As a contribution to enlarging and widening the array of studies available for examination of insect—plant relations, this study contrasts local and regional predictors of species richness and host specificity of leaf—chewing insects feeding on New Guinea trees. I tested the degree to which the variance in local insect variables (species richness, ratio of specialist to generalist chewers, abundance, and biomass) for leaf chewers associated with different species of tropical trees was accounted for by local and regional host—related variables, and by local abundance of potential enemies. Leaf—chewing insects feeding externally on ten species of forest trees were collected over a 1—yr field study in a submontane area in Papua New Guinea. The tree species were taxonomically distant, and the list included species of different successional status, different patterns of leaf production, and different heights. The affinity of chewing insects with particular tree species was ascertained by feeding trials in the laboratory. Local insect species richness varied greatly among tree species, by more than four—fold between the poorest (18 species) and the richest (94 species) tree species. The proportion of specialist insect species associated with particular tree species was significantly different among tree species and ranged from 16 to 72%. Trees supporting a rich fauna usually also supported a high proportion of specialists. The abundance and biomass of leaf—chewing insects per host species was difficult to predict. However, path analyses showed that most of the causal variance in insect species richness and in the ratio of specialist to generalist chewers was accounted for by five properties of the tree species: numbers of young leaves available throughout the year, ant (i.e., enemy) abundance, leaf palatability, leaf water content, and altitudinal range. Most of the variation in local species richness could be predicted from local processes (i.e., food resources and abundance of enemies), not from regional processes. This study and others suggest that in complex environments, such as tropical rain forests, local processes may be comparatively more important in maintaining the local species richness of insect herbivores than in less complex ones, such as temperate woodlands. The implications of these findings are evident: the effects of moderate habitat fragmentation and ecological stress may be more pronounced in tropical systems than in temperate systems and may result in a greater proportional loss of local biodiversity in the former.