New Dunk article
Resizing the Dunk: Russell Engelman’s view
Photo: me and Dunk specimen CMNH 5936, important in Engelman's paper, at Denver Museum of Nature and Science.
Friends,
Russell Engelman’s paper revising the body proportions of Dunkleosteus
terrelli, and thus its relatives in the gang called the arthrodires, made
major waves in palaeoichthyology. We’re
used to seeing Dunkleosteus length estimates anywhere from 6 to 10
meters, 6-8 being most common. For comparison, the largest proven Great White
shark (GWS) is slightly under 7 m. (Since someone will ask, I’m citing Ellis and McCosker.)
The paper involved is "A Devonian Fish Tale: A New Method of Body Length Estimation Suggests Much Smaller Sizes for Dunkleosteus terrelli" in the journal Diversity.
It built on Engelman’s earlier paper, Giant, swimming mouths: oral
dimensions of extant sharks do not accurately predict body size in Dunkleosteus
terrelli (Placodermi: Arthrodira) [PeerJ] ) That one, critiquing a method of
estimating Dunk size, was much less of a bombshell and sailed through peer
review without major objections, See the “Review” section on the paper link just
mentioned.
Now to swipe a bit from the abstract of “A Devonian Fish Tale,”
(footnotes removed, see original). For those of us outside scientific academia,
it’s a long abstract with a lot of technical terminology. Some key sentences:
Dunkleosteus
terrelli is a large arthrodire [placoderm] best known from the latest Devonian
(late Famennian) Cleveland Shale of Ohio, USA…Arthrodires combine an ossified
head and thoracic armor with a mostly cartilaginous post-thoracic skeleton
(which includes the caudal region and major fins). Thus, the head and thoracic
armor of arthrodires are frequently preserved in the fossil record but the rest
of the body is typically lost during fossilization.
Post-thoracic
remains are only known for a handful of arthrodire taxa. The best known of
these is Coccosteus cuspidatus,… but
estimating the size of Dunkleosteus using smaller arthrodires requires a
significant degree of extrapolation. [Coccsosteus
is barely 40 cm long.] The most recent attempt to estimate the size of
Dunkleosteus [used] a proxy known to reliably predict size in large sharks
(upper jaw perimeter) [but] these estimates …fail to control for anatomical
differences between arthrodires and elasmobranchs (specifically, arthrodires
having much larger mouths relative to body size).
Thus, it is clear
a new method is needed to estimate size in Dunkleosteus and other arthrodires.
Specifically, this method must accurately estimate length across fishes in
general (e.g., lampreys, chondrichthyans, and bony fishes), be measurable in
arthrodire fossils, and provide accurate length estimates for arthrodires known
from complete remains. One such potential proxy is orbit-opercular length
(hereafter OOL). This is the length from the anterior margin of the orbit to
the posterior margin of the head. There are several biological reasons to
believe OOL and total length would be highly correlated in fishes. …Here, I
estimate body size (total length) and body mass (weight) in Dunkleosteus using
a broad sample of extant fishes as well as arthrodires for which complete
remains are known…
There are
currently no individuals in the hypodigm of D. terrelli that could potentially
pertain to an individual 4.7 m or greater in length, as would be necessary to
produce the monstrous sizes reported for this species in previous studies.
Indeed, larger (=likely adult) individuals of Dunkleosteus tend to be closer in
size to CMNH 5768 than CMNH 5936. Therefore, although the maximal length of D.
terrelli is likely ~4.1 m, 3.4 m is a more typical adult size for this species.
(Those are specimen numbers for skull/armor
fossils in Cleveland Museum of Natural History, which is basically Dunk
Central.)
Naturally, this created quite a maelstrom. I contacted several experts,
including Engelman. Russell was very generous with his time and answered all my
questions in a series of emails in July 2023.
This article covers that Q&A, condensed for space and to avoid
repetition. This isn’t a point-by-point of clashing viewpoints. I haven’t read
enough responses from other experts yet, and some are in journals I can’t
access. So this is based on an
avocational Dunk lover’s questions.
I began this very skeptical of the assertion that a specimen
estimated to be 8.9 meters long was more likely 4.1 m, even if the weight didn’t
shrink as much. Engelman’s Dunk, rather
resembling a tuna on steroids, didn’t look “right,” but more specifically it
didn’t strike me as well-balanced or controllable enough. It still doesn’t look right, but it’s not like
the oarfish and the mola-mola look right, so I was willing to listen and ask
questions.
