We used regression models and information-theoretic model selection to assess the relative importance of environment, local dispersal and historical contingency as controls of the distributions of 26 common plant species in tropical forest on Barro Colorado Island (BCI), Panama. We censused eighty-eight 0.09-ha plots scattered across the landscape. Environmental control, local dispersal and historical contingency were represented by environmental variables (soil moisture, slope, soil type,

more »

... nce to shore, old-forest presence), a spatial autoregressive parameter (ρ), and four spatial trend variables, respectively. We built regression models, representing all combinations of the three hypotheses, for each species. The probability that the best model included the environmental variables, spatial trend variables and ρ averaged 33%, 64% and 50% across the study species, respectively. The environmental variables, spatial trend variables, ρ, and a simple intercept model received the strongest support for 4, 15, 5 and 2 species, respectively. Comparing the model results to information on species traits showed that species with strong spatial trends produced few and heavy diaspores, while species with strong soil moisture relationships were particularly drought-sensitive. In conclusion, history and local dispersal appeared to be the dominant controls of the distributions of common plant species on BCI.