Both local and regional habitat characteristics influence species richness and community structure. The scale at which communities are studied, however, affects the detection of relationships between habitat characteristics and patterns of habitat selection, species diversity, and species composition, and may obscure observation of differences in how species perceive the scale of environmental variation. To determine how environmental variation at different scales is related to species occurrence and richness, I analyze mist net sampling data on several guilds of forest understory birds. Bird capture, vegetation, and physical environment data come from 23 0.5-ha study sites in primary and secondary forest in Amazonian Ecuador. The percentages of primary forest within concentric circles around each site form forest imbeddedness measures (FIMs), which are evaluated using satellite imagery. Variation in FIM size represents different measurement scales for determining forest cover. Primary forest cover is also analyzed in successively larger tori surrounding sites and is used, after variable reduction with Principal Components Analysis, to summarize variation in forest cover around sites. Linear regression, surface trend analysis, and ordination help to quantify how variation in guild composition and species richness is explained by forest cover, vegetation structure, and physical environment. Species composition is related to variation in primary forest cover, primarily within 200-600 m of study sites. Canonical Correspondence Analysis (CCA) indicates that nectarivores, shrub-layer frugivores, and ant-following birds are captured in areas with relatively low primary forest cover. In contrast. shrub-layer insectivores, shrub-layer omnivores, and birds probing dead foliage for large insects tend to be captured in areas of relatively high primary forest cover. The species richness of insect gleaners, ant followers, and omnivores is statistically related to the percent cover of primary forest within hundreds of meters from the study sites. This suggests that some mechanisms that influence guild composition act over substantial distances. Nonetheless, the small radii of FIMs related to the species richness of dead-leaf probers suggests that local conditions and variation in forest cover over short (<200 m) distances directly or indirectly influence species richness of some primary forest birds. The significant relationship between temperature variation among capture sites and species richness of ant-following birds suggests that these species choose among habitats in a temperature range at which physiological constraints operate, either directly on the birds themselves or on the ants they follow. Species richness within the nectarivores, in contrast, shows no relationship with large-scale variation in primary forest cover. The radius of the FIM most closely associated with species richness differs among guilds, which suggests variation in the scale at which forest cover is associated with guild structure, as well as variation in the strength of the association. Differences in the scale of relationships between environment and species richness among guilds suggest that the mechanisms that influence both species' habitat use and community structure differ among guilds. A single mechanism, operating at a single scale, is inconsistent with these patterns. Independent data from eight additional sites suggest that prediction of species richness from surrounding forest structure could be useful for managing human impacts on tropical avian communities.