Engelman’s silhouette Dunk image (copyright Russell Engelman)
First, some self-correction. In my writing to date, I’ve assumed
the armor mass on a big Dunk added a hundred kilograms or more to the
weight. Engelman pointed me to a figure
I’d never read – 30 kg. Even if that’s too conservative, it’s enough to make me
revise my assumptions about what mass and length of post-armor body is needed to
balance that body and how long a moment arm is needed for the tail to impart
the necessary speed. I think his work also puts the final nail in the coffin of
the old eel-like tail configuration. I doubted this for a long time, based on
the tail surface area for the kind of push a Dunk would need. This was proven
in a 2017 paper and is reinforced by Engelman’s animal, however much the latter
may look like it chased a roadrunner fish into a painted tunnel on a reef.
On to Mr. Engelman’s thinking.
MB: The paper says, “The model must
accurately estimate body size in fishes regardless of phylogeny.” Why? Why
should it not be specialized for something as strange as Dunkleosteus?
Russell:
That is exactly the point. Because Dunkleosteus
doesn't have any close living relatives, estimating its size requires a
variable that near-universally correlates with body size among both living
fishes (bony fishes and sharks) as well as extinct placoderms. Using only
placoderms creates problems because all known species of complete arthrodires
are only about 15-95 cm long, which is much smaller than even the smallest
estimates for Dunkleosteus. Using placoderms only requires significant
extrapolation of the data, and the further data is extrapolated beyond the
range of values spanned by the dataset used to create it, the less accurate the
result is.
Thus, in order to accurately estimate the length of Dunkleosteus,
you need a single line that accurately predicts length in both living fishes as
well as the few placoderms for which we have complete remains. If it worked on
sharks + bony fish but didn't work on arthrodires, it was unlikely to work on Dunkleosteus.
However, as mentioned in the paper, complete arthrodires like Millerosteus,
Coccosteus, Watsonosteus, and Amazichthys all fall along
the same regression line as other fishes.
The original point of the analysis was basically
"statistically carpet bombing." It was very, very unlikely that Dunkleosteus
would show head-body proportions outside the extreme diversity seen in extant
fishes, especially when you have things like opah at one extreme and oarfish at
the other. My original goal was to use this as a reality check to just rule out
the most extreme estimates and give a more reasonable range of possible sizes.
I had no idea beforehand head-body proportions correlated that strongly in
fishes or that by themself they would be useful for approximating size.
This is also why I checked the method by scaling off of various
armor plates or other measurements of Dunkleosteus. I got similar
results no matter what I did. Indeed, if I run the same analysis using only
arthrodires, I get lengths of only 3 m for CMNH 5768 (this is one reason why
extrapolation can be an issue). Dunkleosteus just has an abnormally deep
body compared to other arthrodires, and this is very evident if you scale their
armors to the same size (Figure 11 in my Diversity paper).
MB: I’d always assumed the head looked
like it was logically sized on a reconstruction like the full-size model at the Cleveland Museum of Natural History [the large model at CMNH] which
appears a little shorter than the closest living cartilaginous comparison, the Great
White shark.
CMNH Dunk: see link for credit, license)
Russell:
Because that model is only about 4 m (~13 feet) long. That
model is relatively accurate for the time and place; it didn't take into
account some details of arthrodire comparative anatomy. Specifically, the
pelvic fins are located too far back, they need to be moved up slightly until
they are near the end of the ventral armor. This, in turn, would result in a
shorter caudal peduncle (i.e., distance to base of tail fin), because in basal
fishes (including arthrodires) the pelvic fins are located about halfway along
the total length of the body and are just behind the center of mass. These were
things that weren't well known at the time that model was created… [The body anatomy
in] my paper was helped a lot by the publications of Amazichthys by
Jobbins et al. (2022) and Trinajstic et al. (2022)'s report of
preserved arthrodire organs.
This is one of the big take home messages from the paper(s).
Arthrodires are not sharks. They may show some broad-scale similarities to
sharks, but arthrodires are a distinct group of animals and in fact seem to
have evolved active swimming habits separately from either sharks or bony
fishes. Thus, it probably should have been expected that arthrodires would
differ in body shape and comparative anatomy from either sharks or bony fishes,
because they are independently coming up with solutions to the same
evolutionary problems. And this is kind of what we see: in some respects
arthrodires are similar to sharks, in others they are more similar to bony
fishes, and in some cases arthrodires seem to be doing something unlike almost
any living fish. It's not possible to simply slap shark anatomy onto an
arthrodire and call it a day, because arthrodire and shark body plans are very
different.
