The early armored fish Bothriolepis, which Yara Haridy affectionately refers to as a “beetle mermaid”. Art by Brian Engh,

If I had to sum up my main research program over the past 20+ years, it would be, “Why pneumatic bone?” Or as I typically put it in my talks, most bone has marrow inside, so if you find bone with air inside, someone has some explaining to do (f’rinstance).

One of the reasons I like hanging out with Yara Haridy is that she is interested in an even more fundamental question: “Why bone?” And also “How bone?” And she has a paper out today that gives us new insights into the form and function of bone cells — osteocytes — in some of the earliest vertebrates that had them (Haridy et al. 2021; if you’re in TL;DR mode, here’s the link). 

Bones have multiple functions in vertebrate bodies: they’re a mechanical framework for our muscles, and a mineral reservoir, and form armor in many taxa, and are involved in hormone regulation, and doubtless other things that we are still discovering, even now. To fulfill those functions, bone tissue has to be formed in the first place, it has to be maintained, and it has to be able to be reshaped as an individual grows. Derived extant vertebrates, including humans, have an impressive array of cellular machinery to make all those things happen. Central to most of those operations are osteocytes, the cells inside living bone, which maintain intimate connections to extracellular bone tissue and to other osteocytes via fine, tentacle-like processes.

Individual osteocytes look something like the Flying Spaghetti Monster. The central portion of the FSM, with the meatballs, is the osteocyte body, and the noodly appendages are the processes.

Now imagine that you cloned the FSM many, many times, and the resulting array of FSMs stayed in physical contact with each other via their noodly appendates, forming a network.

Then imagine that you entombed all of the cloned FSMs in concrete. This is more or less what cellular bone — the kind you find in humans, dinosaurs, and even some jawless fish — looks like on the microscopic scale: osteocytes (the FSMs) and their processes (the noodly appendages) embedded in space-filling stuff (the bone matrix). Some critters, including teleost fish, have acellular bone, but I don’t have time for those unbelievers today.

When an animal dies and decomposes, the osteocytes and their processes decay away, leaving behind the spaces that they used to occupy. The big spaces that hold osteocytes are called lacunae, and the little tunnels that hold the osteocyte processes (noodly appendages, in this metaphor) are canaliculi. Collectively, the lacunae and canaliculi form the lacunocanalicular network or LCN.

Those spaces can then be filled by matrix — not extracellular bone matrix, but future rock matrix, like mud and clay. In point of fact, not all of the spaces are filled with matrix. Even in their 420-million-year-old fish, Yara and colleagues found some osteocyte lacunae that had not been filled with matrix, and were filled by air instead. Whether the lacunocanalicular network is filled with matrix or air, its preservation in fossil bone has turned out to be a boon for paleontologists, because we can ‘see’ the sizes and shapes of osteocytes, and their level of connectivity, by studying the lacunae and canaliculi they left behind. 

Histological thin section of bone in the osteostracan Tremataspis mammillata (MB.f.TS.463), imaged with transmitting light microscopy showing osteocyte lacunae (osl) and canaliculi (ca); scale bar, 100 microns. Haridy et al. (2021: fig 2A).

Traditionally osteocyte lacunae and canaliculi in fossil bone have been imaged by taking thin sections of the specimens and looking at them under microscopes.

Synchrotron tomography of bone of Bothriolepis trautscholdi (MB.f.9188a) with the vasculature and osteocytes segmented; scale bar, 0.4 mm. (D) Close-up of tomography in (C) showing the resolution of the osteocyte lacunae volumes; scale bar, 10 microns. bs, bone sample; vs, vasculature channels. Modified from Haridy et al. (2021: fig. 2C-D).

If you’re fancy, you can also do synchrotron tomography, which is fine enough to show osteocyte lacunae — the colored blobs in the image on the right, above.

Those methods have their limitations. Light microscopy will reveal both lacunae and canaliculi in 2D, but it’s hard to get a 3D understanding of the lacunocanalicular network that way (at least in fossils; in modern samples it can be done with confocal miscoscopy). Synchrotron tomography can resolve lacunae in 3D, but not canaliculi, sort of like a map that shows only cities but not the highways that connect them.

Enter FIB-SEM: focused ion beam scanning electron microscopy. An ion gun blasts the specimen with a beam of gallium ions, which vaporizes a slice of the specimen that is less than 1 micron thick, and an SEM images the freshly exposed face. If you do this over and over again, you can build up a 3D model of the stuff that once occupied the volume that got zapped. 

