Thursday, June 18, 2020

Navajoceratops & Terminocavus : new transitional species of horned dinosaurs

Finally the paper is out!

Fowler DW & Freedman Fowler EA (2020) Transitional evolutionary forms in chasmosaurine ceratopsid dinosaurs : evidence from the Campanian of New Mexico, PeerJ 8:e9251 DOI 10.7717/peerj.9251 

This is a big chapter from my 2016 PhD dissertation, and personal milestone.

If you don't know what I'm talking about, read the paper (free) here:

or our press release here:

Before I get into things, I just want to say a thank you to Prof. Thomas Holtz and Todd Johnson for posting the teaser photo on their Twitter and Facebook feeds: all the positive comments and likes from interested folks were very nice to see. It is directly because of that post that I decided to write this the way I did. This is for you!

So what is in the paper?

Key findings:

  • Two new transitional taxa of horned dinosaur (Navajoceratops sullivani, Terminocavus sealeyi), linking Utahceratops, Pentaceratops, and Anchiceratops into an evolutionary lineage evolving from 75.9 to 71.5 million years ago\
  • Significant overhaul of chasmosaurine ceratopsid dinosaur classification and relationships, including informative new characters.
  • Proposal that ~83Ma, vicariance caused cladogenesis of North American chasmosaurines, separating them into a northern Chasmosaurus lineage and southern Pentaceratops lineage, which thereafter evolved by anagenesis. 
  • Proposal of a plausible mechanism to explain differences between northern and southern dinosaur populations: ~83 Ma, high sea level abutting the incipient Rocky Mountains cut off northern and southern dinosaur populations for 1 to 4 million years, causing vicariance and laying the foundation for the apparent differences between northern and southern dinosaur faunas in the Late Cretaceous of western North America.
  • Presence of southern-affiliated Pentaceratops-lineage chasmosaurines (“Pentaceratops aquilonius”; “Chasmosaurusrusselli; Anchiceratops) in Canada reaffirms that northern and southern Laramidia were reconnected by the middle Campanian stage of the late Cretaceous.

More takehomes for the ceratopsid specialists:

  • 2-3 specimens recognized as unique taxa but too incomplete to name (or works in progress)
  • Promise of 2-3 more new named intermediate and end-member taxa in forthcoming works
  • Revision of epiparietal nomenclature and character coding
  • Discussion of the role of heterochrony in evolution of chasmosaurine skull ornamentation


You need to know some background to make sense of our new paper. Experts can jump ahead, but let’s get through it quickly using bullet points:

  • Ceratopsids are the dinosaurs with big frills at the back of their skulls, and big horns or flat bosses on their faces.
  • Triceratops horridus is probably the most famous ceratopsid. ~63 species have been named historically, although many are no longer valid.
  • Species are mostly defined based on slight differences in the shape and size of the adornments of the skull, principally the horns and the frill.
  • Ceratopsids are almost exclusively known from North America (Sinoceratops from China is the exception).
  • Ceratopsids can be divided into the short-frilled Centrosaurinae, and the long-frilled Chasmosaurinae. Our paper only concerns Chasmosaurinae.
  • In chasmosaurines, the most diagnostic bone is the parietal, this is the middle bone of the frill. In fact, it is very hard to name a new species if the skull does not preserve the parietal.  

Subtle differences in the skull shape differentiate ceratopsid species (image by Dr. Robert Boessenecker)

Ok, now we can jump in and look at the main take home points (I’ve written a lot, but then you saw the paper, so you should have been expecting that):

Two new horned dinosaurs, Navajoceratops sullivani and Terminocavus sealeyi

On the face of it this is a description of two fragmentary skulls as new species of dinosaur. No big deal right? The specimens themselves were from some extravagant looking dinosaurs, sure, but even that isn’t all that important. What sets these new taxa apart is that they, well… don’t set themselves apart. In fact they are VERY similar to already existing species, as they are what we might call “transitional species” (an oxymoron, depending on how you define a species, but bear with me here). Their intermediate morphology makes them MUCH MORE interesting since they link up two genera that most people didn’t previously consider as closely related, and tell us some intriguing new implications for how this group evolved.

Transitional species

Back in 1998 Thomas Lehman published a description of a huge skull attributed to Pentaceratops and held at the Sam Noble Museum in Oklahoma.

(As it turns out, this skull doesn’t have the most diagnostic bone preserved (parietal), but let’s not worry about that right now, because that’s not the interesting part).

Hidden away at the end of this description is an intriguing figure which compares the parietal frills of chasmosaurines, specifically looking at the notch or embayment in the center of the posterior edge or bar of the parietal. Lehman had arranged the figure to show the shallow-notched Chasmosaurus on the left, and other more deeply-notched chasmosaurines on the right, roughly arranged stratigraphically.

