The assemblages of root-feeding chrysomelid larvae from 21 locally common tree species were studied in a secondary tropical forest in New Guinea and compared with confamilial larvae and adults feeding on the foliage. Larval host plants were inferred from adults emerging from the soil containing the roots of known tree species. In total, 2495 chrysomelids from 100 species were reared from the roots. Almost 90% of adults in the forest canopy recruited from the species with root-feeding larvae, while species with leaf-feeding larvae represented 1% of individuals (the feeding guild for the remaining 9% was unknown). The root-feeding larvae were thus more important in tropical than temperate forests, possibly because of predation pressure by ants on tropical vegetation. The number of chrysomelids emerging annually from the soil in 1 ha of the forest was approximately 0.2 million. Root-feeding larvae were polyphagous as their modal host range included three or four from the six plant families studied. The lack of correlation between the phylogenetic distance of tree species and the similarity of their chrysomelid assemblages indicated that host choice was not constrained by plant phylogeny. The host range of larvae feeding on roots was as wide as that of the conspecific adults feeding on the foliage. The density and species composition of larval and adult assemblages on the studied trees were not correlated. These results suggest that even studies restricted to adult assemblages, which represent a majority of chrysomelid studies, can be informative, as the composition of adult assemblages is not necessarily constrained by larval host-plant selection.

Frugivorous dacine fruit flies were studied in a lowland tropical rain forest in Papua New Guinea to determine their host specificity, abundance, and the number of species attacking various plant species. Plant species hosted 0–3 fruit fly species at median (1–3 quartile) densities of 1 (0–17) fruit flies per 100 fruits. Fruit flies were mostly specialized to a single plant family (83% species) and within each family to a single genus (88% species), while most of the species (66%) were able to feed on >1 congeneric plant species. Only 30 from the 53 studied plant species were colonized by fruit flies. The plant–fruit fly food web, including these 30 plant species and the total of 29 fruit fly species feeding on them, was divided into 14 compartments, each including 1–8 plant species hosting mutually disjunct assemblages of fruit flies. This structure minimizes indirect interactions among plant species via shared herbivores. The local species pool was estimated at 152±32 (±SE) fruit fly species. Forty per cent of all taxonomically described species known from Papua New Guinea were reared or trapped in our study area. Such a high proportion indicates low beta-diversity of fruit flies. Steiner traps were highly efficient in sampling the lure-responsive fruit fly species as they re-collected 84% of all species trapped in the same area 5 y before. Fruit fly monitoring by these traps is a cheap, simple and efficient method for the study of spatial and temporal changes in rain-forest communities.