Drivers of CS
We estimated the phylogenetic signal in species interactions (PSI) associated to the calculation of CS to investigate the main drivers of the observed cophylogeny and ask whether closely related species interact with similar sets of partners. Following Russo et al. (2018) we used Mantel tests to assess if closely related lemur or plant species were more likely to interact with similar set of partners than expected by chance (e.g. Kahnt et al. 2019). For each of the interacting clades, the distance matrices compared were the between-species phylogenetic distance and the between-species dissimilarity (Jaccard distances) in interactions with the other clade. A significant CS (i.e. overall phylogenetic congruence) as well as a significant PSI for both lemur and plants were interpreted as a case of coevolution. Instead, if phylogenies were congruent but PSI was only significant for one of the interacting clades, this would indicate phylogenetic tracking. Finally, a significant CS but non-significant PSI for both lemurs and plants would suggests that the interactions were little important for the evolution of both clades, highlighting the importance of vicariance or other evolutionary processes. We used the function ‘phylosignal_network’ from the R -package RPANDA to estimate PSI values and estimate significance through permutations (Perez-Lamarque et al. 2022). In order to preserve any influence of the degree distribution of species, we used a conservative approach and set the permutation argument to ‘nbpartners’. This keeps constant the number of partners per species and shuffle at random their identity.