Calculations and statistical analysis
The effect of mixing litter on each response variable was assessed by computing, for each pair of replicates, the deviation between observed (O) and expected (E) values: (O-E)/E. The relative mixture effect (RME; Wardle et al., 1997) was calculated as the deviation between the values in the mixture (O) and the average values of the two single species (E). To detect any species-specific effects of the mixture, the relative individual performance (RIP; Zhou et al., 2020) was calculated as the deviation between the values of mass loss and fungal variables of a species in the mixture (O) and the values of the same single species (E). RME and RIP were considered significantly different from zero (i.e., nonadditive) when the mean was bigger than its 95% confidence limits (CL), i.e., when the 95% CL did not overlap 0 in the graphs, synergistic if the deviation was positive or antagonistic if the deviation was negative (Ball et al., 2008).
To provide a measure of the effect of the mixture on colonisation by each of the biota associated with leaf litter, a ‘RME/RIP decomposers’ and a ‘RME detritivores’ was calculated as above, with the same weight for each of the three fungal variables and of the three shredder variables, respectively. The effect of the mixture across exposure scenarios was assessed with a ‘global RME/RIP’, calculated with the same weight for each of the three exposures to overcome the different number of observations among exposure scenarios. To capture the effect of mixing litter on the process of decomposition, defined as ‘all biological processes contributing to organic matter mass loss and transformation’ (Gessner et al., 2010), an integrated measure containing the effects on mass loss and associated biota - ‘RME/RIP processing’ - was calculated as above with the same weight for mass loss, decomposers, and detritivores.
To examine all sources of variation of the values of the response variables, a mixed model analysis of variance (GLM ANOVA) was used to assess the effect of: 1) mixture [are observed values similar to expected ones?]; 2) exposure scenario [are observed and expected values similar among the three exposure scenarios?]; 3) interaction 1×2 [does the effect of mixture depend on exposure scenario?]. Due to the different number of sampling occasions among exposures, ANOVA was calculated with type III SS (Shaw & Mitchell-Olds, 2003). In case of a significant interaction, one-way ANOVA was carried out as above to detect the effect of the levels of each factor. Because the identity of the sampling days varied among exposures, time was set as a random factor. Tukey’s HSD test was used for pairwise comparisons after a significant exposure scenario effect (Zar, 2010). All data was tested for parametrical assumptions; heteroscedastic variables (Levene’s test) were transformed with the square root (counts) or with the natural logarithm (all other variables) (Zar, 2010). The statistical analyses were performed with the software STATISTICA 13.0 with the level of significance set at p =0.05.