Neural spine bifurcation in sauropods, Part 4: is Suuwassea a juvenile of a known diplodocid?

April 12, 2012

I don’t intend to write a comprehensive treatise on the morphology and phylogeny of Suuwassea. Jerry Harris has already done that, several times over (Harris 2006a, b, c, 2007, Whitlock and Harris 2010). Rather, I want to address the contention of Woodruff and Fowler (2012) that Suuwassea is a juvenile of a known diplodocid, building on the information presented in the first three posts in this series (Part 1, Part 2, Part 3).

In the abstract, Woodruff and Fowler (2012:1) wrote:

On the basis of shallow bifurcation of its cervical and dorsal neural spines, the small diplodocid Suuwassea is more parsimoniously interpreted as an immature specimen of an already recognized diplodocid taxon.

First of all, that’s not what ‘parsimoniously’ means. It’s just not. In a phylogenetic analysis using unweighted characters, there is no such thing as a ‘key’ character — which by the way means that the subtitle of the paper, “a critical phylogenetic character”, is wrong. All characters are equal. Even if the characters were weighted, neural spine bifurcation would have to be weighted pretty darned heavily for it to outweigh all the other characters combined, which is what the sentence quoted above suggests.

Comparisons with known diplodocids

Next problem: if Suuwassea is a juvenile of an already recognized diplodocid, it shouldn’t take long to figure out which one. There aren’t all that many candidates, and we can consider them in turn.  There are loads of characters, especially cranial and appendicular, separating Suuwassea from Apatosaurus, Diplodocus and the rest, and anyone who wants to keep track of all of them is welcome to do so. I care about vertebrae, and I’m prepared to argue that Suuwassea is a distinct taxon based on cervical morphology alone.


Here are the sixth cervical vertebrae of Suuwassea emiliae ANS 21122 (Harris 2006c:Text-fig. 7B) and Diplodocus carnegii CM 84/94 (Hatcher 1901:pl. 3, flipped left-to-right for ease of comparison). They are not to scale–I made the images the same cotyle diameter for ease of comparison.

Elongation first. C6 of S. emilieae has a centrum length of 257 mm, a cotyle diameter of 75 mm, and so an EI of 3.4. C6 of D. carnegii has a centrum length of 442 mm, a cotyle diameter of 99 mm, and an EI of 4.5. So Diplodocus is one third more elongate than Suuwassea. It is true that sauropod cervicals elongate through ontogeny, but the Suuwassea holotype is a decent-sized animal, and would be expected to have attained adult proportions even if it was not fully adult (also, ANS 21122 has more cervical ribs fused than CM 84/94). We know from the juvenile ?Sauroposeidon vertebra YPM 5294 (Wedel et al. 2000:372) that subadult sauropod cervicals attained great elongation: that element is from an animal young enough to have had an unfused neural arch but it has an EI exceeding 5.0.

Then there’s neural spine shape. Yes, it is variable in sauropods, but this is ridiculous. I strongly doubt that any non-pathological Diplodocus cervical anywhere ever has had a neural spine shaped like that of the Suuwassea vertebra.

Also note that the prezygapophyses of the D. carnegii C6 strongly overhang the condyle but are only slightly elevated, whereas those of S. emilieae are right above the condyle but strongly elevated, so that the prezygapophyseal rami might fairly be called pedestals. Such pedestaling of the prezygapophyses is present in some cervicals of Apatosaurus, although perhaps not to the same extreme. Some Apatosaurus cervicals have pretty funky, smokestack-looking neural spines, although–again–not to the same extreme as in Suuwassea. Still, from the mid-centrum on up, S. emiliae looks a bit apatosaur-ish. So let’s try that next.


Here we have C6 of Suuwassea as before, this time with Apatosaurus louisae CM 3018 (Gilmore 1936:pl. 24), again scaled to the same cotyle diameter.

C6 of A. louisae has a centrum length of 440 mm, a cotyle diameter of 150 mm, and an EI of 2.9 (I know it doesn’t look that short, but I’m going off Gilmore’s data, and I trust the measuring tape more than the drafting pen, no matter how skillfully the latter is wielded.)  So this is not a bad match with the value of 3.4 for Suuwassea.

