Oviraptorids and Cranial Morphometrics 2: Reloaded

Hot on the heels of my post on how cranial morphometric analyses of theropods end up excluding oviraptorids from them, a new paper ups the ante. (Of course, I do not mean that these papers actively exclude oviraptorids, but rather have to run the analyses without them in addition to with them, because they produce such a bizarre effect.)

Bhart-Anjan S. Bhullar and colleagues have taken several steps further than Brusatte et al. (2011) in which a PCA analysis of cranial landmarks was undertaken, by not only adding in a range of extant and extinct avialaeans, but also plotting juvenile or perinate skulls into the matrix. Here, the purpose is to see if juvenile skulls influence the plot against phylogeny, and to assess if there are any patterns of ontogenetic trajectories across that phylogeny. The results affirm that, yes, juveniles do not tend to plot with their adult counter parts, but moreover that avian and juvenile skulls tend to be skewed to one side in the plots.

Bhullar et al. (2012) demonstrate heterochronic ontogenetic trajectory in theropod phylogeny. The arrow in both plots originated with Euparkeria, but terminates at Gallus. On the left is the original figure from the paper, on the right a modified form from the supplemental online data.

Observed trends show that one can track phylogeny across the plot, and thus that progressive transformation of the cranium can be observed. By overlaying the morphometric landmark analysis on the plot, it is observable how the rostrum of adult taxa slow becomes shorter, while the postorbital region becomes larger. As the authors observe, the braincase becomes re-oriented somewhat vertically and flexed, so that the cerebrum overlies the cerebellum, while the medulla still projects posteriorly through the foramen magnum. The position of the brain can be seen in the figure above (at left) in yellow.

Bhullar et al.’s trajectory is a simple one, flowing “around the horn” as it were from Euparkeria capensis, up to basal saurischian Herrerasaurus ischigualastensis, and then on to basal coelurosaurs like Guanlong wucaii, and finally into avialaeans. But around that horn, following phylogeny more strictly, we land at basal theropod Coelophysis bauri, the ceratosaur Ceratosaurus nasicornis, then Gorgosaurus libratus (for full-on tyrannosaur action, just in case Guanlong isn’t really), and finally into basal maniraptoriforms like the ornithomimosaur Garudimimus brevipes, and then to Velociraptor mongoliensis (which I’m sure needs no introduction). This whole scenario, with or without a giant tyrannosaur in the middle, loops about a bit in morphospace. But things get stranger when we consider basal maniraptorans, from an offshoot from the Velociraptor node down to oviraptorids or up and back into Guanlong-territory of morphospace to meet Byronosaurus jaffei. The cranial morphology of maniraptoran despite heavy similarities and debates on the nature of whom, exactly, is closer to the origin of Aves.

Perhaps at this point not surprisingly, the authors also considered the effects of excluding oviraptorosaurs; unlike the analysis of Brusatte et al. (2011), they only included Incisivosaurus gauthieri, Citipati osmolskae, and a perinate skull that hasn’t been described referred to the latter. But the travesty of the thing is that they had to see if oviraptorosaurs’ departure affects the outcome of the analysis, as if the only purpose an oviraptorosaur has in an analysis is to be excluded!

(I am, of course, joking.)

Plots from Brusatte et al., 2011 (left) and Bhullar et al., 2012 (right). Plots are shown having been run with oviraptorids (top) and without (bottom). Deeper red fields mark out “herbivorous” taxa, while lighter red circles mark out giant-bodied theropods. These tend to plot consistently.

In the plot above, you will notice the high provincialization of groups of theropods, with limited overlap. An original concept for mucking aboutwith this figure would have nested all the subgroups of Theropoda that were found, but these all take a jumble when around the entirety of non maniraptoran theropods, so it seemed easier to just follow the trend and lay out “herbivorous” and “giant theropod” subgroups. The presence or absence of oviraptorosaurs contributes heavily to the “herbivore” plot, while “giant theropods” remain consistent and together … like chums.

Bhullar et al. differ somewhat from Brusatte et al., especially as in the latter, the authors chose to perform two sets of analyses with a maximum of 32 points on their landmark analysis, while the former went all the way to 45. Presumably, no taxon was left out of the Bhullar analysis. This difference is probably what results in the different array of similar taxa found between the two. Despite this, they are remarkably similar, and one can easily find the “around the horn” C-shaped trajectory towards Avialae in the Brusatte et al. analysis.

