Time to revive this thread, I guess.
First, wanted to archive this post here:
Replying to article cited in
theworldofanimals.proboards.com/post/64494/threadThis is quite clearly factually incorrect though.
The Giganotosaurus holotype has an endocast volume of 275 cm² (Carabajal and Canale 2010), the endocast volume of T. rex FMNH PR 2081 is 414 cm² (Hurlburt et al. 2013), so even at this extreme, it is about 50% more voluminous, not twice as voluminous.
Other T. rex specimens (that are more representative of the species as a whole) have smaller endocast volumes, that of AMNH 5117 is 314 cm² (Hurlburt et al. 2013), and that of AMNH 5029 is 343 cm² (Larsson et al. 2000) or 382 cm² (Hurlburt et al. 2013; I am not sure what exactly the reason for this difference is, as they seem to have used similar methods in both cases, whereas I am disregarding the GDI estimates for brain volumes as they seem to consistently produce overestimates for the same specimens based on Hurlburt et al.’s table), so that is a difference of 15% to 39 %, not 100%.
These two are unfortunately a bit lacking when it comes to size estimates (a general problem with these brain size studies); the only one available is a mass estimate of 4312 kg for AMNH 5117 when compared to Sue’s 5634 kg (based on the outdated stylopodial equations). This is cited as "Erickson pers. comm." by Hurlburt et al., and it is not clear how exactly it was estimated (as the specimen doesn’t have a femur), but is specifically stated to base on the same methodology (so presumably some sort of scaling from a specimen that had its mass estimated from this regression). Based on that, it is about 77% the mass of Sue (i.e., 6.4 t based on Hartman’s 8.4 t sue GDI), which would make this a fairly good representation of your average T. rex (still slightly larger actually, the average T. rex being about 70% the mass of Sue), and thus likely closest to the average T. rex’ brain size.
So while this might shatter some people’s illusions, T. rex likely did not have a brain twice the size of Giganotosaurus’, but rather a brain about 15% larger for typical specimens, possibly up to 50% when comparing a particularly large T. rex and a small Giganotosaurus.
So between normal specimens, that’s an almost negligible difference, of the sort you can easily get between individuals of a single species (in fact the gap between AMNH 5117 and FMNH PR 2081 is far wider than that between the former and Giganotosaurus). Larsson et al. 2000 would suggest there might be a significant difference when it comes to cerebrum size. However, Hurlburt et al. suggest that the claims in this regard aren’t actually reproducible, with relative cerebrum size in Carcharodontosaurus being only slightly smaller than in T. rex, and relative cerebrum size in Allosaurus even being noticeably larger. Also, cerebrum size in Giganotosaurus hasn’t been quantified, with all previous comparisons basing on only Carcharodontosaurus (which has a smaller endocast overall, so naturally is expected to have a smaller cerebrum too). Perhaps this could be tried following Larsson et al.’s ellipsoid method based on the figures in Carabajal and Canale though, but I have my doubts about the solidity of that method, tbh. What’s really needed is a stronger statistical basis, incorporating specimens with more constrained mass estimates (the Giganotosaurus holotype is a lot more useful in this regard than Carcharodontosaurus), because many of the specimens analyzed are sorely lacking in this regard.
Certainly I would agree with the statement that there is a probable trend of T. rex having a slightly bigger brain and slightly bigger cerebrum than giant carcharodontosaurs of similar size, but it is not even close to the claims commonly getting tossed around, and, at this stage, not very statistically significant either. What people on the internet make of this, namely that T. rex was considerably smarter than other large theropods, is more whishful thinking than reality.
More importantly, I think this study warrants some discussion:
First of all, it has of course been properly published (Herculano-Houzel 2023) by now, though the key message stayed the same.
This is an interesting study and it calls attention to some important considerations, but there are a whole number methodological issues due to which I think the conclusions are very questionable.
1) PremiseFirstly, the author straight up dismisses any notion of any relationship between brain and body size, exclusively uses absolute brain size (except where relative brain size is useful to support her findings), and exclusively cites her own work (including opinion pieces) to support this. I won’t dispute that absolute brain size and absolute neuron counts matter, and that encephalization quotients (EQs) have limitations, but the reasoning in Herculano-Houzel’s paper does set off few alarm bells.
EQ alone certainly needs to be interpreted carefully. For example, it appears that very large extant animals, despite often having large brains and high cognitive performances (e.g. whales and elephants) generally tend to have low estimated EQs.
This is a statistical problem:
EQ is based on using regressions of a large sample of animals, used to estimate the average brain size for a given body size. This is routinely done separately for mammals, birds and non-avian reptiles (something Herculano-Houzel seems to ignore, because she cites the different scaling of brain size in these groups as evidence for ignoring the entire concept alltogether).
