Brief thoughts on qualitative and quantitative projects
December 10, 2023
Back in the first post about our recent paper on bifurcated cervical ribs in apatosaurines, I noted:
I’m fond of this one because it’s pleasingly low-tech and traditional. We looked at some fossils, noticed some interesting features, thought about what they mean, wrote it up, illustrated it with specimen photos and diagrams, and called it done. There is certainly a time and place for phylogenetic analysis, geometric morphometrics, and all the other numerical methods that are increasingly common in vertebrate palaeontology, but I genuinely think it’s important that this kind of work doesn’t squeeze out the more foundational process of looking at, and thinking about, fossils.
This has been much on my mind of late, especially as the majority of talks at SVPCA 2023 and many new papers involve numerical methods. Sometimes I feel that Matt and I are in danger of being left behind by a new wave of palaeontology, and it’s definitely true that we could usefully apply (say) geometric morphometrics to our specimens if we had the time to learn how it’s done.
And yet, and yet …
Today I came across a Mastodon thread summarising a preprint (Ploner and Stafford 2021), “How analysis strategy affects analysis results”. Stafford’s summary says, in part:
A host of […] projects have confirmed the worrying conclusion that you can have defensible analyses which produce differing results. But Sebastian and I wanted to pursue further the issue of exactly how wide the spread of results is.
Sebastian computationally generated 1000s of different possible analysis models — permuting possible covariates and interactions — to get a size of the space of possible results. The question was this: how do human teams fill space? Does expertise in analysis mean outcomes cover a restricted, perhaps tiny, zone of the possible outcomes?
The multiverse of computationally generated analyses covered a smaller range than the spread discovered by human teams! Whatever human analysis teams were doing — choosing different possible statistical frameworks, model forms, data recoding, outlier exclusion, etc — it produced more widely varying output than randomly combining covariates in a single model form (mixed model logistic regression, since you asked)
And the punchline:
This result suggests […] that there is a hidden universe of data analysis choices which can both be a) legitimate and b) poorly recorded or recognised by researchers
The important part of this, to me, is Stafford’s in-passing point that the 29 teams whose results differed so widely all made legitimate and defensible data-analysis choices. They all did good work. But they all did different good work with the same dataset, and the outcome was that they all got different results.
When I showed this thread to Matt, his comment was:
That is interesting. And a bit worrying, since a lot of the “big science” to come in future decades will be big analyses of big datasets.
I’m glad to be poking around weird anatomy instead.
I think there is wisdom in that. I’m nervous about the idea that if we did (for example) apply geometric morphometrics to a set of cervical ribs, we might get significantly different results depending on what landmarks we chose, or on other factors.
Whereas the kind of largely descriptive work we did in the recent paper has a different quality. What we wrote has no computer-generated veneer of objectivity: it’s just what we saw and what we thought about it. Our interpretations could be wrong, but that’s fine: they’re written down so they’re refutable. Heck, even our descriptions could be wrong — we might have misinterpreted structures. But that’s OK, too: other people can look at the fossils, reach their own conclusions, and argue their case about why what we wrote is wrong.
But you can’t really argue against the results of a finite element analysis or what have you. All you can do is run another finite element analysis, get different results from the team that did the first one, and say “huh, the computer spat out a different result this this”.
I would find that unsatisfying.
Again, please note: I am not saying that numerical method are without value! I’m not even necessarly saying they are less valuable than we assume (though I do think we should treat the outputs of any given numerical analysis with a bit more scepticism). I’m just saying I’m glad I don’t have to do much of that kind of work.
References
- Ploner, Sebastian, and Tom Stafford. How analysis strategy affects analysis results: assessing results space and structure of Silberzahn et el. (2018) through model specification. PsyArXiv, 8 Dec. 2023. doi: 10.31234/osf.io/b2hm7
- Silberzahn, R., E. L. Uhlmann, D. P. Martin, P. Anselmi, F. Aust, E. Awtrey, Š. Bahník, F. Bai, C. Bannard, E. Bonnier, R. Carlsson, F. Cheung, G. Christensen, R. Clay, M. A. Craig, A. Dalla Rosa, L. Dam, M. H. Evans, I. Flores Cervantes, N. Fong, M. Gamez-Djokic, A. Glenz, S. Gordon-McKeon, T. J. Heaton, K. Hederos, M. Heene, A. J. Hofelich Mohr, F. Högden, K. Hui, M. Johannesson, J. Kalodimos, E. Kaszubowski, D. M. Kennedy, R. Lei, T. A. Lindsay, S. Liverani, C. R. Madan, D. Molden, E. Molleman, R. D. Morey, L. B. Mulder, B. R. Nijstad, N. G. Pope, B. Pope, J. M. Prenoveau, F. Rink, E. Robusto, H. Roderique, A. Sandberg, E. Schlüter, F. D. Schönbrodt, M. F. Sherman, S. A. Sommer, K. Sotak, S. Spain, C. Spörlein, T. Stafford, L. Stefanutti, S. Tauber, J. Ullrich, M. Vianello, E.-J. Wagenmakers, M. Witkowiak, S. Yoon, and B. A. Nosek. 2018. Many analysts, one data set: Making transparent how variations in analytic choices affect results. Advances in Methods and Practices in Psychological Science 1(3):337–356. doi: 10.1177/2515245917747646
Last night a thought occurred to me, and I wrote to Matt:
If birds had gone extinct 66 Mya along with all the other dinosaurs, would it ever have occurred to us that they had flow-through lungs? Is there — can there be, outside of amazing soft-tissue preservation — any way for bone fossils to tell us about this?
(Yes, we have evidence for air-sacs in the pneumatization of vertebrae and other bones, but I doubt that would have led us to the idea of the flow-through lung. I’m not even convinced it would have led us to the idea of air-sacs, if we didn’t have extant birds as a model.)
Matt wrote back and gave me permission to write up his reply into an SV-POW! post, which you are now, obviously, reading. Here’s what he said.
No, we’d have no idea about the flow-through lungs from fossils.
In fact, it’s particularly bad for birds. Big saurischian dinosaurs had lots of postcranial skeletal pneumaticity (PSP), and some extant birds have a lot of PSP, but most Mesozoic birds have limited to zero diagnostic PSP. A few have some external foramina on the vertebrae that might be pneumatic, but might just be lateral foramina for the equatorial arteries. It doesn’t help that most Mesozoic birds are smashed flat and often have other elements overlapping the vertebrae — most often the proximal portions of their own ribs.
