Monday, October 20, 2014

Taxonomic Messes

In his recent Princeton Field Guide to Dinosaurs, Greg Paul did something interesting: he sank most of the Centrosaurinae into Centrosaurus. Thus, instead of Styracosaurus albertensis, you now have Centrosaurus albertensis. Where once there was Pachyrhinosaurus lakustai, there is now only Centrosaurus lakustai. This seems ridiculous on the face of it, but Paul's justification is not entirely off-kilter: he surmises that the Centrosaurinae exhibits as much or less variation than is found between different species of Varanus. But that comparison is, in itself, completely arbitrary. The taxonomy of living animals--and especially living reptiles--is dictated by an entirely separate group of specialists. If paleontologists were in charge of Varanus, it might well be divided into separate genera. Herpetologists (and Greg Paul) are especially fond of subgenera, whereas as I believe that subgenera are confusing and unnecessary.

What Paul is doing is called taxonomic lumping: in a (misguided) effort to simplify things, he has wedged a number of distinct animals under the same generic banner. It's worth pointing out that all taxonomy is ultimately arbitrary anyway, but the point is that people within the field know what you're talking about. The opposite of lumping is splitting, where several very similar animals are placed within their own genera. You can see this currently in the Triceratopsini, a group currently comprised of a number of animals that barely differ from one another.

But there are extreme variations on both: where you split or lump despite what your own data is telling you. I'll give two recent examples.

Just last month, Gates & Scheetz named Rhinorex cordrupus in the Journal of Systematic Palaeontology. At first glance, Rhinorex looks a whole lot like Gryposaurus, which is may overlap in time with. The authors could only find two characters (subtle features of the premaxilla and nasal) that differentiated Rhinorex from Gryposaurus. As there are already three species of Gryposaurus around, you may wonder whether two measly features justify a new genus. Well, Gates & Scheetz's phylogenetic analyses wondered the same thing. They ran two analyses, and here's what the first one said:

Rhinorex falls right outside of Gryposaurus. Well, okay. This was a majority rule Bayesian analysis, for those keeping score. What's the strict consensus tree say?

Well, now Rhinorex is right there within the Gryposaurus group. It could easily be considered Gryposaurus condrupus. In both analyses, it's hard to make a case for the generic distinctiveness of Rhinorex. But that doesn't stop Gates & Scheetz from doing it anyway. I suspect this taxon will collapse into Gryposaurus in future studies, and the name may be abandoned.

The next example presents a very different situation. In the last few months, Nicholas Longrich published a very interesting description of a chasmosaurine frill fragment called the "Manyberries Chasmosaurine." It had been considered Anchiceratops in the past, but Longrich establishes that it does not conform to that genus. Rather, it appears to represent an extreme northern (and much older) occurrence of Pentaceratops. He also wonders if another specimen, the "Williams Fork" chasmosaurine, may also be Pentaceratops (to be fair, the original describers of the WF chasmosaurine all but diagnosed it as Pentaceratops). To find Pentaceratops in Alberta is interesting for a number of reasons, among them the wrench it throws into the Laramidian provincialism hypothesis. 

But that's not the most flummoxing thing Longrich did in his paper. He ran a phylogenetic analysis with Pentaceratops sternbergi, Utahceratops gettyi, his new species, Pentaceratops aquilonius, and the Williams Fork chasmosaurine. Here's what it came up with:

Wait a second, Nick, there's a typo in your analysis. P. aquilonius is suddenly paraphyletic with respect to Utahceratops. If P. aquilonius is outside the Utahceratops + Pentaceratops group, then you have to name it something else. Alternately, you could sink Utahceratops into Pentaceratops. You've got options! Obviously, upon running this analysis and seeing the result, you did one of those things, right? Well, no:

"Given the overall similarity to Pentaceratops, it seems most appropriate to refer the new animal to Pentaceratops. Referring the animal to Pentaceratops does have the effect of making the genus paraphyletic with respect to Utahceratops--but there is not in fact any requirement that genera must be monophyletic, and this move seems preferable to synonymizing Utahceratops and Pentaceratops."*

No! No it doesn't! You're making a mockery of a basic tenant of taxonomy! We keep genera monophyletic because we're trying to reflect the reality of their relationships. If P. aquilonius is not, in fact, closer to Pentaceratops sternbergi than Utahceratops is, then it's not Pentaceratops. Phylogenetic trees are hypotheses about who is related to who, and how closely. To use another ceratopsid example to illustrate why Longrich's scheme makes no sense, imagine that we just embraced this concept of generic paraphyly and renamed Triceratops horridus as Bison alticornis? That's what Marsh originally called it, after all, and since genera don't have to be monophyletic, there's no reason we can't have a goddamn mammal name wedged in with a bunch of big-headed dinosaurs.

But we don't do that, because it would be stupid, but also we want generic names to mean something. To abandon generic monophyly--even if it's an unwritten rule--would render meaningless the hypotheses our phylogenetic analyses produce. In taxonomy, names aren't just things we use to identify specimens or pretty pictures in books. They encompass a formal diagnosis about what makes X different from Y, and furthermore how X and Y are related. Maybe Longrich was trying to avoid pissing off Sampson et al. in sinking their Beehive State taxon (Utahceratops). Well, okay, but Nicholas Longrich is no stranger to reassigning taxa.

Is this Chasmosaurus russelli or Mojoceratops perifania? Longrich and Maidment & Barrett disagree.