By the very nature of being very large, very large animals are relatively poorly represented in such a regression, and thus have a relatively little influence on its shape, so if they tend to deviate from the scaling seen in other animals, this is not properly accounted for. This gets exacerbated by the use of relatively simple (linear or log-linear) regression models here, because common practice is to extrapolate beyond the range of the data, something that always brings with it certain problems to begin with. As a result, EQ calculations are more accurate and relevant for animals similar in size to the majority of the animals for which these regressions were calculated than for animals near the upper end of the data range, or even beyond it. It just so happens that large dinosaurs are far beyond the data range of either the reptile or bird regressions, which are the ones that are relevant for them (and even among mammals,
T. rex-sized taxa are not well-represented). So yes, there are problems with EQ. But does that mean it is irrelevant alltogether, or that relative brain size should be completely ignored in favour of absolute size? Of course not!
Crucially, previous studies (e.g. the various works by Hurlburt, see below) of course did find statistically significant correlations between brain and body size (as did Herculano-Houzel herself in the very paper I am referring to, she just sort of ignores that this might hold any relevance), and why wouldn’t there be? Obviously a larger body is a necessity for a larger brain, but of course there are also small animals that have high cognitive performance, despite never being able to have brains as large es very large ones. Humans don’t have the 8 kg brain of a sperm whale, does that make us less intelligent than sperm whales? The very notion of a small animal being intelligent (e.g. a rodent) becomes difficult to justify if we assume that it is exclusively absolute brain size (or the absolute neuron count, which correlates with it) that enables high cognitive performance, and I haven’t seen this issue addressed properly (in fact all of the issues this raises are summarily ignored in the study, if not worse). Sure, a 9 ton
T. rex can have a brain the size of a baboon’s. But that doesn’t automatically mean it was equally smart to a baboon, other variables than simple absolute size of the brain play a role here, the size of the body being among them.
So I have some issues with the entire premise of the paper, but fair enough; considering absolute instead of relative brain size for a change is still a valid and valuable addition to the debate, even though I disagree that it is the end-all argument it is being presented as in the study.
However, this is far from the only problem I have with the study.
2) Inconsistent and poorly supported reasoning for choice of regression model for telencephalic neuron count ~ brain sizeDespite previously dismissing the scaling of brain size with body size as irrelevant, Herculano-Houzel then does a complete 180 and actually does use regressions for brain size~body mass. It’s just that she doesn’t use relative brain size or EQ as a measure for intelligence, instead she uses the scaling of brain size with body mass to predict how high the neuron density in that brain should be. I.e., she assumes if an animal already has a relatively small brain for its size (like non-avian reptiles), it will necessarily also have a relatively low neuron density, but if it has a large brain for its size (like birds), then it will have a high neuron density.
Based on this she assigns vastly different neuron densities to different dinosaurs based on which regression (non-avian reptiles or various birds) they fall closer to. The problem is that the exact same concerns that can rightfully be levelled against an overreliance on EQ also apply here, because she’s basically using the same concept, merely to predict something else. But truthfully, nobody really knows the neuron density of a
T. rex, what size the brain of an 8 ton bird would be, because there are no 8 ton birds (exact same problem as determining a relevant encephalization quotient here). Not only that, but it’s a bit of an argumentative leap to begin with that just because an animal has a certain brain-body ratio, it also necessarily had a specific neuron density (and if you wanted to establish this, certainly PGLS or Independent Contrasts would be called for to remove phylogenetic signal from the data and demonstrate such a correlation).
So this point in the study (the estimation of neuron densities) I would argue is pretty both self-contradictory with the previous reasoning of the same study, and rests on rather weak foundations regardless.
3) Lack of consideration for residual interspecific differences in brain proportions not explained by brain sizeMoreover, she also gives little regard to the composition or shape of the brain, for example how much of the brain is actually forebrain (this she for some reason has no qualms about estimating with universal scaling relationships), how much of the forebrain is actually cerebrum, and how much of that cerebrum is actually pallium. She just applies universal scaling relationships (which she previously criticized when it came to EQ) and sort of assumes that the total number of telencephalic neurons offers a good approximation of the number of neurons in the pallium, which is, honestly, a bit of a ridiculous proposition. Tyrannosaurids have extremely enlarged olfactory bulbs, even compared to other theropods (which generally have large olfactory bulbs to begin with), and these are also part of the telencephalon (see Zelenitsky et al. 2009 and 2011).
Here is where phylogenetic and multivariate regressions would have been important to provide a more accurate prediction for telencephalon size and neuron count than can be made based on brain size alone.
4) Conflating endocast and brain sizeBesides that, there is also a less systematic, but still relevant issue; The author is mixing various figures of endocast and estimated brain volumes in her dataset without much regard for the difference (she does sort of discuss it briefly in her methods, but then goes right back to ignoring it under the pretext that someone else supposedly did the same…though I think she may have misunderstood the study in question).
For example, her highest estimated neuron count for a dinosaur, that for
T. rex at around 3200 M, is based on a brain size of 343 ccm…the problem being that this is not actually the brain size of
T. rex, it is the endocast volume. The actual brain of the same specimen didn’t fill the entire endocast, since there are no vascular impressions on the endocast walls in most non-avian dinosaurs the way there are in animals where this is the case (e.g. birds), and if extant crocodilians are anything to go by, the brain probably filled less than half of the endocast (Hurlburt et al. 2013; though that is generally another major assumption that leads to additional uncertainty when trying to determine brain size, and down the line in trying to determine metrics of cognitive capacity from it).