So ironically, even if we somehow came up with the stacked notions that (1) PSP implied air sacs, and (2) air sacs implied flow-through lungs, we’d be much more likely to infer flow-through lungs in Diplodocus and Tyrannosaurus than in Archaeopteryx or most other Mesozoic birds.
But wait, it gets worse! The work by Colleen Farmer, Emma Schachner, and colleagues that demonstrated unidirectional flow in the lungs of crocs, monitor lizards, and iguanas would presumably still get done, but those animals have flow-through lungs without PSP and without particularly elevated metabolisms (although monitors are trying hard). Without the example of birds showing us how that primitive flow-through system can be further refined and supercharged to power tachymetabolism, we’d still learn of flow-through lungs, but we’d have no reason to connect them to PSP or any particular metabolic strategy.
I’ve probably mentioned this before, but it really irks me that we assume that birds are the pinnacle of lung evolution. Why? Birds survived the K/Pg extinction because they were small and could hide and eat seeds and grubs for a while, not because they had better lungs than everything else (otherwise mammals, lizards, etc. would have done even worse). To me it would be a heck of a coincidence if the one group of ornithodirans that survived — for reasons unrelated to lung function — just happened to have the most efficient lungs. It’s always been tantalizing to me that extant birds start out with 12 embryonic air sacs, which through development usually merge into the usual 9 (unpaired clavicular, and paired cervical, anterior thoracic, posterior thoracic, and abdominal sacs). This seems like an embryonic footprint of a greater diversity — and possibly even a greater complexity — of respiratory anatomy in the ancestral ornithodiran, saurischian, or theropod (or all of the above).
Matt and I are writing a paper about Barosaurus cervicals (yes, again). Regular readers will recall that the best Barosaurus cervical material we have ever seen was in a prep lab for Western Paleo Labs. We have some pretty good photos, such as this one:
Barosaurus cervical vertebra lying on its right side in anterodorsal view (i.e. with dorsal to the left), showing the distinctive shape of the prezygapophyseal rami.
The problem is that this specimen was privately owned at the time we saw it, and so far as we know it still is. So according to all standard procedures, we should consider it unavailable to science until such time as it is deposited in an accredited museum. (I was pretty sure the SVP has an explicit policy to this effect, but I couldn’t find it on the site. Can anyone?)
So what should we do? All the possible courses of action seem unfortunate.
1. We could go ahead and include photos, drawings and descriptions of these vertebrae in the paper — but that would violate community norms by building an argument on observations that cannot be in general be replicated by other researchers. (For all we know, these vertebrae are now decorating Nicolas Cage’s pool room.)
2. We could omit these vertebrae from the paper, but use the information we gained from examining them in formulating our diagnostic criteria for Barosaurus cervicals — but this would also not really be replicable, plus it would have that horrible “we know something that you don’t” quality.
3. We could act as though these vertebrae do not exist, or as though we had never seen them, writing the paper based only on our observations of inferior material and of the good AMNH 6341 that is not accessible for study or photography — but that would make our characterisation of Barosaurus cervical morphology less helpful than it could be.
4. We could refrain from publishing on Barosaurus cervicals at all until such time as these vertebrae, or similarly well-preserved ones, are available to study at accredited institutions — but that would simply deprive the world of an interesting and exciting study.
Is there a fifth path that we have not seen? And if not, which of these four is the least objectionable?
Pneumatization sites: how does air get into vertebrae?
December 8, 2021
Science doesn’t always get done in the right order.
In the course of the research for my paper with Mike this past spring, “Why is vertebral pneumaticity in sauropod dinosaur so variable?”, published in Qeios in January, I had a couple of epiphanies. The first was that I had collated enough information to map the sites at which arteries and veins enter and exit the vertebrae in most tetrapods. The second was that, having done that, I’d also made a map of (almost) all the places that diverticula enter the vertebrae to pneumatize them. This is obviously related to the thesis we laid out in that paper, that postcranial skeletal pneumaticity is so variable because pneumatic diverticula follow pre-existing blood vessels as they develop, and blood vessels themselves are notoriously variable. In fact, if you had to summarize that thesis in one diagram, it would probably look like the one above, which I drew by hand in my research notebook in early March.
Only that’s not quite correct. I didn’t have those epiphanies “in the course of the research”, I had them after the pneumatic variation paper was done and published. And at the time they felt less like epiphanies and more like a series of “Holy crap” realizations:
- Holy crap, that diagram would have been really helpful when we were writing the pneumatic variation paper, since it establishes, almost tautologically, that diverticula invade vertebrae where blood vessels already have. In a rational world, Mike and I would have done this project first, and the pneumatic variation paper would have stood on its shoulders.
- Holy crap, how have I been working on vertebral pneumaticity for more than two decades without ever creating a map of all the places a vertebra can be pneumatized, or even realizing that such a map would be useful?
- Holy crap, how have I been working on dinosaur bones — and specifically their associated soft tissues — for more than two decades without wondering exactly how the blood was getting into and out of each bone?
Arguably, not only should Mike and I have done this project first, I should have taken a stab at it way back when I was working on my Master’s thesis. Better late than never, I guess.
I used a sauropod caudal as my vertebral archetype because it has all the bits a tetrapod vertebra can have, including the hemal arch or chevron. This was important, because Zurriaguz et al. (2017) demonstrated that the chevrons are pneumatic in some titanosaurs.
For the actual presentation I redrew the vessels on top of a scan of a Camarasaurus caudal from Marsh, which Mike found and cleaned up (modified from Marsh 1896: plate 34, part 4, and plate 39, part 3c).
We deliberately used an unfused caudal to emphasize that ‘ribs’ — technically, costal elements — are present, they just fuse to the neural arch and centrum rather than remaining separate, mobile elements like dorsal ribs.
Anyway, I’m yapping about this now because this project is rolling: Mike and I submitted an abstract on it for the 3rd Palaeontological Virtual Congress, and a short slideshow on the project is now up at the 3PVC site for attendees to look at and comment on. The congress started last Wednesday and runs through Dec. 15, after which I’m sure we’ll submit the abstract and slide deck somewhere as a preprint, and then turn it into a paper as quickly as possible.
