Recent advances in understanding insect communities in tropical forests have contributed little to our knowledge of large-scale patterns of insect diversity, because incomplete taxonomic knowledge of many tropical species hinders the mapping of their distribution records. This impedes an understanding of global biodiversity patterns and explains why tropical insects are under-represented in conservation biology. Our study of approximately 500 species from three herbivorous guilds feeding on foliage (caterpillars, Lepidoptera), wood (ambrosia beetles, Coleoptera) and fruit (fruitflies, Diptera) found a low rate of change in species composition (beta diversity) across 75,000 square kilometres of contiguous lowland rainforest in Papua New Guinea, as most species were widely distributed. For caterpillars feeding on large plant genera, most species fed on multiple host species, so that even locally restricted plant species did not support endemic herbivores. Large plant genera represented a continuously distributed resource easily colonized by moths and butterflies over hundreds of kilometres. Low beta diversity was also documented in groups with differing host specificity (fruitflies and ambrosia beetles), suggesting that dispersal limitation does not have a substantial role in shaping the distribution of insect species in New Guinea lowland rainforests. Similar patterns of low beta diversity can be expected in other tropical lowland rainforests, as they are typically situated in the extensive low basins of major tropical rivers similar to the Sepik-Ramu region of New Guinea studied here.

Escalation (macroevolutionary increase) or divergence (disparity between relatives) in trait values are two frequent outcomes of the plant-herbivore arms race. We studied the defences and caterpillars associated with 21 sympatric New Guinean figs. Herbivore generalists were concentrated on hosts with low protease and oxidative activity. The distribution of specialists correlated with phylogeny, protease and trichomes. Additionally, highly specialised Asota moths used alkaloid rich plants. The evolution of proteases was conserved, alkaloid diversity has escalated across the studied species, oxidative activity has escalated within one clade, and trichomes have diverged across the phylogeny. Herbivore specificity correlated with their response to host defences: escalating traits largely affected generalists and divergent traits specialists; but the effect of escalating traits on extreme specialists was positive. In turn, the evolution of defences in Ficus can be driven towards both escalation and divergence in individual traits, in combination providing protection against a broad spectrum of herbivores.

Spatial and temporal variability in predation risk for herbivores on 13 rainforest species of Ficus (Moraceae) in Papua New Guinea was studied in order to assess whether predator‐free refuges exist on their foliage and if so, whether herbivorous insects concentrate their activity in such refugia. Predation risk from invertebrate predators was measured as the disappearance rate of live termites set up as baits on the foliage. By far the most important predators were ants, accounting for 77% of attacks. No consistent differences in predation rate between Ficus species were found so that tree identity could not be used as an indicator of enemy‐free space. Predation risk was highly variable among conspecific trees and also changed rapidly in time, over periods as short as 10 days. Such short‐term and unpredictable predator‐free refuges may be difficult for herbivores to find and exploit. Predation risk during the day was three times higher than during the night, but abundance of herbivores on the foliage was also higher during the day. Thus, night was confirmed as a relatively enemy‐free time which, however, was not exploited by herbivores.

Abstract. 1. Bark and ambrosia beetles are crucial for woody biomass decomposition in tropical forests worldwide. Despite that, quantitative data on their host specificity are scarce.2. Bark and ambrosia beetles (Scolytinae and Platypodinae) were reared from 13 species of tropical trees representing 11 families from all major lineages of dicotyledonous plants. Standardised samples of beetle-infested twigs, branches, trunks, and roots were taken from three individuals of each tree species growing in a lowland tropical rainforest in Papua New Guinea.3. A total of 81742 beetles from 74 species were reared, 67 of them identified. Local species richness of bark and ambrosia beetles was estimated at 80-92 species.4. Ambrosia beetles were broad generalists as 95% of species did not show any preference for a particular host species or clade. Similarity of ambrosia beetle communities from different tree species was not correlated with phylogenetic distances between tree species. Similarity of ambrosia beetle communities from individual conspecific trees was not higher than that from heterospecific trees and different parts of the trees hosted similar ambrosia beetle communities, as only a few species preferred particular tree parts.5. In contrast, phloeophagous bark beetles showed strict specificity to host plant genus or family. However, this guild was poor in species (12 species) and restricted to only three plant families (Moraceae, Myristicaceae, Sapindaceae).6. Local diversity of both bark and ambrosia beetles is not driven by the local diversity of trees in tropical forests, since ambrosia beetles display no host specificity and bark beetles are species poor and restricted to a few plant families.