The elevational gradient of species richness is often claimed to mirror the latitudinal gradient and has traditionally been explained by assuming a decrease in productivity with elevation and more recently by Rapoport's rule. The influence of area on the pattern has rarely been considered. Analyses of all South American tropical land birds (more than one-fourth of the extant bird species on Earth) are used to examine four species richness/elevation models: null model, Rapoport's rule, and monotonic or hump-shaped productivity/species richness relationships. To quantify the area effect, species-area curves were created for seven elevational zones. Not accounting for area, species richness declined monotonically with elevation, but area accounted for 67%-91% of the variation in species richness per zone. When area was factored out, a hump-shaped pattern emerged, with more species in the 500-1,000-m (P<.005) and 1,000-1,500-m zones (P<.10) than in the 0-500-m zone. Rapoport's rule and the monotonic productivity/species richness relationship were thus not supported. Instead, elevational turnover rates and numbers of shared species between zones suggested that the hump-shaped pattern reflects geometric constraints (as predicted by the null model) imposed by the narrow span of the gradient, and it is suggested that midelevational zones may represent sink habitats.