I’ll probably have more to say on this in a day or so, but for now the comment field is open, and your thoughts are welcome.
References
- Marsh, O.C. 1896. The Dinosaurs of North America. 16th annual report of the U. S. Geological Survey, 1894-95, pt. I. US Government Printing Office, Washington, D.C.
- Taylor, Michael P., and Mathew J. Wedel. 2021. Why is vertebral pneumaticity in sauropod dinosaurs so variable? Qeios 1G6J3Q. doi:10.32388/1G6J3Q
- Zurriaguz, V., Martinelli, A., Rougier, G.W. and Ezcurra, M.D. 2017. A saltasaurine titanosaur (Sauropoda: Titanosauriformes) from the Angostura Colorada Formation (upper Campanian, Cretaceous) of northwestern Patagonia, Argentina. Cretaceous Research 75: 101-114.
In mammals — certainly the most-studied vertebrates — regional differentiation of the vertebral column is distinct and easy to spot. But things aren’t so simple with sauropods. We all know that the neck of any tetrapod is made up of cervical vertebrae, and that the trunk is made up of dorsal vertebrae (subdivided into thoracic and lumbar vertebrae in the case of mammals). But how do we tell whether a given verebra is a posterior cervical or an anterior dorsal?
Here two vertabrae: a dorsal vertebra (D3) and a cervical vertebra (C13) from CM 84, the holotype of Diplodocus carnegii, modified from Hatcher (1901: plates III and VII):
It’s easy to tell these apart, even when as here we have only lateral-view images: the dorsal vertebra is tall, its centrum is short, its neural spine is anteroposteriorly compressed and its parapophysis is up on the dorsal half of the centrum; but the cervical vertebra is relatively low, its centrum is elongated, its neural spine is roughly triangular and its parapophysis hangs down well below the centrum (and has a cervical rib fused to it and the diapophysis).
But things get trickier in the shoulder region because, in sauropods at least, the transition through the last few cervicals to the first few dorsals is gradual — the vertebrae become shorter, taller and broader — and tends to have no very obvious break point. In this respect, they differ from mammals, in which the regional differentiation of the spinal column is more abrupt. (Although even here, things may not be as simple as generally assumed: for example, Gunji and Endo (2016) argued that the 1st thoracic vertebra of the giraffe behaves functionally like an 8th cervical.)
So here are those two vertebrae in context: the sequence D3 D2 D1 C15 C14 C13 in CM 84, the holotype of Diplodocus carnegii, modified from Hatcher (1901: plates III and VII):
Given that the leftmost is obviously a dorsal and the rightmost obviously a cervical, where would you place the break-point?
The most usual definition seems to be that the first dorsal vertebra is the first one that has a free rib, i.e. one not fused to the vertebra: in the illustration above, you can see that the three cervicals on the right all have their cervical ribs fused to their diapophyses and parapophyses, and the three dorsals on the left do not. This definition of the cervical/dorsal distinction seems to be widely assumed, but it is rarely explicitly asserted. (Does anyone know of a paper that lays it out for sauropods, or for dinosaurs more generally?)
But wait!
Hatcher (1903:8) — the same dude — in his Haplocanthosaurus monograph, writes:
The First Dorsal (Plate I., Fig. 1). […] That the vertebra now under consideration was a dorsal is conclusively shown not by the presence of tubercular and capitular rib facets showing that it supported on either side a free rib, for there are in our collections of sauropods, skeletons of other dinosaurs fully adult but, with the posterior cervical, bearing free cervical ribs articulating by both tubercular and capitular facets as do the ribs of the dorsal region. The character in this vertebra distinguishing it as a dorsal is the broadly expanded external border of the anterior branch of the horizontal lamina [i.e. what we would now call the centroprezygapophyseal lamina]. This element has been this modified in this and the succeeding dorsal, no doubt, as is known to be the case in Diplodocus to give greater surface for the attachment of the powerful muscles necessary for the support of the scapula.
Hatcher’s illustrations show this feature, though they don’t make it particularly obvious: here are the last two cervicals and the first dorsal, modified from Hatcher (1903:plate I), with the facet in question highlighted in pink: right lateral view at the top, then anterior, and finally posterior view at the bottom. (The facet is only visible in lateral and anterior views):
Taken at face value, Hatcher’s words here seem to imply that he considers the torso to begin where the scapula first lies alongside the vertebral column. Yet if you go back to the Diplodocus transition earlier in this post, a similar scapular facet is not apparent in the vertebra that he designated D1, and seems to be present only in D2.
Is this scapular-orientation based definition a widespread usage? Can anyone point me to other papers that use it?
Wilson (2002:226) mentions a genetic definition of the cervical/dorsal distinction
Vertebral segment identity may be controlled by a single Hox gene. The cervicodorsal transition in many tetrapods, for instance, appears to be defined by the expression boundary of the Hoxc-6 gene.
But this of course is no use in the case of extinct animals such as sauropods.
So what’s going on here? In 1964, United States Supreme Court Justice Potter Stewart, in describing his threshold test for obscenity, famously said “I shall not today attempt further to define the kinds of material I understand to be embraced within that shorthand description, and perhaps I could never succeed in intelligibly doing so. But I know it when I see it.” Is that all we have for the definition of what makes a vertebra cervicals as opposed to dorsal? We know it when we see it?
Help me out, folks! What should the test for cervical-vs-dorsal be?
References
- Gunji, Mego, and Hideki Endo. 2016. Functional cervicothoracic boundary modified by anatomical shifts in the neck of giraffes. Royal Society Open Science 3:150604. doi:10.1098/rsos.150604
- Hatcher, Jonathan B. 1901. Diplodocus (Marsh): its osteology, taxonomy and probable habits, with a restoration of the skeleton. Memoirs of the Carnegie Museum 1:1-63 and plates I-XIII.
- Hatcher, J. B. 1903b. Osteology of Haplocanthosaurus with description of a new species, and remarks on the probable habits of the Sauropoda and the age and origin of the Atlantosaurus beds; additional remarks on Diplodocus. Memoirs of the Carnegie Museum 2:1-75 and plates I-VI.
- Wilson, Jeffrey A. 2002. Sauropod dinosaur phylogeny: critique and cladistic analysis. Zoological Journal of the Linnean Society 136:217-276.
