How not to do science

Dieser Beitrag auf Deutsch.

So there is a recent paper out in PLOS ONE by Georgi et al. They hypothesise that semicircular ducts scale closely with head size, testing this on dinosaurs.

What’s a semicircular canal? Read up on it at wikipedia. Three of them form a part of your inner ear, and tell you about accelerations of your head, so that you can determine if and in what direction you are moving.

Georgi et al. (2013) find that in the investigated dinosaurs head mass more strongly correlates with anterior semicircular canal size than does body mass, i.e., the anterior semicircular canal’s size is determined mainly by head size, not by body size. Supposedly, it also statistically separates bipedal from quadrupedal taxa, with bipeds having relatively larger canals.

Sounds great, hu? Get a good mass estimate for the skull, get a nice CT scan of a skull, and presto you can tell if your dino was a biped or a quadruped!

Hold it!

There are several issues with the Georgi et al. paper. And one of them involves Plateosaurus. Now, you know I get a bit touchy when people write stuff about Plateosaurus that’s inaccurate, but what really riles me is when people do that although they should know better. For example, if they are proper scientists with full access to the literature, and still have not realised that what they claim in their paper has been debunked repeatedly in the last decade. I mean, you’re free to disagree with a paper, but mention it and address it, don’t  pretend it doesn’t exist!

So, Georgi et al. cite Galton and Upchurch (2004), who say that Plateosaurus was a facultative biped – i.e., it was able to move either on two or on four legs. But…..

….but when you write a paper that deals with animals you’re not familiar with, and you know you do not know the relevant literature by heart, you should simply go looking for recent literature. For example with a Science Direct search for “Plateosaurus locomotion”. Hm, not much help – so if there was any research done on this it wasn’t published with Elsevier! What about Mendeley?

Oh, golly! Look at those hits:

(click to embiggen)

Seems to me that if you’re writing a paper that deals with Plateosaurus locomotion, the first two hits are must-reads! Let’s look at the abstracts only:

Bonnan and Senter (2007): “results indicate that the range of motion in the forelimbs of Plateosaurus and Massospondylus did not allow efficient, habitual quadrupedal locomotion.”

Hm……… This does not sound as if Plateosaurus was a facultative biped, much less a habitual quadruped that occasionally ran on the hind limbs only, as portrayed by Christian and Preuschoft (1996), and as Georgi et al. (2013) write. So, you should read both papers and try to figure out why these two totally conflicting views are out there.

On to Mallison (2010): “Plateosaurus was indeed an obligate biped”. Right there in the abstract. Hmmmmmmmm…………

But what if you did not use Mendeley? I can tell you that GEOREF gives you these results, too, but you need institutional access. However, there also is Google Scholar. Let’s give it a try:

(click to embiggonate)

Here, there are a bunch of not quite as strikingly relevant papers at the top, but they all pertain to the topic. However, searching for “Plateosaurus” in them quickly shows you which are relevant to your work and which aren’t. In this case, you’d quickly filter out all but Christian et al. (1996) and my papers.

But it gets even EASIER! Google is your friend: “Plateosaurus locomotion” gives you, among many other pertinent links – here, let me google that for you!
1st hit: http://palaeo-electronica.org/2010_2/198/index.html – a paper by me on the posture of Plateosaurus; 2nd link is a direct to PDF
3rd hit: http://en.wikipedia.org/wiki/Plateosaurus – has a lot of literature cited on the posture, including above paper and Bonnan and Senter 2007 (see below)
5th hit: http://www.app.pan.pl/archive/published/app55/app20090075.pdf – direct to another paper by me on the range of motion and posture of Plateosaurus.

You can find Bonnan and Senter (2007), who concluded that Plateosaurus was an obligate biped, easily in the wikipedia article. You can there also find earlier papers cited that conclude P. was a facultative biped.

Is this a big deal?

Normally, one would say that one messed up dinosaur among over 20 is not a big problem for a study. Just because it is my beloved Plateosaurus doesn’t make it very important. However, in this case, the error has significant consequences! Georgi et al. (2013) split all their dinosaurs into either quadrupeds (8) or bipeds (14), with only four (4) being classified as ambiguous. These four are
– Plateosaurus – a basal sauropodomorph
– Anchisaurus – another basal sauropodomorph, and a small one at that
– Corythosaurus – a hadrosaur
– Edmontosaurus – another hadrosaur.

