We developed a model of hardwood tree colonization in forest patches. We began with a basic model of species' recruitment density calculated as a function of seed production and juvenile survivorship. Survivorship probability was expressed as a function of seed size, using seed-sowing data for a wide variety of species. To account for dispersal, we used an exponential distancedecay function based on empirical colonization data for species classified by dispersal mechanism and seed mass. The basic model reasonably predicted observed recruitment densities at or near forest edges, except for small seeded, wind-dispersed species with strong establishment constraints, for which it over-predicts. Our dispersal term yielded predictions that were not statistically different from observed colonization. However, species with large seeds and unspecialized dispersal mechanism appear to have distinct thresholds beyond which no dispersal occurs. Further research should better account for establishment constraints among small-seeded species, dispersal constraints among large-seeded species and unspecialized dispersers, and improve the dispersal functions to better reflect vectors such as birds. Nevertheless, the present model is adequate for the prediction of colonization probabilities in fragmented forests, requiring only an estimate of the abundance of source trees of a species and the mean inter-fragment distances.