Extant-only comparative methods fail to recover the disparity preserved in the bird fossil record.

Most extant species are in clades with poor fossil records, and recent studies of comparative methods show they have low power to infer even highly simplified models of trait evolution without fossil data. Birds are a well-studied radiation, yet their early evolutionary patterns are still contentious. The fossil record suggests that birds underwent a rapid ecological radiation after the end-Cretaceous mass extinction, and several smaller, subsequent radiations. This hypothesized series of repeated radiations from fossil data is difficult to test using extant data alone. By uniting morphological and phylogenetic data on 604 extant genera of birds with morphological data on 58 species of extinct birds from 50 million years ago, the "halfway point" of avian evolution, I have been able to test how well extant-only methods predict the diversity of fossil forms. All extant-only methods underestimate the disparity, although the ratio of within- to between-clade disparity does suggest high early rates. The failure of standard models to predict high early disparity suggests that recent radiations are obscuring deep time patterns in the evolution of birds. Metrics from different models can be used in conjunction to provide more valuable insights than simply finding the model with the highest relative fit.