This list gives me the creeps when used for the analysis presented by Georgi et al. (2013). 25% is messed up to begin with, and then there is the ugly fact that Bonnan and Senter (2007) found not only Plateosaurus but also Massospondylus incapable of manus pronation, and thus quadrupedal locomotion. Anchisaurus is very similar to Massopondylus, and may well have been an obligate biped, too.

The two hadrosaurs plot with quadrupeds – well, duh, these animals have unguals shaped like they bore hoofs on the fingers of their hands. They were quadrupeds, and if they indeed sometimes walked on their hindlimbs alone that didn’t change their posture much – if you tilt a hadrosaur body up very far the tail hits the ground, and the tails of hadrosaurs were not very flexible in the vertical direction. Basically, a hadrosaur goes bipedal not by changing posture, but by pulling up its hands. They were effectively bipeds with stilts in front – or without stilts in front. Shouldn’t do anything at all to the inner ear.

So the only interesting part of the “ambiguous” animals is – nothing!

Also, one might argue that it is a lot of work googling and mendeleying 20some dinosaurs. However, I timed how long it took me to find the info that tells me “you must read this”. Thanks to the little excerpts under the titles, it takes all of 93 seconds. 93 seconds and you know that your assumption “Plateosaurus was a a facultative biped” is questionable. 30 times 93 seconds is less than an hour’s work, to spare you the embarrassment. Much less if you do not turn to three or four searched per taxon – and good old standard google will usually suffice. So yes, it is entirely doable and actually the smart thing to do: google them all! Go to wikipedia, maybe – it is not a research tool, and for many taxa there is no good info to be found on wikipedia, but if you are lucky you come across an article that gives you a nice concise summary of some of the literature, and thus an excellent starting point for your own literature research.

(and, for the record, I think the rest of the paper is bollocks, too. But that’s for a rebuttal, if I ever find time).

Adding injury to insult

So Georgi et al. (2013) got my beloved Plateosaurus‘ posture wrong. At least they included it in their study – wouldn’t it be a real pity if data from the most abundant basal sauropodomorph, a species of which we have nearly complete and articulated skeletons (one example here and here) and lots of skulls, didn’t get included properly?

So I did what I always do, and what I can tell most reviewers and readers never do: I checked the sources! Georgi et al (2013) wrote in the main text:

“wherever possible (21 of 29 specimens), preference was given to body masses estimated using Seebacher’s polynomial estimation approach“.

So I took a look at Table S1 of Georgi et a. (2013), to see what specimen of Plateosaurus they included, and where they got their skull length and body mass from. They hide this info in a supplement, instead of putting the table into the paper. I dislike this practice, because it means that it takes more effort to access the info – for no gain. It is not like PLOS ONE is a print journal that wants as much stuff as possible in online supplements to save on printing and shipping costs.

Oh well, here it is – I cropped out a lot of lines for the other species that do not interest us, and a bunch of irrelevant references. I also put a red circle over the problem.

(click to embiggerate)

“Why is that a problem?” – that question is likely that’s going through your head right now. “They used Seebacher where they could, and other sources where they couldn’t.” Well, see Seebacher’s paper and despair:

Say again?

Here is Seebacher (2001), giving a mass estimate and a total length for the very specimen, AMNH 6810, for which Georgi et al. (2013) have skull data – and yet they use a different mass estimate!

epic FAIL!

A special fail, because one look a the two tables shows you that Georgi et al. (2013) used a mass estimate for an animal with a total length of 8 meters, not the 6.5 meters of AMNH 6810. AARRGGHH!!!!!!!!!!