MB: Asked follow-up questions about Dunk
anatomy and the CMNH model, as it’s one of the most widespread popular images of
the Dunk and inspired many others.
Russell:
That model is about the longest Dunkleosteus could
plausibly be based on the head length data; the head/trunk armor is about the
size of CMNH 5768 but the model has a head-body proportion more similar to
Amazichthys (so ~10-20% longer than predicted by head dimensions). However, in
arthrodires, the pelvic girdle is always closely associated with the end of the
ventral armor. In that model there is a significant gap of 1 to 2 feet, Amazichthys
gets away with this because it has a very long ventral armor which allows the
position of the pelvic fins to remain consistent despite the longer body.
Dunkleosteus does not, and in fact has a very deep torso relative to the
dimensions of its armor.
With the description of preserved organs from the Gogo
arthrodires, it has become very apparent that the ventral armor of placoderms
is sort of like the ribcage in tetrapods in that it protects the organ cavity.
The ventral armor then ends very close to the pelvic girdle and anus.
Specifically, think of something like a crocodile, where the rib cage/gastralia
extend all the way to the vent of the animal. Both the anus and the pelvic fins
have a very consistent location in arthrodires, sharks, and basal bony fishes
in terms of their relative length from the tip of the snout in the animal
measured from the snout (~45-55% total length). This would mean in order to
keep the proportions more consistent with what we see in other arthrodires, the
body in that model needs to be shortened a bit such that the pelvic fins are
closer to the ventral armor. I would estimate the model is somewhere between½
to ¾ of a head too long.
The other thing is that the model …was more or less deliberately
designed to look like a shark. The back half has proportions very similar to a
carcharhinid or a lamnid, but because the torso of Dunkleosteus is so
deep in order to maintain a shark-like shape the posterior body has to be
scaled up to be a bit longer in order to maintain the same length-width ratio,
rather than allowing the body to be deeper similar to a lamnid or a thunnin.
Thus, because the reconstruction has a stereotypically shark-y shape, it looks
intuitively “right” and visually appealing to us, even though once you get familiar
with arthrodire body plans the shape looks a bit off…. keep in mind this
"shark-like" depiction was not super popular at the time and well
into the early 2000s most people still depicted Dunkleosteus as a
copy-paste of Coccosteus.
MB: I’ve always thought people relied
too much on Coccosteus. I don’t
know hydrodynamics. I do know something about aerodynamics, and you can’t blow
up a small business jet and get a working airliner.
Russell:
[A] big take home point from the study. We can no longer simply
assume that all arthrodires had the same body shape where every taxon is a
copy-paste of Coccosteus. Arthrodires spanned nearly sixty million years
of geologic time and it stands to reason they would be at least as diverse in
body shape as modern sharks. Even just comparing one to the other Dunkleosteus
clearly has a much deeper trunk armor than Coccosteus, and it seems to
be deeper in height rather than reduced in length.
Even Coccosteus doesn't really look like Coccosteus.
More complete specimens have turned up since the publication of Miles and
Westoll (1968) and they suggest that the reconstruction in Miles and Westoll
might be a little too long in the post-thoracic and tail regions than what is
seen in actual complete specimens.
MB: Is measuring the OOL, the head to
the back edge of the operculum, always an incorrect practice if it does not
line up with the gill chamber?
Russell:
OOL is literally just "head length minus snout length."
Head length has a very consistent definition in ichthyology, measured from the
snout to the end of the braincase or gill chamber, whichever is more posterior.
The posterior regions of the head (braincase and gill chamber) have some pretty
severe restrictions on how big they can be relative to the body, which means
they should have a relatively good correlation with body size. By contrast, the
snout can vary a bit more based on ecological habits (think the long snout of a
trumpetfish). The other issue is placoderms have systematically shorter snouts
than other fishes. Thus, using total head length results in systematic
underestimates of placoderm body length, whereas OOL produces values closer to
the actual specimens.
Additionally, I hadn't been aware of this at the time of my
publication but Carr et al. (2009) did some work that suggested the
heads of arthrodires aren't constructed that differently from other fishes.
Specifically, the anteroposterior locations of the vagus and glossopharyngeal
nerves are the same between placoderms and other fishes, but in placoderms the
cranial skeleton has expanded backwards such that the gill chamber is covered
by the cranium and cheek plates from the top and sides. Thus, all the major
organs are located in the same place between arthrodires and other fishes, it's
just that the external covering has been rearranged, and placoderms show pretty
much the same pattern as other fishes.
