Summary of mixed-effects model analyses including a combination of measured covariates to assess effects resource availability and leaf damage by herbivores on total iridoid glycoside, aucubin and catalpol concentrations.

Models were fitted by stepwise inclusion of terms. Order of covariates was determined through step-wise model extension starting with separate models for each covariate. Models were compared using Akaike's information criterion (AIC), and the model with the lowest AIC containing one of the covariates was selected. In further steps the model was extended until all covariates were fitted, and the experimental factors were entered in the same order as in models without additional covariates (see Table 1). Listed are the results of likelihood ratio tests that were applied to assess model improvement ( = L ratio) and the statistical significance of the explanatory terms ( = p values). Note that contrasts for the presence/absence of legumes and grasses were entered separately in a series of analyses. Arrows indicate a significant increase (↑) or decrease (↓) of variables in relation to measured covariates or the experimental factors of the biodiversity experiment, R indicates higher values in reproductive shoots, and Y indicates higher values in the younger leaf cohort.