When I first read Matthew Baron’s new dinosaur study, I actually gasped.

For most of my life, I’ve believed that the dinosaurs fell into two major groups: the lizard-hipped saurischians, which included the meat-eating theropods like Tyrannosaurus and long-necked sauropodomorphs like Brontosaurus; and the bird-hipped ornithischians, which included horned species like Triceratops and armored ones like Stegosaurus. That’s how dinosaurs have been divided since 1887. It’s what I learned as a kid. It’s what all the textbooks and museums have always said. And according to Baron, a Ph.D. student at the University of Cambridge, it’s wrong.

By thoroughly comparing 74 early dinosaurs and their relatives, Baron has radically redrawn the two major branches of the dinosaur family tree. Defying 130 years of accepted dogma, he splits the saurischians apart, leaving the sauropods in one branch, and placing the theropods with the ornthischians on the other. Put it this way: This is like someone telling you that neither cats nor dogs are what you thought they were, and some of the animals you call “cats” are actually dogs.

Dinosaur images by Dmitry Bogdanov, Torley, Durbed, under a Creative Commons Attribution-Share Alike 3.0 License.

“Luckily, most of what we’ve pieced together about dinosaurs—how they fed, breathed, moved, reproduced, grew up, and socialized—will stand unchanged,” says Lindsay Zanno from the North Carolina Museum of Natural Sciences, who was not involved in the study. Still, “these conclusions lead us to question the most basic structure of the entire dinosaur family tree, which we have used as the backbone of our research for over a century. If confirmed by independent studies, the changes will shake dinosaur paleontology to its core.”

“You guys are brave,” I told Baron on the phone.

“Or stupid,” he says, laughing, with a tinge of nervousness.

If Baron is right, how could such a well-established fact about dinosaurs turn out to be wrong?

Partly, people may have been blinded by their preconceptions. The split between saurischians and ornithischians was first proposed in 1887 by British paleontologist Harry Seeley, who divided the groups based on the shape of their hips—a seductively obvious difference. “When you look at specimens, it’s very quick and easy to say: that’s a saurischian, that’s an ornithischian, and never the two shall meet,” says Baron. “People then go into their studies with that mindset, and Seeley’s idea has never been rigorously tested.”

That’s not to say that previous researchers did sloppy science. They worked with what they had. It certainly didn’t help them that the earliest dinosaurs, which tell us most about how the group first evolved, are also the rarest.

If you collapsed the entire reign of the dinosaurs into a single calendar year, icons like Tyrannosaurus and Triceratops only appeared after Christmas, Velociraptor was only alive in mid-December, and Stegosaurus and Diplodocus arose in early July. There are far fewer fossils from the first half of the dinosaurs’ history in the Triassic and early Jurassic periods. And among the species from those pioneering millennia, the saurischians commanded most of the attention. Early ornithischians are seldom unearthed.

Those were the species Baron originally set out to study. As he pored over the few existing specimens, he got a sense that they were oddly theropod-ish. “Other people had been sniffing around this but because the [saurischian/ornithischian split] is so dogmatic, they didn’t follow it up,” says Baron. “When I talked to David Norman, my supervisor, he had this wry smile, and confessed that he always had the same hunch. And he said: Wouldn’t it be interesting to test that?

Baron spent three years gathering a huge data set on 74 species of dinosaurs and dinosaur-adjacent species. He largely ignored the late-arriving celebrities and focused from the Triassic and early Jurassic. For each species, Baron painstakingly noted and measured 457 physical characteristics. He then used these traits to create his own family tree. “The more species that are included in these analyses, the more robust the result,” says Susannah Maidment, from the University of Brighton. And Baron “examined many, many more species than have previously been used,” with computing power that just wasn’t available in previous decades.  

Baron’s tree rewrites the first chapters of the dinosaur story in several ways. It suggests that they first arose around 247 million years ago, slightly earlier than the 231 to 243 million year range that’s typically cited. It hints that they might have originated in the northern half of the world rather than the southern half. And most importantly, it says that the ancestral dinosaurs split into two major groups—just not the ones we traditionally recognize.

One group, which Baron bills as Saurischia v2.0, lumps the long-necked sauropodomorphs with an obscure group of ancient meat-eaters called the herrerasaurids.  The bigger surprise was the union between ornithischians and theropods, in a new group that Baron calls Ornithoscelida. This would mean that Tyrannosaurus is more closely related to Triceratops than it is to Brontosaurus, even though scientists have always believed that it's the other way around. Unlike the distinctive bird hips of the ornithischians, there’s no single character that distinguishes the ornithoscelidans. Rather, they’re characterized by a hodgepodge of at least 21 hard-to-define features. The first of them were probably small, agile, two-legged runners with large, grasping hands and a varied diet.

