1 The spatial distribution of 30 woody species (15 species each of Euphorbi
aceae and Moraceae) and their associated leaf-chewing communities (Orthopte
ra, Phasmatodea, Coleoptera and Lepidoptera) were studied in coastal, river
ine and rain forest habitats. A successional series, from abandoned gardens
to primary forest, was examined. Host plant records for more than 27 000 i
nsects, all verified by feeding experiments, and spatial distribution of al
most 9000 plant specimens were evaluated.
2 Phylogenetic (taxonomic) relatedness of host plants explained 56% of the
variability in the composition of their herbivore communities, while the ec
ological (distribution) similarity of plants explained only 4%.
3 The successional optimum of plant species was not an important determinan
t of the composition of their herbivore communities.
4 Neither plant successional optimum nor plant palatability to a generalist
herbivore were correlated with the number of species, abundance or host sp
ecificity of its herbivores, nor was there a correlation between a plant's
palatability to a generalist herbivore and its successional optimum.
5 Herbivore communities became dominated increasingly by a few abundant spe
cies in later stages of succession.
6 On average, Ficus species had lower palatability and supported more speci
es of herbivores than species of Euphorbiaceae. The abundance of herbivores
and their dominance index were not significantly different between the two
7 These results contradict several previous studies of successional trends
in temperate regions. Many tropical successions, however, start with pionee
r trees, rather than with annual herbs, and may present a permanent and pre
dictable habitat for insects even at the earliest stages, with no advantage
for polyphagous species. Numerous pioneer trees in the tropics possess ant
i-herbivore defences, resulting in their low palatability to generalists, i
ncreased host specificity of herbivores, and often idiosyncratic compositio
n of herbivore communities. Even plant traits such as species richness of t
heir herbivores or palatability may have a phylogenetic component which sho
uld not be ignored.