Surveys in 1974 of bumble bee species distributions along elevational gradients (Pyke 1982) were revisited to reevaluate the original conclusion that coexistence of bumble bee species can be ascribed to niche differentiation, primarily on the basis of proboscis lengths and the associated corolla lengths of visited flowers. Each bee species largely visited a few plant species, which were preferred relative to other species. Bee proboscis length was correlated with average corolla length of visited flowers, but not when species with relatively long and short proboscises were considered separately. Bumble bee abundance was affected by presence or absence of major plant species and, contrary to the interpretation of Pyke (1982), elevation, with neither factor dominating. Multimodal distributions of proboscis lengths and altitudinal replacement of bee species of similar proboscis length were consistent with the original hypothesis that bumble bee species compete for floral resources, especially nectar, and cannot coexist if proboscis lengths are too similar, unless one species is a "nectar robber" and hence has exclusive use of some floral resources. However, observed overlap in elevational distributions of bumble bee species with similar proboscis length cannot be reconciled with this hypothesis unless other phenomena are invoked.