Indeed, that’s how Baron sees the first dinosaurs as a whole. “They were probably generalists, which served as a big advantage during the middle and late Triassic—arid and very harsh times to be alive,” he says. “But that’s not going to be the most popular idea. A lot of people thought that dinosaurs were ancestrally carnivorous, and it’s just the ornithischians being weird.” Instead, he suggests that they began as omnivores, and two separate groups—the theropods and the herrerasaurids—separately took to a fully carnivorous lifestyle.

The new tree also changes the origin story of feathers. There are plenty of feathered dinosaurs, whose coverings range from simple filaments (“dinofuzz”) to true flight-capable plumes. The majority of these species are theropods—the group that includes Tyrannosaurus, Velociraptor, and all modern birds, and that was previously part of the saurischian line. But some ornithischians also had fuzz or quills, including the horned dinosaur Psittacosaurus and the small, agile Tianyulong and Kulindadromeus.

So, if representatives from both major dinosaur lineages had feathers, perhaps the earliest dinosaurs did. Perhaps proto-feathers were part of the group’s default anatomy, only to be lost by some members, like the long-necked sauropodomorphs. But in Baron’s tree, all the feathered species are part of the newly christened ornithoscelidan group. In this tree, feathers evolved not at the dawn of the dinosaurs, but sometime after the sauropodomorphs split away from the rest. They were an ornithoscelidan invention.

Dinosaur specialists typically hear about new discoveries long before they make a public splash, either at conferences or on the grapevine. But Baron and his co-authors, David Norman and Paul Barrett, have mostly kept their study quiet. “This is going to be a bit of a bombshell moment,” he says. “I don’t know how people are going to react.”

Pretty positively, it turns out. “This paper has kept me up at night!” Darla Zelenitsky, from the University of Calgary, told me. “This research is revolutionary, and the newly proposed relationships will fundamentally change the way we as paleontologists think about dinosaurs.” Thomas Holtz Jr., from the University of Maryland, adds, “I teach two dinosaur courses, and may run one using this new hypothesis as the preferred model.”

Others are sitting on the fence. “I wouldn’t start rewriting the textbooks just yet,” says Steve Brusatte, from the University of Edinburgh. “This is just one analysis, and lots of recent studies recovered the more traditional grouping. Since this new result contradicts such a vast legacy of research, I think the bar [to accepting it] should be high.”

Building family trees using fossils is not straightforward. The results depend on which methods you use, which species you include, which traits you measure, and how you interpret the specimens. The discovery of new fossils, or a different take on existing ones, could change the tree dramatically.

For example, despite his attempts to focus on early dinosaurs, Baron only included two ornithischians from the Triassic period, “and both have issues,” says Thomas Holtz Jr. One of them—Pisanosaurus—is known from just a single, incomplete, and highly fragmented skeleton. Some scientists aren’t even sure if it’s a dinosaur at all. The second—Eocursor—is definitely an ornithischian, but one group of researchers recently suggested that it lived later than previously thought, in the early Jurassic instead of the Triassic.

The dinosaur-adjacent species in the study—those that were closely related to dinosaurs, but weren’t dinosaurs themselves—are also crucial. They hint at what traits were primitive to the dinosaurs, and thus which species were closer to the ancestral templates. “We are leaps and bounds ahead of where we were 15 years ago, but we still have a fairly limited understanding of [these groups],” says Nathan Smith, from the Natural History Museum of Los Angeles County.

But Baron counters that his new tree doesn’t hinge on any single oddball species. He and his colleagues checked what would happen if they took out this dinosaur, or that one, or all the early ornithischians, or all the early theropods, and so on. “We tested and tested and then some, and it’s a robust result,” he says. At one point, he and his colleagues also forced their software to do everything it took to produce the old saurischian/ornithischian tree. “We needed 20 extra steps,” Baron says. “Twenty physical traits would need to be chalked up to complete coincidence before we’d think that this new tree is incorrect.”

“All such analyses rise or fall on the quality of the data, so what we need to do now is to forensically examine the new dataset to see why it’s giving different results from previous studies,” says Brusatte. “I’m naturally skeptical, but sometimes paradigm shifts do happen.”

Indeed, major shake-ups in animal classification happen more often than you might think. Even now, new studies are still redrawing the relationships between major groups of birds and mammals—living species that we’re familiar with. “Even though we have complete genomes and entire bodies, we still can't resolve if elephants are closer to sloths than to horses, for example,” says Holtz.

“Even 40 years ago it was still uncertain if dinosaurs were a single united group, or if birds came from them,” says John Hutchinson, from the Royal Veterinary College. “We’ve moved on since then and will continue to—maybe this paper will retrospectively be looked upon as a turning point. It is by rationally challenging any dogma that science gets stronger.”

“Knowing [the authors], I’m fairly certain that none of them is slapping their hands and saying, ‘Problem solved—we can all go home!’” adds Smith.