FIB-SEM tomography imaging and processing of the fossil jawless vertebrate Tremataspis mammillata (MB.f.9025). (A and B) FIB-SEM setup showing the FIB in relation to the SEM both aimed at the region of interest. (C) Bone surface with an excavate area made by the FIB. (D) Internal wall of the excavated area lined with small black dots that are the fossil osteocyte lacunae. (E) Single osteocyte lacuna from the surface that is scanned; the single SEM image shows the lacunae and canaliculi in black and the mineralized bone in gray; scale bar, 5 microns. (F and G) An image stack is obtained, and 3D made of fossil LCN can be made. Haridy et al. (2021: fig 3).

FIB-SEM is fine enough to resolve both osteocyte lacunae and canaliculi — the lacunocanalicular network or LCN — in three dimensions, in fossil specimens where confocal light microscopy doesn’t always work very well. And the resolution is pretty insane. The rough edges on the 3D models of the LCN aren’t sampling artifacts, they’re accurately reflecting the real morphology of the walls of the lacunae and canaliculi as they were preserved in the fossil bone.

But wait — that’s not all! Not only can FIB-SEM show us osteocyte lacunae and canaliculi in incredible detail in three dimensions, it can also help us figure out at least some of what osteocytes were doing. Together, osteocytes and their processes can sense mechanical strain in bone, trigger bone remodeling, and resorb and lay down bone from the walls of the lacunae and canaliculi. That last process starts with osteocytic osteolysis — the resorption of bone matrix (= osteolysis) from the lacunae and canaliculi by the osteocytes themselves (as opposed to the more familiar destruction of bone at a larger spatial scale by osteoclasts), which is typically followed by the replacement of new matrix where the old bone used to be. Lots of extant vertebrates do osteocytic osteolysis, especially those that have a high demand for calcium and phosphorus in physiologically challenging times. Examples including migrating salmon, lactating mice, and lactating humans. But when did that capacity evolve — did the earliest osteocytes already have the ability to resorb and replace bone? As Yara said to me when she was telling me about her new paper, “We think we know how things work by looking at extant animals, but we’re looking at this highly pruned tree, and we can’t just assume that things worked the same way earlier in our evolutionary history.” 

Yara wanted to investigate when osteolytic osteolysis first evolved when she started her dissertation in 2018, but she didn’t know that FIB-SEM existed. Then she was visiting a neutron tomography facility in Berlin and she saw a poster on the wall about people using FIB-SEM to image corrosion in batteries on ultra-fine scales. She thought, “Wow, the corrosion pits in the batteries look like osteocytes!” The rest you probably figured out faster than it’s taking me to write this sentence: together with her collaborators, she got some samples of bone from the jawless fish Tremataspis and Bothriolepis and zapped them with the FIB-SEM.

Osteocytic osteolysis as a mechanism for early mineral metabolism. (A to C) Illustrations depicting the process of osteocytic osteolysis; the phases are stasis phase, dissolution phase, and redeposition phase, respectively. (D) Single SEM image from the FIB-SEM acquisition showing the air-filled osteocyte lacunae and canaliculi of T. mammillata. (E) Same SEM image as in (D) with contrast shifted to show the demineralized zone surrounding the lacunae. (F and G) 3D render of the stack of images from (D and E) T. mammillata (MB.f.9025). The 3D model shows several osteocytes and their canaliculi, with the red areas showing where the “areas of low density” were found. ald, area of low density; os, osteocyte. Scale bar, 5 microns.

And, wonder of wonders, some of the osteocyte lacunae in Tremataspis were surrounded by a halo of less-dense bone, which is evidence for osteocytic osteolysis. Now, Yara and colleagues can’t be sure whether the bone is less dense because it was being resorbed when the animal died — the actual lytic or bone-destructive phase — or because new bone was being laid down after the old bone had been resorbed; naturally the new bone is less dense as it being formed than it will be when it is complete. They also can’t be sure why the process was occurring in that one individual Tremataspis. Mice only do osteocytic osteolysis when they’re lactating, and salmon only do it when they’re migrating, so the presence of osteocytic osteolysis might indicate that the Tremataspis in question was doing something stressful related to its ecology or life history — both topics we know almost nothing about.

