This is the third post in a series on neural spine bifurcation in sauropods, inspired by Woodruff and Fowler (2012). In the first post, I looked at neural spine bifurcation in Morrison sauropod genera based on the classic monographic descriptions. In the second post, I showed that size is an unreliable criterion for assessing age and that serial variation can mimic ontogenetic change in sauropod cervicals. In this post I look at the evidence for ontogenetic changes in neural spine bifurcation presented by Woodruff and Fowler (2012). This posts builds on the last two, so please refer back to them as needed.

Another opening digression, on the OMNH baby sauropod material this time

Nearly all of the Morrison Formation material in the OMNH collections comes from Black Mesa in the Oklahoma panhandle. It was collected in the 1930s by WPA crews working under the direction of J. Willis Stovall. Adequate tools and training for fossil preparation were in short supply. A lot of the prep was done by unskilled laborers using hammers, chisels, pen-knives, and sandpaper (apologies if you have experience with fossil preparation and are now feeling a bit ill). Uncommonly for the Morrison, the bones are very similar in color to the rock matrix, and the prep guys sometimes didn’t realize that they were sanding through bone until they got through the cortex and  into the trabeculae. Consequently, a lot of interesting morphology on the OMNH Morrison material has been sanded right off, especially some of the more delicate processes on the vertebrae. This will become important later on.

Do the ‘ontogenetic’ series in Woodruff and Fowler (2012) actually show increasing bifurcation through development?

In the Materials and Methods, Woodruff and Fowler (2012:2) stated:

Study specimens comprise 38 cervical, eight dorsal, and two caudal vertebrae from 18 immature and one adult diplodocid (Diplodocus sp., Apatosaurus sp., and Barosaurus sp.), and two immature macronarians (both Camarasaurus sp.).

However, their Table 1 and Supplementary Information list only 15 specimens, not 18. Of the 15, one is probably not a diplodocid (SMA 0009 ‘Baby Toni’) — a fact that, oddly, the authors knew, as stated in the Supplementary Information.  Of the remaining 14 specimens, 11 are isolated vertebrae, so only three represent reasonably complete probably-diplodocoid series (MOR 592, AMNH 7535, and CM 555). From CM 555 they discuss only one vertebra, the C6; and AMNH 7535 is not mentioned at all outside of Table 1 and a passing mention the Supplementary Information, so the subadult data actually used in the paper consist of isolated vertebrae and one articulated series, MOR 592. (For the sake of comparison, in the first post on this topic I looked up 10 articulated series, only two of which–Diplodocus carnegii CM 84/94 and Camarasaurus lentus CM 11338–are even mentioned in Woodruff and Fowler [2012].)

In light of the previous post, on serial variation, the dangers of using isolated vertebrae should by now be apparent. Recall that even adult diplodocids are expected to have completely unsplit spines as far back as C5 (Apatosaurus) or C8 (Barosaurus) and as far forward as D7 (Apatosaurus) or D6 (Barosaurus), and only partially split spines in the adjacent positions. Furthermore, size is a notoriously unreliable criterion of age; MOR 790 8-10-96-204 from Figure 2 in Woodruff and Fowler (2012) also appears in their Figure 3 as the second-smallest vertebra in this ‘ontogenetic’ series, despite most likely coming from a well-fused adult approximately the same size as the D. carnegii individual that represents the end of the series. So without any evidence other than sheer size (if that size overlaps with the adult size range) and degree of neural spine bifurcation (which cannot help but overlap with the adult range, since the adult range encompasses all possible states), simply picking small vertebrae with unsplit spines and calling them juvenile is at best circular and at worst completely wrong–as in the case of MOR 790 8-10-96-204 examined in the last post.

Unfortunately it is not possible to tell what criteria Woodruff and Fowler (2012) used to infer age in their specimens, because they don’t say. Neural arch fusion is discussed in general terms in the Supplementary Information, but in the text and in the figures everything is discussed simply in terms of size. For example:

In the next largest specimen (MOR 790 7-26-96-89, vertebral arch 9.9 cm high), the neural spine is relatively longer still and widens at the apex…

The Supplementary Information provides more evidence that Woodruff and Fowler (2012) did not consider the confounding effects of size, serial position, and ontogenetic stage. In the section on the Mother’s Day Quarry in the Supplementary Information, they wrote:

Because of this size distribution it is not surprising that there are also different ontogenetic stages present which result in cervical centrum lengths varying between 12 and 30 cm.

