The Aerosteon saga, Part 2: Overinflation and undercitation

October 5, 2008

In the last post I introduced Aerosteon, which has been touted as providing the first solid evidence for bird-like air sacs in non-avian dinosaurs, and I explained a little about how we know what we think we know about dinosaur air sacs. Aerosteon is legitimately cool and does show all of the features that Sereno et al. (2008) claim, and then some (see below). But the rest of the paper bugs me for several reasons: bad anatomy, bad phylogenetic reasoning, unfair criticism of other workers, and misleading citation of previous work. I’ll document every bit of this, but don’t forget that you can download almost all of the relevant papers for free. Don’t take my word for it, and don’t assume that any forthcoming rebuttals are accurate. Read the papers for yourself and make up your own mind.

Bad Anatomy #1: All cervical all the time (again)

In Part 1 we saw how Pat O’Connor and Leon Claessens (2005) cleared up about a century’s worth of confusion about how pneumatic bones map to specific parts of the avian respiratory system. In short, the cervical air sacs do not ever pneumatize any part of the skeleton past the mid-thoracic vertebrae, and the posterior thoracic, sacral, and caudal vertebrae are only pneumatized by diverticula of abdominal air sacs. Many non-avian dinosaurs have pneumatized vertebrae in the neck, thorax, sacrum, and tail, which are the diagnostic skeletal traces of air sacs both anterior and posterior to the lung, which are crucial for flow-through breathing like that of birds. O’Connor and Claessens never claimed that non-avian dinosaurs must have had flow-through breathing–that is unknowable for now–but they made a good case that most theropods, at least, had the requisite air sacs.

If there’s already great support for bird-like air sacs in non-avian theropods and has been for years, that would diminish the perceived importance of Aerosteon. Which may explain why Sereno et al. try to discredit O’Connor’s and Claessens’s work generally, and to discredit the importance of vertebral pneumaticity specifically. To do that, they resurrect the old canard about the cervical air sac pneumatizing the entire vertebral column.

Well, this is really something. I mean, O’Connor and Claessens (2005) looked at literally hundreds of birds and never found that once. What amazing new evidence do Sereno et al. bring to the table? Let’s find out:

“Second, cervical air sacs have been observed extending to the posterior end of the vertebral column in birds. Several authors have described cervical air sacs extending posteriorly beyond the abdominal air sacs in the ostrich (Struthio camelus) [21,36].” (p. 3)

So the case turns out to rest on refs 21 and 26, which are (thumbs through paper) McLelland (1989) [21] and Bezuidenhout et al. (1999) [36]. You’ll remember our old friend McLelland, who actually wrote:

“What can be stated with certainty is that in birds generally the cervical air sac aerates the cervical and thoracic vertebrae (Fig. 5. 22) and the vertebral ribs; the clavicular air sac aerates the sternum, sternal ribs, pectoral girdle and humerus (Fig. 5. 23); and the abdominal air sac aerates the synsacrum, pelvis and femur.” (pp. 271-272)

McLelland also mentions:

“In the Ostrich (Struthio camelus) a paired diverticulum arises from the vertebral diverticulum and extends through the peritoneal cavity beyond the abdominal air sac.” (p. 260)

The peritoneal cavity is the space containing the viscera, so these peritoneal diverticula are not going to the vertebrae, they’re going to the cloaca. Their existence provides no support for the contention of Sereno et al. that the entire vertebral column can be pneumatized by cervical air sacs, which is explicitly contradicted by the earlier McLelland quote (not to mention by the gigantic pile o’ dead birds cut up by O’Connor and Claessens).

Let’s go on to Bezuidenhout et al., who wrote:

“The caudal vertebral diverticula [of the cervical air sacs] extend caudally along the vertebral column, invading the spaces between the vertebrae and the vertebral ribs, and between the vertebrae and the oesophagus dorsally to the lungs, up to the level of the sixth vertebral ribs.” (p. 324)

“The cranial extension [of the diverticula perirenalia of the abdominal air sacs] formed secondary diverticula that invaded the spaces between the heads and tubercles of vertebral ribs 7-9 and the corresponding thoracic vertebrae (Fig. 3, 4, and 5). These diverticula were related to the dorsal border of the lung. The caudal extension was situated dorsally to the kidney. It formed secondary diverticula that invaded the spaces between the synsacrum and ilium (Fig. 3, 4, and 5).” (p. 322)

So far, so good. The vertebral diverticula of the cervical air sacs only go back as far as the middle of the thorax, and synsacrum is pneumatized by abdominal air sacs. But wait:

“Roche (1888) and McLelland (1989) describe paired extensions of the vertebral diverticula of the cervical air sacs in the ostrich that extend caudally into the peritoneal cavity beyond the abdominal air sacs. Although these extensions of the vertebral diverticula were not specifically identified in the present study, similar structures have been observed by the authors in post mortem material that was not part of this study.” (pp. 324-325)

So the story from Bezuidenhout et al. is precisely the same as from McLelland: the diverticula of the cervical air sacs that pneumatize the vertebral column only go back to the mid-thorax; secondary diverticula from the vertebral diverticula do extend all the way to the cloaca, but do so within the peritoneal cavity, where they are no longer in contact with the vertebrae; and the synsacrum is pneumatized by diverticula of the abdominal air sacs, which are in contact with the vertebrae.

Please, please, can we as a community drop the idea that the cervical air sacs can pneumatize the back half of the vertebral column? Nobody’s ever seen it happen, a gigantic search effort found no evidence that it’s possible (O’Connor and Claessens, 2005), and the refs that people keep citing on it–McLelland (1989) and Bezuidenhout et al. (1999)–actually talk about something completely different.

Bad Anatomy #2: Pleurocoels and the spectre of continuous pneumatization

“Pleurocoel” is an old term for the big pneumatic cavities that are often present in the vertebral centra of sauropods and theropods. For about a decade and a half, almost everyone who has worked on pneumaticity in dinosaurs has advocated abandoning the term, and calling these structures either pneumatic fossae (if they don’t lead to internal chambers) or pneumatic foramina (if they do)–see Britt (1993, 1997), Wedel et al. (2000), Wedel (2003b), O’Connor (2006), Taylor and Naish (2007), among others. Why? Because from the start the term “pleurocoel” was not rigorously defined, so it was variously used to mean pneumatic foramina, pneumatic fossae, internal pneumatic chambers, or some combination of the above. It is much better to just describe the actual morphology using informative terms.

