Tree species used in agroforestry systems can be selected based on which species support higher numbers of plant-animal interactions. Integrating network theory with conservation science is a relatively new aim and several approaches have appeared in the literature that use network theory to identify priority or key plant species in empirical networks. Here we use data from bird-plant networks in coffee agroforests to compare two popular approaches: modularity analysis and the centrality index approach. The dataset includes a total of 834 interactions between 28 plant and 64 bird species. We constructed four quantitative networks, a full and reduced network and a dry and rainy season network. All networks were significantly modular. Modularity analysis identified only two plant species as holding core topological roles in a network, i.e., connectors or network hubs. The centrality index approach identified 12 plant species as priority or key species. The contrasting outcomes from the two approaches demonstrates that more work is needed to merge network theory and conservation science. Before farmers and practitioners can implement science-based management practices to have the greatest impact on biodiversity conservation, scientists will have to decide which approach to follow, or if the two approaches can be merged. We propose one idea for merging the two approaches would be to recommend that farmers or practitioners select plant species with the highest centrality score from each module of the network as well as all plant species identified as holding core topological roles from modularity analysis.