Amazing things are out there waiting to be noticed
March 22, 2021
It is said that, some time around 1590 AD, Galileo Galilei dropped two spheres of different masses from the Leaning Tower of Pisa[1], thereby demonstrating that they fell at the same rate. This was a big deal because it contradicted Aristotle’s theory of gravity, in which objects are supposed to fall at a speed proportional to their mass.
Aristotle lived from 384–322 BC, which means his observably incorrect theory had been scientific orthodoxy for more than 1,900 years before being overturned[2].
How did this happen? For nearly two millennia, every scientist had it in his power to hold a little stone in one hand and a rock in the other, drop them both, and see with his own eyes that they fell at the same speed. Aristotle’s theory was obviously wrong, yet that obviously wrong theory remained orthodox for eighty generations.
My take is that it happened because people — even scientists — have a strong tendency to trust respected predecessors, and not even to look to see whether their observations and theories are correct. I am guessing that in that 1,900 years, plenty of scientists did indeed do the stone-and-rock experiment, but discounted their own observations because they had too much respect for Aristotle.
But even truly great scientists can be wrong.
Now, here is the same story, told on a much much smaller scale.
Well into the 2010s, it was well known that in sauropods, caudal vertebrae past the first handful are pneumatized only in diplodocines and in saltasaurine titanosaurs. As a bright young sauropod researcher, for example, I knew this from the codings in important and respected phylogenetic analysis such as those of Wilson (2002) and Upchurch et al. (2004).
Until the day I visited the Museum für Naturkunde Berlin and actually, you know, looked at the big mounted Giraffatitan skeleton in the atrium. And this is what I saw:
That’s caudal vertebrae 24–26 in left lateral view, and you could not wish to see a nicer, clearer pneumatic feature than the double foramen in caudal 25.
That observation led directly to Matt’s and my 2013 paper on caudal pneumaticity in Giraffatitan and Apatosaurus (Wedel and Taylor 2013) and clued us into how much more common pneumatic hiatuses are then we’d realised. It also birthed the notion of “cryptic diverticula” — those whose traces are not directly recorded in the fossils, but whose presence can be inferred by traces on other vertebrae. And that led to our most recent paper on pneumatic variation in sauropods (Taylor and Wedel 2021) — from which you might recognise the photo above, since a cleaned-up version of it appears there as Figure 5.
The moral
Just because “everyone knows” something is true, it doesn’t necessarily mean that it actually is true. Verify. Use your own eyes. Even Aristotle can be wrong about gravity. Even Jeff Wilson and Paul Upchurch can be wrong about caudal pneumaticity in non-diplodocines. That shouldn’t in any way undermine the rightly excellent reputations they have built. But we sometimes need to look past reputations, however well earned, to see what’s right in front of us.
Go and look at fossils. Does what you see contradict what “everyone knows”? Good! You’ve discovered something!
References
- Taylor, Michael P., and Mathew J. Wedel. 2021. Why is vertebral pneumaticity in sauropod dinosaurs so variable? (version 5) Qeios 1G6J3Q.5. doi:10.32388/1G6J3Q.5
- Upchurch, Paul, Paul M. Barrett and Peter Dodson. 2004. Sauropoda. pp. 259–322 in D. B. Weishampel, P. Dodson and H. Osmólska (eds.), The Dinosauria, 2nd edition. University of California Press, Berkeley and Los Angeles. 861 pp.
- Wedel, Mathew J., and Michael P. Taylor 2013. Caudal pneumaticity and pneumatic hiatuses in the sauropod dinosaurs Giraffatitan and Apatosaurus. PLOS ONE 8(10):e78213. 14 pages. doi: 10.1371/journal.pone.0078213
- Wilson, Jeffrey A. 2002. Sauropod dinosaur phylogeny: critique and cladistic analysis. Zoological Journal of the Linnean Society 136:217–276.
Notes
1. There is some skepticism about whether Galileo’s experiment really took place, or was merely a thought experiment. But since the experiment was described by Galileo’s pupil Vincenzo Viviani in a biography written in 1654, I am inclined to trust the contemporary account ahead of the unfounded scepticism of moderns. Also, Viviani’s wording, translated as “Galileo showed this by repeated experiments made from the height of the Leaning Tower of Pisa in the presence of other professors and all the students” reads like a documentary account rather than a romanticization. And a thought experiment, with no observable result, would not have demonstrated anything.
2. Earlier experiments had similarly shown that Aristotle’s gravitational theory was wrong, including in the works of John Philoponus in the sixth century — but Aristotle’s orthodoxy nevertheless survived until Galileo.
We’ve noted many times over the years how inconsistent pneumatic features are in sauropod vertebra. Fossae and formamina vary between individuals of the same species, and along the spinal column, and even between the sides of individual vertebrae. Here’s an example that we touched on in Wedel and Taylor (2013), but which is seen in all its glory here:
Taylor and Wedel (2021: Figure 5). Giraffatitan brancai tail MB.R.5000, part of the mounted skeleton at the Museum für Naturkunde Berlin. Caudal vertebrae 24–26 in left lateral view. While caudal 26 has no pneumatic features, caudal 25 has two distinct pneumatic fossae, likely excavated around two distinct vascular foramina carrying an artery and a vein. Caudal 24 is more shallowly excavated than 25, but may also exhibit two separate fossae.
But bone is usually the least variable material in the vertebrate body. Muscles vary more, nerves more again, and blood vessels most of all. So why are the vertebrae of sauropods so much more variable than other bones?
Our new paper, published today (Taylor and Wedel 2021) proposes an answer! Please read it for the details, but here’s the summary:
- Early in ontogenly, the blood supply to vertebrae comes from arteries that initially served the spinal cord, penetrating the bone of the neural canal.
- Later in ontegeny, additional arteries penetrate the centra, leaving vascular foramina (small holes carrying blood vessels).
- This hand-off does not always run to completion, due to the variability of blood vessels.
- In extant birds, when pneumatic diverticula enter the bone they do so via vascular foramina, alongside blood vessels.
- The same was probaby true in sauropods.
- So in vertebrae that got all their blood supply from vascular foramina in the neural canal, diverticula were unable to enter the centra from the outside.
- So those centra were never pneumatized from the outside, and no externally visible pneumatic cavities were formed.
