Because teeth are the most easily preserved part of the vertebrate skeleton and are particularly morphologically variable in mammals, studies of fossil mammals rely heavily on dental morphology. Dental morphology is used both for systematics and phylogeny as well as for inferences about paleoecology, diet in particular. We analyze the influence of evolutionary history on our ability to reconstruct diet from dental morphology in the mammalian order Carnivora, and we find that much of our understanding of diet in carnivorans is dependent on the phylogenetic constraints on diet in this clade. Substantial error in estimating diet from dental morphology is present regardless of the morphological data used to make the inference, although more extensive morphological datasets are more accurate in predicting diet than more limited character sets. Unfortunately, including phylogeny in making dietary inferences actually decreases the accuracy of these predictions, showing that dietary predictions from morphology are substantially dependent on the evolutionary constraints on carnivore diet and tooth shape. The “evolutionary ratchet” that drives lineages of carnivorans to evolve greater degrees of hypercarnivory through time actually plays a role in allowing dietary inference from tooth shape, but consequently requires caution in interpreting dietary inference from the teeth fossil carnivores. These difficulties are another reminder of the differences in evolutionary tempo and mode between morphology and ecology.

Dental dimensions measured using digital calipers from specimens in the UC Museum of Vertebrate Zoology, with taxonomy corrected to MSW3 for consistency with phylogenetic data. Specimen-level data includes 2 specimens of each species where available; species-level data is simply one of the specimens with the most complete set of measurements or randomly chosen if both were equally complete. Diet data collected from primary literature as described in paper.

Missing values in measurement data represent specimens for which the dimension could not be measured, either because the image we took of it was not of adequate quality or because that part of the specimen was missing or damaged. See ReadMe file for additional information.