Yara and colleagues didn’t find any evidence of osteocytic osteolysis in their Bothriolepis sample, but this is one of those ‘absence of evidence is not evidence of absence’ things — you wouldn’t find evidence of osteocytic osteolysis in my skeleton either, despite a long ancestral history, because I’m skeletally healthy, not fasting or migrating, and not lactating. Possibly other Bothriolepis individuals that were going through a rough patch, metabolically speaking, would show osteocytic osteolysis. So far, as a species, we’ve only looked at the one sample, from the one individual.

I asked Yara what she wanted people to take away from her new paper. Her response:

  1. We have technology can image fossil bone cells at the same resolution that we can see modern bone cells.
  2. Bone metabolism was going on 420 million years ago, in the earliest osteocytes, the same way it happens in modern mammals, including humans. 

I expect that we will see a lot more FIB-SEM papers on fossils in years to come. That research program started today, with the publication of Haridy et al. (2021). I often sign off posts with “stay tuned”, and this time I really mean it.


Yara Haridy, Markus Osenberg, André Hilger, Ingo Manke, Donald Davesne, and Florian Witzmann. 2021. Bone metabolism and evolutionary origin of osteocytes: Novel application of FIB-SEM tomography. Science Advances 7(14): eabb9113. DOI: 10.1126/sciadv.abb9113.

Accidental anaglyphs

October 16, 2020

Everyone knows that the very first thing you should do to improve your specimen photography is to use a tripod: it eliminates hand-shake and gives you much crisper photos. In most respects, my photographs have got much, much better since I’ve been habitually using a tripod.

But it has meant I’ve not been able to benefit from happy accidents like the one that gave me this 3D anaglyph of the Archbishop‘s Cervical S in dorsal view:

(Do you have red-cyan glasses? Yes? Good! You will be able to appreciate all the delicious morphological information in this photo. No? Go and order some right now — they cost literally a dollar.)

The reason I was able to make this very useful image is because back in the old pre-tripod days I would sometimes accidentally move a little bit between taking two more-or-less identical photographs. Here are the two images that I was able to composite into the anaglyph above:

Each of them is pretty uninformative alone: who can tell one nondescript area of brown bone from another? But when combined, they are extraordinarily more informative. If you don’t have 3D glasses then (A) get some! and (B) you can get some idea of how helpful the 3D information is from the crude wigglegram below, which simply switches back and forth between the two images.

And I can’t overstate how enormously helpful I have found these accidentally sourced anaglyphs as I write the descriptive part of the Archbishop manuscript. Even at this level of crudity, they have shown me several important points of morphology that I would certainly have missed if I’d been working only from my orthogonal-view photos, and saved me from more than one misinterpretation.

The moral is twofold:

  1. When taking specimen photographs, use a tripod — but deliberately get some pairs of shots where the camera is moved to the side by about 7 cm (the distance between the pupils in an average human).
  2. If you don’t have any red-cyan glasses, get some!

Daniel Vidal et al.’s new paper in Scientific Reports (Vidal et al. 2020) has been out for a couple of days now. Dealing as it does with sauropod neck posture, it’s obviously of interest to me, and to Matt. (See our earlier relevant papers Taylor et al. 2009, Taylor and Wedel 2013 and Taylor 2014.)


To brutally over-summarise Vidal et al.’s paper, it comes down to this: they digitized the beautifully preserved and nearly complete skeleton of Spinophorosaurus, and digitally articulated the scans of the bones to make a virtual skeletal mount. In doing this, they were careful to consider the neutral pose of consecutive vertebrae in isolation, looking at only one pair at a time, so as to avoid any unconscious biases as to how the articulated column “should” look.

Then they took the resulting pose, objectively arrived at — shown above in their figure 1 — and looked to see what it told them. And as you can well see, it showed a dramatically different pose from that of the original reconstruction.

Original skeletal reconstruction of Spinophorosaurus nigerensis (Remes et al. 2009:figure 5, reversed for ease of comparison). Dimensions are based on GCP-CV-4229/NMB-1699-R, elements that are not represented are shaded. Scale bar = 1 m.

In particular, they found that as the sacrum is distinctly “wedged” (i.e. its anteroposterior length is greater ventrally than it is dorsally, giving it a functionally trapezoidal shape, shown in their figure 1A), so that the column of the torso is inclined 20 degrees dorsally relative to that of the tail. They also found lesser but still significant wedging in the last two dorsal vertebrae (figure 1B) and apparently some slight wedging in the first dorsal (figure 1C) and last cervical (figure 1D).

