Recent work indicates that the structure of mutualistic networks can provide stability to communities and ecosystems. Network theory predicts that changes in network structure may thus influence ecosystem stability. The structure of interaction networks (e.g. how many mutualist species, how many interactions, how many interactions per species, how are generalists and specialists linked to each other) determines network susceptibility to habitat perturbations and environmental change. Few empirical studies have directly tested the effects of disturbance on network structure. To test predictions of network theory, we compared mutualist partner communities of ants and plants in continuous lowland Amazonian rainforest to those networks in nearby forest fragments. We calculated network structure using both qualitative (presence-absence) and quantitative network metrics. To explore the roles of generalist (ants that also nest on other structures) and obligate ants, we considered the structure of the entire ant-plant network, plus the network of obligate ants and their host plants alone. We hypothesized that ant-plant networks would be negatively affected by fragmentation. We expected: 1) fewer mutualist species, 2) fewer interactions between ants and plants (connectance), and 3) changes in the interactions between specialists and generalists (nestedness) in forest fragments.

Results/Conclusions

Large-scale landscape disturbance had little effect on the structure of the ant-plant mutualistic network we studied. Qualitative and quantitative measures of connectance of both the whole ant-plant network and the subset of obligate ants was not significantly different between treatments, although means for continuous forest plots were consistently higher. Nestedness (based on presence-absence) indicated no significant difference in the nested structure between continuous forest sites and forest fragments. An alternate measure based on weighted interactions (frequency matrices) was also not statistically different. These results reflect that generalists and specialist ants were lost at similar, low rates, on average, and no novel generalists were identified in forest fragments. This sort of nested interaction structure provides protection against reductions in the abundance of specialists because if specialists are lost from the network, the core of interacting generalists remains unchanged. In some other studies, perturbation resulted in increased nestedness when generalist invaders become the species with the most links in the network. Our results indicate that the structure of this ant-plant mutualistic network was resistant to the effects of large-scale disturbance.