Now, there may be different ontogenetic stages present in the quarry, and the cervicals in the quarry may vary in length by a factor of 2.5, but the latter does not demonstrate the former. In D. carnegii CM 84/94 the longest postaxial cervical (C14, 642 mm) is 2.6 times the length of the shortest (C3, 243 mm; data from Hatcher 1901). The size range reported as evidence of multiple ontogenetic stages by Woodruff and Fowler (2012) turns out to be slightly less than that expected in a single individual.

With that in mind, let’s look at each of the putative ontogenetic sequences in Woodruff and Fowler (2012):

Anterior cervical vertebrae

Woodruff and Fowler (2012:fig. 3)

The proposed ontogenetic series used by Woodruff and Fowler (2012) for anterior cervical vertebrae consists of:

• CMC VP7944, an isolated ?Diplodocus vertebra from the Mother’s Day site, which is described in the text but not pictured;
• MOR 790 7-30-96-132, an isolated vertebra from the same site;
• MOR 790 8-10-96-204, another isolated vertebra from the same site;
• MOR 592, from a partial cervical series of a subadult Diplodocus but with the serial position unspecified;
• ANS 21122, C6 of Suuwassea (included in Fig. 3, but not discussed as evidence in the accompanying text)
• CM 555, C6 of a nearly complete (C2-C14) cervical series of a subadult Apatosaurus;
• CM 84/94, C7 of Diplodocus carnegii

CMC VP7944 is not pictured, but from the description in the text it’s perfectly possible that it represents a C3, C4, or C5, all of which have undivided spines even in adult diplodocids. It therefore contributes no information: the hypothesis that the spine is undivided because of ontogeny is not yet demonstrated, and the hypothesis that the spine is undivided because of serial position is not yet falsified.

MOR 790 7-30-96-132 is shown only from the front, so the centrum proportions and the shape of the neural spine cannot be assessed. The neural arch appears to be fused, but the cervical ribs are not. Again, we cannot rule out the possibility that it comes from an very anterior cervical and therefore its undivided spine could be an artifact of its serial position. It therefore contributes no information on possible ontogenetic changes in neural spine bifurcation.

As shown in the previous post, MOR 790 8-10-96-204 is probably a C4 or C5 of an adult or near-adult Diplodocus about the same size as or only slightly smaller than D. carnegii CM 84/94. It is small and has an undivded spine because it is an anterior cervical, not because it is from a juvenile. It therefore contributes no support to the ontogenetic bifurcation hypothesis.

The pictured vertebra of MOR 592 has a shallow notch in the tip of the spine, which is expected in C6 in Apatosaurus and Diplodocus and in C9 and C10 in Barosaurus. The serial position of the vertebra is not stated in the paper, but about half of the anterior cervicals even in an adult diplodocid are expected to have unsplit or shallowly split spines based on serial position alone. Based on the evidence presented, we cannot rule out the possibility that the shallow cleft in the pictured vertebra is an artifact of serial position rather than ontogeny. It therefore contributes no support to the ontogenetic bifurcation hypothesis.

ANS 21122 has an incompletely divided neural spine, which is in fact expected for the sixth cervical in adult diplodocids as shown by A. parvus CM 563/UWGM (in which C6 is missing but C5 has an unsplit spine and C7 a deeply bifid spine) and D. carnegii CM 84/94 (in which C6 is also shallowly bifid). A. ajax NMST-PV 20375 has a wider split in the spine of C6, but the exact point of splitting appears to vary by a position or two among diplodocids. The hypothesis that the spine of ANS 21122 C6 is already as split as it would ever have gotten cannot be falsified on the basis of the available evidence.

CM 555 C6: see the previous paragraph. Note that in ANS 21122 the neural arch and cervical ribs are fused in C6, and in C6 of CM 555 they are not.

CM 84/94 C7 has a deeply split spine, but this expected at that position. C6 of the same series has a much more shallow cleft, and C5 would be predicted to have no cleft at all (recall from the first post that the neural spines of C3-C5 of this specimen are sculptures). So any trend toward increasing bifurcation is highly dependent on serial position; if serial position cannot be specified then it is not possible to say anything useful about the degree of bifurcation in a given vertebra.

Summary. CMC VP7944 and MOR 790 7-30-96-132 could be very anterior vertebrae, C3-C5, in which bifurcation is not expected even in adults. Since they are isolated elements, that hypothesis is very difficult to falsify. MOR 790 8-10-96-204 is almost certainly a C4 or C5 of an adult or near-adult Diplodocus. ANS 21122 and CM 555 C6 are incompletely divided, as expected for vertebrae in that position even in adults. CM 84/94 has a shallowly divided spine in C6 and more deeply bifid spines from C7 onward, just like CM 555.