The upshot of all this is that their new reconstruction of Spinophorosaurus has a strongly inclined dorsal column, and consequently an strongly inclined cervical column in neutral pose.

Vidal et al. also note that all eusauropods have wedged sacra to a greater or lesser extent, and conclude that to varying degrees all eusauropods had a more inclined torso and neck than we have been used to reconstructing them with.


I have to be careful about this paper, because its results flatter my preconceptions. I have always been a raised-neck advocate, and there is a temptation to leap onto any paper that reaches the same conclusion and see it as corroboration of my position.

The first thing to say is that the core observation is absolutely right, — and it’s one of those things that once it’s pointed out it’s so obvious that you wonder why you never made anything of it yourself. Yes, it’s true that sauropod sacra are wedged. It’s often difficult to see in lateral view because the ilia are usually fused to the sacral ribs, but when you see them in three dimensions it’s obvious. Occasionally you find a sacrum without its ilium, and then the wedging can hardly be missed … yet somehow, we’ve all been missing its implications for a century and a half.

Sacrum of Diplodocus AMNH 516 in left lateral and (for our purposes irrelevant) ventral views. (Osborn 1904 figure 3)

Of course this means that, other thing being equal, the tail and torso will not be parallel with each other, but will project in such a way that the angle between them, measured dorsally, is less than 180 degrees. And to be fair, Greg Paul has long been illustrating diplodocids with an upward kink to the tail, and some other palaeoartists have picked up on this — notably Scott Hartman with his very uncomfortable-looking Mamenchisaurus.

But I do have three important caveats that mean I can’t just take the conclusions of the Vidal et al. paper at face value.

1. Intervertebral cartilage

I know that we have rather banged on about this (Taylor and Wedel 2013, Taylor 2014) but it remains true that bones alone can tell us almost nothing about how vertebrae articulated. Unless we incorporate intervertebral cartilage into our models, they can only mislead us. To their credit, Vidal et al. are aware of this — though you wouldn’t know it from the actual paper, whose single mention of cartilage is in respect of a hypothesised cartilaginous suprascapula. But buried away the supplementary information is this rather despairing paragraph:

Cartilaginous Neutral Pose (CNP): the term was coined by Taylor for “the pose found when intervertebral cartilage [that separates the centra of adjacent vertebrae] is included”. Since the amount of inter-vertebral space cannot be certainly known for most fossil vertebrate taxa, true CNP will likely remain unknown for most taxa or always based on estimates.

Now this is true, so far as it goes: it’s usually impossible to know how much cartilage there was, and what shape it took, as only very unusual preservational conditions give us this information. But I don’t think that lets us out from the duty of recognising how crucial that cartilage is. It’s not enough just to say “It’s too hard to measure” and assume it didn’t exist. We need to be saying “Here are the results if we assume zero-thickness cartilage, here’s what we get if we assume cartilage thickness equal to 5% centrum length, and here’s what we get if we assume 10%”.

I really don’t think it’s good enough in 2020 to say “We know there was some intervertebral cartilage, but since we don’t know exactly how much we’re going to assume there was none at all”.

The thing about incorporating cartilage into articulating models is that we would, quite possibly, get crazy results. I refer you to the disturbing figure 4 in my 2014 paper:

Figure 4. Effect of adding cartilage to the neutral pose of the neck of Diplodocus carnegii CM 84. Images of vertebra from Hatcher (1901:plate III). At the bottom, the vertebrae are composed in a horizontal posture. Superimposed, the same vertebrae are shown inclined by the additional extension angles indicated in Table 2.

I imagine that taking cartilage into account for the Spinophorosaurus reconstruction might have given rise to equally crazy “neutral” postures. I can see why Vidal et al. might have been reluctant to open that can of worms; but the thing is, it’s a can that really needs opening.

2. Sacrum orientation

As Vidal et al.’s figure 1A clearly shows, the sacrum of Spinophorosaurus is indeed wedge-shaped, with the anterior articular surface of the first sacral forming an angle of 20 degrees relative to the posterior articular surface of the last:

But I don’t see why it follows that “the coalesced sacrum is situated so that the posterior face of the last sacral centrum is sub-vertical. This makes the presacral series to slope dorsally and the tail to be subhorizontal (Figs. 1 and 4S)”. Vidal et al. justify this with the claim by saying:

Since a subhorizontal tail has been known to be present in the majority of known sauropods[27, 28, 29], the [osteologically induced curvature] of the tail of Spinophorosaurus is therefore compatible with this condition.