Sereno et al. use the term pleurocoel. And the result is predictable: anatomical confusion.

“We are inclined to support the latter, more conservative interpretation that pleurocoels in nonavian dinosaurs are a product of paraxial cervical air sacs and provide, at best, ambiguous evidence for intrathoracic ventilatory air sacs. First, pleurocoels are rare in birds, and no living bird has an unbroken cervical-to-caudal series of pleurocoels as occurs in some nonavian dinosaurs, including the one we describe below [31]. As Wedel [26] has underscored, pleurocoels extend posteriorly in the axial column of saurischian dinosaurs to a variable extent, but neither adults nor juveniles of any species show an apneumatic gap. Allotting an unbroken series of pleurocoels of graded form, as in the case we describe below, to three different pneumatic sources (cervical air sacs, lung diverticulae, abdominal air sacs) is difficult to defend. Drawing a direct analogy based on birds for the source(s) of pneumaticity in the posterior axial column in nonavian dinosaurs [22,31,33], thus, is problematic.” (p. 3)

That all sounds pretty convincing . . . until you realize that “pleurocoels” are pneumatic foramina, in which case the entire paragraph becomes misleading at best and flatly incorrect at worst. I’ll show you–here are the key sentences with “pleurocoel” replaced with “pneumatic foramen”:

“First, [pneumatic foramina] are rare in birds, and no living bird has an unbroken cervical-to-caudal series of [pneumatic foramina] as occurs in some nonavian dinosaurs, including the one we describe below [31].”

Just flat wrong. In many birds the vertebral column is continuously pneumatized. That the pneumatic foramina of birds are typically small and tucked up inside the cervical rib loops (unlike the much larger pneumatic foramina of most saurischians) is beside the point. There certainly are interesting questions about why the size of pneumatic foramina varies among taxa and over evolutionary time, but in this context distinguishing between small pneumatic foramina and big “pleurocoels” only obscures the similarity of vertebral pneumatization in birds and other saurischians.

I really don’t want to talk about my own work here, but since it’s cited misleadingly, I have to do so for a while:

“As Wedel [26] has underscored, [pneumatic foramina] extend posteriorly in the axial column of saurischian dinosaurs to a variable extent, but neither adults nor juveniles of any species show an apneumatic gap.”

Neither do most adult birds, as I took pains to point out in the paper they cite (Wedel 2003a). Because the vertebral column of birds is invaded by up to three sets of diverticula–from the cervical air sacs, lungs, and abdominal air sacs–juveniles often have these apneumatic gaps. But in adults the diverticula from the different sources often meet up and anastomose, so you get a continuous series of pneumatic foramina right down the column. Oh, speaking of which:

“Allotting an unbroken series of [pneumatic foramina] of graded form, as in the case we describe below, to three different pneumatic sources (cervical air sacs, lung diverticulae, abdominal air sacs) is difficult to defend.”

Really? Because that’s exactly what happens in most birds.

The absence of an apneumatic gap (or “pneumatic hiatus” [Wedel 2003a]) in dinosaurs is not evidence of anything. If we found one, it would be additional evidence for pneumatization of the vertebral column from multiple sources, but that’s already well-supported. And if we don’t find one, we can’t assume they didn’t have multiple routes of pneumatization, because many birds also lack pneumatic hiatuses as adults. Even in birds that typically do have pneumatic hiatuses, like chickens, the hiatuses are not usually present in all adults. The absence of pneumatic hiatuses in described non-avian dinosaurs could easily be a side-effect of inadequate sampling, for three reasons:

  1. Pneumatic hiatuses are most likely to show up in very young individuals, before the diverticula from different sources have time to meet up. We have very few complete and well-preserved baby saurischians–and by well-preserved, I mean preserved well enough to be able to trace pneumatic features right down the column. The pretty but pancaked theropods from Liaoning won’t cut it (I know, I’ve looked).
  2. Nobody knew to be on the lookout for pneumatic hiatuses until 2003, when I coined the term and pointed out their importance, so there hasn’t been very much time in which to find them.
  3. Not everybody works on pneumaticity, and those of us who do have only seen a small fraction of the world’s dinosaurs. It is possible that pneumatic hiatuses are present in well-known dinosaurs and no one has noticed–yet.

Let’s finish up with the pleurocoel mess:

“Drawing a direct analogy based on birds for the source(s) of pneumaticity in the posterior axial column in nonavian dinosaurs [22,31,33], thus, is problematic.”

No, it’s not. But you can certainly make it seem problematic if you use ambiguous, uninformative terms that obscure the fundamental similarity in vertebral pneumatization in birds and other saurischians.

Bad Phylogenetic Reasoning

“Second, cervical air sacs have been observed extending to the posterior end of the vertebral column in birds. Several authors have described cervical air sacs extending posteriorly beyond the abdominal air sacs in the ostrich (Struthio camelus) [21,36]. Ratites have relatively smaller abdominal sacs than in other birds and, as nonvolant basal avians, serve as better analogs for nonavian saurischians than volant neognaths [37].” (p. 3)

We’ve already dealt with the first two sentences, now let’s handle the third:

“Ratites have relatively smaller abdominal sacs than in other birds”

True, but irrelevant. First, it’s not the size of the abdominal air sacs that’s at stake, it’s their ability to pneumatize the posterior part of the vertebral column. We’ve already seen that both of the papers Sereno et al. cite actually say explicitly that the posterior vertebral column of birds is pneumatized by diverticula of the abdominal air sac, even in the ostrich. Second, we can’t assume that the small abdominal air sacs of ratites represent the primitive condition. There are many cases of reduction, loss, or fusion of air sacs in birds, and the abdominal air sac is no exception. Since Sereno et al. like McLelland (1989) so much, we’ll see what he had to say:

“[The abdominal air sac] is reported to be very poorly developed in the Accipitridae, Fulica, Fregata and ratites (Groebbels, 1932); penguins (Spheniscidae), rheas (Rheidae) and loons (Gaviidae) (Duncker, 1971); and large passeriform species and parrots (Psittacidae) (Schulze, 1910). It appears to be especially small in hummingbirds (Trochilidae) (Stanislaus, 1937) and in Casuarius and Apteryx (Groebbels, 1932).” (p. 264)

If small abdominal air sacs are the criterion for choosing extant analogues, maybe we should be comparing sauropods to hummingbirds. Moving on:

“and, as nonvolant basal avians, serve as better analogs for nonavian saurischians than volant neognaths [37].”