One of the more significant effects of the analysis is the result that ontogenetic trajectories in this plot follow a juvenile to adult ontogeny as we move from positive to negative along the PC1 axis in Bhullar et al. (which compares to PC2 in Brusatte et al.; seriously, just rotate the one analysis around, and they align quite nicely). This led to the tantalizing title of the paper, “Birds have paedomorphic dinosaur skulls.” And by this, it means that birds conserved the early ontogenetic shape of the skulls of earlier theropod dinosaurs well into their adulthood, to the point that by the time we get to birds like Confuciusornis sanctus, adult bird skulls match up quite well with juvenile Alligator mississippiensis.

As the figure says.Following the first image above, Bhullar et al. project that the development of ontogeny follows a phylogenetic trend, and when one considers oviraptorosaurs, we find that this trend diverges somewhere around Avialae. In fact, the trend plots oviraptorosaurs closer to birds than are dromaeosaurids or troodontids, both of which cluster in the “crazy experimentation” and loop-de-loop of “the horn.” Let’s look at that again:

Ontogenetic trajectory and paedomorphic trends in Theropoda. Modified after Bhullar et al., 2012.

Here oviraptorosaurs are arrayed more or less in vertical morphospace, meaning that their transformation through ontogeny (where the perinate Citiapti osmolskae skull is closer to Incisivosaurus gauthieri than to the adult way at the bottom) is largely off the ontogenetic trend, and instead along the PC2 axis, which largely corresponds to snout size. Meaning … adult oviraptorids are more like one another as they age than they are other taxa even closely related (ornithomimosaurs, dromaeosaurids, troodontids, or even therizinosauroids, which were also included in the analysis).

This actually suggests some things when it comes to diet. First, the shapes of the skull bones are largely consistent through oviraptorid ontogeny, such that muscle positions can be predicted to stay in largely the same place; second, that the jaws remain mechanically similar; and third, that diet through age is a matter of scale. As the oviraptorid grows, it may have developed different tastes, generalized more or less, but the diet remained largely the same, and that only the size of what could be fit in the mouth differed. It remains to be seen if this is true.

But the oddest thing about this paper is something buried in the supplemental online information (warning! pdf):

From Bhullar et al., 2012 (SOM).

Eoraptor lunensis, Herrerasaurus ischigualastensis and Coelophysis bauri are by no means the same dinosaur. In some phylogenies, they are not even that closely related to one another: Coelophysis bauri is a basal theropod, while both Herrerasaurus ischigualastensis and Eoraptor lunensis may lie outside Theropoda sensu stricto, or at the least may be stem-saurischians. Despite this, the comparison of Eoraptor to the juvenile Coelophysis is so uncanny one must wonder; further, that Herrerasaurus, an intermediate in some phylogenies between the two, looks nothing like either. The suggestion then is that Eoraptor lunensis represents a paedomorphic adult, or is in fact a juvenile. While the holotype skull this is based on is, in fact, a subadult specimen, it has not been determined what degree of changes would occur at adulthood.

So, there you have it. Three papers in two years, whose statistical analysis has had to consider the morphometric effects of cranial variation in theropods, and two of which had to pretend oviraptorosaurs don’t exist to test their results. Oviraptorosaurs need a massive treatment, just at the least in a variety of morphometric directions but also and especially in dynamic modeling of jaw function and bite force estimation with respect to reconstructions that consider certain oddities about their jaws; as it is, most analyses consider only the margins of the jaws when estimating biting, but oviraptorids (at least, just them) used a fair bit more than just the edges of their jaws. More on that later.

Update (7/23/12): Mickey Mortimer noted that the plotting of landmarks onto the skulls of the specimens specified by Bhullar et al. did not seem to be in keeping with their wireframes. There, Mickey Mortimer suggested that the Confuciusornis skull was dramatically different from the map provided by Bhullar, and so revised the map and overlayed it onto the alligator skull as provided by the authors. They do not line up. In the comments below, Bhullar responds (as well as at Mickey’s blog) that the figure overlaying a Confuciusornis onto an Alligator skull does not reflect a direct overlay of the original wireframes, but PCA-corrected maps using a program (tpsRelw, for thin-plate spline Relative warps, available here — for free! — from F. James Rolff and SUNY Stony Brook) which orients the perspective of the frame when plotting the wireframes for thin-plate spline analysis. To be exact, in virtually all figures in this post and the paper, TPS has modified the original wireframes to produce warped frames, so that they are more concordant with one another. Bhullar affirms that his original landmark images are identical to Mickey’s.