But those three numbered references are to Gilmore 1932, Coombs 1975 and Bakker 1968 — three venerable papers, all over fifty years old, dating from a period long before the current understanding of sauropod posture. What’s more, each of those three was about disproving the previously widespread assumption of tail-dragging in sauropods, but the wedged sacrum of Spinophorosaurus if anything suggests the opposite posture.

So my question is, given that the dorsal and caudal portions of the vertebral column are at some specific angle to each other, how do we decide which (if either) is horizontal, and which is inclined?

Three interpretations of the wedged sacrum of Spinophorosaurus, in right lateral view. In all three, the green line represents the trajectory of the dorsal column in the torso, and the red line that of the caudal column. At the top, the tail is horizontal (as favoured by Vidal et al. 2020) resulting in an inclined torso; at the bottom, the torso is horizontal, resulting in a dorsally inclined tail; in the middle, an intermediate posture shows both the torso and the tail slightly inclined.

I am not convinced that the evidence presented by Vidal et al. persuasively favours any of these possibilities over the others. (They restore the forequarters of Spinophorosaurus with a very vertical and ventrally positioned scapula in order to enable the forefeet to reach the ground; this may be correct or it may not, but it’s by no means certain — especially as the humeri are cross-scaled from a referred specimen and the radius, ulna and manus completely unknown.)

3. Distortion

Finally, we should mention the problem of distortion. This is not really a criticism of the paper, just a warning that sacra as preserved should not be taken as gospel. I have no statistics or even systematic observations to back up this assertion, but the impression I have, from having looked closely at quite a lot of sauropod vertebra, is the sacra are perhaps more prone to distortion than most vertebrae. So, for example, the very extreme almost 30-degree wedging that Vidal et al. observed in the sacrum of the Brachiosaurus altithorax holotype FMNH PR 25107 should perhaps not be taken at face value.

Now what?

Vidal el al. are obviously onto something. Sauropod sacra are screwy, and I’m glad they have drawn attention in a systematic way to something that had only been alluded to in passing previously, and often in a way that made it seems as though the wedging they describe was unique to a few special specimens. So it’s good that this paper is out there.

But we really do need to see it as only a beginning. Some of the things I want to see:

  • Taking cartilage into account. If this results in silly postures, we need to understand why that is the case, not just pretend the problem doesn’t exist.
  • Comparison of sauropod sacra with those of other animals — most important, extant animals whose actual posture we can observe. This might be able to tell us whether wedging really has the implications for posture that we’re assuming.
  • Better justification of the claim that the torso rather than the tail was inclined.
  • An emerging consensus on sauropod shoulder articulation, since this also bears on torso orientation. (I don’t really have a position on this, but I think Matt does.)
  • The digital Spinophorosaurus model used in this study. (The paper says “The digital fossils used to build the virtual skeleton are deposited and accessioned at the Museo Paleontológico de Elche” but there is no link, I can’t easily find them on the website and they really should be published alongside the paper.)

Anyway, this is a good beginning. Onward and upward!


  • Bakker, Robert T. 1968. The Superiority of Dinosaurs. Discovery 3:11–22.
  • Coombs, Walter P. 1975. Sauropod habits and habitats. Palaeogeography, Palaeoclimatology, Palaeoecology 17:1-33.
  • Gilmore, Charles W. 1932. On a newly mounted skeleton of Diplodocus in the United States National Museum. Proceedings of the United States National Museum 81:1-21.
  • Hatcher, John Bell. 1901. Diplodocus (Marsh): its osteology, taxonomy, and probable habits, with a restoration of the skeleton. Memoirs of the Carnegie Museum 1:1-63.
  • Osborn, Henry F. 1904. Manus, sacrum and caudals of Sauropoda. Bulletin of the American Museum of Natural History 20:181-190.
  • Taylor, Michael P. 2014. Quantifying the effect of intervertebral cartilage on neutral posture in the necks of sauropod dinosaurs. PeerJ 2:e712. doi:10.7717/peerj.712
  • Taylor, Michael P., and Mathew J. Wedel. 2013c. The effect of intervertebral cartilage on neutral posture and range of motion in the necks of sauropod dinosaurs. PLOS ONE 8(10):e78214. 17 pages. doi:10.1371/journal.pone.0078214
  • Taylor, Michael P., Mathew J. Wedel and Darren Naish. 2009. Head and neck posture in sauropod dinosaurs inferred from extant animals. Acta Palaeontologica Polonica 54(2):213-230.
  • Vidal, Daniel, P Mocho, A. Aberasturi, J. L. Sanz and F. Ortega. 2020. High browsing skeletal adaptations in Spinophorosaurus reveal an evolutionary innovation in sauropod dinosaurs. Scientific Reports 10(6638). Indispensible supplementary information at