The curious figure from Lehman 1998. Chasmosaurus (1-7) on the left, Agujaceratops (8), Pentaceratops n. sp. (9, 10), and Anchiceratops (11, 12) on the right.

Lehman had shown Anchiceratops (which lived ~71.5 million years ago) next to what was then called Pentaceratops (which lived ~74.5 million years ago). Both bore an unusual pair of spikes (epiparietal 1) on the midline of the frill that were turned outwards like the wings of a butterfly (more so in Anchiceratops than Pentaceratops). The paper did not explicitly suggest a close relationship between Anchiceratops and Pentaceratops, but the possibility was intriguing. There were other physical similarities: both species had long brow horns, both had large triangular epiossifications (spikes) on the edge of the parietal frill… but then Anchiceratops didn’t have a notch in its posterior bar, and its frill borders were constructed from broad flat plates of bone, contrasting the thin struts of Pentaceratops’ frill. The hypothesis seemed like a stretch of the available data, and it wasn’t a topology that was retrieved by phylogenetic analyses.

I had encountered Lehman’s 1998 figure while researching Pentaceratops; my interest stemming from our discovery in 2002 of a partial skull of a chasmosaurine ceratopsid at Ahshislepah Wash in New Mexico (during one of Dr. Robert Sullivan’s expeditions). We had provisionally identified it as a Pentaceratops, but it looked slightly different, with a deeper notch in the middle of the parietal posterior bar, and wavy lateral rami (struts) of the posterior bar that got thicker as you moved towards the lateral margins. Furthermore, the various skulls attributed to Pentaceratops sternbergii had been collected from the top of the Fruitland Formation (~75.3 Ma), which is slightly older than our new skull, which came from the lower part of the overlying Kirtland Formation (~75.1 Ma). The Pentaceratops-Anchiceratops link implied by Lehman’s figure seemed to be corroborated by the new skull, but I was still in two minds about it.

SMP VP-1500: Holotype parietal of Navajoceratops sullivani in the field. R Sullivan foot for scale.

SMP VP-1500: Holotype parietal of Navajoceratops sullivani

Some time later I was shown photos of a parietal from a partial skull that had been collected by NMMNH volunteer Paul Sealey in 1997 from even higher in the Kirtland Formation. Its shape was astonishing! The midline notch was even deeper and almost closed in completely, it had broad, flat, plate-like rami of the posterior bar, and the spikes on the frill were large and triangular.

NMMNH P-27468: Holotype parietal of Terminocavus sealyi

Of course, these are what we would later name Navajoceratops sullivani, and Terminocavus selaeyi. Taken together, the two new skulls really looked like missing links between Pentaceratops and Anchiceratops.

At this point I started to look in further detail at the frills of all chasmosaurines and came up with a number of new or altered characters that would help better define how the shape varied ever so slightly between species. These new characters (e.g. the posteriormost point of the parietal posterior bar) defined some of the existing species more clearly, and (cutting a long story short) allowed us to differentiate another two skulls (now identified as aff. Pentaceratops n. sp.) from the holotype series of Pentaceratops (although this has its own issues: see the paper, especially the supp info!).

At the same time I had recalibrated over a hundred previously published radiometric dates from various dinosaur-bearing formations to get them all measured to the same standard, and used this to correlate rock formations across the continent, and plot dinosaur species occurrences in time (this was published as Fowler, 2017, free to read at PLoS).

Combining data from the radiometric dates, taxonomy, phylogenetics and morphometric analysis, we ended up with a lineage comprising stratigraphically successive species:

75.9 Ma - Utahceratops gettyi
75.3 Ma - Pentaceratops sternbergii
75.2 Ma - aff. Pentaceratops sp. nov.
75.1 Ma - Navajoceratops sullivani
74.6 Ma - Terminocavus sealeyi
73.7 Ma - (probably) Kirtland “Taxon C”
71.5 Ma - Anchiceratops ornatus

Navajoceratops and Terminocavus form intermediates in a lineage leading to Anchiceratops. Artwork by Ville Sinkkonen

The parietal morphology of each species is subtly different from the species that precedes it, and the species which follows. Detailed descriptions of the changes are in the paper, but can be summarized as deepening and gradual enclosure of the midline parietal embayment (with associated change in shape of the posterior bar and position of the epiparietal loci), coupled with expansion of the median, lateral, and posterior bars from thin and strap-like (Utahceratops, Pentaceratops), to more broad and plate-like (Terminocavus, and especially Anchiceratops).

