Studies of habitat fragmentation have been restricted primarily to anthropogenically-altered habitats, with most research conducted 60'90 years post-fragmentation. It is unclear whether patterns in older systems concur with results from these dynamic landscapes, and hence the long-term viability of populations inhabiting habitat fragments remains largely unexplored. I focused on resident birds in fragments of humid pine-oak forest in Oaxaca, southern Mexico, isolated over 5000 years ago by climate-change. Seventeen fragments, ranging from 2 ha to over 150,000 ha were sampled in 1997 and 1998 yielding 141 species, of which 60 residents were used for analysis. Avian assemblages exhibited a highly nested structure and, with several notable exceptions, assemblages of birds in low-richness fragments were predictable subsets of those in more diverse fragments. Patch-scale factors'area, shape, elevation, habitat diversity and fractal dimension of edge'all exerted strong univariate influence on avian richness but were so closely inter-related that none had a significant independent effect. Thus, larger fragments were more complex in shape, included higher peaks, supported more diverse forests, and contained higher diversities of resident species. In contrast, the landscape-scale index used'distance from nearest large fragment (>50,000 ha)'had little effect on richness. This was reinforced by species-level analyses'one species was significantly influenced by isolation, compared with 31 species that displayed significant minimum-area distributions, restricted to patches larger than a particular threshold value. In terms of autecology, vagility, relative abundance and elevational breadth were closely related to distribution'those species with greater mobility, higher abundances and broader elevational tolerances were consistently more widespread. I suggest that more abundant species were less prone to extinction initially, more vagile species were betterdispersers and species with broader elevational tolerances more likely to be successful colonists. As with previous research from older landscapes, patch-scale factors were consistently found to be influential, with high quality fragments supporting diverse communities regardless of landscape context. This suggests that the influence of landscape-scale factors noted in younger, anthropogenically fragmented systems may be transitory, overwhelmed by patch-scale factors with time. Which patch attributes are most influential could not be resolved, however, indicating that even thousands of years after fragmentation, they affect diversity patterns in concert. Rather than differentiating effects of area from habitat heterogeneity and other patch-level factors, I advocate resource-based approaches to understand and manage diversity in habitat fragments.