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 BarosaurusDiplodocus 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 DiplodocusBarosaurus 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.



  • 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


If you were curious about the Wedel et al. presentation on the Snowmass Haplocanthosaurus at the 1st Palaeo Virtual Congress but didn’t attend the event, it is now preserved for posterity and freely available to the world as a PeerJ Preprint (as promised). Here’s the link.

I’ll have much more to say about this going forward, but for now here are slides 20 and 21 on the intervertebral joint spaces. This is obviously just the same vert cloned three times and articulated with itself. With the digital rearticulation of the reconstructed and retrodeformed caudal series still in progress, we cloned caudal 3, the only vertebra that preserves both sets of zygapophyses, to get a rough estimate of the sizes and shapes of the soft tissues that filled the intervertebral spaces and neural canal.

The reconstructed intervertebral discs (in blue) are very crude and diagrammatic. The reason I’m putting these particular slides up is to get the cited references out in the open on the blog, to start correcting the misapprehension that all non-mammalian amniotes have exclusively synovial intervertebral joints (see the discussion in the comments on this post). In the list below I’m including Banerji (1957), which is not cited in the presentation but which I did cite in that comment thread; it’s an important source and at least for now it is a free download. These refs are just the tip of a very big iceberg. One of my goals for 2019 is to do a series of posts reviewing the extensive literature on amphiarthrodial (fibrocartilaginous) intervertebral joints in living lepidosaurs and birds. Stay tuned!

And please go have a look at the presentation if you are at all interested or curious. As we said in the next to last slide, “this research is ongoing, and we welcome your input. If there are facts or hypotheses we haven’t considered but should, please let us know!”


Diplodocus goes digital

August 21, 2018

No time for a proper post, so here’s a screenshot from Amira of Diplodocus caudal MWC 8239 (the one you saw being CT scanned last post) about to be digitally hemisected. Trust me, you’ll want to click through for the big version. Many thanks to Thierra Nalley for the Amira help.

Further bulletins as time and opportunity allow.

John Yasmer, DO (right) and me getting ready to scan MWC 8239, a caudal vertebra of Diplodocus on loan from Dinosaur Journey, at Hemet Valley Imaging yesterday.

Alignment lasers – it’s always fun watching them flow over the bone as a specimen slides through the tube (for alignment purposes, obviously, not scanning – nobody’s in the room for that).

Lateral scout. I wonder, who will be the first to correctly identify the genus and species of the two stinkin’ mammals trailing the Diplo caudal?

A model we generated at the imaging center. This is just a cell phone photo of a single window on a big monitor. The actual model is much better, but I am in a brief temporal lacuna where I can’t screenshot it.

What am I doing with this thing? All will be revealed soon.

Preserved bits of the Snowmass Haplocanthosaurus, MWC 8028, with me for scale. Modified from Wedel (2009: fig. 10), but not much – MWC 8028 was about the same size as CM 879.

Let’s say you had a critter with weird neural canals and super-deeply-dished-in centrum-ends, and you wanted to digitally rearticulate the vertebrae and reconstruct the spinal cord and intervertebral cartilages, in a project that would bring together a bunch of arcane stuff that you’d been noodling about for years. Your process might include an imposing number of steps, and help from a LOT of people along the way:

1. Drive to Dinosaur Journey in Fruita, Colorado, to pick up the fossils and bring them back to SoCal. (Thank you Paige Wiren, John Foster, and Rebecca Hunt-Foster for an excuse to come to the Moab area, thank you Brian Engh for the awesome road trip, and thank you Julia McHugh for access to specimens and help packing them up!).