Of course, the glaring problem with suggesting that Suuwassea is a juvenile Apatosaurus is that it has normal-sized cervical ribs, not the insane scythes of doom that hang below the centrum of every post-axial Apatosaurus cervical (see these posts [#1, #2, #3] for some crazy examples, and this post for more pictures and discussion). The giant cervical ribs are present even in very juvenile Apatosaurus cervicals, such as the large collection of juvenile apatosaurs in the BYU collection from Cactus Park (albeit unfused; the immense parapophyses still point the way even if the ribs themselves are missing).

I know, I know, I just said that there is no such thing as a key character. But all of the known species of Apatosaurus have giant cervical ribs, and indeed are often identified in the field as Apatosaurus on that basis alone. I suppose it’s not impossible that Suuwassea is nested within the other Apatosaurus species, based on some bizarre combination of as-yet undiscovered characters and intermediate specimens, and lost the giant cervical ribs along the way, but now we’re into angels dancing on the heads of non-existent pins. If Suuwassea is an apatosaurine but outside the clade of giant-cervical-rib-bearing Apatosaurus, then whether we call it a species of Apatosaurus or a separate genus–say, Suuwassea–is more a matter of taste than anything else.  Note that Lovelace et al. (2008) recovered Suuwassea as an apatosaurine, but not as Apatosaurus.

Lest anyone without access to the paper think I’m cheating by hiding serial variation, here are the other well-preserved cervicals of Suuwassea, to scale:

Suuwassea emilieae cervicals 3, 5, and 6 in left lateral view, from Harris (2006c:Text-figs. 5, 6, and 7)

So, if Suuwassea is a juvenile of a known diplodocid but it’s not Diplodocus or Apatosaurus, what’s left?


Barosaurus lentus AMNH 6341 cervicals 8-16 in left lateral view, from McIntosh (2005:fig. 2.1)

Probably not.



“Amphicoelias brontodiplodocus”

“Amphicoelias brontodiplodocus” cervicals 7-10 in left lateral view, from Galiano and Albersdorfer (2010:fig 10a)

Okay, now I’m just messing with you.

Is Suuwassea even a juvenile?

By now it is probably obvious, even from cervical morphology alone, that if ANS 21122 is a juvenile of anything, it’s a juvenile Suuwassea. But is it in fact a juvenile?

We-ell. The cervical neural arches are all fused, but not all of the cervical ribs are. Jerry did a fine job of describing exactly what was going on at each serial position (Harris 2006c). In C3, the left cervical rib is not attached, and the right one is attached at the parapophysis but not fused. In C5, the ribs are attached, not fused at the parapophyses, and fused at the diapophyses*. In C6, the ribs are fused at both attachment points. C7 lacks the ribs, but their absence appears to be caused by breakage rather than lack of fusion. One fragmentary posterior cervical of uncertain position is missing the diapophyses but has one rib fused at the parapophysis.

* This is cool because it is the first time that I know of that anyone has documented which of the two attachment points fused first within a single cervical rib. I wonder if other sauropods did it the same way?

So based on cervicals alone, we would infer that Suuwassea was not fully mature. However–and this is absolutely crucial for the synonymization hypothesis–the Suuwassea holotype ANS 21122 already has a greater degree of cervical element fusion than Diplodocus carnegii holotype CM 84/94 (which has unfused ribs back to C5) and Apatosaurus CM 555 (which has unfused arches back to C8 and unfused ribs throughout), both of which have attained essentially ‘adult’ morphology. So if Woodruff and Fowler (2012) are correct, the ontogenetic clock has to run forward from CM 555 and CM 84/94, through a Suuwassea-like stage, and then back to normal Apatosaurus or Diplodocus morphology.

But we don’t have to rely on cervicals alone, because ANS 21122 also includes some dorsals and caudals. And the caudals are very interesting in that the neural arches are not fused through most of the series. Harris (2006c:1107):

Of all the caudal vertebrae preserved in ANS 21122, only the distal, ‘whiplash’ caudals are complete. All the remaining vertebrae consist only of vertebral bodies that lack all phylogenetically informative portions of their respective arches. On the proximal and middle caudals, this absence is due to lack of fusion as evidenced by the deeply fluted articular surfaces for the arches on the bodies. In contrast, the arches on the most distal vertebrae that retain them are seamlessly fused, but everything dorsal to the bases of the corporozygapophyseal laminae are broken.

