Biogeographical studies of the Caribbean have generated numerous controversies most of which swirl around geological history and issues regarding land bridges, vicariance and dispersal. However, the botanical perspective largely has been wanting. We approach the problem from an ecological perspective rather than a strictly historical one and use orchids as our model system. There are approximately 710 species in 134 genera in the West Indies. The frequency of single island endemics is high (44%) and these are non-randomly distributed among genera. Of the five genera that are represented by more than 25 species, Lepanthes has the highest incidence of endemism (98% of 95 species) whereas Epidendrum has the lowest (27% of 56 species). Nevertheless, all members of the family except Vanilla have dust-like seeds with a strong potential for long-distance dispersal. We used two methods to reveal biogeographical patterns of only the wind dispersed orchids: one involves a previous study of parsimony analysis of distributions (PAD) and the other employs species-area relationships. PAD examines affinities among regions or islands based only on those species that they share. Our data matrix includes 52 regions or areas and 346 species. We found that groupings of islands were congruent with their gross ecological features either from similar geomorphology or common physiography. From these results we hypothesized that there should be strong species-area relationships for all orchid species, a prediction based on ecological probabilities. We found a highly significant species-area relationship for 26 mountainous islands, but the slope for 19 relatively homogeneous, calcareous islands was marginally non-significant. The species-area regression for the latter explained only 16% of the variation and had a slope of 0.25, a value consistent with studies of other organisms, including those of the Caribbean. On the other hand, a much more dramatic relationship between species numbers and island areas was revealed for the islands with substantial topography. The regression explained 79% of the variation and had a very steep slope (² = 0.51). Whereas only three of the 19 calcareous islands had endemic species, 9 of the 26 montane islands had endemics. Species/area relationships shown by the latter indicated there was a strong, significant correlation similar to the analysis involving all species and all 26 islands. We conclude that orchid species richness and distributions are somewhat predictable and are especially influenced by ecological conditions of the islands.