Re: Speed Potential in Tyrannosaurs (long) [resent plain text]

[John R Hutchinson <jrhutch@stanford.edu>]

[message got truncated before, I hear, so here's plain text version, which 
hopefully will make it thru]:

Hi folks,
        I'll provide a reply to GS Paul's comments here, but don't intend 
to get drawn into another longwinded debate.  As I've said before, the 
details as usual will be worked out over the next few years/decades in 
peer-reviewed journals.  But it's always important to talk informally to 
non-specialists and specialists alike about scientific debates.  To me, the 
paper is ancient history already, and I have long since moved on to other 
projects, including more realistic 3D models of dinosaur locomotion, and 
the usual experimental + modeling studies on living animals too.  But I 
understand if people want to have their say about this controversy, and I 
welcome disagreement or agreement.  [Sure, I like the latter more than the 
former; I'm only human!]
        First off, I appreciate Greg's point of view.  Again, I welcome 
debate as long as it is about the science itself and is presented in a 
collegial fashion.  No one in their right mind has fun when scientific 
arguments turn nastily sarcastic or personal.  I like Greg's hypothesis 
that tyrannosaurs could run fast; it's provocative and creative, and has 
some data consistent with it, and Greg has done a good job of marshaling 
evidence and presenting logical arguments.  As I've said often, I'll be 
ready to admit I'm wrong if someone shows me data that prove me wrong.  I 
haven't seen it yet, though, not in Greg's post or any others since our 
paper came out.  Maybe Bakker's SVP talk will show some unambiguous 
fast-running tracks, whose speeds can be bounded with a confidence interval 
to lie at or above 11m/s for a large tyrannosaur.  I won't be at SVP this 
year as I am attending biomechanics/biology conferences for once this year, 
but it sounds like a wild time there in OK.
        OK, on to business.  I still think the fast-running tyrannosaur 
hypothesis is incorrect, and new independent studies I've done continue to 
show support for what Mariano and I found in the Nature paper.  I agree 
with a lot more of Greg's comments in his post than he probably thinks, but 
we do disagree on the fundamentals, it seems.  The new paper, which should 
be out sometime next year, is longer and explains things a lot more 
clearly, with more data and taxa too.  It should help people understand 
what we did; most of the arguments I've seen against our paper stem from a 
lack of understanding and/or an a priori dismissal of our methods.  And 
yes, my conclusions do not change significantly, but I also have some new 
conclusions that are broader in scope.
        Specific replies to Greg's post:

1. Small tyrannosaur: I think Greg is overemphasizing this "error."  As we 
noted in the paper, and have since substantiated with new approaches, the 
rather high T value for the small tyrannosaur was for a crouched pose; T 
dropped to 5% for a more straight-legged pose, and I am now sure it could 
be even lower than that.  The Coelophysis model could drop to less than 1% 
via the same strategy, but we showed how Tyrannosaurus couldn't get nearly 
that low. This, and the models of a human, chicken, and gator already 
fulfilled (to our satisfaction, and the reviewers') Greg's demand for 
"noncontroversial examples" and that "the method should have been adjusted 
until reasonable results were obtained."  I've done the ostrich model now 
and, surprise, it's a runner!  Glad I did it, though, now we have a 3D 
ostrich model based on a real animal, and many more to come.  Greg comments 
that "the belief that increased leg flexion results in increased leg muscle 
mass has yet to be demonstrated by measurements, especially of animals of 
similar size and locomotary potential."   I'm not sure what he means here; 
surely if animals bend their legs their muscle mass does not increase, 
violating the laws of energy/mass conservation.  What happens when animals 
bend their legs, as Andrew Biewener and others have shown, is that the 
ratio of the GRF moment arm (R) to the muscle moment arm (r) increases, 
decreasing effective mechanical advantage and increasing the effort 
required from the muscles.  This is extremely well substantiated in the 
literature; I know of no contradictory evidence and the method is soundly 
based on first-principles.  Now, increased muscle effort (force) could 
require more muscle mass if the muscles were exerting themselves maximally 
before the limbs were bent, but increased muscle mass can only be attained 
through exercise or evolution.  Humans "Groucho running" with bent legs do 
require ~50% more energy precisely because the ratio of R/r is higher; look 
into it and see; it follows directly from the biomechanics.  Thus it's not 
"meaningless", even though it is not "optimal".

