Estimates of global species diversity have varied widely, primarily based on variation in the numbers derived from different inventory methods of arthropods and other small invertebrates. Within vertebrates, current diversity metrics for fishes, amphibians, and reptiles are known to be poor estimators, whereas those for birds and mammals are often assumed to be relatively well established. We show that avian evolutionary diversity is significantly underestimated due to a taxonomic tradition not found in most other taxonomic groups. Using a sample of 200 species taken from a list of 9159 biological species determined primarily by morphological criteria, we applied a diagnostic, evolutionary species concept to a morphological and distributional data set that resulted in an estimate of 18,043 species of birds worldwide, with a 95% confidence interval of 15,845 to 20,470. In a second, independent analysis, we examined intraspecific genetic data from 437 traditional avian species, finding an average of 2.4 evolutionary units per species, which can be considered proxies for phylogenetic species. Comparing recent lists of species to that used in this study (based primarily on morphology) revealed that taxonomic changes in the past 25 years have led to an increase of only 9%, well below what our results predict. Therefore, our molecular and morphological results suggest that the current taxonomy of birds understimates avian species diversity by at least a factor of two. We suggest that a revised taxonomy that better captures avian species diversity will enhance the quantification and analysis of global patterns of diversity and distribution, as well as provide a more appropriate framework for understanding the evolutionary history of birds.