Nonvolant here is a red herring; like non-avian saurischians, ratites are flightless, but they’re definitely secondarily flightless, so the apparent similarity is homoplasy, not plesiomorphy.

Sereno et al. think they’ve found a single taxon–the ostrich–in which the cervical air sac does pneumatize the entire vertebral column, and for their “all cervical all the time” model of dinosaurian pneumaticity to fly, they need to make the case that ostriches are the best possible models for all non-avian saurischians. But the similarities they use to argue this are either homoplastic throughout Aves (small abdominal air sacs) or homoplastic between ratites and non-avian saurischians (flightlessness). So it would be a lousy phylogenetic argument even if their anatomical assertion about the ostrich was correct–which it’s not.

Bad Behavior #1: Unfair Criticism

I urge you to just go read Sereno et al. and see how thoroughly dismissive they are of everything ever written by O’Connor and Claessens. I’ll only work through a couple of examples, but there are plenty of others.

Here’s a choice passage:

“Recently O’Connor proposed that axial pneumaticity in the abelisaurid theropod Majungasaurus can be used ‘‘to refine inferences related to pulmonary structure’’ [35: 159], because ‘‘it shows a reduction in the pneumaticity in the last two dorsal neural arches, with enhanced pneumaticity in sacral aches [31: 22]. Specifically, the ‘‘size and number of neural arch foramina’’ are reduced in dorsal vertebrae 12 and 13, whereas the same are ‘‘enhanced’’ in sacral vertebrae, indicating ‘‘two different sources of pneumatization’’ [33: 253]. The actual differences, however, were not described, and dorsal vertebrae 12 and 13 were not figured until recently [35: figs. 3, 12, 13].” (p. 11)

See, this is funny, because in the last sentence Sereno et al. are picking on O’Connor for the lag of a whole two years between the Nature paper (O’Connor and Claessens 2005) and the monographic description of the Majungasaurus vertebrae (O’Connor 2007). It’s funny because almost all of the many critters named by Sereno and various sets of coauthors have never received any morphological description beyond the original 3-5 page writeups in the weeklies. Here’s a short, non-exhaustive list, with original years of publication:

Eoraptor – 1993

Afrovenator – 1994

Deltadromeus – 1996

Suchomimus – 1998 (okay, this one did get its furcula described in 2007)

Jobaria – 1999

Rugops – 2004

Next to any of those, a two-year turnaround is practically instantaneous. Anyway, Sereno et al. go on to try to explain how the diminution of pneumaticity in the middle of the vertebral column of Majungasaurus is totally not like a pneumatic hiatus at all. Then it’s on to Aerosteon:

“The situation in Aerosteon is instructive for the contrast that it provides across the same vertebral transition. In this case, pneumaticity appears to peak in the last dorsal, with a large pneumatic canal in the transverse process that is not present in sacral vertebrae (Figure 4C). The pleurocoels, in addition, develop a posterodorsally inclined partition in the posteriormost dorsal vertebrae that passes into the sacral series unchanged. The axial column of Aerosteon does not suggest a clean partitioning based on the number or size of pneumatic spaces, but rather a gradation in pleurocoel form that extends from the anterior cervical vertebrae through the distal caudal vertebrae.” (p. 12)

Three things to note here. First, on the previous page, Sereno et al. bashed on O’Connor and Claessens for not figuring all of the relevant vertebrae in their first, short paper. But here they mention a partition in the pneumatic fossae that passes into the sacral vertebrae, without actually figuring any sacral vertebrae in the entire paper. PLoS ONE is online-only and doesn’t charge for length or figures, so there is really no reason not to show those vertebrae if they’re important. Second, their own figures are at odds with their argument. Figures 5 and 6 show cervical centra with big pneumatic foramina, and Figure 9 shows prominent pneumatic foramina in the caudal centra. But Figures 7 and 8 show that in the posterior dorsals the “pleurocoels” are reduced to shallow fossae. It’s been clear for a while that fossae are developmentally and evolutionarily antecedent to foramina (for a recent graphic example, see Wedel 2007:text-fig 8), so those middle dorsals appear to be stuck at a less pneumatized stage than the vertebrae on either side.* Which is exactly what O’Connor and Claessens (2005) described for Majungasaurus. Finally, it’s easy to miss the shift from foramina to fossae and back in the vertebral column of Aerosteon because they’re all collectively called pleurocoels–another case of an important morphological similarity being obscured by the use of an ambiguous term.

*It is true that the posterior dorsals of Aerosteon have highly pneumatic neural spines, but there is some evidence in sauropods that the vertebral centra and neural spines are pneumatized independently (see page 215 here). And regardless of what is going on in the neural spines, the “pneumatic diminution” (my term, newly coined) in the centra of Aerosteon is still interesting and worthy of comment. Odd that they didn’t mention it.

In short, Sereno et al. knock O’Connor and Claessens (2005) for

  • letting a little time pass between the short paper and the monograph, when Sereno and previous sets of coauthors hardly monograph anything;
  • not figuring all of the relevant vertebrae in the short paper, when they don’t, either, in their much longer paper; and
  • making a big deal about the “pneumatic diminution” in Majungasaurus, when the vertebral centra of Aerosteon show something very similar.

There are lots of things that I could say here. The driest and least loaded is that it’s difficult to take the criticisms of Sereno et al. seriously when they are guilty of the same or worse on every single point.

Bad Behavior #2: Misleading use of citations

Here’s another bit that requires some explanation:

“The posterior thoracic, synsacral, and caudal vertebrae, in contrast, are pneumatized by diverticula extending directly from the lung or from abdominal air sacs [1, 16, 19, 21, 22].”