2. Energetic arguments:  Our paper focused on the mechanics, not 
energetics, of running tyrannosaurs, although as a side comment we did note 
that a fast-walking tyrannosaur would have needed a surprisingly high % of 
active muscle volume.  Regardless, the paper stands or falls based on the 
mechanical arguments: in order to run fast, an animal's muscles must be 
able to exert enough force to prevent limb collapse, and that force is 
dependent on muscle cross-sectional area, and hence mass.  I don't think we 
know enough to say that "If there is one thing we can be certain of, it is 
that adult Tyrannosaurus moved with the same, low level of mass specific 
power output observed in elephants."  Read the energetics scaling 
literature (which has a better size range for walking than for running, 
unfortunately) and notice the strong secondary signal for many 
animals.  They don't all fall on the line.  Therefore we should not expect 
scaling "laws" to dictate exactly what the metabolic cost of locomotion was 
in a particular animal, although with an energetics-based model Greg could 
estimate it and do sensitivity analysis to see how sure he is.

3. Scaling:  Size factors into the scaling equation for the muscle mass to 
run fast, because if you collapse our equation 1, it reads simply as T ~ 
(L*R)/r. Therefore, T should scale as two linear dimensions divided by 
third linear dimension, or hence T ~ mbody^0.33.  In a tricky way, then, T 
is not so much dependent on the exact body mass value that we input, but 
rather on the linear dimensions of the animal.  So size matters, but body 
mass value does not matter in our equation, in answer to previous 
queries.  Again, this all follows from simple mechanics and validation from 
decades of animal research.  As we noted on our webpage 
(http://tam.cornell.edu/students/garcia/.trex_www/naturepaper.html) (caught 
it too late for fixing in the paper), the giant chicken scaling line was 
indeed off base a bit, because two joint angles in the giant chicken were 
accidentally put 5 degrees different from the small chicken (learned my 
lesson about having too many Excel files at once!).  The scaling is 
relatively linear (not exactly, because of significant figure rounding) 
when the correction we noted is made.  We've checked through all of the 
math in the paper now many times, and none of the corrections change our 
main conclusions.  The scaling of extensor muscle fiber lengths (L) is an 
interesting question and we are looking at it; it's more complex than Greg 
portrays it.

4. Absolute speeds vs. relative speeds:  Our paper was mainly about 
relative speeds (Froude number).  While in some cases in mammals, maximum 
absolute speed does not change much with body mass, maximum relative speed 
clearly does decline.  The interesting question to me is, when does max 
relative speed decline so much that a certain gait or absolute speed 
becomes unattainable due to biomechanical constraints?  That's what our 
paper touched on, and new papers will continue to investigate.  Greg says 
"Its up to those who think otherwise to produce data showing that relative 
leg muscle mass increases in animals of similar top speed as size 
increases. Good luck." When I hear statements like this, I can't help but 
think that people are too lazy or dogmatic to collect their own data and 
test their own assumptions and ideas.  I think the burden is just as much 
on Greg et al. as it is on me to dissect real animals and show how leg 
muscle mass changes with speed/size.  Looking at photos, running around the 
living room, or watching Hatari or PBS isn't a rigorous way to collect data 
or test hypotheses that involve the complex relationship between anatomy 
and dynamics.  I agree, though, that ceratomorphs would be a wonderful 
group to look at in detail; it's been on my mind for years.  Hopefully 
someday.  As a biologist, I've always tried to base my work on a foundation 
of studies of living (or recently dead) animals; it's what most 
paleontologists do the worst job of.  Hence I've worked hands-on in 
locomotor experiments with live birds, crocodiles, and elephants, and have 
dissected more tetrapods than I can shake a drumstick at.  We'll see what 
can be done in the few decades that my career will span.  [Correction to 
Greg's post: for short distances, human sprinters can reach 12m/s top 
speed, not 10, although the average speed over 100m is the very recent 
world record by Tim Montgomery, 9.78 m/s.]