2. Take the fossils to the Hemet Valley Medical Center for CT scanning. (Thank you John Yasmer and team.)

3. Find a colleague who would help you generate 3D models from the CT scans. (Thank you Thierra Nalley.)

4. Talk it over with your university’s 3D vizualization team, who suggest a cunning plan: (Thank you Gary Wisser, Jeff Macalino, and Sunami Chun at WesternU.)

5. They print the best-preserved vertebra at 75% scale. (50% scale resin print shown here.)

6. You and a collaborator physically sculpt in the missing bits with some Super Sculpey. (Thank you Jessie Atterholt for sculpting, and thank you Jeremiah Scott for documenting the process.)

(7.) The 3D-viz team use their fancy optical scanner (basically a photogrammetry machine) to make:

  • a second-generation digital model (digital)
  • from the sculpted-over 3D print (physical)
  • of the first-generation digital model (digital)
  • made from the CT scans (digital)
  • of the original fossil material (physical).

(8.) With some copying, pasting, and retro-deforming, use that model of the restored vert as a template for restoring the rest of the vertebrae, stretching, mirroring, and otherwise hole-filling as needed. (Prelim 2D hand-drawn version of caudal 1 shown here.)

(9.) Test-articulate the restored vertebrae to see if and how they fit, and revise the models as necessary. (Low-fi speculative 2D version from January shown here.)

(10.) Once the model vertebrae are digitally rearticulated, model the negative spaces between the centra and inside the neural canals to reconstruct the intervertebral cartilages and spinal cord.

(11.) Push the university’s 3D printers to the limit attempting to fabricate an articulated vertebral series complete with cartilages and cord in different colors and possibly different materials, thereby making a third-generation physical object that embodies the original idea you had back in January.

(12.) Report your findings, publish the CT scans and 3D models (original and restored), let the world replicate or repudiate your results. And maaaybe: be mildly astonished if people care about the weird butt of the most-roadkilled specimen of the small obscure sauropod that has somehow become your regular dance partner.

We did number 6 yesterday, so just counting the arbitrarily-numbered steps (and ignoring the fact that 7-12 get progressively more complicated and time-consuming), we’re halfway done. Yay! I’ll keep you posted on how it goes from here.

The most complete caudal vertebra of the Snowmass Haplocanthosaurus (Foster and Wedel 2014) in right lateral view: specimen photo, CT scout, 3D model, 3D print at 50% scale. The photos of the specimen and the 3D print probably match the worst with the others, because they are subject to perspective distortions that the digital reconstructions are free from.

Here’s one nice thing about having a 3D print of a specimen that you’re working on: you can hand it to other anatomists and paleontologists and get their take on its weird features, and it’s small enough and light enough that you can bring it halfway across the country to show in person to an entirely different set of colleagues. For all that we hear about humans being a visual species, we are also a tactile one, and in my admittedly limited experience, grokking morphology by handling 3D printed fossils is almost as good as – and for big, heavy, fragile sauropod vertebrae, sometimes better than – handling the real thing.

Many thanks to Julia McHugh at Dinosaur Journey for access to the specimen, John Yasmer at the Hemet Valley Medical Center for CT scanning, Thierra Nalley at Western University of Health Sciences for help with segmenting and visualization in Amira, and Gary Wisser, WesternU’s 3D visualization specialist, for the sweet print. Further bulletins as events warrant.


Foster, J.R., and Wedel, M.J. 2014. Haplocanthosaurus (Saurischia: Sauropoda) from the lower Morrison Formation (Upper Jurassic) near Snowmass, Colorado. Volumina Jurassica 12(2): 197–210. DOI: 10.5604/17313708 .1130144

Peter Falkingham and Nick Gardner independently put me onto Sketchfab: a website that provides a way to view and navigate 3D models without needing to download any software beyond the browser that you’re already running.

So get yourself over to the live Xenoposeidon model! Verify for yourself that the laminae are as I described them, that the posterior margin of the neural arch really does grade into the posterior articular surface of the centum, etc. Really, this is worth ten times whatever set of illustrations I might have provided.

Truly, we are living in the future!

UPDATE, 23 November 2017: see also this beautiful 3d model of the skull of Triceratops horridus, photogrammetrised from images taken at the Museum National d’Histoire Naturelle, Paris, France, by Benoît Rogez; and the same creator’s Nanotyrannus lancensis model, also from MNHN photos. And, most astonishingly, his model of the whole MNHN palaeontology gallery!