One thing I just flat hate about a lot of “high-impact” journals is that they use numbered references instead of parenthetical citations by authors’ names. It makes it really easy to just read a paper without seeing who is being cited and who isn’t. And that can be a problem. The five refs cited in this sentence are King (1966), Muller (1908), Duncker (1971), McLelland (1989), and O’Connor (2004). Of those, King and McLelland are review papers; only Muller, Duncker, and O’Connor present the results of original research. But Sereno et al. do cite O’Connor here, so what’s my beef? They only cite his 2004 paper, which–crucially–does not include the devastating falsification of the “all cervical all the time model” that one can find in O’Connor and Claessens (2005) and O’Connor (2006). So even though the latter two papers are more recent, more comprehensive, and more relevant, they’re not cited here. Hmm. Is it because they contradict (with shedloads of evidence) the “all cervical all the time” model that Sereno et al. are trying to develop for non-avian saurischians?

The very next sentence:

“Some authors have concluded, therefore, that the lung and abdominal air sacs must also be responsible for pneumaticity in the posterior half of the axial column in nonavian dinosaurs and, on this basis, have packed the thoracic cavity of theropods with a full complement of avian ventilatory air sacs [33].”

If you’re keeping track at home, ref 33 is O’Connor and Claessens (2005). Now that Sereno et al. have something to slate them for, it’s time for a citation.

But wait. O’Connor and Claessens (2005) did not “pack the thoracic cavity with a full complement” of air sacs; they were very explicit in the text about having only found evidence for some of the air sacs (namely cervical and abdominal) from both the anterior and posterior functional sets. Their figure 4 shows a Majungasaurus with all of the regular avian air sacs, but they say in the caption that showing the other air sacs in light grey “represents tertiary-level inferences emphasizing the uncertainty surrounding the reconstruction of soft tissues not constrained by osteological evidence.” Hardly the reckless abandon one would assume from the rhetoric of Sereno et al. (This gets better–see Update 2 below.)

And it gets worse. O’Connor and Claessens (2005) are not the only authors who have inferred that all the basic components of the avian respiratory system were present in some or all non-avian saurischians. Brooks Britt (1993, 1997) came to the same conclusion. So have I (Wedel et al. 2000, Wedel 2003a,b, 2005, 2007). Sereno et al. don’t give Britt the same snide treatment they give O’Connor and Claessens, possibly because one of the authors is currently collaborating with Britt on describing a couple of new sauropods. They spare me for a different reason. Next sentence:

“An opposing view is that the continuous series of pleurocoels observed in many nonavian dinosaurs suggests that the nonventilatory, paraxial cervical air sacs extended posteriorly along the column [26,34]. We are inclined to support the latter, more conservative interpretation . . .”

Refs 26 and 34 are Wedel (2003a) and Chinsamy and Hillenius (2004), respectively. We’ve already seen Chinsamy and Hillenius in the last post; they recycled the mistaken text from Ruben et al. (2003) mis-citing McLelland (1989). And now my situation is clearer, too: Sereno et al. don’t include me in their slam of O’Connor and Claessens–even though I am every bit as ‘guilty’–because I am supposed to represent the counterargument.

But if you actually go read Wedel (2003a) you’ll see that the paper is about as pro-abdominal-air-sac as the available evidence allowed me to be. I raised the possiblity of “all cervical all the time” because this was back in the dark ages when no one knew for sure whether it happened in birds or not–i.e., before O’Connor and Claessens (2005) straightened everything out–but I didn’t sell it. In fact, in that paper I came up with pneumatic hiatuses as a way to falsify the “all cervical all the time” model in fossil taxa; the fact that no-one has published any yet doesn’t mean the hypothesis can’t be falsified from another route. And it has been, by O’Connor and Claessens (2005)–the very paper that Sereno et al. are trying to use my paper against!

So here’s the short version: Chinsamy and Hillenius (2004) were wrong, because they borrowed their text from Ruben et al. (2003), who misquoted McLelland (1989) (ironically, since Sereno et al. make precisely the same mistake). And I was wrong (for the purposes to which Sereno et al. put my work) because I thought that pneumatization of the whole vertebral column by the cervical air sacs was at least a possibility, before O’Connor and Claessens (2005) showed that it is not, for living saurischians at least. Sereno et al.’s critique of O’Connor and Claessens is empty fluff, and their counterargument is based on arguments that were either wrong in the first place or have already been falsified–by O’Connor and Claessens.


There are plenty of other places where Sereno et al. unfairly bash on O’Connor and Claessens and conveniently under-cite those authors who got to the pneumaticity party before them. I’d carefully explain them all, but life is too short and I’ve satisfied my conscience by exposing some of their worst excesses.

Sereno et al. are wrong about avian anatomy. Their phylogenetic inferences are wrong. They use selective citation to suppress genuine contributions and resurrect falsified hypotheses, but those hypotheses remain falsified. They use obfuscatory terminology to obscure important similarities between birds and non-avian saurischians, including Aerosteon. Their new model of avian lung evolution is based on old misconceptions about pneumatization in birds, and flatly contradicted by the very papers they cite to support it.


Thanks for slogging through all this. Here’s your sauropod vertebra:

UPDATE: I’ve been bad (or not)

I have been privately accused of ethical misconduct for something I said in this post, so I will now preeptively fisk myself by way of explanation.

A critic, whose name I will not mention (but it’s not Brooks Britt), accused me of publicly divulging information exchanged in confidence, in this section:

“Sereno et al. don’t give Britt the same snide treatment they give O’Connor and Claessens, possibly because one of the authors is currently collaborating with Britt on describing a couple of new sauropods.”

The critic claims that I learned of the collaboration between Britt and one of the authors of the Aerosteon paper in a private conversation, and that I was “out of line” in posting it in a public forum. The critic is partly right but mostly wrong. I did hear about the collaboration in a private conversation in either late 2006 or early 2007, but it was information I already knew from another, more public source. Brooks Britt delivered the talk for the Chure et al. (2006) abstract at the SVP meeting in October, 2006. At the podium, in front of a few hundred people, he mentioned that one of the Sereno et al. (2008) authors was collaborating with him on the description, and then showed a phylogenetic analysis generated by that collaborator (no prizes for guessing which Sereno et al. author). So all of the literally hundreds of people who were in that session knew about the collaboration, although I can’t say how many of them have remembered that it was mentioned.

