So I have a new paper out and inevitably I’m going to talk a bit about it on the Musings. While I’ve had a few abstracts and the odd short paper out as sole author, this is pretty much my first proper effort in a major journal where I’m the only author. Not that I didn’t have help of course (which is what the acknowledgements are for) and I do especially want to take the opportunity up front to thank various people for their contributions and help, but most especially Susie Maidment for her help in data collection.
Right, onto the actual paper. Way back in 2010 I was looking at pterosaur tails in connection to an anuroganthid that turned up with (for one of them) an unusually long tail. This got me thinking about dinosaur tails and it struck me that while we obviously had some taxa with short tails (like Caudipteryx) and some looked pretty long (like Diplodocus) that no one seemed to have looked at just what kind of variation there was. Moreover, the more I thought about it, and the more I looked through papers and collections (and then later on asked various colleagues) the more often I came across ‘complete’ specimens that were nothing but when it came to the distal caudals. And so began my investigation into the tail lengths of the non-avian dinosaurs (though admittedly Archeopteryx sneaks into the paper as do the scansoriopterigids). Obviously the paper is there to be read, but hopefully this will serve as a quick summary and discussion of the basic points for those who can’t get it or don’t want to read it.
The first thing to note is that actually we really do have very few dinosaur fossils with complete tails. Despite a good hunt through the literature, a couple of collections, and exchanges with a number of colleagues I was able to track down very few specimens where every caudal was known. Even in things from localities like the Jehol and Solnhofen where skeletons are preserved in beautiful condition and soft tissues are common, there are actually very few specimens with every caudal vertebra preserved. Sure the sauropods might expect to do badly given how incomplete they always seem to be, and we’ve got more than a few dinosaurs known from only fragmentary remains. However, on the other hand we now have thousands of dinosaur fossils, and some species are known from dozens or even hundreds of good specimens and many of these are from sites of excellent preservation. But for all my searching and asking, I found less than 20 dinosaur specimens in total that have every caudal preserved. That’s really very low. Even things like ankylosaurs and dromaeosaurs with those lovely reinforced tails don’t seem to do any better either, complete tails are really, really rare.
Now there are a good number that are probably close to being complete with only a few distal ones missing, but obviously quite how true this may be is hard to determine. Sure there tends to be a general tapering of the size of the caudals which can give you a reasonable guess as to where it likely ends, especially if they are very small when they stop, but things like Diplodocus with it’s near endless rod-like caudals or the sudden stop in Nomingia means you could easily be wrong. In short, while a specimen like Sue we can probably have a pretty good guess how long the tail was and quite how much was missing, for plenty of other species it’s not going to be so easy. And things get worse from here.
Not only are there few dinosaurs with complete tails, but in one wonderfully illustrative case we have some major intraspecific variation. Two specimens of Leptoceratops are preserved side by side and so we can be confident that these aren’t just the same species, but are even from the same population. The problem is, one has 10 more caudals than the other, and their tails are proportionally rather different in length too. There’s quite a bit of intraspecific variation there, and indeed a look across other amniotes suggests that this is quite common – caudal counts and caudal lengths can vary a lot in tetrapod species. Tail length is sexually dimorphic in some snakes for example, and can vary a lot even in mammals.
Interspecific variation can be high too, which means it may not be safe to reconstruct missing tails from even close relatives. The wonderful little Epidexipteryx has the joint shortest tail known for any dinosaur that I found, but it’s sister taxon, Epidendrosaurus, has one of the longest tail known (and that one is incomplete and would have been longer still). While this might be an unusual case, there’s a decent bit of variation seen in a couple of other clades too.
All of this means that we need to be a fair bit more careful when talking about dinosaur tails and especially when it comes to recounting their size in terms of length. The length of a dinosaur is absolutely ubiquitous in the media as a measure of size and it turns up in a few papers too. However while some taxa are of course known absolutely in terms of their length, and many are probably about right despite being not entirely complete, others would seem to be little more than a best guess – and a best guess based on not very much to be honest. The data for sauropods in particular seems to be incredibly sparse and accounting for the inter- and intraspecific variation seen, I don’t think I’d be confident in reconstructing the tail of something like Argentinosaurs to within even a 50% error – it could be really long or very short and there’s no way of picking one over the other. Even ignoring some of the outliers, there’s a fair bit of variation there and can have quite an effect on the appearance of an animal.
