The influence of functional traits on CS
We obtained data on primate (body size; degree of frugivory and dietary guild) and plant (fruit length and seed diameter) (Table S2) to test whether they affect the magnitude of the CS, by characterizing each interaction by its Procrustean residual. We accounted for the possible influence of the sample size to control for sampling bias (Supplementary Material S3).
Fruit length and seed diameter are traits that tend to be conserved along the phylogenetic tree (Jordano 1995, Moles et al. 2005), thus we used average values obtained for each plant genera as measures. We tested the model for multicollinearity of the variables by using the variance inflation factor (VIF). Multicollinearity occurs when two or more predictors are correlated and provide redundant information about the response. VIF values for body size was found to be high (VIF=8.39, tolerance = 0.08), which were also correlated with model intercept (0.9). Facing the likely confounding property of body size added to the collinearity produced, we decided to suppress it from our model (see explanation in Supplementary Material S3).
We fitted our final model with variables describing primate biology in terms of the amount of fruit intake and functional role played according to dietary category, such as the frugivory degree (percentage of the diet that corresponds to fruit pulp, excluding seed predatory interactions) and dietary guild (percentage of each food item in the diet). The former adds information on the real amount of fruit pulp in diet, whereas the latter informs about the effects of morphological variation resulting from adaptations of the digestive tract (Hawes and Peres 2014), with relevant consequences for the ability of seeds to germinate after being defecated (Fuzessy et al. 2016). Our GLMM included then Procrustean residuals as a response variable, primate species and plant genera as random effects, and functional traits (frugivory degree and dietary guilds for primates; fruit length and seed diameter for plants) as fixed effects, using the ‘lmerTest’ package (Kuznetsova et al. 2017).