News,
An interview with palaeontologist Hady George, a PhD student in Professor Emily Rayfield’s group at the University of Bristol, who has worked on the endocasts of stem-mammals (George et al., 2024) and dinosaurs including T. rex (Caspar et al., 2024). In this article series I will summarise some of what we know, how we know it, and why it matters with the help of a few experts in the field.
The brain of an adult coelacanth fills less than 1% of its skull cavity. And yet, from fossils alone, we may have an idea of what a T. rex brain looked like. How? The answer lies in a field called palaeoneurology. To investigate nervous systems or brains of the past, we have to study indirect evidence, the footprints they leave behind, in the form of a cranial cavity where the brain used to be. What palaeontologists do these days is investigate the cavity that would house the brain. They create a representation of that cavity called an endocast. This endocast can then potentially represent the actual brain shape. How accurate is it though?
“It depends on the organism. In modern mammals and birds, the brain is so inflated that it almost entirely occupies the endocranial cavity, thus the endocast very accurately reflects brain shape and size. Often, they are such a close match that you can even clearly see the various gyri and sulci,” Hady George told me.
“In reptiles such as crocodilians for example, the endocasts of adults are a far cry from their true brains. In adult coelacanth fish, the discrepancy is ridiculous: just under 1% of the endocranial cavity is occupied by the brain.”
This is a big problem and limitation of this method. Because we are using endocasts to investigate, for instance, the T. rex brain, we cannot be sure about the size of the brain, as it may be multiple times smaller than the endocast itself. Depending on the particular species we are investigating, we can only make assumptions based on living relatives. In T. rex’s case, these are crocodilians and birds. As Hady mentioned, there can be a large difference, like in the case of the crocodilians, or a small one like in birds.
“Interestingly, the endocasts of various juvenile reptiles and coelacanths much more closely match their brains in terms of size and shape, so perhaps the utility of the endocast as a proxy for the brain generally decreases as development progresses.”
“This suggests that at some point in the evolutionary history of both mammals and birds, as their brains enlarged, they eventually evolved endocasts that are great matches for their brains.”
Does the endocast of theropods, for instance, like T. rex, match nicely with the brain size? Did the brain size/endocast ratio shift in theropod dinos, or did it happen in birds later? Hady pointed to where this shift becomes visible in the fossil record, but acknowledged the fundamental limitation:
“Of course, as actual brain tissue almost never fossilises, we might never prove exactly when this happened.”
There is some good news, however. In some cases, the brain has enlarged so much that the fingerprints of the brain show up on the inside of the cavity. That can be evidence that the endocast represents brain size relatively accurately.
“We can see where enlarged forebrains and creases representing sulci or similar structures begin to show up in endocasts, and we can infer this can only occur if the margins of the brain are (indirectly thanks to meningeal tissues) pressing up against the surrounding skull bones.”
“With regards to birds, we start to see endocasts with these traits in dinosaurs closely related to ‘raptors’, or more technically, in the Maniraptoriformes lineage. With regards to mammals, we see this in the Probainognathia lineage.”
In other words, a brain pressing against its skull leaves a readable signature in stone. And that insight, that the brain’s contours can be read from the inside of the skull, is precisely where this field began.
The evolution of endocast modelling
In the 1920s, a researcher by the name of Tilly Edinger (1897–1967) essentially singlehandedly founded this field of palaeoneurology (Buchholtz and Seyfarth, 2001). She was a Jewish German-American palaeontologist working in the museum in Frankfurt up until she fled from Nazi Germany to the US, where she continued her work at the Museum of Comparative Zoology, Harvard. Imagine the difficulty she faced as a Jewish woman in Nazi Germany, and despite all that upheaval, the discipline she built survived. An achievement very few can claim.
Her main insight into palaeontology was that brains leave an imprint on the inside of the cranial cavity. This could be studied across species and time periods for purposes of comparative zoology, to further understand the evolution of nervous systems. However, natural endocasts are an extreme rarity. They occur when sediment fills the inside of the cavity and mineralises. Some researchers made artificial endocasts by pouring solutions, like latex or plaster, into cavities (Balanoff et al., 2016). Up until the 1980s, researchers actually had to crack open the fossil skulls to investigate the cranial cavity. This was a harsh method which severely damaged rare fossils. Then, a technological revolution entered the field: computed tomography (CT).
In the 1980s, Tate and Cann (1982) and Conroy and Vannier (1984) pioneered the use of CT scans on fossils. This resolved the destruction problem: cavities could be mapped without damaging the fossil. CT lets researchers see inside a fossil without touching it, specimens stay intact and museum-ready, and scans can be revisited forever. The method uses digital ‘segmentation’ of the cranial cavity from scan slices and interpolation between them to build a complete 3D model of the cranial cavity (Balanoff et al., 2016). Brochu (2000) produced the first digitally rendered endocast of a T. rex skull, a fitting mascot for the technique’s public appeal, as this is the most famous dinosaur. By this time, CT scanning had become routine in palaeontology.
But a digital mould is still a mould. The question of accuracy, whether the endocast represents the brain anatomy, remained. To answer it, researchers turned to living animals.
“An approach scientists generally use is staining specimens of living animals with chemicals such as iodine (often referred to as Lugol’s solution), phosphotungstic acid (PTA), and phosphomolybdic acid (PMA) before scanning them.”
“These stains cause soft tissues to show up in the scan data, allowing researchers to create digital models of brains, which can be compared directly to endocasts.”
“In super rare cases, brain tissue can fossilise, allowing us to even make such comparisons in extinct animals.”
If we find fossils to fill these gaps, we may be able to infer when the brain-size/endocast ratio shifted in downstream lineages. However, even if researchers manage to do that, how much can we find out about the brain anatomy/physiology?
