Cophylogenetic signal (CS)
To conduct a direct assessment of CS between Angiosperms and primates, we implemented a Procrustean Approach to Cophylogeny (PACo; Balbuena et al. 2013). The PACo approach addresses cophylogeny by optimizing the phylogeny fit using interaction graphs of a given matrix. The cophylogenetic signal of each individual interaction is given by the squared residual distance, r, between the two corresponding points in the phylogenetic trees. PACo thus returns a quantification of the global fit of the phylogenetic objects based on observed interactions as the sum of squared residual distances (R = ∑ r2) between phylogenetic-interaction graphs (Balbuena et al. 2013). As in any regression analysis, the smaller the residual distances, the better the fit of the two phylogenies to each other and the more support for a hypothesis of CS as reflected by the extant interactions (Hutchinson et al. 2017).
A high CS illustrates a strong congruency of the phylogenies of the two partner groups. Procrustean residuals measure the variation in interactions that is not explained by the co-phylogenetic structure of the interaction matrix, so that interactions with small Procrustean residuals contribute the most to CS, while those deviating more from the expectation derived from the shared phylogenetic history contribute the less to CS. CS was considered to be significant when it was smaller than 95% of the values obtained from 1,000 randomizations of the aggregated interaction dataset. Further details are available in Suplementary Material S5.
To ensure that the cophylogenetic pattern at the continental scale in the whole Neotropics was not due to distinct geographical subsets of primate species and plant genera coexisting in different regions, we split our dataset into the following Neotropical biogeographic forests (Morrone 2014): Atlantic forest, Amazon forest and Mesoamerican tropical forest. We then filtered our dataset and obtained, for each region, specific phylogenies of plants and primates, and their respective seed dispersal interactions. We then measured the CS and tested for its significance following the same statistical approach as for the continent-wide assessment.