Secondary succession is well‐understood, to the point of being predictable for plant communities, but the successional changes in plant‐herbivore interactions remains poorly explored. This is particularly true for tropical forests, despite the increasing importance of early successional stages in tropical landscapes. Deriving expectations from successional theory, we examine properties of plant‐herbivore interaction networks while accounting for host phylogenetic structure along a succession chronosequence in montane rainforest in Papua New Guinea. We present one of the most comprehensive successional investigations of interaction networks, equating to >40 person years of field sampling, and one of the few focused on montane tropical forests. We use a series of nine 0.2ha forest plots across young secondary, mature secondary and primary montane forest, sampled almost completely for woody plants and larval leaf chewers (Lepidoptera), using forest felling. These networks comprised of 12,357 plant‐herbivore interactions and were analysed using quantitative network metrics, a phylogenetically controlled host‐use index and a qualitative network beta diversity measure. Network structural changes were low and specialisation metrics surprisingly similar throughout succession, despite high network beta diversity. Herbivore abundance was greatest in the earliest stages, and hosts here had more species‐rich herbivore assemblages, presumably reflecting higher palatability due to lower defensive investment. All herbivore communities were highly specialised, using a phylogenetically narrow set of hosts, while host phylogenetic diversity itself decreased throughout the chronosequence. Relatively high phylogenetic diversity, and thus high diversity of plant defenses, in early succession forest may result in herbivores feeding on fewer hosts than expected. Successional theory, derived primarily from temperate systems, is limited in predicting tropical host‐herbivore interactions. All succession stages harbour diverse and unique interaction networks, which together with largely similar network structures and consistent host use patterns, suggests general rules of assembly may apply to these systems.

Quantitative Plant-Herbivore interaction matricesNine txt files with quantitative interaction matrices. Files ABCD-Primary forest plots, HJK-Mature secondary forest plots and FG- Young secondary forest plotsDATA_MATRICES.zipYawan_Tree_PhylogenyA Newick file of the Host Tree Phylogeny used for analysis of host community phylogenetic diversity and Distance Based Specialisation Index (DSI)Yawan_Phylogeny.trePlant Community Data and accompanying environmental data used in Ordination analysisPlant Community Data from nine 0.2ha plots. Plots ABCD-Primary forest, HJK- Mature Secondary Forest and FG- Young Secondary Forest. Data are from plants above 5cm dbh. Values represent individuals.Ecography Ordination Plant Communtiy Data.xlsx