Scott Hartman has been good enough to make this for me – a Spinosaurus with a short, ‘normal’ and long tail. All of these kinds of lengths can be seen in various theropods and to my mind are all plausible – indeed, we’ve been quite conservative here and could easily have copped off another hatful of caudals or plugged on a good few more and the results would still be quite plausible and within the bounds seen by other theropods. Of course note that while the length of that tail in each varies enormously, and as such, so too does the total length of the animal, the mass would not change that much. A 16 m long Spinosaurus sounds massive compared to a 12 m one, but if the only difference is in tail length, then in terms of mass there might not be much in it, just a few tens of kilos in a multi-ton animal.
So, estimating the length of a dinosaur without a mostly complete tail could give you a rather inaccurate number. There does seem to have been a fair bit of inaccurate information out there in the literature in the past with people giving ranges of caudal counts for groups when individuals were known with much higher values, and clades being described as having ‘long’ tails when they didn’t (or there was no real way to tell). However, there is a little more to this, I also did an analysis where (as far as possible, which admittedly wasn’t that far) the variation in tail length was compared to snout-vent length.
When examining living species, most biologists use snout-vent length as a proxy for how large animals are. After all, the tail length can vary a lot as we’ve seen, and even weight isn’t a great measure for a lot of living animals as it can fluctuate a lot on an annual basis, and of course isn’t available for specimens in museums. So a measure from the tip of the snout to the vent / anus is a common measure of size but we don’t seem to use it much in palaeontology (and certainly not for dinosaurs). In short therefore, we’re using a measure which not only includes a lot of variation in the tail that might screw up the results (and that most of the time we don’t know for sure anyway), but it’s not compatible with other datasets on extant taxa. The question is though, would the equivalent be any better for dinosaurs?
My simple analysis suggests so – that from the available data, tails are rather more variable in dinosaurs than the body. As for the vent, well, that we obviously don’t know exactly as a decidedly soft tissue structure so I plumped for the last sacral being a point that would be close to the vent and an unambiguous point on the skeleton that would be easily identified and would likely be preserved. This measure (snout-sacrum) is one I suggest we should start using when we want to talk about dinosaurs sizes in terms of length.
So there you have it. We don’t seem to have too many dinosaur tails, those we have suggest much inter- and intraspecific variation and so estimates of total length or using total length may not be very reliable. Snout-sacrum length is probably more reliable and in any case would bring the data in line with that used by most biologists. My final note though is an appeal – despite the work I did trying to uncover dinosaurs with complete tails, I’m sure I’ve missed some. Perhaps they’ve simply not been described, or are squirreled away in obscure journals, or are only listed as paratypes etc. I have seen a couple of things published since this work was finalised that look like the tail is complete but where the paper doesn’t actually say and it’s not entirely clear from the figures. I can’t believe that some of those massed ranks of undescribed Psittacosaurus, Protoceratops and various massed ranks of hadrosaurs and iguanodontians don’t have a few more complete ones lying around that can be measured. So if you do know of any specimens out there with complete tails (and better yet, totally complete specimens in terms of the skull and vertebral column) do please let me know. I’ve exhausted all the easily available avenues to date, but I’d love to do the analysis again with much more data. One day.
Dave Hone's Archosaur Musings 
I describe a chasmosaurine (cf. Anchiceratops) postcranium with complete tail in the recent ceratopsian volume.
Oh, very interesting, thanks! Err, don’t suppose you could send me some data / a PDF / photos / etc. could you? I really do want to start collecting this data more fully as I want to boost the power of the study and I’m sure there is more to do (though I still don’t think there’s any pachycephalosaur tails).
UA 2 (Stegoceras) might have a complete tail, but I can’t remember off the top of my head.
I’ll email the data shortly.
Really? Cool! That was one group where everywhere I looked and asked I was lucky to find something with a handful of caudals. Thanks on both counts.
Sadly, UALVP 2 does not have a complete tail (it has some anterior and posterior caudals).
Ah damn. Still, it makes the point, you have to check. And I have come across descriptions of specimens that talk about ‘complete’ tails which when you have the material in hand means “a good amount of tail preserved’.
Dave, what about counter-balancing the weight of the animal? This is just from my untrained eye, but the short tail Spinosaurus might a have balance issues.
Zhen, the shorter tail does push the center of gravity forward a bit, but that can be addressed by changes in locomotion easily enough (changes that I did not make in the image for Dave, as I wanted the tail differences to stick out).