The 2006 SVP abstract book carries this warning: “Observers are reminded that the technical content of the SVP sessions is not to be reported in any medium (print, electronic, or Internet) without the prior permission of the authors.” I’m not sure if knowing who is collaborating with whom counts as technical content or not. And I don’t really care. It’s not top-secret research results, and literally hundreds of people know about it, or did for a few minutes back in 2006 (probably just until Mary Schweitzer kicked us in the brainpan with her T. rex histo). And more importantly, when authors engage in selective citation, singling out some for praise and harshly condemning others when we were all about equally “guilty”, they shouldn’t be surprised if some of those so used publicly speculate about their motives in doing so. If nothing else, it may persuade them not to behave that way in the future.

UPDATE 2: Packing in the air sacs

There is an amusing coda to the bit where Sereno et al. (2008) accuse O’Connor and Claessens (2005) of having “packed the thoracic cavity of theropods with a full complement of avian ventilatory air sacs”. You’ll recall that O’Connor and Claessens did in fact show all of the avian air sacs, but greyed out the clavicular, anterior thoracic, and posterior thoracic sacs for which they had no direct evidence, and put a huge disclaimer about those air sacs in the figure caption.

I ignored the non-blog coverage of the Aerosteon story until recently, which is a shame, because it’s most interesting. Here’s the full-color Aerosteon restoration that went out to the media outlets from Sereno’s Project Exploration (borrowed from the National Geographic News page):

Notice that all of the air sacs are colored in, including those for which even Aerosteon has no direct evidence (i.e., the anterior and posterior thoracic air sacs). But wait–could the pneumatic gastralia be evidence of those phantom sacs? Not according to Sereno et al.:

“The external (ventral) position of the pneumatopores suggests that the pneumatic diverticulae lay in superficial tracts outside the gastral cuirass. It seems unlikely that pneumatic diverticulae would penetrated the ventral thoracic wall to access external pneumatopores, when entering the gastralia directly from their internal (dorsal) surface would be much easier. A plausible explanation may be that these ventral pneumatic tracts are part of a subcutaneous system, which is present to varying degrees in birds and is composed of diverticulae from cervical, clavicular, and abdominal air sacs [1,21,22]. Subcutaneous diverticulae usually exit the thoracic cavity and extend under the skin to distant body surfaces. In the brown pelican (Pelecanus occidentalis), for example, diverticulae of the clavicular air sac exit the thoracic cavity dorsally and extend under the skin to reach the entire ventral surface of the thorax [61].” (p. 13)

I have no quibbles with any of that. It’s just curious that Sereno et al. would eviscerate O’Connor and Claessens for going overboard on air sacs (when O&C were actually quite careful) and then do the same thing, sans caveats, in their press release.

Parting thought, from the National Geographic News story (emphasis added):

“The fossil [Aerosteon] provides the first evidence of dinosaur air sacs, which pump air into the lungs and are used by modern-day birds, said Paul Sereno, the project’s lead researcher and a National Geographic explorer-in-residence.” [NOTE: This overstatement is not in the Aerosteon press release, and may have been hyperbole by an underinformed journalist.]


49 Responses to “The Aerosteon saga, Part 2: Overinflation and undercitation”

  1. […] features that Sereno et al. claim (and even some that they don’t–stay tuned for Part 2). But there are parts of the paper that I disagree with, and to understand why, I have to tell you […]

  2. William Miller Says:

    Wow. This is AMAZINGLY comprehensive, and very fun to read (I especially liked “The old confusion has been swept away by a blood-dimmed tide of bird carcasses and good science.”)

    It’s truly amazing how interesting you can make things that those uninitiated into the mysteries of vertebrae (i. e., people who don’t read this blog) would consider boring.

  3. Adam Yates Says:

    Hey Matt!

    Brilliant post. And extremely timeous. I’ve only just got back into the paper writing groove and just spent all weekend on the paper we discussed last year. Sadly my poorer knowledge of pneumaticity had let the Aerosteon paper lead me off track somewhat (specifically I accepted unquestionably that cervical airsacs were pneumatising the posterior dorsals in ostriches – and wondered how come O’Connor and Claessens didn’t find it). So now its rewrite time!
    Two questions: a) I wonder if given that birds are but one surviving subgroup of saurischians that it still might be a logical possibility for cervical airsacs to invade te whole dorsal column in more basal taxa while birds have lost the ability.(not that I am defending the possibility – there is no evidence for it).
    b) In one of your own papers you cite Turkeys as another example of a taxon where cervical airsacs pneumatise the posterior dorsals. I wonder was this an example of confusion caused by the anastomosis of cervical and abdomnal systems? Indeed if the two can join up does one need to use early ontogenetic stages to determine which system is doing the pneumatisation? Did OConnor and Claessens use young, pre-anastomised individuals?

  4. Adam Yates Says:

    Okay scratch question ‘a’, I just read your first Aerosteon post and see that you tackle it there.

  5. […] The Aerosteon saga, Part 2: Overinflation and undercitation Ottobre 5, 2008 […]

  6. Jay Says:

    Wow – a good & detailed critique.
    Matt, why don’t you publish this (somethings like a “comments on Sereno et al., etc etc bla bla”. PlOS One is free, and i think this sort of scientific debunking is is very informative (especially giving credit to O’Connor * Claesson’s findings

  7. […] pneumaticity and raises some criticisms of Sereno et al.’s paper describing Aerosteon in two posts over on […]

  8. Matt. Really, you should publish this as a rebuttal. If PlosOne is free, then why not? You already have most of the work done.

  9. Nathan Myers Says:

    I must take issue with one statement here: “Not everybody works on pneumaticity…” How could anyone find out about pneumaticity and then not not work on it? The notion beggars belief, and must surely be, simply, mistaken.

