This is a nice write-up by Laura Geggel of a current exchange of comments in Nature about dinosaur phylogeny: “Dino Family Tree Overturned? Not Quite, But Changes May Lie Ahead”.
The upshot is that last spring, Matthew Baron and colleagues (2017) claimed that the traditional groupings of dinosaurs were all wrong. For more than a hundred years, paleontologists have grouped theropods together with sauropods, as “saurischians”, based on pelvic morphology. Baron et al. suggested that the theropods are instead relatives of the ornithischians—including duckbills and ceratopsians.
These branches are within the deepest part of the dinosaur phylogeny, and many of the fossil groups in the dataset lived much later and have many derived traits that would have been absent in their common ancestors. This makes it harder test their relationships than one might expect. The problem is analogous to determining relationships among the very deepest nodes of the mammal phylogeny—for example, do we group together primates, bats, and rodents into a higher level taxon, and are insectivores really a single group? Paleontologists have radically revised some ideas about early mammal diversification in the wake of genetic comparisons of living species, because these relationships just are not well reflected by morphological traits. For dinosaurs, there are no genetic comparisons, and we shouldn’t be very surprised that morphology might not be a straightforward indication of the deepest relationships.
But the new exchange of comments, initiated by Max Langer and colleagues, shows that the dinosaur phylogeny is not going to be overturned easily. In their assessment, Baron and coworkers scored some characters incorrectly. They suggest that the correct data still support the traditional hypothesis that connects the theropods and sauropods.
I don’t have any deep insight about dinosaur phylogeny. But I am interested in the case because it reflects a singular problem with phylogenetic analyses that we are also seeing expressed in the study of hominin relationships.
Many empirical sciences are going through a “replication crisis”, as statisticians are showing that studies are systematically underpowered and results driven by false positives and p-hacking. We can’t precisely compare phylogenetic methods to the kind of statistical analyses underlie many hypothesis tests in other branches of science.
But something very similar is true in phylogenetics. Scientists working on fossil relationships are working with sparse data matrices, many key taxa are very poorly represented, with samples that often include only a single individual, and many interesting questions involve deep nodes. The advent of genetics in the phylogenetics of mammals, birds, and many other groups has shown just how badly morphological data represent deep relationships.
The adoption of Bayesian methods has helped a bit, in that the Bayes factor provides at least a way of saying that the data don’t clearly distinguish hypotheses from each other. I think that today many scientists working on hominin relationships have a fairly healthy attitude, that we just do not know how some key species should be arranged in a phylogeny.
Certainly we face that problem with species like Homo naledi and Australopithecus sediba. These species are exceptionally well represented across the skeleton by fossils, but their placement cannot be determined with any confidence except in very broad terms.
For dinosaurs, I expect that this phylogenetic problem will continue for quite a while, as the current exchange shows that the phylogenetic methods are very sensitive to small changes in the datasets.
Baron, M. G., Norman, D. B., & Barrett, P. M. (2017). A new hypothesis of dinosaur relationships and early dinosaur evolution. Nature, 543(7646), 501-506.
Langer, M. C., et al. (2017). Untangling the dinosaur family tree. Nature 551, E1–E3.