Somehow that pretty straightforward argument ended up running to eleven pages. I guess that’s what you get when you reference your thoughts thoroughly, illustrate them in detail, and discuss the implications. But the heart of the paper is that little bullet-list.
Taylor and Wedel (2021: Figure 6). Domestic duck Anas platyrhynchos, dorsal vertebrae 2–7 in left lateral view. Note that the two anteriormost vertebrae (D2 and D3) each have a shallow pneumatic fossa penetrated by numerous small foramina.
(What is the relevance of these duck dorsals? You will need to read the discussion in the paper to find out!)
Our choice of publication venue
The world moves fast. It’s strange to think that only eleven years ago my Brachiosaurus revision (Taylor 2009) was in the Journal of Vertebrate Palaeontology, a journal that now feels very retro. Since then, Matt and I have both published several times in PeerJ, which we love. More recently, we’ve been posting preprints of our papers — and indeed I have three papers stalled in peer-review revisions that are all available as preprints (two Taylor and Wedels and a single sole-authored one). But this time we’re pushing on even further into the Shiny Digital Future.
We’ve published at Qeios. (It’s pronounced “chaos”, but the site doesn’t tell you that; I discovered it on Twitter.) If you’ve not heard of it — I was only very vaguely aware of it myself until this evening — it runs on the same model as the better known F1000 Research, with this very important difference: it’s free. Also, it looks rather slicker.
That model is: publish first, then filter. This is the opposite of the traditional scholarly publishing flow where you filter first — by peer reviewers erecting a series of obstacles to getting your work out — and only after negotiating that course to do get to see your work published. At Qeios, you go right ahead and publish: it’s available right off the bat, but clearly marked as awaiting peer-review:
And then it undergoes review. Who reviews it? Anyone! Ideally, of course, people with some expertise in the relevant fields. We can then post any number of revised versions in response to the reviews — each revision having its own DOI and being fixed and permanent.
How will this work out? We don’t know. It is, in part, an experiment. What will make it work — what will impute credibility to our paper — is good, solid reviews. So if you have any relevant expertise, we do invite you to get over there and write a review.
And finally …
Matt noted that I first sent him the link to the Qeios site at 7:44 pm my time. I think that was the first time he’d heard of it. He and I had plenty of back and forth on where to publish this paper before I pushed on and did it at Qeios. And I tweeted that our paper was available for review at 8:44 — one hour exactly after Matt learned that the venue existed. Now here we are at 12:04 my time, three hours and 20 minutes later, and it’s already been viewed 126 times and downloaded 60 times. I think that’s pretty awesome.
References
- Taylor, Michael P. 2009. A re-evaluation of Brachiosaurus altithorax Riggs 1903 (Dinosauria, Sauropoda) and its generic separation from Giraffatitan brancai (Janensch 1914). Journal of Vertebrate Paleontology 29(3):787-806. [PDF]
- Taylor, Michael P., and Mathew J. Wedel. 2021. Why is vertebral pneumaticity in sauropod dinosaurs so variable? Qeios 1G6J3Q. doi: 10.32388/1G6J3Q [PDF]
- Wedel, Mathew J., and Michael P. Taylor 2013b. Caudal pneumaticity and pneumatic hiatuses in the sauropod dinosaurs Giraffatitan and Apatosaurus. PLOS ONE 8(10):e78213. 14 pages. doi: 10.1371/journal.pone.0078213 [PDF]
Supersaurus, Ultrasaurus and Dystylosaurus in 2019, part 5: what actually is Supersaurus?
June 25, 2019
When I started this series, it wasn’t going to be a series at all. I thought it was going to be a single post, hence the title that refers to all three of Jensen’s 1985 sauropods even though most of the posts so far have been only about Supersaurus. The tale seems to have grown in the telling. But we really are getting towards the end now. This should be the last post that is only about Supersaurus, and then we should be able to finish with one more that covers all three animals.
Supersaurus skeletal reconstruction at NAMAL, based in part on preserved fossil material. Mike Taylor for scale, lying in front of the referred scapulocoracoid BYU 12962.
So: what actually is Supersaurus?
Is Supersaurus the same thing as Barosaurus?
As we established previously, a lot of material has been referred not only to Supersaurus in general, but to the type individual in particular: a cervical, two dorsals, four sacrals, 20 caudals, two scapulocoracoids, an ulna, a carpal, right ilium and pubis, both ischia, and a phalanx. (After Jensen’s original papers, Curtice and his collaborators did much of the work to assemble this list.) And remember, too, that Lovelace et al. (2008) described a completely separate Supersaurus specimen from Wyoming.
So: a problem arises: Matt and I are about as certain as we can be that the big cervical verebra BYU 9024 is Barosaurus. That means there are two possibilities: either the cervical been wrongly referred to the Supersaurus type individual, and our conception of Supersaurus needs to change accordingly; or it was correctly referred, which means that Supersaurus is merely a very big Barosaurus, and the name should be sunk.
I would be a lot more confident about which of these is the right thing to do if Matt and I had had time to look at all the sacral, caudal and appendicular material of Supersaurus during the Sauropocalypse. But our time was very limited (seven museums in nine days) and we had to focus on the presacrals.
What we really want is a solid assessment of all the putative Supersaurus material and a judgement of whether the differences between it and regular Barosaurus might be size- or age-related. We can’t have that (at least, not unless someone with more time on their hands than Matt or me takes it on).
But we are not left without hope. We have the published literature.
Pylogenetic analyses
Lovelace et al. (2008:figure 14). Strict consensus tree resulting from the addition of Supersaurus and “Seismosaurus” into a modified matrix from Harris & Dodson (2004).
First, Lovelace et. al’s (2008) description of Jimbo, the WDC’s referred Supersaurus specimen, included a phylogenetic analysis. This recovered Supersaurus as the sister taxon to Apatosaurus, with Suuwassea as its outgroup, and the Barosaurus–Diplodocus clade sister to that broader grouping. That finding would argue against Supersaurus being Barosaurus. (They commented that “It is possible that some similarities between Supersaurus and other apatosaurines result from a size-coupled increase in robustness, but it is worth noting that apatosaurine robustness does not correlate with size, and large diplodocines like Seismosaurus do not exhibit markedly more robust pelvic or costal elements.)