This is part of the reason I tested with several genera of
coccosteomorphs and Amazichthys. If the method worked on all of these
taxa, then the chances it would produce inaccurate lengths in Dunkleosteus
are low. And that's pretty much what happened; the method based on extant
fishes predicted length in these other taxa relatively accurately, and the
error suggested arthrodires showed similar patterns of variation to living
fishes.
The operculum not lining up with the gill chamber is an issue
restricted to groupers and a few other fish, and that's mostly because they
have a huge skin flap that extends its margins. Counting that would be like
measuring the skull of a basset hound including the ears. It's not really an
issue with placoderms.
MB: asked about some of the more
interesting models on the market.
1. The dual model on Etsy from HellbenderMuseum with
a Dunk in both body plans. (No review by me yet: )
Russell:
I think what the seller is calling the "conservative"
model is probably more accurate. The smaller model seems to have a head-body
ratio of 1:4.5, whereas the conservative one has a head-body ratio of 1:5.5,
which are the proportions I had gotten with my model.
2. The PNSO model, which I praised highly except for
dorsal fin placement, and the Paleozoo model, my all-time favorite. Matt's
Sci/Tech Blog: Dunkleosteus Model: Gorgeous New One from PNSO
(mattbille.blogspot.com) Matt's
Sci/Tech Blog: Review: A wonderful Dunkleosteus model from Paleozoo
(mattbille.blogspot.com)
Russell:
I think what's going on is that people are trying to fit the known
dimensions of the Dunk specimens with previously cited lengths of 5-6 or more
meters. The problem is if you try to fit the known dimensions of Dunkleosteus
specimens to previously cited lengths you get this extremely elongated,
eel-like animal, which looks very strange and doesn't have anatomical data to
support the unusual body shape.
I actually noticed several laypeople had noticed something was odd
with Dunkleosteus' proportions when I was reviewing previous size estimates of
Dunkleosteus because they also produced these extremely eel-y reconstructions.
However, all of them just kind of assumed their estimation had to be wrong
because 5-6 m was the typically cited length for Dunkleosteus, or that the
larger length estimates must have been based on some giant, unpublished
specimens in the bowels of the CMNH that served as the basis for those
suggested lengths of 8-10 m.
MB: asked about two interesting models, from PNSO and Paleozoo. The latter is my favorite for accuracy, but I praised PNSO highly except for placing the dorsal fin too far back.(PNSO credit Matt Bille, Paelozoo by the manufacturer)
Russell:
The PNSO [model] likely used the length estimates from Ferron et
al. for CMNH 5768 and used that to determine proportions. PNSO is usually
pretty rigorous about anatomical inferences when they make models and Ferron et
al. 2017 has been the only paper directly concerning Dunkleosteus'
appearance since 1932. [Edit: he added, "There's one more, Carr et al., 2010."]
[MB: The Ferron paper is here: https://doi.org/10.7717/peerj.4081:
I have not seen these authors’ response to Engelman’s critique of their methods
in his “Swimming Mouths” paper [https://doi.org/10.7717/peerj.15131] but following the branches of critiqued and cited papers knowledgably requires
a very long article by a true expert; as for me, “Man’s got to know his
limitations.” ]
MB: I wondered about placement of the
dorsal in your graphic: I understand unpublished Moroccan fossils show the
dorsal fin starts right behind the dorsal armor.
Russell:
Define “right behind.” The dorsal fin in my reconstruction is as
close to the trunk armor as physically possible, it just looks like it is
further posterior than it actually is due to the shorter body. The dorsal fin
in Dunkleosteus is limited in how far forward it can be placed by the
huge carinal process sticking out of the back of the mediodorsal. If the fin is
too far forward the basal (which are also a midline structure) conflict for
space with the big, bony spike and associated submedian dorsal plate extending out
of the trunk.
Also note that the dorsal fin position in my reconstruction is
considerably further forward than it is in specimens of Coccosteus,
Incisoscutum, etc. In most of the coccosteomorph arthrodires the dorsal fin
begins behind the pelvis, whereas in this reconstruction of Dunkleosteus
it begins before the pelvis, more similar to the aspinothoracidans we have
dorsal fins for. So in a sense it is further forward than in other arthrodires.
MB: I sent him my article “Dunkleosteus:
First King of the Oceans,” Prehistoric Times #126, Summer 2018.
Russell:
There are a few things you might be interested in given the
material in the Prehistoric Times article.