This is the man who, in 2010, sank Chasmosaurus kaiseni as nondiagnostic, then took a bunch of material previously assigned to it (and some from C. russelli, from what I can tell) and named it Mojoceratops perifania. Yet he retained C. russelli as a valid taxon. The whole thing is confusing to me, honestly. When Maidment & Barrett disagreed with the separation of Mojoceratops from C. russelli in 2011, Longrich responded in his 2013 Judiceratops description and reaffirmed the unique status of Mojoceratops

Also in 2010, Longrich looked at OMNH 10165, a chasmosaurine previously described by Thomas Lehman as a particularly impressive specimen of Pentaceratops, and didn't just rename it Titanoceratops ouranos, but he also raised its position with the Chasmosaurinae to be the basalmost member of the "Triceratopsini." So far, I have not seen a formal response to this reclassification--maybe I missed it. 

This Pentaceratops skeletal is from Wikipedia, but clearly taken from Lehman (1993). Nick Longrich says it's Titanoceratops ouranos. I haven't seen a formal response. 

I'm just asking for consistency. You can't defend Mojoceratops as generically distinct from Chasmosaurus but then argue that P. aquilonius should be within Pentaceratops even when your own phylogenetic analysis tells you it's not. I'm not sure what to do with Titanoceratops. He makes some good points about the specimen's distinctness, but I'd be more comfortable giving it new species name within Pentaceratops. I can't say whether the features supposedly shared by Triceratops and Torosaurus are truly unique or variable.

These sorts of taxonomic messes--Rhinorex and Pentaceratops aquilonius--are bad for paleontology. Remember the mess that the Bone Wars got us into? People spent the next century untangling the Gordian Knot of taxa that came out of Marsh & Cope's petty competition to out-name each other. Heck, Nedoceratops is still a problem. Although I lean towards taxonomic lumping, I'm all for splitting where the evidence tells you to split. That is not the case with Rhinorex, but certainly appears to be the case for P. aquilonius (or you could just sink Utahceratops). The jury appears to be out on Mojoceratops, and I could probably go either way. Longrich's 2010 description posits Mojoceratops as a sister taxon to Agujaceratops, but in 2014, he found it to cluster with Chasmosaurus. Given the nature of my Secret Project, I'll be interested to see where it ends up.

That's enough for now. Again, what started out as a short post became overlong. If any of you have insight into the Titanoceratops/Mojoceratops situation, I'd love to hear it in the comments.

*One wonders how this particular taxonomic act got through peer review.

Postscript: I'm learning that Blogger hates dinosaur names.


  1. Genera HAVE to be paraphyletic if populations that are ancestral to diverse clades aren't to remain genera-less.

    1. Exactly (although Utahceratops is not exactly a diverse clade). The system, as it stands, cannot work with all genera being monophyletic. (Just try to do it with stem-humans, for example.)

      The best option would be to abandon binomial nomenclature. Unfortunately we're too wedded to it.

    2. The best solution would be to simply make genera and species into clades, and then have these clades be able to fit into each other. Simply replace the species in definitions with specimens. This does however create the problem of having to re-name a ton of stuff currently specific names, as species can have the same name at that level. This would obviously create a gradual transition. The big downside to any abandonment of binomial nomenclenture is looking back through old papers, thought.

    3. Species and clade are distinct concepts and should be treated as such, despite the fuzziness of the species concept.

      Paraphyly and polyphyly are also distinct concepts--placing a dinosaur within the genus Bison would (hopefully) make the genus polyphyletic (i.e. not being based on a clade at all), rather than paraphyletic (i.e. being a clade minus some members), whereas Longrich et al made Pentaceratops paraphyletic, which means Pentaceratops would still have some taxonomic use (e.g. "all members of the Pentaceratops clade not a member of the Utahceratops clade")

    4. But that's a specific usage. You still end up with the basic idea of taxonomy - sorting the species into groups based on what's related to what - broken.

  2. John, you're right in that I may have misused "paraphyletic" and "polyphyletic."

    Your definition of a Pentaceratops clade apart from Utahceratops goes away if we just sink Utahceratops into Pentaceratops (P. gettyi). Then Longrich's tree diagram still works out. That's what I'm getting at. What he did is needlessly complicating.

    1. I prefer that approach, too, and I've done that on PhyloPic:

      But the general point still holds. It's impossible to make all genera monophyletic. Suppose, for example, that Zuniceratops christopheri is actually an ancestor of Ceratopsidae. It would be impossible to place it into a monophyletic genus unless that genus also included all ceratopsids.

    2. But that's why cladogenesis is the default assumption--we can't know who's directly ancestral to who.

    3. There isn't really a null hypothesis here, although the way cladograms are drawn makes it look that way. If something has no apomorphies relative to a given node and precedes all other known descendants of that node, it's a potential ancestor. We shouldn't assume it is or isn't.

    4. Besides, there are cases where the record is more complete and we can see actual populations shifting over time. (Mesozoic dinosaurs are admittedly not the best case to illustrate this.)

    5. The more I work on my Secret Project, though, the more open I am to anagenesis among certain clades of Laramidian dinosaurs, and I think it was pretty conclusively demonstrated in Triceratops (between T. horridus and T. prorsus, not Torosaurus).

    6. That's cool. Achelousaurus => Pachyrhinosaurus could be an example that involves generic boundaries. If A. horneri is really ancestral to Pachyrhinosaurus, then Achelousaurus can't possibly be monophyletic.