Evolution toward increased specificity in pollination systems is thought to have played a central role in the diversification of angiosperms. Theory predicts that the presence of trade-offs in adapting to different pollinator types will favor specialization, yet few studies have attempted to characterize such interactions in nature. I conducted flight cage experiments with bats, hummingbirds, and artificial flowers to examine effects of corolla width on pollination. I videotaped visits to analyze pollinator behavior and counted pollen grains transferred to stigmas. Results demonstrated that flower-pollinator fit is critical to effective pollination; wide corollas guided bat snouts better, and narrow corollas guided hummingbird bills better. Poor fit resulted in variable entry angles and decreased pollen transfer. A model using these results predicts that wide corollas will be selected for when bats make more than 44% of the visits and narrow corollas when they make fewer. Intermediate corollas are never favored (i.e., generalization is always suboptimal). This is the first study to clearly document a pollinator-mediated fitness trade-off in floral morphology.