Whitlock (2011:figure 7). Phylogenetic hypothesis presented in this analysis. Cladogram represents a strict consensus of three equally parsimonious trees (273 steps), labelled with relevant clade names. Decay indices reported below each node.
Whitlock’s (2011) more detailed phylogenetic analysis recovered Supersaurus is a somewhat more traditional position, closer to Barosaurus than to Apatosaurus. But still not very close. Supersaurus is here the most basal diplodocine, the outgroup to Dinheirosaurus, Torneria and the Barosaurus+Diplodocus pair. It’s not a result that would immediately make you want to synonymise Supersaurus with Barosaurus.
One problem with both Lovelace et al.’s and Whitlock’s analyses is that they took as read that the WDC specimen really is Supersaurus — the same thing as the BYU specimen. What if it isn’t? Maybe the WDC animal is something different that’s more closely related to Apatosaurus, while the BYU specimen is a big Barosaurus? Is that possible?
Enter Tschopp et al. (2015), whose monumental specimen-level analysis separated Jimbo out from BYU Supersaurus — and so they tested the hypothesis that these two specimens are the same thing, instead of assuming it. Here’s what they found:
Tschopp et al. (2015:figure 118). Reduced consensus tree obtained by implied weighting. Eight OTUs were deleted a posteriori. Numbers at the nodes indicate the number of changes between the two branches departing from the node (for the apomorphy count), where they differ from the trees under equal weights.
As you can see, BYU Supersaurus and the WDC specimen came out as sister taxa in every most parsimonious tree. And Tschopp et al.’s (2015) figure 115 shows that this is true under equal-weights parsimony as well as under implied weighting. So this gives us confidence that the WDC team’s referral of Jimbo to Supersaurus probably is correct after all.
But that Supersaurus duo comes out some way away from Barosaurus, being well outside the Diplodocus–Barosaurus node.
These are the only three phylogenetic analyses I am aware of to have included Supersaurus — though if there are others, please shout in the comments. In none of them do Supersaurus and Barosaurus come out as sister taxa, and in fact they are separated by multiple nodes in all three analyses.
More compellingly, Andrea Cau re-ran Tschopp et al.’s (2015) analysis with Supersaurus and Barosaurus constrained to be sister groups (thanks, Andrea!) and found that the best resulting trees were 18 steps longer than the unenforced trees (1994 steps vs 1976). This is convincing evidence that the totality of the Supersaurus material is not Barosaurus.
Is BYU Supersaurus a chimaera?
All of this strongly suggests — it comes close to conclusively proving — that Supersaurus (as defined by all the BYU and WDC material) is not Barosaurus. But if Matt and I are right that BYU 9024 is a vertebra of Barosaurus, then it follows that this cervical doesn’t belong to Supersaurus.
And that, I think, throws the whole material list of BYU Supersaurus into question. Because if the big cervical belongs to something different, then it follows that there are (at least) two big diplodocids mixed up in the Dry Mesa quarry, contra Curtice et al.’s (2001) assertion that all the big bones there can be referred to two individuals, one diplodocid and one brachiosaur.
In which case, how can we know which of the elements belongs to which of the animals?
Are the scapulocoracoids from the same individual?
Can we even trust the assumption that the two scapulocoracoids were from the same animal? Maybe not. In favour of that possibility, the two elements are similar sizes, and were found close together. But there are reasons to be sceptical.
Based on our photos in the earlier post, I was coming to the conclusion that Scap B is much less sculpted than Scap A. But I started to change my mind once I was able to make a weak anaglyph of Scap B. Now, thanks to Heinrich Mallison and the magic of photogrammetry, my set of bad photos have become a 3D model, which is far more informative again.
Here, then, is a comparative anaglyph of the two scapulocoracoids.
Red-cyan anaglyps of both scapulocoracoids of Supersaurus from BYU’s Dry Mesa Quarry, Utah. Top: the holotype BYU 9025, left scapulocoracoid (“Scap A”); Bottom: referred specimen BYU 12962, right scapulocoracoid (“Scap B”), reversed for easier comparison. Scap B rendered from a 3D model created by Heinrich Mallison. Scaled to the same length. (We could not scale them in correct proportion, since the true current lengths of both are unknown.)
These are not obviously from the same individual, or from the same species, or even necessarily the same “subfamily”. A few of the more obvious morphological differences:
- In Scap A, the acromion process projects posterodorsosally, whereas in Scap B it projects dorsally (i.e. at right angles to the long axis of the scap.)
- In Scap A, the acromion process is positioned close to mid-length of the whole element, whereas in Scap B it is closer to the proximal end.
- In Scap A, the acromion process comes to a point, whereas in Scap B is it lobe-shaped.
- In Scap A, the ridge running running up to the acromion process is broad and becomes rugose dorsally, whereas in Scap B it is narrow and remains smooth along its whole length.
- Scap B has a distinct ventral bump around midlength, which Scap A lacks (or at most has in a much reduced form).
- In Scap B, the ventral border below the acromion process distinctly curves down to the glenoid, but in Scap B this ventral margin is almost straight.
- In Scap A, the glenoid margin is gently curved, nearly straight, whereas in Scap B it has a well defined “corner”, with distinct scapular and coracoid contributions that are at right angles to each other.
- In Scap A, the dorsal margin of the coracoid is well defined and has a low laterally protruding ridge. This is absent in Scap B, where the coracoid’s dorsal margin is poorly defined.
Now, much of this is quite possibly due to damage — as (I assume) is the excavation in the dorsal margin of the distal part of the scapular blade in Scap A. But when you put it all together, I think they really are rather different, even allowing for variation in limb-girdle bones. Certainly if you found them both in different quarries, you would not leap to the conclusion that they belong to the same species. Jensen’s (1985:701) description of Scap B (BYU 5001 of his usage) as “same as Holotype, BYU 5500” is difficult to justify.
The possibility that the two scaps are from different individuals is also weakly supported by the fact that the preservation looks very different between the two elements — dark and rough for Scap A but light and smooth for Scap B. But I don’t trust that line of evidence as much as I might for two reasons. First, different photography conditions can give strikingly different coloured casts to photos, making similar bones appear different. And second, I know from experience that bones from a single specimen can vary in colour and preservation much more than you’d expect.