Bhullar, B.-A. S., Marugán-Lobón, J., Racimo, F., Bever, G. S., Rowe, T. B. & Norell, M. A. 2012. Birds have paedomorphic skulls. Nature 487:223-226.
Brusatte, S. L., Sakamoto, M., Montanari, S. & Harcourt Smith, W. E. H. 2012. The evolution of cranial form and function in theropod dinosaurs: Insights from geometric morphometrics. Journal of Evolutionary Biology 25:365-377.

About these ads
This entry was posted in Morphometrics, Paleobiology, Paleontology, Science Reporting and tagged , , , . Bookmark the permalink.

8 Responses to Oviraptorids and Cranial Morphometrics 2: Reloaded

  1. Mickey Mortimer says:

    I’m skeptical of some aspects of this paper. How did they get a shape for Yixianornis when the skull is flattened in ventral view? Does the resemblence between a juvenile croc and Confuciusornis have much meaning considering the latter has a skull which is very unusual for basal birds, with its robust construction, highly reduced antorbital area, etc.?

    • It’s quite likely that the authors estimated the shapes for several taxa with fragementary skulls. It is not explained. The process in Brusatte et al. at least moderated this by running two sets of analyses because of the fragmentary nature of the material. Despite this, the two analyses plot most of their taxa in common in the same regions. Thus, I think that even if some of their taxa are more questionable than others, this shouldn’t influence the overall plot of many of the subsidiary analyses assessing paedomorphic trends, the most important of which use complete skulls or composites of them.

  2. Bhart-Anjan Bhullar says:

    Mickey, quick responses: “Yixianornis,” as noted in the supplement, was a reconstruction based on it and other Yanornis/Yixianornis-type birds. It wasn’t great, and it wasn’t in the original study — added based on a reviewer’s comments. The results are very robust and stand up to the elimination of taxa quite well. Confuciusornis was just one example; we could as easily have used one of the enantiornithines, for instance, and the skulls would have been almost as similar. The early avialans in general cluster with the juvenile/embryonic archosaurs. Moreover, apart from the premaxilla, which is a very separate developmental module derived from a different embryonic primordium and patterned by different genes, Confuciusornis does _not_ have an especially reduced antorbital region compared to other early avialans. Rather, it’s square in the middle, shorter than Archaeopteryx and similar to a juvenile enantiornithine. Modern birds have shorter antorbital regions still.

    • Thanks for the reply. I’ve responded in greater depth at my blog, where I’ve found even worse problems, as the wireframes don’t even roughly match the skulls they are based on. See http://theropoddatabase.blogspot.com/2012/07/does-confuciusornis-really-have-skull.html for more details.

      • Bhart-Anjan Bhullar says:

        Just reposting this — I made a stupid omission from the captions. The wireframes from the paper were generated in the program TPSRelW by using the ‘camera’ option and dropping a pointer onto the points representing different taxa. Now I realize I should have specified that the program plots the position using only PCs 1 and 2 when those two axes are chosen ( they explain the vast majority of variation in the data), because my Confuciusornis landmarks are quite similar to yours, as are my Alligator embryo landmarks. However, the PCA plot is ‘correct’ by testing in both MorphoJ and TPS, so the relations we show, and the results of the minimum-spanning-tree etc. shouldn’t be affecting by this. The one difference I noticed with the Alligator data is that for the posterodorsal corner of the lateral temporal fenestra (landmark 36), I realized I was using the edge of the squamosal to determine the position of this landmark. I should have reworded it as such, but you can see here that there is a disconnect between the squamosal position and the position on the quadratojugal.

        So instead of precise wireframes of the specimens, they instead represent how the program sees them as plotted just in the space of PC 1 and PC 2. I should have specified this in the figure captions. Does that make some sense?

Leave a Reply

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

WordPress.com Logo

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

Twitter picture

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

Facebook photo

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

Google+ photo

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

Connecting to %s