Verdict: no ontogenetic change has been demonstrated.

Posterior cervical vertebrae

Woodruff and Fowler (2012:Fig. 4A)

The proposed ontogenetic series includes:

• OMNH 1267 and 1270
• MOR 790 7-26-96-89
• MOR 592
• CM 84/94

OMNH 1267 and 1270 are isolated neural arches of baby sauropods from the Black Mesa quarries. OMNH 1267 does not appear to be bifurcated, but it has a very low neural spine and it was probably sanded during preparation, so who knows what might have been lost. OMNH 1270 actually shows a bifurcation–Woodruff and Fowler (2012:3) describe it as having “a small excavated area”–but again it is not clear that the spines are as intact now as they were in life. More seriously,  since these are isolated elements (you can all join in with the refrain) their serial position cannot be determined with any accuracy, and therefore they are not much use in determining ontogenetic change. Although they are anteroposteriorly short, that does not necessarily make them posterior cervicals. The cervical vertebrae of all sauropods start out proportionally shorter and broader than they end up (Wedel et al. 2000:368-369), and the possibility that these are actually from anterior cervicals–not all of which are expected to have bifurcations–is difficult to rule out.

The other three vertebrae in the series have deeply bifurcated spines. In the text, Woodruff and Fowler (2012:3) make the case that the bifurcation in MOR 592 is deeper than in the preceding vertebra, MOR 790 7-26-96-89. However, the proportions of the two vertebrae are very different, suggesting that they are from different serial positions, and the centrum of MOR 790 7-26-96-89 is actually larger in diameter than that of the representative vertebra from MOR 592. So unless centrum size decreased through ontogeny, these vertebrae are not comparable. As usual, we don’t know where in the neck the isolated MOR 790 vertebra belongs, and we only see it in anterior view. Nothing presented in the paper rules out possibility that is actually an anterior cervical, and in fact the very low neural spines suggest that that is the case.

Allowing for lateral crushing, the vertebra from MOR 592 (again, we are not told which one it is) looks very similar to the D. carnegii CM 84/94 vertebra (C15–again, I had to look it up in Hatcher), and is probably from a similar position in the neck. In comparing the two, Woodruff and Fowler (2012:4) say that in CM 84/94, “the bifurcated area has broadened considerably”, but this clearly an illusion caused by the lateral compression of the MOR 592 vertebra — its centrum is also only half as wide proportionally as in the CM 84/94 vertebra.

Summary. The OMNH vertebrae are of unknown serial position and probably lost at least some  surface bone during preparation, so their original degree of bifurcation is hard to determine. The other three vertebrae in the series all have deeply bifid spines, but they are out of order by centrum size, MOR 790 7-26-96-89 might be an anterior cervical based on its low neural spines, and the “broadening” of the trough between MOR 792 and CM 84/94 is an artifact of crushing.

Verdict: no ontogenetic change has been demonstrated.

Anterior dorsal vertebrae

Woodruff and Fowler (2012:Fig. 5A)

The ontogenetic series here consists of:

• MOR 790 7-17-96-45
• MOR 592
• CM 84/94

As usual, the serial positions of the MOR 592 and CM 84/94 vertebrae are presumably known but not stated in the paper. The D. carnegii CM 84/94 vertebra is D4. Comparisons to the MOR 592 vertebra are not helped by the fact that it is shown in oblique posterior view. Nevertheless, the two vertebrae are very similar and, based on the plates in Hatcher (1901), the MOR 592 vertebra is most likely a D4 or D5 of Diplodocus. The spines in the larger two vertebrae are equally bifurcated, so the inference of ontogenetic increase in bifurcation rests on the smallest of the three vertebrae, MOR 790 7-17-96-45.

MOR 790 7-17-96-45 is an isolated unfused neural arch, clearly from a juvenile. Its serial position is hard to determine, but it is probably not from as far back as D4 or D5 because it appears to lack a hypantrum and shows no sign of the parapophyses, which migrate up onto the neural arch through the cervico-dorsal transition. The element is only figured in anterior view, so it is hard to tell how long it is proportionally. Still, based on the single photo in the paper (which is helpfully shown at larger scale in Fig. 5B), it seems to be reasonably long, with the prezygapophyses, transverse processes, neural spines, and postzygapophyses well separated from anterior to posterior. In fact, I see no strong evidence that it is a dorsal neural arch at all–the arch of a posterior cervical would look the same in anterior view.

