Understanding processes underlying spatial distribution of tree species is fundamental to studying species coexistence and diversity. This study modeled point patterns of tree distribution, expressed by Cartesian coordinates of individual trees within a mapped forest stand, for the purpose of identifying processes that may generate spatial patterns of tree communities. We used four primary point pattern processes (homogeneous Poisson process, inhomogeneous Poisson process, homogeneous Thomas process, and inhomogeneous Thomas process) to model tree distribution in two stem-mapped forests in Taiwan, Republic of China. These four models simulate spatial processes of habitat association and seed dispersal, allowing us to evaluate the potential contribution of habitat heterogeneity and dispersal limitation to the formation of spatial patterns of tree species. The results showed that the inhomogeneous Thomas process was the best fit model and described most of the species studied, suggesting that spatial patterns of tree species might be formed by the joint effects of habitat associations and dispersal limitation. The homogeneous Thomas process that models the effect of dispersal limitation was the second best model. We also found that the best fit models could be predicted by species attributes, including species abundance and dispersal mode. The significant traits, however, differed between the two study plots and demonstrated site-specific patterns. This study indicated that the interactive operation of niche-based (habitat heterogeneity) and neutral-based (dispersal limitation) may be important in generating spatial patterns of tree species in forest communities.