    Regardless, I’m wondering about the backstory here. Maybe you’ve deliberately avoided the topic in the interest of setting the record straight, but I wonder Dr. Sereno was up to. Msplcd dfns f hs dvsr’s hbbyhrs? Prsnl cnflct? Wnt f nthng ls ntrstng t wrt bt rstn? Clssc rct-crnl nvrsn?

  10. Matt Wedel Says:

    I wonder if given that birds are but one surviving subgroup of saurischians that it still might be a logical possibility for cervical airsacs to invade te whole dorsal column in more basal taxa while birds have lost the ability.(not that I am defending the possibility – there is no evidence for it).

    It’s possible. But I think it’s not likely, and here’s why: the sequence of pneumatization of the avian vertebral column during development today is the same as the sequence of pneumatization of the saurischian vertebral column during the evolution of those lineages. I have no idea why something so non-critical would be so highly conserved, but that seems to be the case.

    b) In one of your own papers you cite Turkeys as another example of a taxon where cervical airsacs pneumatise the posterior dorsals.

    Yep, that’s Wedel (2003a), linked above.

    I wonder was this an example of confusion caused by the anastomosis of cervical and abdomnal systems? Indeed if the two can join up does one need to use early ontogenetic stages to determine which system is doing the pneumatisation? Did OConnor and Claessens use young, pre-anastomised individuals?

    Eeeen-teresting. I am embarrassed to admit that this had never occurred to me. I am also embarrassed to admit that I am the wrong person to talk to about what actually happens in birds. I mean, I can read the papers without _completely_ misinterpreting them, which is apparently a minor achievement, but I’ve only scanned and dissected a handful of real live dead birds, to Pat O’Connor’s hundreds. So he’d be the person to ask.

    One thing I am very curious to know is whether the anastomoses are apparent or real. Okay, so you’ve got diverticula coming along the vertebral column from multiple sources: cervical air sacs, lungs, and abdominal air sacs. What if those diverticula bump into each other but don’t actually unite? What if there is a double layer of super-thin epithelium separating the adjacent diverticula? That would be effectively impossible to see in a CT scan, and it might be a real cast iron you-know-what to detect even in a dissection. I’m curious because I’ve always wondered if air is actively circulated in the diverticula or if it just diffuses around.

    To the many people who have asked why I don’t post this stuff as commentary at PLoS ONE: thanks sincerely, and please be patient. SVP is just a week away, and I want to chat up some others to see if they’d like to do that themselves or do it jointly with me. In the meantime, all of the relevant scientific points are easily available in the published record; all I’ve done is bring them to light. Now that the info is out there, whatever happens, happens. But it would be a courtesy if you waited a couple of weeks before posting this at PLoS ONE, to give those of us who are, ahem, closer to the situation a chance to talk it over and decide on a course.

    Regardless, I’m wondering about the backstory here. Maybe you’ve deliberately avoided the topic in the interest of setting the record straight

    I did avoid that as much as I could, because it’s mostly irrelevant. Almost everything you need to know about this is in the literature (and the rest was announced to a huge crowd a couple of years ago–see the update above). And in the interest of keeping this at least within hailing distance of the high ground, I’m going to disemvowel the rest of your comment. Anything else in this vein, by anyone, will get deleted. I’m not trying to harsh on you personally, you’re just the first one to step over our invisible acceptability line.

    So, anyone think that sauropod vertebra is kinda BIG for a mid-caudal?

  11. Mike Taylor Says:

    Not only is that caudal big, it’s also opisthocoelous, which is pretty darned unusual. So, what’s the deal?

  12. Nathan Myers Says:

    Matt: Impersonal harshitation notwithstanding, I’ll still buy you a pitcher at the place of your choosing if you ever find your way to Berkeley, and you can tell them I said so.

  13. Matt Wedel Says:

    It’s a date.

  14. Graeme Worth Says:

    All of this (plus some previous instances from the same journal set) raises the question as to whether this sort of non-peer-reviewed publication will really serve the interests of good science. Specifically, in this instance, I’d be interested in whether or not you think this paper, in it’s current manifestation, would have been published without extensive review in a peer-reviewed journal?

  15. Mike Taylor Says:

    Woah! PLoS ONE is definitely a peer-reviewed journal — a point made clearly on its home-page at and explained in a little more detail at

    It’s true that the Sereno et al. article, unusually, doesn’t mention the reviewers in the acknowledgments, but that’s just an oversight, or else both reviewers asked not to be acknowledged. The journal is an excellent one, for all sorts of reasons (open access and outstanding figure quality being two of them) and I’d hate its reputation to suffer because of one questionable paper.

  16. Graeme Worth Says:

    Thanks for correcting me on the peer-review status, Mike. I had been under the impression that it wasn’t, an impression that had been strengthened by at least 2 other papers that were either not reviewed or reviewed so badly that they might as well not have bothered (in fact I think 1 of the papers was eventually withdrawn). I also think that the aims of the PLoS journals is admirable for the reasons you mention, but it is necessary to maintain reader confidence otherwise at some point authors will view the journal as a “last resort” publisher, and there are enough of those around already.
    Perhaps I should rephrase my earlier question – would this paper have appeared as it is if submitted to, say, JVP, Paleontology, Science etc?

  17. In two words: Also brilliant.

  18. Kevin Padian Says:

    I’m a little puzzled by Mike Taylor’s comment that PLoS ONE is “definitely a peer-reviewed journal.” From the website, it seems that it CAN be, in the traditional sense; but this is only one of three possibilities, and you cannot seem to tell which one from a published article.

    To wit, an Academic Editor of PLoS ONE “can conduct the peer review themselves, based on their own knowledge and experience; They can take further advice through discussion with other members of the editorial board; [or] They can solicit reports from further referees.”

    This is unlike any other first-tier academic journal that I know. The rationale seems to be (again from their website) that “Unlike many journals which attempt to use the peer review process to determine whether or not an article reaches the level of ‘importance’ required by a given journal, PLoS ONE uses peer review to determine whether a paper is technically sound and worthy of inclusion in the published scientific record. Once the work is published in PLoS ONE, the broader community is then able to discuss and evaluate the significance of the article.”