Well that’s often said, but my understanding is that the muscles are far more important than tail lengths or pretty much every bird would fall over when you look at how their skeletons are set up, or for that matter, thinks like Caudipteryx.
The tailfeathers should also be considered when thinking about counterbalancing
Nobody will be surprised to read that distal tail incompleteness isn’t limited to dinosaurs. I’m working on a matrix where the presence of tail fin rays is a character. Lots of taxa were scored as lacking them; that’s probably correct as far as the living animals are concerned, but the fossils simply don’t show that for most of them; I had to change their scores to unknown.
(Indeed, a bunch of taxa for which the entire tail is unknown were scored as lacking tail fin rays. But that’s another kind of sloppiness, I think.)
Birds famously have largely horizontal femora, bringing the knees instead of the hips close to the center of gravity; it’s not just muscles.
Right, but in context my point was it’s not just the skeleton / gross mass distribution either and just chopping a few feet off the end of Spinosaurs would not make it fall forwards instantly – that the muscles have a major role if the load distribution relative to the femur / feet isn’t even.
Not that I’ve seen them in person but these Tyrrell panel mounts of Ornithomimus (?) RTMP 95.110.1 http://en.wikipedia.org/wiki/File:Ornithomimus_edmontonicus.jpg and Gorgosaurus RTMP 91.36.500 http://www.samizdat.qc.ca/cosmos/origines/Deluge_de_Noe/medias/Gorgosaurus_petit.jpg both seem to be sporting complete tails. As far as I know these specimens have only been published in papers dealing with cranial anatomy rather than caudal, thogh both are partially figured in the Persons-Currie paper on T.rex caudal muculature.
The Gorgosaurus got a look in (though scaled from a less than perfect source) though the Ornithomimus does look good, thanks.
I think CMN 2120 (Gorgosaurus) has a complete tail, too.
“Similarly, the proportional length of individual caudal vertebra also varies, as exemplified by Epidendrosaurus when compared with Sinosauropteryx. The former has 40 caudals (Zhang et al., 2002) preserved with a proportional tail length of 7.3, but the latter has at least 59 caudals but a proportionally shorter tail, as a result of much shorter centra.”
Zhang et al. 2002 never stated that Epidendrosaurus has 40 caudals. The line drawing (fig.1b) indicates that three proximal caudals are preserved, and even these are difficult to verify in the photo (fig.1a), at least in the resolution of the online edition. Most of the tail length is represented by a “tail imprint” with no vertebrae identified. Is “tail imprint” the only possible identification for this structure? I can think of other hypotheses (impression of elongate tail feathers similar to Epidexipteryx; trace of a shorter tail’s movement formed as the body was dragged into its final position; coincidental burrowing trace unrelated to Epidendrosaurus…), but having never seen the specimen I can’t comment on the plausibility of any of these.
Zhang et al. 2002 did publish an online supplement with a skeletal reconstruction that gives Epidendrosaurus about 27 caudals (supp.4), which could simply be a guess based on its phylogenetic position.
In Scansoriopteryx, a potential synonym of Epidendrosaurus, 22 caudals are reported but the tail is incomplete (Czerkas and Yuan 2002). So we know that Scansoriopteryx/Epidendrosaurus had a longer tail than Epidexipteryx, but I’m far from convinced that it had the proportionately longest tail of any dinosaur.
Or another alternative: What if the “proximal caudals” of Epidendrosaurus are actually displaced mid-dorsals, and your measurement for the “tail imprint” length is actually the tail + sacrum + posterior dorsals?
“but having never seen the specimen I can’t comment on the plausibility of any of these. ”
But I have and this data was taken from the holotype, and indeed I’m an author on a paper that is slowly making it way through review with a new (if short) description of the material and new interpretations. We consider this a tail and nothing else. Now we could be wrong, but (obviously) I don’t think so, and assuming we are, of the available data, it’s the longest tail going.
The citation of 40 is wrong for Zhang et al (my error) but we *have* counted them and there are that many there.
Great! I’m looking forward to reading that some day.
In what sense is Epidendrosaurus’ tail the longest, though? Tail vs. estimated snout-sacrum length? In your table of tail vs. femur length, Epidendrosaurus comes in second to Scutellosaurus.