So that obviously results in an incorrect estimation of brain size, which in turn leads to an incorrect estimation of the neuron count (especially considering that relative brain size was also used as a way to judge how to determine neuron density).
5) Inaccurate mass estimatesAdded on top of that, we do of course have other issues that much of the other research on dinosaur encephalization also has (see also above): Reliable mass estimates are often unavailable for the same specimens whose endocasts are available for such studies, or are available but aren’t used (e.g. the body masses for the
T. rex and
Allosaurus specimens in Herculano-Houzel’s study are more than doubtful, she scores sue with two body mass estimates, 5 t and 7.4 t, which are likely both underestimates). I’d take that as a good argument to focus on absolute brain size…but it doesn’t really apply when you use absolute brain size, but then go back to using relative brain size in order to predict neuron counts from it. So that makes this latter part even less reliable than it already is.
Summary:Overall, there are just so many argumentative leaps and assumptions in that paper that are either outright inaccurate, or at least very speculative:
>that you can accurately predict brain size from endocast size or that the difference doesn’t matter
>that you can accurately predict neuron density from relative brain size
>that the assumed brain and body masses used to inform that prediction are accurate (which they are not)
>that the universal scaling relationships she applies to overall brain mass to estimate the number of telencephalic neurons applies, despite differences in brain composition and telencephalon shape
>And that, in spite of the latter, the number of telencephalic neurons is, after all that, still an accurate predictor of cognitive capacity
And as a result, I think the author’s confidence in the end result is simply inappropriate.
As I said before, I don’t mean to imply giant dinosaurs were unintelligent. However, claiming to be able to know their neuron count with precision, and claiming to have solid evidence enabling a quantitative statement that they had primate-like intelligence is overconfident. COULD
T. rex have been as smart as a baboon? Yeah, probably. But do we have solid evidence to say that it definitely was? Far from it.
Nevertheless there are some interesting things to take away from this study (or from other, previous studies, for anyone who has been able to pay attention then already), namely that overreliance on EQ does create a bias against large animals, and that we should probably expect some of the larger dinosaurs to be among the most intelligent ones, or at least quite intelligent simply by virtue of overall brain size.
OK, so this actually got longer than I thought it would; re-reading that paper I actually noticed several issues in addition to the ones I recalled. But anyway, here’s my take on it.
lionclaws Tagging you, since you’ve cited this study twice recently and I still owed you my explanation for what I stated before regarding it not being taken very seriously among paleontologists.
---
Carabajal, A.P. and Canale, J.I. 2010. Cranial endocast of the carcharodontosaurid theropod Giganotosaurus carolinii Coria & Salgado, 1995. Neues Jahrbuch für Geologie und Paläontologie-Abhandlungen 258 (2): 249–256.
Larsson, H.C., Sereno, P.C. and Wilson, J.A. 2000. Forebrain enlargement among nonavian theropod dinosaurs. Journal of vertebrate Paleontology 20 (3): 615–618.
Herculano-Houzel, S. 2023. Theropod dinosaurs had primate-like numbers of telencephalic neurons. The Journal of Comparative Neurology.
pubmed.ncbi.nlm.nih.gov/36603059/Hurlburt, G. 1999. Comparison of body mass estimation techniques, using recent reptiles and the pelycosaur Edaphosaurus boanerges. Journal of Vertebrate Paleontology 19 (2): 338–350.
Hurlburt, G.R. 1991. New encephalization quotients for dinosaurs and other extinct vertebrates. Journal of Vertebrate Paleontology 11: 57A.
Hurlburt, G.R. 1995. Reptile, bird and mammal encephalization quotients (EQ’s) of dinosaurs and other extinct vertebrates. Journal of Vertebrate Paleontology 15: 37A-37A.
Hurlburt, G.R. 1996. Relative brain size in recent and fossil amniotes: determination and interpretation.PhD Dissertation, University of Toronto, Toronto, Canada, 250pp.
Hurlburt, G.R., Ridgely, R.C. and Witmer, L.M. 2013. Relative size of brain and cerebrum in tyrannosaurid dinosaurs: an analysis using brain-endocast quantitative relationships in extant alligators. In: Tyrannosaurid Paleobiology, 1–21.
Zelenitsky, D.K., Therrien, F. and Kobayashi, Y. 2009. Olfactory acuity in theropods: palaeobiological and evolutionary implications. Proceedings of the Royal Society B: Biological Sciences 276 (1657): 667–673.
Zelenitsky, D.K., Therrien, F., Ridgely, R.C., McGee, A.R. and Witmer, L.M. 2011. Evolution of olfaction in non-avian theropod dinosaurs and birds. Proceedings of the Royal Society B: Biological Sciences 278 (1725): 3625–3634.