Given that MOR 7-17-96-45 lacks a hypantrum and parapophyses, it is not directly comparable to the two larger vertebrae. Although we cannot determine its position in the presacral series, its spine is shallowly bifurcated, to about half the distince from the metapophyses to the postzygapophyses. In Apatosaurus louisae CM 3018, the notch in D3 is about equally deep, and in C15 it is only slightly deeper, still ending above the level of postzygapophyses. So there is some variation in the depth of the bifurcation in the posterior cervicals and anterior dorsals in the North American diplodocids. Without knowing the precise serial position of MOR 7-17-96-45, it is difficult to derive inferences about the ontogeny of neural spine bifurcation.

Diplodocid anterior dorsal vertebrae. Left and right, dorsal vertebrae 3 and 4 of adult Apatosaurus louisae holotype CM 3018, from Gilmore (1936: plate XXV). Center, juvenile neural arch MOR 7-17-96-45, modified from Woodruff and Fowler (2012: fig. 5B), corrected for shearing and scaled up.

What this element does conclusively demonstrate is that the neural arches of posterior cervicals or anterior dorsals in even small, unfused juvenile diplodocids were in fact bifurcated to to a degree intermediate between  D3 and D4 in the large adult Apatosaurus louisae CM3018 — in fact, so far as neural cleft depth is concerned, it makes rather a nice intermediate between them.  (It differs in other respects, most notable that it is proportionally broad, lacks a hypantrum and parapophyses, etc.)

Summary. The two larger specimens in the ‘ontogenetic series’ are from similar serial positions and show the same degree of bifurcation. MOR 7-17-96-45 is from a more anterior position, based on its lack of hypantrum and parapophyses.  Although it is a juvenile, its degree of bifurcation is similar to that of anterior dorsal vertebrae in adult Apatosaurus (and that of C15 in A. louisae CM 3018, if MOR 7-17-96-45 is, in fact, a cervical).

Verdict: no ontogenetic change has been demonstrated.

Posterior dorsal vertebrae

Woodruff and Fowler (2012:Fig. 6A)

The ontogenetic series consists of:

• OMNH 1261
• MOR 592
• CM 84/94

The D. carnegii CM 84/94 vertebra is D6, and based on its almost identical morphology the MOR 592 vertebra is probably from the same serial position. They show equivalent degrees of bifurcation.

OMNH 1261 is another isolated juvenile neural arch. The portion of the spine that remains is unbifurcated. However, the spine is very short and it is possible that some material is missing from the tip. More importantly, the last 3-4 dorsals in Apatosaurus, Diplodocus, and Barosaurus typically have extremely shallow notches in the neural spines or no notches at all. If OMNH 1261 is a very posterior dorsal, it would not be expected to show a notch even when fully mature.

Verdict: no ontogenetic change has been demonstrated.

Woodruff and Fowler (2012:Fig. 7)

Caudal vertebrae

The ontogenetic series here consists of:

• MOR 592
• CM 84/94

The first thing to note is that the ‘bifurcation’ in MOR 592 is at right angles to that in the proximal caudals of D. carnegiiCM 84/94, so the one can hardly be antecedent to the other.

More importantly, antero-posterior ‘bifurcations’ like that in MOR 592 are occasionally seen in the caudal vertebrae of adult sauropods. Below are two examples, caudals 7 and 8 of A. parvus CM 563/UWGM 15556. In other words, in this character MOR 592 already displays adult morphology.

Verdict: no ontogenetic change has been demonstrated.

A. parvus CM 563/UWGM 15556 caudals 8 and 7 in right lateral view, from Gilmore (1936:pl.. 33)

Camarasaurus

The ontogenetic series here consists of:

• OMNH 1417
• AMNH 5761

OMNH 1417 is an isolated cervical neural spine, and the pictured vertebra of Camarasaurus supremus AMNH 5761 is a posterior cervical. In C. grandis and C. lewisi, all of the cervical vertebrae eventually develop at least a shallow notch in the tip of the neural spine, but as shown in the previous post there seems to be some variation between Camarasaurus species, and, likely, between individuals. In the absence of information about its serial position and the species to which it belonged, the lack of bifurcation in OMNH 1417 is uninformative; it could belong to an anterior cervical of C. supremus that would not be expected to develop a bifurcation.

Verdict: no ontogenetic change has been demonstrated. There is evidence that neural spine bifurcation developed ontogenetically in Camarasaurus, but it comes from the juvenile C. lentus CM 11338, described by Gilmore (1925), and the geriatric C. lewisi, described by McIntosh, Miller et al. (1996)–see the first post in this series for discussion.

