The decomposition rates of plant litter mixtures often deviate from the averaged rates of monocultures of their component litter species. The mechanisms behind these non-additive effects in decomposition of litter mixtures are lively debated. One plausible explanation for non-additive effects is given by the improved microenvironmental condition (IMC) theory. According to this theory, plant litter species, whose physical characteristics improve the microclimatic conditions for decomposers, will promote the decomposition of their co-occurring litter species. We tested the IMC theory in relation to leaf litter and soil moisture in two contrasting moisture conditions in a dry subarctic mountain birch forest with vascular plant leaf litters of poor and high quality. The non-additive effects in mass loss of litter mixtures increased when moisture conditions in litter and soil became more favourable for plant litter decomposition. The sign of this increase (antagonistic or synergistic) in non-additive effects was more predictable for litter mixtures of poor litter quality. Although the specific mechanisms underlying the IMC theory depended on the litter quality of the litter mixtures, a standardized water holding capacity (WHC) was the litter trait most closely related to the non-additive effects in mixtures of both poor and high quality litter types. Furthermore, we found that higher dissimilarity in WHC traits between the component litter species in a mixture increased synergistic effects in litter mixtures under limiting moisture conditions. However, under improved moisture conditions, increased antagonistic effects were observed. Thus, we found clear support for the IMC theory and showed that climatic conditions and leaf litter physical traits determine whether the non-additive effects in litter mixtures are antagonistic or synergistic. Our study emphasizes the need to include litter physical traits into predictive models of mixing effects on plant litter decomposition and in general suggests climate specificity into these models.