Now this is pretty darned interesting, because it shows that neural arch fusion in Suuwassea was not a simple zipper that ran from back to front (as in crocs [Brochu 1996] and phytosaurs [Irmis 2007*]) or front to back. We can’t really say, based on this one specimen, what the sequence was, but we can say for certain that the anterior and middle caudals came last. Oh, and for what it’s worth, the scap-coracoid joint is also unfused (Harris 2007), but we know that that’s often the case for substantially “adult” sauropods such as the mounted Berlin Giraffatitan.

* Relevant to this entire post series are the wise words of my homeboy and former Padian labmate Randy Irmis, who wrote in the abstract of his 2007 neurocentral fusion paper:

A preliminary survey indicates that there is considerable variation of both the sequence and timing of neurocentral suture closure in other archosaur clades. Therefore, it is unwise to apply a priori the crocodylian pattern to other archosaur groups to determine ontogenetic stage. Currently, apart from histological data, there are few if any reliable independent criteria for determining ontogenetic stage. I propose that histology be integrated with independent ontogenetic criteria (such as neurocentral suture closure) and morphometric data to provide a better understanding of archosaur ontogeny.

The unfused arches in the Suuwassea caudals are especially interesting because, for the first time that I know of, we have a sauropod with cervical neural arches and at least some cervical ribs fused, but with unfused neural arches elsewhere in the body. This is in contrast to D. carnegii CM 84/94, in which all the neural arches are fused but the anterior cervical ribs are not. So the developmental timing in Suuwassea is dramatically different than in D. carnegii, at least, which is one more problem for the synonymization hypothesis. Two more problems, actually, in that (1) Suuwassea probably isn’t Diplodocus, and (2) it doesn’t belong in the same ontogenetic series as Diplodocus, contra Woodruff and Fowler (2012:Figs. 3 and 9)–if the timing of the various fusions differs between the taxa, there is no basis for assuming that the hypothetical ontogenetic bifurcation would follow the same rules.

And speaking of ontogenetic bifurcation, a final point about the ‘bifurcations’ in Suuwassea.

Woodruff and Fowler (2012:Fig. 9)

The first line of the caption is misleading. Two of these vertebrae have weakly bifurcated neural spines because they are sixth cervicals (Suuwassea in B, Apatosaurus in D), and that’s what you expect in C6 in adult diplodocids. One of them, the C5 of Suuwassea in C, isn’t bifurcated at all: it’s broken. Harris (2006c:1099):

The spinous process expands mediolaterally toward its apex, attaining maximal width just proximal to its terminus. A long, narrow crack at the distal end gives the appearance of bifurcation, but the collinear dorsal margin indicates that no true split was present.

As for the final vertebra, MOR 592 in A, who knows? Woodruff and Fowler (2012) do not say what serial position it is from. Based on the shallow notch in the spine, I’ll bet it’s either a C6 or very close to it–and if so, no deeper split is expected.

So the entire rationale for the taxonomic side of Woodruff and Fowler (2012)–that Suuwassea has incompletely bifurcated neural spines because it is a juvenile –turns out be an illusion caused by not taking serial variation into account. Suuwassea ANS 21122 probably is a subadult, based on the unfused caudal neural arches, but its cervical vertebrae already show the expected adult morphology in neural arch fusion, cervical rib fusion, and–most importantly–neural spine bifurcation.


The evidence that Suuwassea is not a juvenile of a known diplodocid is not in this post. It’s in the hard work, comprehensive descriptions, and detailed, thoughtful comparisons by Harris and Dodson (2004), Harris (2006a, b, c, 2007), Lovelace et al. (2008), Whitlock and Harris (2010), and Whitlock (2011). This post is just an arrow scratched in the dirt. Please, go read those papers. And then read all the monographs I cited in the first post in this series (and am too lazy to cite again here). Give those people their due by taking their work seriously and learning from it.