Conclusions

The ‘ontogenetic’ series of Woodruff and Fowler (2012) are not really ontogenetic series. In all of the diplodocid presacral vertebrae and in Camarasaurus, the smallest elements in the series are isolated vertebrae or neural arches for which the serial position is almost impossible to determine (and for the reader, completely impossible given the limited information in the paper) and even the taxonomic identifications are suspect (e.g., the OMNH material–how one reliably distinguishes the Apatosaurus and Camarasaurus neural arches is beyond me). The larger vertebrae in the presacral series are all compromised in various ways: one includes an adult masquerading as a juvenile (MOR 790 8-10-96-204 in the anterior cervicals), one is out of order by centrum size (MOR 790 7-26-96-89 and MOR 592 in the posterior cervicals), and two show no change in degree of bifurcation from the middle of the series to the upper end (MOR 592 and CM 84/94 in the anterior and posterior dorsals). The shallow longitudinal bifurcation in the MOR 592 caudal vertebra is similar to those found in caudal vertebrae of adult diplodocids, and is not antecedent to the transverse bifurcations discussed in the rest of the paper.

Crucially, when information on size and serial position is taken into account, none of the ‘ontogenetic series’ in the paper show any convincing evidence that neural spine bifurcation increases over ontogeny. The best evidence that bifurcation does increase over ontogeny comes from Camarasaurus, specifically the juvenile C. lentus CM 11338 described by Gilmore (1925) and geriatric C. lewisi BYU 9047 described by McIntosh et al. (1996), it was already recognized prior to Woodruff and Fowler (2012), and it has not caused any taxonomic confusion.

There is an asymmetry of interference here. To call into question the conclusions of Woodruff and Fowler (2012), all one has to do is show that the evidence could be explained by serial, intraspecific, or interspecific variation, taphonomy, damage during preparation, and so on. But to demonstrate that bifurcation develops over ontogeny, one has to falsify all of the competing hypotheses. I know of only one way to do that: find a presacral vertebral column that is (1) articulated, (2) from an individual that is clearly juvenile based on criteria other than size and degree of bifurcation, which (3) can be confidently referred to one of the known genera, and then show that it has unbifurcated spines in the same serial positions where adult vertebrae have bifurcated spines. Isolated vertebrae are not enough, bones from non-juveniles are not enough, and juvenile bones that might pertain to new taxa are not enough. It may be that this is not yet possible because the necessary fossils just haven’t been found yet. I am not suggesting that we stop doing science, or that the ontogenetic hypothesis of neural spine bifurcation is unreasonable. It’s perfectly possible that it’s true (though MOR 7-17-96-45 ironically suggests otherwise). But it’s not yet been demonstrated, at least for diplodocids, and to the extent that the taxonomic hypotheses of Woodruff and Fowler (2012) rely on an ontogenetic increase in bifurcation in diplodocids, they are suspect. That will be the subject of the next post.

The rest of the series

Links to all of the posts in this series:

• Part 1: what we knew a month ago
• Part 2: why serial position matters
• Part 3: the evidence from ontogenetic series
• Part 4: is Suuwassea a juvenile of a known diplodocid?
• Part 5: is Haplocanthosaurus a juvenile of a known diplodocid?
• Part 6: more reasons why Haplocanthosaurus is not a juvenile of a known diplodocid

and the post that started it all:

• Changes through growth in sauropods and ornithopods

References

• Gilmore, C.W. 1925. A nearly complete articulated skeleton of Camarasaurus, a saurischian dinosaur from the Dinosaur National Monument. Memoirs of the Carnegie Museum 10:347-384.
• Gilmore, C.W. 1936. Osteology of Apatosaurus with special reference to specimens in the Carnegie Museum. Memoirs of the Carnegie Museum 11:175-300.
• Hatcher, J.B. 1901. Diplodocus (Marsh): its osteology, taxonomy, and probable habits, with a restoration of the skeleton. Memoirs of the Carnegie Museum 1:1-63.
• McIntosh, J.S., Miller, W.E., Stadtman, K.L., and Gillette, D.D. 1996. The osteology of Camarasaurus lewisi (Jensen, 1988). BYU Geology Studies 41:73-115.
• Wedel, M.J., Cifelli, R.L., and Sanders, R.K. 2000. Osteology, paleobiology, and relationships of the sauropod dinosaur Sauroposeidon. Acta Palaeontologica Polonica 45(4):343-388.
• Woodruff, D.C, and Fowler, D.W. 2012. Ontogenetic influence on neural spine bifurcation in Diplodocoidea (Dinosauria: Sauropoda): a critical phylogenetic character. Journal of Morphology, online ahead of print.