Early-diverging lineages provide little information about ancestors

Large phylogenetic trees are often unbalanced near the root, with a species-depauperate lineage sister to a far more species-rich clade. The species-poor branches, sometimes referred to as "early diverging," are often viewed as key sources of information about deep evolutionary history. As a result, these lineages occupy a central role in debates about ancestral states. Using simulations and likelihood-based models of trait evolution, we show that the information contributed by species-depauperate lineages to ancestral state reconstructions is surprisingly limited. Instead, in unbalanced phylogenetic trees, the accuracy of ancestral state reconstructions at the root node is determined primarily by the species-rich clade and, more specifically, the evolutionary distance separating that clade from the root, rather than by the character state or evolutionary rate of the "early-diverging" sister lineage. This pattern reflects a general property of how phylogenetic inference aggregates information across descendant lineages: ancestral state reconstructions combine evidence from both descendant lineages rather than applying disproportionate weight to one of the descendants simply because fewer branching events separate it from the ancestral node. Our results highlight the limits of using extant, species-depauperate lineages to infer deep ancestors and the importance of fossils and other evidence that more directly sample early evolutionary history.