Ah yes. That comes from not checking my own paper properly when writing a blog post (which i’ll now edit accordingly). 😉 Still, I’d correct myself to say “could well be the longest” as while (obviously, given it’s a point I make here repeatedly) it’s possible that it terminates right after what we have here and only one caudal is missing (not least when it’s the kind of thing Epidexipteryx does do), given how it tapers really quite gradually and it relatively tall when it peters out on the slab, I’d suspect there’s quite a bit more there. There are other candidates too naturally, not least Leaellynasaura which based on the information available sounds like it’s massive.
Still, at the bare minimum Epidendrosaurus is clearly one of the longest and that’s still very significant in the context of it’s sister taxon.
I had a discussion with Mickey Mortimer on the possible length of tail in Epidexipteryx hui, which unfortunately leads to some confusion: The tail is preserved in pieces, rather than intact; it is assumed by most readers and the authors that what is preserved is, in fact, the whole tail, disarticulation meaning little when no other caudals are present on the slabs. That said, the specimen clusters in phylogenetic analysis with taxa which have at least twice the number of caudal vertebrae, and this coupled with the disarticulation suggest that we should not be saying anything definitive about the proportional length of the tail.
However, there is also Grillo & Azevedo (2011), who used polynomial regression and assumptions of caudal count from close relatives to estimate caudal vertebral sequence completeness and likely missing elements and positions for preserved bodies. This may help in relieving researchers of issues in considering absolute completeness metrics and estimates.
“The tail is preserved in pieces, rather than intact”
Well sort of. Looking at my own photos and the original paper, it appears to be in all of two pieces, with the second lying only about a single caudals width from the proximal part which is continuous with the sacrum. Everything else on the specimen is basically articulated and intact (even the distal tail) suggesting that likely little disarticulation had gone on. It’s not like it’s a bunch of pieces are scattered across the slab and where all manner of bits are likely missing between the two pieces, or the specimen as a whole is a mess.
As such really I don’t think it’s an unreasonable interpretation that the tail is basically complete. One caudal could be missing, but that’s about it.
This was largely Mickey’s position. Mine is that the apparent missing material need not be the smallest possible amount of missing material, save when considering the material on it’s own. My concern is that we shouldn’t considering this species by itself, but rather in context to its phylogeny and apparent relation to Scansoriopteryx heilmannii (aka, Epidendrosaurus ningchengensis). The null hypothesis should be that the tail isn’t short: What does it take to disconfirm this hypothesis?
But you’ve also got to use the avialble evidence as best you can. I think it much more parsimonious (given the completeness and articulation of the material otherwise) that part of the tail has dissociated and moved a few mm that a long tail broke in the middle twice, the middle bit moved away and the end bit drifted to the point that it nearly joined up with the proximal part and conveniently has a centrum of the same diameter.
But even in the context of oter realted species, there;s soem decent differences in length between say the various closely realted oviraptorosaurs. We see the same thing in pterosaurs and even in cats (think of the various hyper-short tailed cats like lynx and Smilodon compared to the others) and we categorically see major differences within species (like Leptoceratops). So I don’t see the apparent problem of sister taxa with very different tail lengths.
Ah yes, since Carpenter et al. have published the description of MPC 100/1305 (the mounted Saichania skeleton), there is at least one other ankylosaurid specimen with a complete tail.
Cool, thanks for the alert! 🙂
Does the variation in caudal count / length tend to be located behind the caudofemoral attachment locations?
I didn’t look in that level of detail I’m afraid, given the inaccessibilty of the material. This is the kind of thing you could only really ook at in the two duplicates – in the case of Psittcaosaurus, I had only one specimen to work from and the other was a set of measurements in a paper, and in the case of the Leptoceratops, both are literally visible only in dorsal view and again the data came from a paper, not examination of the material.
One other piece to add to the puzzle: tails are biomechanically important organs for many extant taxa that walk, run, swim, or fly. Briefly, tail morphology and passive mechanics affect overall vertebral column oscillation frequency & amplitude. For big dinosaurs, think lateral oscillations, the side-to-side body bends associated with walking or running. Put simply, tails that are either longer, or more massive, or more flexible tend to *reduce* the natural oscillation frequency of the mechanical system to which they are attached, while tails which are shorter, lighter, or stiffer tend to increase same. This follows from a couple of basic relations that roughly hold true for beams (read: tails or vertebral columns): the natural frequency (W_n) of an oscillating beam is proportional to the square root of {the flexural stiffness, , divided by the mass, , of the beam}; the flexural stiffness of a beam is proportional to EI, where E is the Young’s modulus and I is the 2nd moment of area <which goes as L^4, the L being the distance from the neutral bending axis of the beam. (The effects on oscillation amplitude can also be predicted, though that is trickier.) One can obtain reasonable values for these and plug them in to get rough ideas for oscillation frequencies one would expect for biological beams. The main point is, tail length variation has mechanical consequences.