Are these really new species? Why aren’t they just individual variation?
A few people probably might say that these different species are all within an expected range for individual variation. Indeed, if you found all of these specimens in the same bonebed, then you might think that they were all variants within a single species. However, they are stratigraphically separated, and show incremental and directional morphological change, supported by the morphometric analysis. We write a section in the supp info about why these are named as new genera (basically for convenience of reference, and avoiding various priority and paraphyly issues)

Significant overhaul of chasmosaurine ceratopsid dinosaur classification and relationships, including informative new characters and revisions of existing characters.
The paper includes a lot of review of Campanian-age chasmosaurines; mainly focusing on specimens which preserve the diagnostic parietal, and reassessing them in the light of some of the new characters.

Of particular note are characters:

  • Point of maximum constriction of the median bar (center, or posterior third)
  • Position of the posteriormost point of the posterior bar
  • Position of epiparietal locus 1
  • Posterior bar strap like (evenly narrow thickness) or narrow medially but broad laterally

We also revise the numbering system for chasmosaurine epiparietals. This should improve phylogenetic analysis as it removes some coding issues with false homology, based on taxa that evolved multiple new epiparietal spikes on the middle of the posterior bar. If you are particularly interested in this aspect you should look through some of the peer review history.

Reconfiguration of the parietal spike numbering system of chasmosaurines was required.

There is still much work to do on the phylogenetic dataset.

One issue is that we lack a outgroup for chasmosaurines. Someone needs to find a chasmosaurine from before the split between Chasmosaurus and Pentaceratops lineages. Without this, the basal morphology of the more derived chasmosaurines is not rooted to anything, which may explain why the cladogram has a tendency to “flip” upside down, with the youngest taxa being forced basally. I am not especially convinced by the cladograms so far.

In our revisions, a few taxa are considered nomen dubia. This includes the recently named Bravoceratops, from the lowermost Javelina Formation of Texas.

The holotype parietal of Bravoceratops was probably reconstructed back-to-font. It is no longer considered diagnostic.

Bravoceratops is still important however, as the narrowness and lack of lateral flanges on the parietal midline bar suggests that Bravoceratops is less derived than Taxon C (Denazin Mbr, Kirtland Fm, ~73.9-73.4 Ma) from our main paper (which possesses moderately developed lateral flanges). This could suggest that the lowermost Javelina Formation is no younger than 73.9 Ma. This is of interest because most people think of the Javelina as a K-Pg boundary unit (i.e. close to 66 Ma), but in fact there is a U-Pb date from the middle of the Javelina which is 69Ma (+/- 0.9Ma), and recently described hadrosaur remains from the lowermost part were ascribed to Kritosaurus sp., a taxon otherwise known only from the late Campanian. Given that the morphology of Bravoceratops also looks similar to middle to late Campanian chasmosaurines, then it would seem that the lowermost Javelina may indeed be late Campanian in age. As such, age of the Javelina may be more complex than typically presented. This might seem of interest only to a stratigraphy wonk like me, but if you like your dinosaurs then it’s worth keeping an eye on because there are some intriguing specimens being found down there!

A cladogenetic split ~83 million years ago
This is one of the most significant new findings. The morphometrics, stratigraphy, and phylogeny support a basal split of chasmosaurines into a Chasmosaurus-lineage and a Pentaceratops-lineage. After this initial cladogenetic (splitting) event, both of the resultant lineages comprise stacks of stratigraphically separated morphospecies, which suggests that they further evolved by anagenesis (linear evolution) rather than cladogenesis (there is probably at least one further cladogenetic event suggested by later occurring fossils, but our paper does not delve into that).

The Chasmosaurus (left) and Pentaceratops (right) lineages split by cladogenesis some time before 77Ma.

The fact that the oldest member of either lineage occurs at ~77 Ma, means that the cladogenetic event must have occurred before this time. We are further informed by the fact that representatives of the Chasmosaurus lineage are more abundant in the north, whereas the Pentaceratops-lineage are more abundant in the south. We can therefore start looking at what was going on geographically in North America before 77Ma that might provide an isolating mechanism for this cladogenetic event; which leads us to…

A mechanism for creating different faunas between northern and southern regions of North America
A few previous researchers had noted differences between northern and southern dinosaur faunas in the upper middle to late Campanian stage (78-72 Ma). Some even went so far as to suggest that there might be different faunas within individual basins. The basin-scale endemism hypothesis has never been well-supported by data (see Fowler, 2017 and references therein). However, at a broader scale, there are some intriguing faunal patterns suggesting that indeed, northern and southern faunas are slightly different, at least proportionally (and in more than just dinosaurs).

Our new analysis suggests that something happened before 77 Ma to separate the southern Chasmosaurus and northern Pentaceratops lineages. Previous workers investigating the biogeographic hypothesis had suggested that no physical barrier existed between north and south. If you consider only the upper half of the Campanian (78-72 Ma), this is probably true. However, our work indicates we need to look earlier than this, and herein lay some interesting interplay of sealevel and orogeny (mountain building).