Article: There is an exceptional Dunk
specimen in the Cleveland Museum of Natural History estimated to have been 8.9m
long in life, making it the size of a large male orca.
Russell:
That's CMNH 5936. That's the same specimen I re-evaluated in the
length(s) papers and came up with an estimate of about 4.1 m. Depending on the
error bars, something as high as 4.5ish meters might be possible.
Article: There is some dispute about
the similar-sized Titanichthys being a filter feeder.
Russell:
From what I am aware of, there is no dispute over the
dietary habits of Titanichthys. Dunkle and Bungart (1942) suggested it
could have eaten benthic invertebrates, but then Dunkle and Bungart (1942) also
suggested it had a prehensile lip like an elephant (which no one seriously
considers anymore). More specifically, it has been suggested on and off that Titanichthys
was a filter-feeder since its discovery, but no one had actually tested this
hypothesis until Coatham et al. (2020) ran finite element analyses on
the lower jaw and found it was more similar to megaplanktonivores like basking
sharks than predatory species. Exactly how Titanichthys fed is still a
mystery, since there are multiple types of feeding strategies for
megaplanktonivores.
Article: Dunkleosteus was
larger than any of the extant sharks. Russell agreed but added interesting
stuff.
Russell:
There are some really frigging big sharks known from the later
Paleozoic. Edestus probably got around 6-7 m in length (and this seems
fairly reliable). Helicoprion and some Helicoprion-adjacent taxa have
been suggested to reach lengths of 12 m, but then again these taxa were
basically Paleozoic sperm whales. Some of the ctenacanths got fairly large, but
exactly how large isn't clear. All of this kind of relates to a broader issue
in that there are a lot of "Paleozoic fish stories," where
fragmentary remains get extrapolated to colossal sizes and it's really hard to
double-check them simply because there are so few complete specimens it is hard
to tell what the actual proportions were. This is especially the case when
claims of giant sizes are based on 50–100-year-old estimates on specimens
(i.e., back when paleontology was less quantitative and people liked to
exaggerate) that are now missing or lost and have never been revisited based on
new data.
Article: Described Dunk as strictly
marine.
Russell:
There are some passing reports of dunkleosteids (probably not Dunkleosteus)
from estuarine/freshwater sediments in Pennsylvania and West Virginia. Some are
clearly estuarine but I think at least one is in freshwater sediments. It's not
entirely clear what the osmotic regulation of arthrodires was, but it's
possible they cared a lot less about salinity than modern sharks do (more like
bony fish) as we know Coccosteus was apparently able to colonize the
Achanarras lakes from the ocean. The idea of "what is salt water in the
Devonian, really" is actually very controversial for a number of reasons.
Chondrichthyans are very picky with their salinity, they tend to stick with
either freshwater (xenacanths, freshwater rays, freshwater hybodonts) or
saltwater (everything else) and never leave due to their wacky biochemistry
(i.e., changes in osmotic tolerance seem to be hard to evolve and difficult to
reverse). Again, sharks are weird and it's not entirely clear if their
physiology should be taken as standard for placoderms, especially given sharks do
some wacky stuff with their biochemistry that circumstantial fossil evidence
suggests arthrodires may not have.
My comments:
Engelman has reopened, in a big way, the discussion of how large a
very prominent species is and what it looked like. He got his entire community talking,
(That’s a huge feat for any grad student: Colleen Canavaugh’s realization of
the symbiosis in giant tube worms comes to mind.) I agree we should consider the
Dunk as usually pictured is too long, but how much too long? I don’t think that’s
certain yet.
The easiest way to explain my current thinking is to look once
again at the models. The PNSO model, which I praised highly, is still the most
lifelike model but now looks to me like it may have an extra meter of Dunk in
there. I think the Paleozoo dunk, which places the dorsal fin further forward
and doesn’t extend the body as much, is the most authentic Dunk model out there (even better than the CollectA model, which has the dorsal and tail off but is the best-looking, or the Mojo),
and I still think so after talking to Engelman. I think there is still a bit of
wiggle room, despite the way so many fish fit close to Engelman’s line, because
we don’t have enough material from big arthodires, and other experts are having
their say on the OOL measurement. But I could be wrong. Following further scientific
debate here is going to be fascinating.
Interested in zoology? Read Of Books and Beasts: A Cryptozoologist's Library.!
Don't forget to sign up for the blog.
www.mattbilleauthor.com // Twitter: MattWriter // mattsciwriter@protonmail.com.