Further, if one thinks of a vertebral column (really, the whole body mass loaded onto and along the vertebral column) as an oscillating beam with a natural oscillation frequency, then anything else (e.g., appendages, like legs) that (1) attaches to the vertebral column, and (2) possess a length & stiffness & mass (& so it's own natural oscillation frequency) will interact with the oscillating backbone. For a first-order example, one would expect that the stereotypical oscillation frequencies of a dinosaur's vertebral column and legs operate in phase, i.e., left-right stride frequency and body-bending frequency would match, further the relative phase (when the leg swings during the body bending cycle) controls how forces are applied to (or are accepted from) the ground. Since Alexander (and others) give stride frequencies based on effective leg lengths, one could work out those numbers to see if the expected stride frequencies for the known leg length match up with expected body bending frequencies. At the very least, one might rule out crazy disparities, and begin to independently home in on useful ballpark values for tails based on those for legs.
One expects to see tail oscillation frequencies (i.e., cycling body bending during walking) that match – so, one might start with leg lengths, work out the most likely frequency range, then assume that same frequency for body bending during walking, and, using a density of 1 for flesh/bone (choose whatever your volume would be to get total mass), using a low-end young's modulus of 1MPa for muscle under tonus, you should be able to solve for tail length.
Put another way, you've got two independent ways to arrive at a natural frequency of operation for a given dinosaur, and the leg length number serves as a check for the tail length number. Cheers
” At the very least, one might rule out crazy disparities, and begin to independently home in on useful ballpark values for tails based on those for legs. ”
Probably, but then you’d need a good dataset for tails to do that with, and I don’t think we have one. Well, not yet anyway. And I think in at least a couple of cases you could be truly miles out – again, there’s some huge disparity between some closely-related taxa. And while I do know of some ongoing work on stiffness etc. that’ll be hard to account for in most tails, and of course there’s other factors like ossified tendons in ornithischians, or the rods of dromaeosaurs or the fused tails of ankylosaurs. I think this is nice in theory, but I really doubt you could effectively put it into practice and get any more reliable results.
At the Royal Belgian Institute of Natural Sciences in Brussels, they have 29 complete Iguanodon skeletons. This can hopefully increase your sample size.
The number is from a non scientific source, but you can maybe verify it by checking the collection: https://darwin.naturalsciences.be/
Hi Wim,
I do know about the Bernissat collections and I checked a couple of papers on them during my research for the paper, but no one seemed to be able to confirm if they really are totally complete. As noted above, many things are advertised as such but a few caudals are often missing and ‘complete’ turns out to mean ‘nearly complete’ which sadly is not quite good enough. I do want to check these out, but I’m not yet convinced there’s actually new data there.
So…the 16m Spinosaurus is fake?
No, nothing like that, there’s jsut a ton of uncertainty over the true length. But that’s also true of basically every dinosaur without a near-complete tail, so don’t use ‘total length’ as any kind of reliable measure of size in anything (and that goes for Giganotosaurus, Carcharodontosaurs etc. etc.).
From your post, I’m gathering that it’s most normal to find dinosaur skeletons without a tail, or just in pieces, and now in Mexico they find the opposite, a tail without a dinosaur. Looks pretty intact in the photos here. Would this be like your Spinosaur? As a rank amateur, the tail looks very aquatic, to me. I’d be interested in your take, Dave.
Well lots of skeletons are very fragmentary and with a lot of bits missing and of course the tail is a bit long unit so it’s almost inevitable that *some* bits will have gone, and that’s why they are so rare. Even the new Mexican one looks incomplete to me from what I have seen, even if it is nicely preserved.
It does *look* aquatic yes (and part of the reason people used to think hadrosaurs were semi-aquatic) but the animals lack any other apparent aquatic adaptations and more importantly although you can’t see it here, the tail would be bound up in ossified tendons that would stop it moving side-to-side that much so it would be a very poor paddle.