5. Anatomy: "Not much faster than elephants" is the key phrase here.  The 
question is, how important is anatomy, especially so-called "cursorial 
adaptations"?  Does the anatomy of a tyrannosaur demand that it must have 
been able to run 10%, 50%, or 300% faster than an elephant of similar 
size?  I am totally unconvinced by the anatomical arguments that put 
tyrannosaur speed at 11-20 m/s; they are based on a weak, correlation and 
analogy-based reasoning that has many untested assumptions about the 
relationship of anatomy and locomotor performance. The few biomechanical 
studies I've seen that put tyrannosaurs/big theropods at those speeds have 
worse flaws than our paper did; I've gone through the math and I understand 
where they went wrong.  Our model included all of the tyrannosaur anatomy 
that was needed, and T.rex was still found wanting.  Long legs, big 
muscles, big cnemial crests, big pelvis, strings and sealing wax and other 
fancy stuff...  They did not change T enough to make tyrannosaurs into 
speed demons.  My feeling is that people who have mainly studied anatomy 
have been seduced by tunnel vision into thinking that anatomy is a 
failsafe, or at least a very reliable, indicator of locomotor 
performance.  Anatomy and performance are correlated, I am sure, but I see 
weak links in many parts of the correlation when I look at what 
quantitative effect the anatomy has.  I think biomechanics is what is 
needed to test anatomical specializations and see how important they really 
are; the work has not been done to my satisfaction.

6. Wrapup: I hear criticism of computer models most frequently from people 
that do not understand them, and second most frequently from people that do 
not use them, or do not like quantification or math.  Greg's post leads me 
to suspect that he does not understand how our model works.  I doubt that 
he could accurately explain it, which makes me wonder how he can doubt its 
results despite the seemingly contradictory evidence he cites.  Before I 
doubt a model, I go through it and see how it works out on paper (e.g., 
draw my Fig. 1 and make sure I can understand where the parameters fit in), 
and which parameters/assumptions are most influential (e.g., sensitivity 
analysis).  I do not only use models; I use every line of evidence that I 
can find, including anatomy.  I do not want to avoid or exclude any 
approach.  But time is limited, and I take the approaches that I deem to be 
most relevant and efficient.  I agree that we badly need more data from 
living animals, that's what my research program is all about.  Computer 
models in isolation are not that great.  But computer models will always be 
a useful supplement to any other line of evidence, particularly when it 
comes to extinct animals, because they allow you to test hypotheses 
indirectly without relying on time machines, analogy, or mere speculation 
and anecdotes.  You just need to be cautious when using them.  The fact 
that scientists have used the same methods we used in the Nature paper for 
decades of studying living animals gives me strong confidence in the 
applicability of these modeling methods to extinct animals, particularly 
when carefully approached with sensitivity analysis.  Failsafe?  No 
way!  I'm not that deluded.  I'm probably more anal-retentive about 
scrutinizing my models these days than I should be.  No line of evidence is 
faultless, but modeling is not useless or untrustworthy either.  Like any 
line of evidence, it's the Way that the evidence is collected, used, and 
evaluated that matters, and I'm glad that at least both Greg and I are 
committed to using science as our Way.

Have at it, folks.   :-)
John


===========================================
John R Hutchinson
NSF Postdoctoral Research Fellow
Biomechanical Engineering Division
Stanford University
Durand 209, BME
Stanford, CA 94305-4038
(650) 736-0804 lab
(415) 871-6437 cell
(650) 725-1587 fax
===========================================