects tend to be idiosyncratic due to the importance of effects of individual species in the mixture. Most importantly, our results show that non-additive litter-mixing effects change greatly across environmental gradients, meaning that the biotic and abiotic characteristics of an ecosystem can be a powerful driver of the magnitude and even the direction of litter-mixing effects on ecosystem processes.