    So in short, the editor can look it over and if he thinks it’s interesting he can publish it and see what the community thinks, instead of the reverse order. This is peer review in a sense, but is relatively soft, because considerations of whether an article is on a “hot topic” or could raise the profile of the journal can come into play. In other journals, the editor might examine the references (and others) to see who has published work on the topic and would be most knowledgeable to flag problems in methods or interpretations (few scientists as editors would be so versatile). As noted, that’s a possibility in PLoS ONE, but only one.

    I do not judge whether or not this is a good model, but it is a different editorial process from those of Science, Nature, PNAS, Proceedings B, JVP, Palaeontology, etc., etc. And I would not conclude from this that the normal peer-review process is a necessary part of publishing in PLoS ONE.

    This, by the way, makes it tricky for academic review committees to determine whether articles in such publications are actually peer-reviewed. They may have to depend in future on documentation from the author that they were, and perhaps on feedback from the PLoS site and blogs such as these to evaluate them.

  19. Mike Taylor Says:

    I like Kevin’s rather polite way of saying “I’m a little puzzled by Mike Taylor’s comment” when he might just as easily have said “Mike Taylor is talking a load of arse” :-)

    Well, now, I suppose this shows that the concept of peer-review is a little more slippery than I’d appreciated. As Kevin notes, an Academic Editor of PLoS ONE “can conduct the peer review themselves, based on their own knowledge and experience”. That, to me, looks like a legitimate form of peer review, and my understanding is that it’s not uncommon in other journals for editors to act as one of the reviewers for articles that they handle. The only way that would be unacceptable would be if the editor in question was one of the authors of the paper; I know that Sereno is a PLoS editor, but _surely_ he didn’t “peer review” his own paper.

    I have to take issue with Kevin’s summary of the PLoS system for one other important reason: “the editor can look it over and if he thinks it’s interesting he can publish it”. The whole distinctive point of PLoS is that the decision to publish or not isn’t supposed to be determined by how “interesting” the article is, but only on whether it’s sound science. In this, it is (presumably deliberately) the opposite of the Science’n’Nature tabloids, where selection is largely on how gee-whiz the specimen is, and where papers that are merely good science don’t stand a chance. I think the PLoS approach is great step forward over the rock-star Science’n’Nature approach, and I’d hate to see them forced into changing that approach by situations like the one we’re looking at here.

    In the final analysis, the failure of peer review in this case is nothing to do with PLoS’s policy of accepting sound science whether or not it’s sexy — the failure was simply in assessing whether the science was sound. It’s shame we don’t know who the reviewers were, because it does seem that they didn’t do their job right.

  20. Graeme Worth Says:

    In addition, it is not clear from the guidelines whether or not papers referred to reviewers are, from the reviewers perspective, anonymous (if the editor himself chooses to review the paper it clearly will not be anonymous). Previous studies of the peer-review process has demonstrated that failure to ensure anonymity can lead to papers being accepted on name and past reputation rather than the specific content of the paper in question. And to get back to the Aerosteon paper, it seems clear from the critique by Matt Wedel that it either did not go to peer review, or, if it did, the process clearly failed and the journal should be reviewing its processes in the light of such failure. It really is important that Matt’s critique or something similar does go to the journal ASAP.

  21. Darren Naish Says:

    As some of you might know, Mark Witton and I published in PLoS ONE earlier this year. Our ms was reviewed by two entirely relevant, acknowledged experts in the field and we – having survived the process – then had to make the usual suggested/required changes, as is normal. My impression from this was that PLoS ONE worked like all other proper journals, but I evidently did not do my homework as I was unaware of the other possible ‘review avenues’. I had assumed until now that all PLoS ONE submissions had gone through the same, two-reviewer process that Mark and I did. Interesting.

  22. Mike Taylor Says:

    … but it’s worth noting that the Witton and Naish azhdarchid paper in PLoS ONE doesn’t mention the reviewers (by name, or even to note that they existed) in its acknowledgements. Neither did a couple more PLoS papers that I checked — so perhaps this is some bizarre (and inadvisable) PLoS style oddity. Anyway, it seems that at least we shouldn’t deduce anything from the lack of mention of reviewers in the acknowledgements of a PLoS paper.

  23. 220mya Says:

    As someone who has participated in the peer-review process as an author, reviewer, and editor, I can state that while it is not uncommon for the editor to provide comments on an MS, this is always in addition to external peer reviewers. In contrast, it seems that PLoS ONE allows the editor the option of being the sole peer-reviewer.

    That having been said, I don’t think the average quality of PLoS ONE papers is any better or worse than those in Nature, Science, or PNAS. And it should be noted that papers submitted by NAS fellows do not get the same peer-review treatment at PNAS as do those submitted by regular scientists.

  24. Andy Says:

    If the reviewers are/were anonymous, some journals don’t let you list them. I recall that on one of my recent papers (Journal of Morphology, maybe?) the editors took out the line where I thanked two anonymous reviewers (they checked with my first, but it was presented as a standard for the journal). Also. . .this practice varies by field. My wife, who is a physicist, was shocked when she saw that I thanked anonymous reviewers. The only case where she has seen it cited in her branch of physics is if one of the reviewers suggests a major set of experiments that wouldn’t have been tried otherwise. So, perhaps this is what is happening with PLoS.

  25. Nathan Myers Says:

    Let us not neglect to note that SV-POW! has once again demonstrated its supreme awesomeness. Authors of other blogs — with a few exceptions! — must hang their collective head in shame.

  26. Darren Naish Says:

    Yes, I hear there’s one particularly awesome blog on scienceblogs. I forget its name.

  27. Nathan Myers Says:

    Nah, there’s no way Pharyngula can compete with (not claiming exhaustiveness) the likes of SV-POW!, TetZoo, or Catalog of Organisms.

  28. William Miller Says:

    Yeah. Pharyngula cheats on comment-number by including politics…

  29. Nathan Myers Says:

    Sure, P. gets the hits, but we were talking about awesomeness. On that topic I feel compelled to mention Not Exactly Rocket Science even though it (unlike SV-POW!) sometimes mentions topics neither sauropodal nor vertebral.

