Quantifying the units of coevolution in pollination networks
Coevolution, the idea that pairs of interacting species evolve by reciprocal selection is well documented and particularly evident in the case of pollination mutualisms. Unfortunately, the coevolutionary narrative becomes murkier at a broader, network-wide scale despite substantive empirical evidence of pairwise coevolution. However, it has been argued that modules---groups of species in a network that tend to interact---may be the direct consequence of coevolutionary processes in pollination networks. While this notion provides an intuitive explanation, the core hypothesis remains to be explicitly tested in a manner that definitively links observed plant-pollinator interactions to the phylogenies of the two groups. To investigate the possible coevolutionary origin of modules in pollination networks, we analysed 59 globally distributed networks, identified the most parsimonious modules within each network, and compiled robustly dated phylogenies for both plants and pollinators in each community. To test for coevolutionary signal in these data, we adapted a recently proposed approach designed to test the congruence of phylogenetic trees in antagonistic networks. Our analysis allowed us to evaluate the dependency of each phylogeny upon the other as well as the contribution of individual interactions, and modules, to overall network coevolution.
Despite some differences between networks, we found significant evidence of coevolution in a majority of the network datasets we investigated. In addition, our analyses indicated that individual interactions varied in their contribution to the overall phylogenetic congruence of their community’s network. Notably, a network’s underlying modular structure also tended to be a significant predictor of interaction-level contributions to coevolutionary signal. This implies that modules may well represent the unit of coevolution in pollination networks. By definitively linking network structure to the evolutionary history of both plants and pollinators, we provide some of the first unambiguous empirical evidence in support of the coevolutionary origin of modularity in pollination networks. Moreover, we observed that particular modules contributed more strongly to network coevolution than others. However, the mechanisms explaining the differing contribution to coevolutionary signal remain to be seen. One fundamental issue in our understanding of mutualistic networks remains --- who ultimately drives coevolution?