And it gets worse before it gets better: Georgi et al. (2013) took the skull length from Sereno et al. (2007). Which makes me wonder: if they had access to AMNH 6810’s skull for scanning, so they could do their inner ear thingy, why didn’t they just measure the length of the thing? The skull is disarticulated, but there are casts made before it was taken apart. Or you can scan the entire thing and simply digitally mount it! In any case, there is not reason to take skull length for this beast from Sereno et al. (2007). Here’s why:

See that? Sereno et al. (2007) did not measure the skull of AMNH 6810, they took the length from Galton (1985). So what should a thorough and diligent researcher now do? She or he should now get hold of the original source, Galton (1985), and check that there was no transcription error. Because, you know, transcription errors happen. Really, they do!
As a reviewer of a paper I always pick out a few data points or text citations at random and do a background check for source citing. In this case, the check would fail the paper – here’s the relevant part of Galton (1985):

Georgi et al. (2013) use “basicranial length” – I guess that’s supposed to say “length to the end of the basicranium”, not “length of the basicranium”. Now, please, go and find me a basicranial length of 0.36 m, as given by Sereno et al. (2007), in Peter M. Galton’s 1985 table!

classic FAIL!

I’ve pointed out the problem with the Plateosaurus data in a comment on the paper, you can find it here. Great that PLOS ONE gives me this opportunity!

And, while we talk about checking things: where is the inner ear data? In extenso? Where are the extracted 3D models? And where is the info on where I can get the CT scans?
Oh, they didn’t publish their raw data, so nobody can easily go and check if they messed up some measurement, typoed some coefficient or otherwise just got things wrong. Way to go, PLOS ONE – you let people publish a flawed paper without the necessary data. UGH! Figure 2 shows one(!) example – and the caption doesn’t even say which specimen this is!

All we do get is Figure 3 that shows plots – but many data points aren’t labelled! And each dot is big enough to cover an order of magnitude (disclaimer: satirical exaggeration). WTF????? Sorry, but this is a lame excuse of a figure! Give me a supplement table with your raw data, or do not publish!

In the end, the paper is ruined by the doubt this all casts on all the other data points. If the very first thing I test fails, and I follow it through and find more failure, then I do not trust the rest of the stuff at all. And I can’t be bothered to check it all – that’s for the authors to do right in the first place. Unless I will one day write a paper that’s close enough to have to read and think it all through, I won’t. Especially because some of the data I’d need to have is missing.

So, overall, I am quite disappointed by Georgi et al. (2013). My colleague Armin Schmitt is doing similar work at Bonn University; before I decide what I think about semicircular canals in dinosaurs I guess I’ll better wait for his publications.

References

Bonnan, M.F., Senter, P. 2007. Were the basal sauropodomorph dinosaurs Plateosaurus and Massospondylus habitual quadrupeds? In: Barrett, P.M., Batten, D.J., editors. Evolution and Palaeobiology of Early Sauropodomorph Dinosaurs. Special Papers in Palaeontology 77:139–155

Christian, A., Preuschoft, H. 1996. Deducing the body posture of extinct large vertebrates from the shape of the vertebral column. Palaeontology 39: 801–812

Galton, P.M. 1985. Cranial anatomy of the prosauropod dinosaur Plateosaurus from the Knollenmergel (Middle Keuper, Upper Triassic) of Germany. II. All the cranial material and details of soft−part anatomy. Geologica et Palaeontologica 19:119–159

Galton, P.M., Upchurch, P. 2004. Prosauropoda. In: Weishampel, D.B., Dodson, P., Osmólska, H., editors. The Dinosauria. Berkeley, California: University of California Press. pp. 232–258

Georgi, J.A., Sipla, J.S., Forster, C.A. 2013. Turning semicircular canal function on its head: Dinosaurs and a novel vestibular analysis. PLoS ONE 8(3): e58517. doi:10.1371/journal.pone.0058517

Mallison, H. 2010a. The digital Plateosaurus I: body mass, mass distribution and posture assessed using CAD and CAE on a digitally mounted complete skeleton. Palaeontologia Electronica 13.2.8A http://www.palaeo-electronica.org/2010_2/198/index.html

Mallison, H. 2010b. The digital Plateosaurus II: an assessment of the range of motion of the limbs and vertebral column and of previous reconstructions using a digital skeletal mount. Acta Palaeontologica Polonica 55(3):433-458. doi:10.4202/app.2009.0075, http://www.app.pan.pl/article/item/app20090075.html

Seebacher, F. 2001. A new method to calculate allometric length-mass relationships of dinosaurs. Journal of Vertebrate Paleontology 21:51–60.

Sereno, P.C., Zhao, X., Brown, L., Lin, T. 2007. New psittacosaurid highlights skull enlargement in horned dinosaurs. Acta Palaeontologica Polonica 52:275–284

Advertisement