  30. mAnasa-taraMgiNI Says:

    Thanks for these two detailed posts. I was surprised by the dismissal of what seemed to be a very thorough work by O’Connor and Claessens. I first thought may be I missed something but was utter confounded by the claim regarding the cervical sacs going all the way posteriorly to pneumatize structures beyond the mid-thorax. I began wondering if Sereno had some personal axe to grind with O’C and C. This clarification by Dr. Wedel is really wonderful.

  31. Vertebrat Says:

    So, anyone think that sauropod vertebra is kinda BIG for a mid-caudal?

    I’m just glad I was even able to identify it as a caudal without peeking at the tags. (BTW, is it missing its parapophyses and part of its prezygs?)

    I suppose that means it’s unpneumatized, unlike the nominal topic of the post. Must weigh a ton. Where’s a diverticulum of the abdominal air sac when you need one?

  32. Nathan Myers Says:

    That one’s for working out, so it needs to stay heavy.

  33. Jaime A. Headden Says:

    Vertebrat, caudals have by this point not only merged their parapophyses into the diapophyses, they are determined at the end of the transverse process, which as you can see is that triangular little knob. This is a typcial morphology for what looks like a titanosauriform caudal. The opisthocoelous condition is particularly interesting, though.

    On the articles, while I have nothing to say on Sereno’s methodology, it is clear he tensed his work on the subject as a point of refuttal, and chose a path that makes the taxon more distinct. This seems common for Nature papers, but odd when one is not competing for “ooohs” and “ahhhs”. I should think, however, that some of the cited work may be ambiguous on whether ostriches invaginate the sacrum and caudals, when the cited sources do not clearly indicate that the diverticula entering the pelvic region do NOT contact the vertebrae. Given the argument that diverticulae follow the vertebrae, but suddenly diverge (without clear cited terminology to indicate otherwise) it might be ambiguous. It is also clear that some works (McLelland) were cited without being read themselves. Why, then, positively cite Ruben et al. on the topic of vertebral pneumaticity, when that paper is particularly false on virtually most of its conclusions based on the preconceived notion of avian ancestry?

  34. Zach Miller Says:

    Wonderful posts! I know next to nothing about pnuacticity and diverticulae and avian breathing…mechanics, so I took the Sereno paper more or less at face value. So I’m happy you wrote this post!

    Have you gotten any flack and/or response from the authors about your particular points?

  35. Matt Wedel Says:

    On the multiple review tracks at PLoS ONE: I really wish they’d just state the review track for each article–i.e., solo editor approved, multiple editor approved, or externally reviewed. Normally the first two wouldn’t even be called ‘peer-reviewed’. I also hope that authors are allowed to preferentially request ‘tougher’ review tracks, but ideally not easier ones. Suppose you really need a peer-reviewed article out in a certain amount of time and you choose PLoS ONE because it’s open access and fast. What if they decide your work is sexy enough to publish without external review? Could be good for them (getting the hot stuff out quickly) but bad for you. Finally, if the journal won’t explicitly say which review track each paper is on, it would be nice if the authors were told so they could put that information in the paper if they so chose.

    And as long as I’m dreaming I’d like to find a two-meter Amphicoelias cervical in my backyard.

    Have you gotten any flack and/or response from the authors about your particular points?

    Nope, but that’s okay. This is a blog, not a journal, and they’re not compelled to respond. Even if they have a full rebuttal ready, they may prefer to wait for these criticisms to appear in a more formal venue, so they can rebut them only once and definitively rather than getting into another one o’ them endless debates on the intert00bz.

  36. Andrea Cau Says:

    Great couple of posts!
    After reading them, I’ve decided to correct part of my recent post about Aerosteon on my theropod blog (I’m sorry, but it’s written in Italian), and also added a new post called “Give to Aerosteon what is Aerosteon’s, and give to Majungasaurus what is Majungasaurus’s”.

  37. Andy Says:

    To follow up on Matt’s comment, another thing I’d like to know is the logic behind the posting/not posting of early reviews of the paper. I’ve found that many papers have them (e.g., the Palau “pygmies”, impedence matching in Paleozoic reptiles, etc.), but others (e.g., Aerosteon) don’t. Is this something the authors or the reviewers decide, or both? I’ve mixed feelings on posting the reviews – they are always of the first (submitted) draft of the paper – but on the other hand, it might be nice to see what critiques were raised early on (and then addressed, or not addressed).

  38. Mike Taylor Says:

    Andy, I think that’s very unusual in palaeo. I only know of one paper whose reviews have been published, and that’s this one.

  39. DD Says:

    one word: superawesomeness.

  40. […] 12, 2008 What with all the fuss over Aerostron, it seems that we missed SV-POW!’s first birthday. Yes, it’s been just […]

  41. […] The Aerosteon saga, Part 2: Overinflation and undercitation Ottobre 5, 2008 […]

  42. […] which specializes on the weird vertebrae of sauropods, paleontologist Matt Wedel wrote a detailed critique of the Aerosteon paper. Beyond debating the anatomical analysis of the new dinosaur, Wedel charged […]

  43. […] is done to science as long as the follow-up descriptions do eventually appear (sometimes they don’t). But Calvo et al. (2008) put out a longer description of Futalognkosaurus the very next year, for […]

  44. […] reference supporting all this is number 20 (remember how much I like numbered references?), and here’s the full text (Stevens and Parrish 1999: p. 800): 20. J. M. Parrish and K. […]

  45. […] down in the comments. For a more nuanced take on the variety of review options at PLoS ONE, see here.] The only holdup there is that the post-hoc review by commenters is not working out quite like […]

  46. […] The world of dinosaur palaeontology is probably just as aware of Matt’s series of Aerosteon response articles here on SV-POW! as it would be if he’d put those together into a paper that was […]

  47. […] has been noted in comments on this blog, this first form, in which the editor makes the decision alone, is “unlike any other […]

  48. triceratopshorridus Says:

    I think it should be noted for those who do not know, that Aerosteon (and all megaraptorans for that matter) are considered tyrannosaurids nowadays, instead of carcharodontosaurids or neovenatorids. So the high degree of pneumatization really isn’t all that surprising.

Leave a Reply

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

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

Twitter picture

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

Facebook photo

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

Connecting to %s

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

%d bloggers like this: