Resolving the “ontogeny problem” in vertebrate paleontology

Abstract Ontogenetic change is a major source of phenotypic variation among members of a species and is often of greater magnitude than the anatomical differences that distinguish closely related species. Ontogeny has therefore become a problematic confounding variable in vertebrate paleontology, especially in study systems distant from extant crown clades, rendering taxonomic hypothesis testing (a fundamental process in evolutionary biology) rife with difficulty. Paleontologists have adopted quantitative methods to compensate for the perception that juvenile specimens lack diagnostic apomorphies seen in their adult conspecifics. Here, I critically evaluate these methods and the assumptions that guide their interpretation using a μCT dataset comprising growth series of American and Chinese alligator. I find that several widespread assumptions are scientifically unjustifiable and that two popular methods—geometric morphometrics and cladistic analysis of ontogeny—have unacceptably high rates of type II error and present numerous procedural difficulties. However, I also identify a suite of ontogenetically invariant characters that differentiate the living species of Alligator throughout ontogeny. These characters overwhelmingly correspond to anatomical systems that develop before (and play a signaling role in) the development of the cranial skeleton itself, suggesting that their ontogenetic invariance is a consequence of the widely conserved vertebrate developmental program. These observations suggest that the architecture of the cranium is fixed early in embryonic development and that ontogenetic remodeling does not alter the topological relationships of the cranial bones or the soft tissue structures they house. I propose a general model for future taxonomic hypothesis tests in the fossil record, in which the hypothesis that two specimens are different ontogenetic stages of a single species can be falsified by the discovery of character differences that cannot be attributed plausibly to ontogenetic variation.