At any rate, I certainly don’t think it’s a given that the two scapulae belonged to to the same individual as Curtice and Stadtman (2001) stated. And of course if they do not, then the issue of which is the holotype takes on greater importance — which is why we spent so long on figuring that out.
So what are we left with?
We know — or at least we are confident — that one of the referred BYU Supersaurus elements is Barosaurus. We don’t think the whole animal is Barosaurus, due to the evidence of three phylogenetic analyses. So we think there are at least two big diplodocoids in the BYU quarry, and we can’t know which of the elements belongs to which animal. We can’t even be confident that the two scapulocoracoids belong to the same animal.
As a result, the only bone that we can confidently state belongs to Supersaurus is the holotype — BYU 9025, which we called “Scap A”. All bets are off regarding all the other Dry Mesa diplodocoid elements. They might belong the Scap A taxon, or to Barosaurus. (Or indeed to something else, but we’ll ignore that possibility as multiplying entities without necessity.)
So to the next question: is the holotype element even diagnostic, beyond the level of “big diplodocoid”? I’m not sure it is, but this is where I’d welcome input from people who are more familiar with sauropod appendicular material than I am. At any rate, Jensen’s (1985:701) original diagnosis based on the holotype scap is useless: “Scapula long but not robust; distal end expanding moderately; shaft not severely constricted in midsection”.
The emended diagnosis of Lovelace et al. (2008:530) says of the scapulocoracoid only “scapular blade expanded dorsally; deltoid ridge perpendicular to the acromian[sic] ridge”. but they also include a more comprehensive assessment of the BYU scapulae (p. 534) as follows:
The only known pectoral elements for Supersaurus are the scapulocoracoids from Dry Mesa (Fig.10). Scapulocoracoid BYU 9025 demonstrates a deltoid ridge that is perpendicular to the acromian ridge and the scapular blade is one-half the entire length of the scapulocoracoid. Both of these features are seen in Apatosaurus but not in Diplodocus or Barosaurus, which have relatively short scapular blades, and an acute angle between the deltoid ridge and the acromian ridge. This angle is much stronger in Barosaurus than it is in Diplodocus. The apatosaurine nature of the scapulocoracoids further reinforces the referral of BYU elements to the type scapula, as well as our referral of WDC DMJ-021 to Supersaurus.
This is a helpful discussion (although note that Lovelace et al. are not consistent about which of the scaps they think is BYU 9025). But, notably, nothing here suggests any unique characters of the scapulocoracoid that could serve to diagnose Supersaurus by its holotype.
Putting it all together, it seems that BYU 9025 is the only bone in the world that unambiguously belongs to Supersaurus (because it is the the holotype, and all referrals are uncertain); and that bone is non-diagnostic. I think it must follow, then, that Supersaurus is currently a nomen dubium.
I say “currently”, because there are at least three possible ways for the name to survive. (Four, if you count everyone just ignoring this sequence of blog-posts.) Next time, we’ll talk about those options.
References
- Curtice, Brian D. and Kenneth L. Stadtman. 2001. The demise of Dystylosaurus edwini and a revision of Supersaurus vivianae. Western Association of Vertebrate Paleontologists and Mesa Southwest Museum and Southwest Paleontologists Symposium, Bulletin 8:33-40.
- Harris, Jerald D., and Peter Dodson. 2004. A new diplodocoid sauropod dinosaur from the Upper Jurassic Morrison Formation of Montana, USA. Acta Palaeontologica Polonica 49:197–210.
- Jensen, James A. 1985. Three new sauropod dinosaurs from the Upper Jurassic of Colorado. Great Basin Naturalist 45(4):697–709.
- Lovelace, David M., Scott A. Hartman and William R. Wahl. 2008. Morphology of a specimen of Supersaurus (Dinosauria, Sauropoda) from the Morrison Formation of Wyoming, and a re-evaluation of diplodocid phylogeny. Arquivos do Museu Nacional, Rio de Janeiro 65(4):527–544.
- Tschopp, Emanuel, Octávio Mateus and Roger B. J. Benson. 2015. A specimen-level phylogenetic analysis and taxonomic revision of Diplodocidae (Dinosauria, Sauropoda). PeerJ 2:e857. doi:10.7717/peerj.857
- Whitlock, John A. 2011. A phylogenetic analysis of Diplodocoidea (Saurischia: Sauropoda). Zoological Journal of the Linnean Society 161(4):872-915. doi:10.1111/j.1096-3642.2010.00665.x
In my recent visit to the LACM herpetology collection, I was interested to note that almost every croc, lizard, and snake vertebra I saw had a pair of neurovascular foramina on either side of the centrum, in “pleurocoel” position. You can see these in the baby Tomistoma tail, above. Some vertebrae have a big foramen, some have a small foramen, and some have no visible foramen at all. Somehow I’d never noticed this before.
This is particularly interesting in light of the observation from birds that pneumatic diverticula tend to follow nerves and vessels as they spread through the body. Maybe we find pneumatic features where we do in dinosaurs and pterosaurs because that’s where the blood vessels were going in the babies. Also, these neurovascular foramina in extant reptiles are highly variable in size and often asymmetric – sound familiar?
It should. Caudal pneumaticity in the tail of Giraffatitan MB.R.5000. Dark blue vertebrae are pneumatic on both sides, light blue vertebrae only have fossae on the right side. Wedel and Taylor (2013b: Figure 4).
I am starting to wonder if some of the variability we associate with pneumaticity is just the variability of soft tissue, full stop. Or if pneumaticity is variable because it developmentally follows in the footsteps of the blood vessels, which are themselves inherently variable. That seems like a promising line of inquiry. And also something I should have though of a lot sooner.
UPDATE in 2023: A promising line of inquiry indeed! This spawned a paper in 2021, a conference presentation later in 2021, which will become a paper in time, and the tentacles of this idea — that diverticula following blood vessels has a lot of explanatory power — are wound through a LOT of my current and upcoming projects.
Vertebral orientation, part 3: Matt weighs in
October 5, 2018
WOW! I knew I was dragging a bit on getting around to this vertebral orientation problem, but I didn’t realize a whole month had passed. Yikes. Thanks to everyone who has commented so far, and thanks to Mike for getting the ball rolling on this. Previous posts in this series are here and here.
First up, this may seem like a pointlessly picky thing to even worry about. Can’t we just orient the vertebrae in whichever way feels the most natural, or is easiest? Do we have to think about this?