“Even when endocasts are a great match for the brain, the internal anatomy of the brain cannot be determined. So, identifying anatomical components such as individual neural circuits is essentially impossible.”
“Additionally, some of the walls of an endocranial cavity are often not ossified. In stem-mammals I have worked on, the ventral (=bottom) surface of their braincases was cartilaginous, and consequently didn’t fossilise.”
“In such cases, we have to estimate how far ventrally does the endocast descend.”
“Of course, there are other relevant limitations that are not unique to endocast reconstruction. Many fossils are badly crushed or sheared, which can distort the shape of the endocast.”
How much faith can we put in a mould?
So, the core evidence about ancient brains is the endocast, which is not always an accurate representation of the brain or its morphology. It requires collaborations across fields and expertise to develop further methodologies and technologies, on top of new findings. However, at this moment, endocasts are the strongest tool to create a window to the past and elucidate nervous system evolution. If researchers can create many more, faster, we might gain the information needed to build a more accurate picture.
“By automating the method by which we create digital models from scan data using AI tools, we might be able to vastly increase the efficiency at which we produce digital endocasts. Sooner than we know it, we may have huge databases of endocasts to work with.”
Imagine being able to analyse an enormous number of endocasts from different species. We would be able to make complex analyses that could elucidate patterns hidden to individual investigations.
“Furthermore, deep-learning techniques are also beginning to automate how we quantitatively analyse endocasts, particularly through geometric morphometrics, a technique that involves characterising a shape through digital landmarks with coordinate data. These techniques are still in their infancy, but AI will no doubt play a role in the future of this subfield like it will in other sciences and beyond.”
However, technology is only one piece of the puzzle. Who should wield it? Perhaps more of us neuroscientists, for instance, could contribute with our findings and expertise in collaborations with palaeontologists.
“Palaeoneurology is a fascinating discipline that brings together people from all sorts of academic backgrounds. Palaeontologists, neuroscientists, and zoologists interested in animal behaviour must combine forces to achieve a holistic understanding of the cognition and intelligence of animals lost to deep time.”
“We are starting to see more of this interdisciplinary collaboration, and hopefully we can eventually start finding links between behaviours of modern animals and their brains’ neural circuits that we can eventually indirectly link to endocast-based data. From there, we can potentially extrapolate to the past to make educated guesses about the evolutionary history of intelligence, sociality, problem solving skills, etc. So far, the water is still muddy, especially since we have a fairly limited understanding of how variation in neurons is related to behaviours in living animals. There’s very little we can currently definitively say about the cognition of extinct animals, but we can be hopeful this will change.”
Because of the destructive nature of time and the conditions that remains undergo across millions of years, fossils that perfectly preserve an animal’s anatomy as it was in life are an extreme rarity. We are left with indirect evidence that unfortunately does not quite satiate our appetite for answers. It remains to be seen where the field takes us further, and what the next unearthed fossil may elucidate.
In Part II, I explore with Dr Kai Caspar what happens when this imperfect evidence meets our deepest curiosity: just how smart was T. rex, really?
Disclaimer: The interviews were conducted by email, and quotes are reproduced verbatim as approved by the interviewees. Their views are their own and not necessarily those of their institutions. Summaries of scientific studies are based on the original publications cited in the reference list, and readers are encouraged to consult those sources directly. The author is a neuroscientist by training and, as noted in the text, palaeontology falls outside his area of expertise. This article is an honest attempt to report the field accurately, not a comprehensive review of it.
Declaration of AI use: This article was written with the assistance of Lumo AI for grammar, spell checking, and proofreading.
References
Balanoff, A.M., Bever, G.S., Colbert, M.W., Clarke, J.A. and Field, D.J., 2016. Best practices for digitally constructing endocranial casts: examples from birds and their dinosaurian relatives. Journal of Anatomy 229(1), 173-190. https://doi.org/10.1111/joa.12378
Brochu, C.A., 2000. A digitally rendered endocast for Tyrannosaurus rex. Journal of Vertebrate Paleontology 20(1), 1-6.
Buchholtz, E.A. and Seyfarth, E.-A., 2001. The Study of “Fossil Brains”: Tilly Edinger (1897-1967) and the Beginnings of Paleoneurology. BioScience 51(8), 674-682. https://doi.org/10.1641/0006-3568(2001)051[0674:TSOFBT]2.0.CO;2
Caspar, K.R., Gutiérrez-Ibáñez, C., Bertrand, O.C., Carr, T.D., Colbourne, J.A.D., Erb, A., George, H., Holtz, T.R. Jr., Naish, D., Wylie, D.R. and Hurlburt, G.R., 2024. How smart was T. rex? Testing claims of exceptional cognition in dinosaurs and the application of neuron count estimates in palaeontological research. The Anatomical Record 307(12), 3685-3716. https://doi.org/10.1002/ar.25459
Conroy, G.C. and Vannier, M.W., 1984. Noninvasive three-dimensional computer imaging of matrix-filled fossil skulls by high-resolution computed tomography. Science 226(4673), 456-458. https://doi.org/10.1126/science.226.4673.456
George, H., Kammerer, C.F., Foffa, D., Clark, N.D.L. and Brusatte, S.L., 2024. Micro-CT data reveal new information on the craniomandibular and neuroanatomy of the dicynodont Gordonia (Therapsida: Anomodontia) from the late Permian of Scotland. Zoological Journal of the Linnean Society 203(3), zlae065.https://doi.org/10.1093/zoolinnean/zlae065
Tate, J.R. and Cann, C.E., 1982. High-resolution computed tomography for the comparative study of fossil and extant bone. American Journal of Physical Anthropology 58(1), 67-73. https://doi.org/10.1002/ajpa.1330580108
Leave a Reply