Host specificity and interaction networks of insects feeding on seeds and fruits in tropical rainforests

In the tropics, antagonistic seed predation networks may have different properties than mutualistic pollination and seed dispersal networks, but the former have been considerably less studied. We tested whether the structure of antagonistic tripartite networks composed of host plants, insects developing within seeds and fruits, and their insect parasitoids could be predicted from plant phylogenetic distance and plant traits. We considered subsets of the networks (‘subnetworks') at three rainforest locations (Panama, Thailand, Papua New Guinea), based on insect families, plant families or plant functional groups. We recorded 3197 interactions and observed a low percentage of realized interactions, especially in Panama, where insect host specificity was higher than in Thailand or New Guinea. Several factors may explain this, including insect faunal composition, incidence of dry fruits, high fruit production and high occurrence of Fabaceae at the Panamanian site. Host specificity was greater among seed‐eaters than pulp‐eaters and for insects feeding on dry fruits as opposed to insects feeding on fleshy fruits. Plant species richness within plant families did not influence insect host specificity, but site characteristics may be important in this regard. Most subnetworks were extremely specialized, such as those including Tortricidae and Bruchinae in Panama. Plant phylogenetic distance, plant basal area and plant traits (fruit length, number of seeds per fruit) had important effects on several network statistics in regressions weighted by sampling effort. A path analysis revealed a weak direct influence of plant phylogenetic distance on parasitoid richness, indicating limited support for the ‘nasty host hypothesis'. Our study emphasizes the duality between seed dispersal and seed predation networks in the tropics, as key plant species differ and host specificity tends to be low in the former and higher in the latter. This underlines the need to study both types of networks for sound practices of forest regeneration and conservation.