During the Cretaceous period a great interior seaway had flooded North America from north to south, separating it into eastern (Appalachia) and western (Laramidia) subcontinents. A paper published in 1990 by Lillegraven and Ostresh looked at the position of the western shoreline of this Western Interior Seaway. They had studied the location of marine rocks, and had aligned this with ammonite fossils contained in the same rocks, which were biostratigraphically informative. They thus produced a fabulous serious of mini maps showing the seaway shoreline moving back and forth from the middle Santonian (~85Ma) through to the K-Pg boundary (66 Ma). Lillegraven and Ostresh also had the foresight to draw in the current thrust (uplift) front of the Rocky Mountains onto their maps.

Lillegraven and Ostresh (1990) had somehow been overlooked by previous workers, but it is tremendously detailed and insightful. It shows that during a period of unusually high sea level around 85-81 Ma, the shoreline of the seaway butted up against the young Rocky Mountains around the region of what is now northern Utah. For hundreds of miles between northern Utah and northern Montana, there would have been very little habitat space between the mountain front and the sea, perhaps as little as a few kilometers.

The shoreline of the Western Interior Seaway (bold line) sometimes abutted the thrust front (weak wiggly line) of the incipient Rocky Mountains. Ammonite zones named at bottom; stratigraphic moving arrow part from Fowler (2017). Map adapted from Lillegraven & Ostresh (1990).

Dinosaurs living in Canada would have been cut off from those in southern Utah (and further south).  This is the perfect recipe for vicariance, and is probably what prompted the cladogenesis that created the Pentaceratops and Chasmosaurus lineages. Indeed, I suspect it may have been the cause of speciation events in other dinosaur groups and non-dinosaurian taxa.

The story develops further in the middle Campanian when the seaway recedes a little. This would have made it possible for northern and southern faunas to mix again. This explains why we see a few southern-allied taxa starting to appear in the north at this time, where they begin to gain a foothold.

So, why should you care?
What we are showing here, and a theme running through a few other recent papers (e.g. Freedman Fowler and Horner, 2015; Wilson et al.,2020), is that recognition of (probable) anagenesis in dinosaurs really starts to break down notions of “diversity” in Late Cretaceous North America.

We can get too hung up on whether or not it is possible to “prove” or falsify one mode of evolution or the other, but for me it is much more interesting to consider the possibility that we really do see occasional cladogenesis and lots of anagenesis. I call this “lineage thinking”. Is there a fundamental difference between the processes that drive simple linear anagenesis, compared to a “true” cladogenetic event that splits a continent-wide population into two? I think that there probably is; that there may be specific points in time when we see pulses of speciation, or indeed lineage extinctions, and these might be tied to external factors, some profound, like a meteorite impact, and others rhythmic and mostly passive, like sea level rise and fall.

It is in many ways too easy to suggest a cladogenetic splitting event, without any thought given to the external abiotic factors that might cause such an event. Cladogenesis requires vicariance driven by an isolating mechanism. In our paper, we propose what I think is a very feasible isolating mechanism – high sea level pushing up against the incipient Rocky Mountains. If we are right, then unusual numbers of lineages in north America in the Campanian might really be just an accident of geography; the interplay of sea level and orogeny, and not indicative of an adaptive radiation into new niches (etc). Indeed, it is suggestive that perhaps a lot of the Campanian diversity seen in dinosaurs of Laramidia might simply be an artifact of Laramidia being a long narrow subcontinent subject to the occasional whims of Neptune, rather than saying anything profound about diversity, radiations, or species-richness decline towards the K-Pg boundary.

Another offshoot of lineage thinking is the utility of dinosaurs for biostratigraphy. When I suggested this in a recent paper on correlating North American terrestrial formations, it was scoffed at by a couple of reviewers… but here’s the thing, because of their rapid evolution ceratopsid dinosaurs potentially have a stratigraphic resolution of ~200-300 thousand years (or less). That’s much better than you get with mammals, and worthy of consideration. Indeed, dinosaurs were used for biostratigraphy in the original North American Land Vertebrate Ages (NALVAs), but were left behind when these became narrowed into North American Land Mammal Ages (NALMAs).

There is a viewpoint that dinosaurs are only “fun” fossil taxa – interesting to read about, but not worth a seat at the high table of biology. Dinosaurs might not have as good of a sample size as for other fossil taxa, and however interesting our chosen clade may be, they are not the bellwethers of global trends in the way that, say, plankton or plants are. Still, it would be wrong to write off dinosaurs as uninformative – they have much to tell us about the past in ways that perhaps other taxa cannot; these are after all, the most successful large-bodied land animals ever to have existed. I hope that our new paper shows that some vibrant new ideas and discoveries can be made if we just adapt our way of thinking a bit… and go out and find new specimens!