Effective conservation and habitat restoration strategies in human-dominated landscapes must be based on an understanding of the ways that habitat loss and fragmen- tation affect native species. We studied avian foraging behavior and patterns of occurrence in the highly fragmented agricultural landscape of the Kellerberrin district of Western Australia to better understand the factors underlying species declines and losses. We con- ducted three surveys of 30 wandoo woodland patches that ranged in size from 1.3 to 101.3 ha. Some patches were part of larger remnants of native vegetation, ranging in size from 3.5 to 1204.8 ha and including other habitat types. We examined the extent to which patterns of species richness, the occurrence and composition of foraging guilds, and the occurrence of individual species varied with features of woodland patches, remnants, and the sur- rounding landscape. Using multiple regression analyses, the best model for species richness included terms for the log of remnant area, the patch diversity of each remnant, and woodland patch condition. We delineated eight foraging guilds based on similarities in the substrate/method dimension and also using multidimensional scaling analyses. The best model for the number of foraging guilds present in a patch included a single variable representing shrub density. Analyses of nestedness based on guilds and on individual species were both significant; and for the latter, 11 of 13 species made a significant contribution to the overall pattern. We derived separate models for the occurrence of each of eight species and one guild using multiple logistic regression. Significant models included, either separately or in combi- nation, the following variables: the log of remnant area, patch area, the total area of wood- land in a remnant, and the distance to other woodland patches. For four species that had sufficient records to examine shifts in foraging behavior, we observed significant differences in both foraging methods and substrates with changes in remnant size and/or the presence of other species or guilds. Our analyses indicated that remnant area was the best single variable for measures of community structure, in part because it was strongly correlated with other variables, such as total woodland area, patch area, remnant patch diversity, number of corridor connections, and measures of isolation. For foraging guilds and for individual species, variables other than remnant area assumed greater importance. The strong patterns of nestedness for for- aging guilds by remnant area may reflect the diminished availability of certain prey items in small, degraded remnants. This notion is reinforced by the relatively high species and guild richness recorded in small patches that were either fenced from grazing or embedded in large remnants. The nested pattern of species within some foraging guilds, however, indicates the importance of additional aspects of their ecology. A focus on richness alone may mask the unique responses of bird species to fragmentation and may divert attention from important considerations in the development of land-use policy and reserve acquisition.