Predictive distribution modeling is an important tool in ecological research and conservation. Spatial autocorrelation in the residuals of models solely based on environmental predictors, but not on 124 Abstracts AOU-SCO/SOC 2004 spatial structure, indicate inadequate fit and invalid estimates for statistics. Furthermore, spatial autocorrelation in the dependent variable carries additional information, which can improve predictions of distribution models, when appropriate spatially explicit modeling techniques are employed. Selecting appropriate models requires a thorough understanding of the underlying causes of autocorrelation. A primary cause of spatial autocorrelation in species’ distributions is correlated environmental variables. If it were the only cause, traditional, non-spatial models would be adequate; when all contributing environmental factors are included in a model they implicitly capture the spatial structure in the species’ distribution. However, we suggest that dispersal also leads to autocorrelation in species distributions, which then would necessitates spatially explicit distribution modeling. Since dispersal data are hard to collect at large extents, we used several well-established ecological theories (e.g., Taylor power law for population dynamics, metapopulation theory, hierarchy theory, density dependence) to predict relative dispersal rates for 28 species of birds using data from the Breeding Bird Survey. We then compared these dispersal rates to the relative strength of residual autocorrelation in the species’ distributions. Our preliminary results indicate that dispersal plays a significant role in shaping the selected bird distributions and that spatially explicit distribution models outperform traditional models as determined by Akaike’s Information Criterion.