The effects that changes in forest structure and composition have on wildlife have often been considered independently, such that the potential for interactive effects has received relatively little attention. We investigated the importance of vegetation structure, floristic composition, and their interaction for predicting bird distribution in mixed broadleaf—conifer forest. We collected vegetation and bird data at 979 stations in a watershed in southern Oregon in the spring of 2001. At each station, we described the vegetation using measures of structure (total vegetation volume) and of floristic composition (broadleaf-conifer composition). We then used logistic regression to model the probability of occurrence of bird species as a function of these 2 variables, their quadratic terms, and their interaction. Using stepwise model selection we identified the best model for each species and used area under the curve (AUC) scores to evaluate model performance. Of the 44 bird species we investigated, 20 had models with AUC scores ≥0.70. Of the best models for these 20 species, 1 included vegetation composition alone, 12 included just the main effects of vegetation composition and structure, and 7 included both the main effects and interaction terms. In summer of 2001 a wildfire burned 2,500 ha of the study area, resulting in substantial changes in the vegetation structure and composition. We used 4 yr of postfire bird and vegetation surveys to test the predictive performance of the habitat models for 9 species that occurred at >15% of the burned stations. Models for 3 of these species performed poorly (AUC < 0.7) in all 4 yr. For the other 6 species, predictive performance was low in the first year after fire and improved during subsequent years, suggesting a lagged response to changes in vegetation structure and composition. Quantifying the interactive effects of vegetation structure and composition improves our understanding of how birds respond to forest management and large-scale disturbances, such as wildfires.