The alarmingly 3D pelvis of the mounted brontosaur at the AMNH. Note that sauropod pubes are usually illustrated lying flat, so what usually passes for ‘lateral’ view would be roughly from the point of view of the animal’s knee.
I think we do. For sauropods, vertebrae are usually oriented for illustration purposes in one of two ways. The first is however they sit most easily on their pallets. This is similar to the problem Mike and I found for ‘lateral’ views of sauropod pelvic elements when were on our AMNH/Yale trip in 2012. In an articulated skeleton, the pubes and ischia usually lean inward by 30-45 degrees from their articulations with the ilia, so they can meet on the midline, but when people illustrate the “lateral view” of a sauropod pubis or ischium, it’s often the ventro-lateral aspect that is face-up when the element is lying on a shelf or a pallet. Photographic lateral does not equal biological lateral for those elements. Similarly, if I’m trying to answer biological questions about vertebrae (see below), I need to know something about their orientation in the body, not just how they sit comfortably on a pallet.
The other way that vertebrae are commonly oriented is according to what we might call the “visual long axis” of the centrum—so for example, dorsoventrally tall but craniocaudally short proximal caudals get oriented with the centrum ‘upright’, whereas dorsoventrally short but craniocaudally long distal caudals get oriented with the centrum ‘horizontal’, even if they’re in the same tail and doing so makes the neural canals or articular faces be oriented inconsistently down the column. (I’m not going to name names, because it seems mean to pick on people for something I just started thinking about myself, but if you go plow through a bunch of sauropod descriptions, you’ll see what I’m talking about.)
Are there biological questions where this matters? You bet! There are some questions that we can’t answer unless we have the vertebrae correctly oriented first. One that comes to mind is measuring the cross-sectional area of the neural canal, which Emily Giffin did a lot of back in the 90s. Especially for the Snowmass Haplocanthosaurus, what counts as the cross-sectional area of the neural canal depends on whether we are looking at the verts orthogonal to their articular faces, or in alignment with the course of the canal. I think the latter is pretty obviously the way to go if we are measuring the cross-sectional area of the canal to try and infer the diameter of the spinal cord—we’d want to see the canal the same way the cord ‘sees’ it as it passes through—but it’s less obvious if we’re measuring, say, the surface area of the articular face of the vertebra to figure out, say, cartilage stress. It doesn’t seem unreasonable to me that we might want to define a ‘neural axis’ for dealing with spinal-cord-related questions, and a ‘biomechanical axis’ for dealing with articulation-related questions.
Caudal 3 of the Snowmass Haplocanthosaurus, hemisected 3D model.
With all that in mind, here are some points.
To me, asking “how do we know if a vertebra is horizontal” is an odd phrasing of the problem, because “horizontal” doesn’t have any biological meaning. I think it makes more sense to couch the question as, “how do we define cranial and caudal for a vertebra?” Normally both the articular surfaces and the neural canal are “aimed” head- and tail-wards, so the question doesn’t come up. Our question is, how do we deal with vertebrae for which the articular surfaces and neural canal give different answers?
For example. Varanus komodoensis caudal.
(And by the way, I’m totally fine using “anterior” and “posterior” for quadrupedal animals like sauropods. I don’t think it causes any confusion, any more than people are confused by “superior” and “inferior” for human vertebrae. But precisely because we’re angling for a universal solution here, I think using “cranial” and “caudal” makes the most sense, just this once. That said, when I made the image above, I used anterior and posterior, and I’m too lazy now to change it.)
I think if we couch the question as “how do we define cranial and caudal”, it sets up a different set of possible answers than Mike proposed in the first post in this series: (1) define cranial and caudal according to the neural canal, and then describe the articular surfaces as inclined or tilted relative to that axis; (2) vice versa—realizing that using the articular surfaces to define the anatomical directions may admit a range of possible solutions, which might resurrect some of the array of possible methods from our first-draft abstract; (3) define cranial and caudal along the long axis of the centrum, which is potentially different from either of the above; (4) we can imagine a range of other possibilities, like “use the zygs” or “make the transverse processes horizontal” (both of which are subsets of Mike’s method C) but I don’t think most of those other possibilities are sufficiently compelling to be worthy of lengthy discussion.
IF we accept “neural canal”, “articular surfaces”, and “centrum long axis” as our strongest contenders, I think it makes most sense to go with the neural canal, for several reasons:
- In a causative sense, the neural tube/spinal cord does define the cranial/caudal axis for the developing skeleton. EDIT: Actually, that’s a bit backwards. It’s the notochord, which is later replaced by the vertebral column, that induces the formation of the brain and spinal cord from the neural plate. But it’s still true that the vertebrae form around the spinal cord, so it’s not wrong to talk about the spinal cord as a defining bit of soft tissue for the developing vertebrae to accommodate.
- The neural canal works equally well for isolated vertebrae and for articulated series. Regardless of how the vertebral column is oriented in life, the neural canal is relatively smooth—it may bend, but it doesn’t kink. So if we line up a series of vertebrae so that their neural canals are aligned, we’re probably pretty close to the actual alignment in life, even before we look at the articular surfaces or zygs.
- The articulated tails of Opisthocoelicaudia and big varanids show that sometimes the articular surfaces simply are tilted to anything that we might reasonably consider to be the cranio-caudal axis or long axis of the vertebra. In those cases, the articular surfaces aren’t orthogonal to horizontal OR to cranio-caudal. So I think articular surfaces are ruled out because they break down in the kinds of edge cases that led us to ask the question in the first places.
Opistocoelicaudia caudals 6-8, stereopair, Borsuk-Bialynicka (1977:plate 5).
“Orient vertebrae, isolated or in series, so that their neural canals define the cranio-caudal axis” may seem like kind of a ‘duh’ conclusion (if you accept that it’s correct; you may not!), but as discussed up top, often vertebrae from a single individual are oriented inconsistently in descriptive works, and orientation does actually matter for answering some kinds of questions. So regardless of which conclusion we settle on, there is a need to sort out this problem.
That’s where I’m at with my thinking. A lot of this has been percolating in my hindbrain over the last few weeks—I figured out most of this while I was writing this very post. Is it compelling? Am I talking nonsense? Let me know in the comments.
Sauropod Vertebra Picture of the Week 