The rest of the series

Links to all of the posts in this series:

and the post that started it all:


19 Responses to “Neural spine bifurcation in sauropods, Part 4: is Suuwassea a juvenile of a known diplodocid?”

  1. ech Says:

    Layman here: “All characters are equal”—Why is this? It *seems* obvious that some characteristics would be more likely than others to evolve convergently, particularly if they aren’t very distinctive or complex or something, so it would seem that those characters would weigh less in determining the shape of the tree.

  2. Because you cannot assume that one character outweighs another a priori. Doing so introduces an excludable bias which is not then noticed (or further ignored). By presuming a priori that all characters are equal, representation of a character in a matrix is presumed to be of even quality relative to another. When we weight characters, we do so primarily to test a previous handled result, not to presume the value of a feature, in order to assess homoplasy, convergence and parallelism in our datasets.

  3. Michael Richmond Says:

    Another layman here: so, does that mean that by choosing the characters in a particular manner, I can cause my analysis to yield whatever result I wish? To be specific, I could choose characters as silly as “length”, “color”, “number of broken-off-bits”, and so forth, and weight those equally with characters such as “number of teeth in lower jaw”, “fused sacral vertebrae”, etc. — right?

  4. Mike Taylor Says:

    Yes, Michael, it does mean that. For example, I could run a phylogenetic analysis containing a croc, a lion, a tiger and a zebra with only one character, “has stripes”, and get the result that zebras are the closest relations of tigers. That’s why character selection for phylogenetic analysis is a science in itself — reams have been written about it.

  5. David Marjanović Says:

    not a simple zipper that ran from back to front (as in crocs [Brochu 1996] and phytosaurs [Irmis 2007*])

    There’s a good reason why crocs and phytosaurs have this zipper pattern: they use(d) their tails for locomotion, so it makes sense to stabilize them first.

    That’s also the pattern of ossification of the vertebrae of Eusthenopteron, presumably for the exact same reason.

    It *seems* obvious that some characteristics would be more likely than others to evolve convergently, particularly if they aren’t very distinctive or complex or something, so it would seem that those characters would weigh less in determining the shape of the tree.

    What Jaime said. Yes, it’s highly likely that not all morphological characters are equally prone to homoplasy – but, in stark contrast to molecular characters, we have just about no clue about which characters are how much more prone to homoplasy than which others.

    There’s always reweighting: you take a tree made with equally weighted characters, weight each character according to its consistency index ( = how often it undergoes homoplasy according to that tree), and run the analysis anew. That’s sometimes done, and it’s implemented in PAUP*, but it’s a bit circular, isn’t it? If your matrix is bad, reweighting is unlikely to save it.

    Equal weighting is a sort of helpless default assumption: throw all the characters in and let PAUP* or TNT sort them out. As you add characters, the signal adds up, and the noise cancels itself out…

    so, does that mean that by choosing the characters in a particular manner, I can cause my analysis to yield whatever result I wish?

    Yes. But you’re a scientist, so you won’t do that. :-) Also, peer review.

    (…except… …except that extremely few reviewers or editors ever read supplementary information, appendices or even tables. I once submitted a manuscript about this problem to a journal; Table 1 was the coding of one taxon in a data matrix, simply lacked the first 45 characters or so because I had forgotten to take care of a Word glitch, and neither the reviewers nor the editor noticed. That was the first time that I replied to a rejection notice, harr harr.)

    The characters in a data matrix for phylogenetic analysis* need to contain phylogenetic signal (they mustn’t be constant throughout the matrix or all of it minus one taxon, and they mustn’t vary too chaotically), and they mustn’t be correlated** (two correlated characters are the same as one character with double weight).

    * Not necessarily for other analyses.
    ** Means: if a taxon has a particular state of one character, it automatically also has a particular state of another character. Characters that predict each other, even partially, are correlated.

  6. David Marjanović Says:

    one character, “has stripes” […] character selection for phylogenetic analysis is a science in itself —

    Quite. One common mistake phylogeneticists make is to define only one state of a character. I’m currently working on a published matrix that is full of “state that takes two or three lines to precisely define: absent (0); present (1)”, where state 0 lumps a wide range of distinguishable states and thus hides phylogenetic signal.

