Geometrid moths were collected on Mt. Kinabalu (Sabah, Malaysia) along two habitat gradients, ranging from primary rain forest to cultivated areas. During 135 nights' trapping in 1997, 4585 individuals representing 500 species were attracted by light. Primary forest samples and those from old-grown regenerated forest exhibited high diversity (Fisher's α = 75–128), while agricultural areas as well as most secondary forests had a significantly lower diversity (α = 34–61). One 15-y-old secondary forest with a rich undergrowth vegetation also housed a diverse geometrid community (α = 89). In three paired samples, diversity of geometrid moths in the canopy was equal to or lower than in the understorey. Of six habitat variables tested, only undergrowth plant species diversity emerged as a significant predictor of geometrid diversity. The NESS index, in combination with multi-dimensional scaling, was used to investigate patterns of between-habitat diversity. Of two dimensions extracted, one represented the degree of habitat disturbance, while the other separated the two study areas. Geometrid samples of strongly disturbed habitats did not converge between sites, indicating that apart from the degree of human interference the regional species pool was also important in determining similarity among local communities.