    Indeed, stripes in crocs aren’t unheard of, but they’re less conspicuous than in tigers or zebras and AFAIK limited to part of the tail…

  7. Mike Taylor Says:

    “The characters in a data matrix for phylogenetic analysis […] mustn’t be correlated”.

    Really? Even if it’s correlation without causation? What if I code up my matrix and it just happens to fall out that “Shelf-like area or fossa (narial fossa) on premaxilla and maxilla lateral to external nares” is scored exactly the same across my taxa as “Collateral ligament foveae on non-ungual pedal phalanges” — should I get rid of one of the characters? How should I choose which one gets the bullet?

    Surely correlated characters are only a sign of something fishy in that the correlation can clue you in that there’s a previously unnoticed causation? If my looking at correlations in my matrix I notice that I have coded both “femur longer than tibia” and “tibia shorter than femur” then of course and I can and must delete one.

  8. Tor Bertin Says:

    Apologies if my lack of knowledge of applied statistics in phylogenetic analyses is blinding me to an obvious answer, but can you assess the importance of individual characters in structuring phylogenetic trees by removing individual characters and examining the resulting consensus tree/trees?

    Specifically, I’m thinking of a phylogenetic equivalent of elasticity analyses used in wildlife population demographic analysis, where individual parameters with an observed mean and variance are each altered in a step-wise fashion by a given percentage, and the resulting population dynamics examined. Elasticity analyses allow you to see which parameters have the strongest effect on the resultant population projections.

    Is there a way to do something similar when constructing phylogenetic trees?

  9. Mike Taylor Says:

    Yes, something like what you describe could certainly be done. (Bootstrap analysis, which is pretty common, works by doing something similar.) But all that would tell you is which characters happen to draw you most strongly towards the most parsimonious tree in that analysis. It doesn’t tell you anything about the intrinsic importance of that character. So you may find that size of deltoid crest is very significant in your sauropod analysis, but in a theropod analysis it might not be.

  10. […] Last time around, Matt walked through a lot of the detailed cervical morphology of Suuwassea and known diplodocids to show that, contra the suggestion of Woodruff and Fowler (2012), Suuwassea is distinct and can’t be explained away as an ontogenomorph of a previously known genus. […]

  11. […] recognized diplodocid taxon”.  Without getting into the subject of Suuwassea again — Matt pretty much wrapped that up in part 4 — the point here is that the word “parsimony” has a particular meaning in studies […]

  12. […] Neural spine bifurcation in sauropods, Part 4: is Suuwassea a juvenile of a known diplodocid? &laquo… Says: April 12, 2012 at 7:45 am […]

  13. Mickey Mortimer Says:

    “Really? Even if it’s correlation without causation?”

    Well no. The problem is that when phylogeneticists say two characters are correlated, they really mean they’re causally correlated. So the ‘correlation is not causation’ saying really doesn’t work in our terminology. ;)

  14. […] of the paper arising from our response to Woodruff and Fowler (2012) [part 1, part 2, part 3, part 4, part 5, part 6].  Here is an oddity. Sacra of Haplocanthosaurus. Top, H. utterbacki holotype CM […]

  15. […] a bit of a fashion these days for drive-by synonymisation of dinosaurs, and sure enough no sooner had Brian Switek written about Kaatedocus for his new […]

  16. I have no idea if anybody’s still reading this thread, but it seems relevant again with Woodruff’s new paper on functional anatomy of the nuchal ligament (D. Cary Woodruff (2016): Nuchal ligament reconstructions in diplodocid sauropods support horizontal neck feeding postures, Historical Biology, DOI: 10.1080/08912963.2016.1158257). It refers to Suuwassea as a juvenile of Apatosaurus with no further explanation than a citation to Woodruff and Fowler 2012. It also claims to have “re-identified” bifurcation in several sauropods, including Tazoudasaurus (actually broken), Trigonosaurus (looks broken to me from the figure), Bonitasaura (a tiny midline depression on D3), and possibly Alamosaurus (without a figure, reference, or any explanation), so I’m skeptical of that too.

  17. […] all know that apatosaurines have big honkin’ cervical ribs (well, most of us know that). But did they also have unusually large neural […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: