Article https://doi.org/10.1038/s41467-023-37574-x Evolution of cortical geometry and its link to function, behaviour and ecology Received: 14 March 2022 Ernst Schwartz 1, Karl-Heinz Nenning1,2, Katja Heuer 3, Nathan Jeffery4, Ornella C. Bertrand 5,6, Roberto Toro3, Gregor Kasprian1, Daniela Prayer 1 & Accepted: 22 March 2023 Georg Langs 1,7 Check for updates Studies in comparative neuroanatomy and of the fossil record demonstrate the influence of socio-ecological niches on the morphology of the cerebral 1234567890():,; cortex, but have led to oftentimes conflicting theories about its evolution. 1234567890():,; Here, we study the relationship between the shape of the cerebral cortex and the topography of its function. We establish a joint geometric representation of the cerebral cortices of ninety species of extant Euarchontoglires, including commonly used experimental model organisms. We show that variability in surface geometry relates to species’ ecology and behaviour, independent of overall brain size. Notably, ancestral shape reconstruction of the cortical sur- face and its change during evolution enables us to trace the evolutionary history of localised cortical expansions, modal segregation of brain function, and their association to behaviour and cognition. We find that individual cortical regions follow different sequences of area increase during evolu- tionary adaptations to dynamic socio-ecological niches. Anatomical correlates of this sequence of events are still observable in extant species, and relate to their current behaviour and ecology. We decompose the deep evolutionary history of the shape of the human cortical surface into spatially and temporally conscribed components with highly interpretable functional associations, highlighting the importance of considering the evolutionary history of cortical regions when studying their anatomy and function. In mammals, the cerebral cortex represents the most important ana- myeloarchitecture6, as well as structural7 and functional connectivity8 tomical structure for cognition1. As such, it is the focal point of a large are sufficiently similar to establish correspondences across species. body of neuroscientific research, investigating its numerous beha- These commonalities coincide with a remarkably stable macroscale vioural correlates as well as its exceptional evolutionary expansion topography of the mammalian cerebral cortex9–11 characterised by and progressively extended ontogeny in humans2,3. Despite dramatic anatomically and functionally defined cortical areas that evolved and evolutionary changes and diversification in the mammalian developed under strong genetic12 and ecological13 constraints. lineage4, the organisation of the cerebral cortex retains a range of However, the relationships between these organisational princi- common underlying attributes. For example, folding wavelength5, ples of cortical structure and function are yet poorly understood14. 1 Department of Biomedical Imaging and Image-guided Therapy, Computational Imaging Research Lab, Medical University of Vienna, Vienna, Austria. 2Center for Biomedical Imaging and Neuromodulation, Nathan Kline Institute, Orangeburg, NY, USA. 3Institut Pasteur, Université Paris Cité, Unité de Neuroanatomie Appliquée et Théorique, F-75015 Paris, France. 4Institute of Life Course & Medical Sciences, University of Liverpool, Liverpool, England. 5Institut Català de Paleontologia Miquel Crusafont, Universitat Autònoma de Barcelona Edifici ICTA-ICP, c/ Columnes s/n, Campus de la UAB, 08193 Cerdanyola del Vallès., Barcelona, Spain. 6School of GeoSciences, University of Edinburgh, Grant Institute, Edinburgh, Scotland EH9 3FE, United Kingdom. 7Computer Science and Artificial Intelligence Lab, Massachusetts Institute of Technology, Cambridge, MA, USA. e-mail: georg.langs@meduniwien.ac.at Nature Communications | (2023)14:2252 1 Article https://doi.org/10.1038/s41467-023-37574-x Aside from insights into evolution, understanding the nature and realisations of a tip-dated Bayesian model of divergence times39 on causes of the brains’ phenotypic variability is relevant when translating which we estimated the ancestral states of the area of the cortical neuroscientific results in model animals to humans. Here, studying the surfaces and their convex hulls40 as proxies for cortical morphology evolutionary history of the brain poses a promising avenue of (Supplementary Figs. 1 and 2). We then performed pairwise matching research15,16. between the cortical surfaces of sister species, and approximated their The heterogeneous change of the extent and number of indivi- ancestral state by interpolating in the space of smooth shells41,42. There, dual cortical areas during evolution17 has spurred a debate on whether we used the relative differences in the global shape features to these changes of function and topography occurred in a concerted determine the interpolation factor. The relative position of the esti- manner18 or are better described as a mosaic of dynamically changing mated ancestral and observed daughter species in the nonlinear space areas19. Comparative studies of both extinct and extant species have of smooth shells thus corresponds to the euclidean distance between been used to study the evolutionary history of the cerebral cortex. Due the corresponding global shape features (Fig. 2). We iterated these to taphonomic processes, the estimation of this history with reference steps, progressing backwards in the phylogenetic tree. Upon reaching to extinct taxa has come to heavily rely on measures derived from its root node, we resampled all surfaces to a common topology by cranial endocasts. Such measures of endocranial volume3, shape20, or placing surface vertices at geometrically corresponding locations43. osteological landmarks21 all represent the outer meningeal surface and Details of the modelling and analytic workflow are provided in thus lack fidelity regarding the true cortical surface. This loss in ana- the Supplementary Methods. We evaluated the appropriateness of the tomic precision limits the translation across both extant and extinct resulting common reference frame by relating measurements taxa and has led to conflicting theories and continued controversy obtained from the inherent group correspondences to established regarding, for example, primate brain evolution22–24. knowledge from evolutionary neuroscience. The recent availability of large open image datasets of rodents25, non-human primates26,27, human subjects28,29 and fossil taxa30 has Neuroecology of cortical morphology facilitated comparative analyses of brain organisation and function31,32. The relationship between the morphology of a-priori defined cortical However, studies are either limited to phylogenetically closely related areas and sensory specialisation has been studied extensively in a species33,34, a-priori defined spatial homologies6,35, or avoid establish- variety of species, both in their natural state as well as by manipulation ing correspondences between morphologically and functionally such as post-natal enucleation44. Using the common cortical reference divergent domains altogether by focusing on global descriptions of space, we performed a data-driven analysis to study if the relationship brain organisation5,36–38. between the morphology of cortical areas and sensory specialisation Addressing these shortcomings and adequately studying the dis- can be observed at the whole-brain level. tributed patterns and sources of cortical variability and evolution First, we performed phylogenetic principal component analysis requires a common analytical framework31 that bridges the gaps (pPCA)45 to reduce the dimensionality of our dataset and tested for the between the analysis of humans, extant mammals and the fossil record. association between the extracted modes of variation describing the Here, we link phylogenetic and geometric methods to establish a shape of the cerebral cortex and ecological and behavioural niches. We common reference frame for the cerebral cortex of Euarchontoglires, found 5 pPCA components out of 21 (Supplementary Fig. 2) that showed the supra-order containing both humans as well as commonly used significant (q < 0.05, Wilcoxon-Rank-Sum/Mann-Whitney U-Test, FDR neuroscientific model animals such as mice, rats, macaques and mar- corrected) association for any of the observed ecological and beha- mosets. By establishing group-wise correspondences between all vioural variables of arboreality, terrestriality, fossoriality or large group cortical surfaces in a dataset of ninety different species, we show how size. Despite the strong overlap with group size, no component was ecological adaptations are reflected in the shape of the surfaces of associated with diurnality. We then aggregated the information con- cerebral cortices of living members of Euarchontoglires. We use spatial tained in the pPCA modes that showed a significant relationship to statistics to quantify the process of modal segregation of brain func- individual ecological factors (Supplementary Fig. 2). Based on these tion. Finally, we combine ancestral state reconstruction with statistical factors, we estimated the areal differences corresponding to each decoding of brain activation to derive a historical time-line of modules ecological variable and decoded the resulting expansion maps into of evolutionary adaptations of the cortical shape in the deep ancestral neuroscientific concepts (Fig. 3; Supplementary Data 3). human lineage since the last common ancestor of rodents and pri- By mapping the relative expansion patterns into a functionally mates. The resulting aligned surface meshes for all extant species, their defined parcellation of the human cerebral cortex46, we found sig- coordinates for commonly used template spaces and tools for map- nificant differences (Kruskall-Wallis, p < 0.001, Supplementary Data 5) ping between them as well as their estimated ancestral shapes are in the cortical expansion patterns associated with every investigated available at https://github.com/cirmuw/EvolutionOfCorticalShape. ecological and behavioural variable except diurnality. The largest effects were observed for ecological variables of habitat (fossorial: Results p < 1e-6, η2 = 0.21, terrestrial: p < 1e-6, η2 = 0.14, arboreal: p < 1e-6, An evolutionary common reference frame of cortical geometry η2 = 0.18) whereas large group size showed lower effect (p < 1e- for comparative neuroscience 6, η2 = 0.1). We established a common reference frame to map correspondences Post-hoc tests revealed significantly differing expansion patterns between densely sampled cortical surface models of ninety extant associated with both ecological and behavioural variables. Visual and species of Primates (58), Rodentia (28), Lagomorpha (2), Dermoptera dorsal attention areas are most strongly affected by habitat, where (1) and Scandentia (1) (Fig. 1; Supplementary Data 1a). For this, we first they are reduced in fossorial and expanded in arboreal species. Their performed manual segmentations of the cortical surfaces on available relative expansion is also positively associated with sociality (i.e., large MRI (75), DiceCT (7) and serial histology (8) data. We computed the social groups), while limbic areas show a negative relationship (Sup- convex hull of these segmentations as the simplest covering of the plementary Fig. 5). We found a significant correlation between the outer surface without any folds or creases. The surface areas of the pairwise effect measures for all variables related to habitat (fossor- resulting as well as the original model serve as global features of the ial:terrestrial: r(19) = 0.592, p = 7.74e-3, 95% CI = [0.145, 0.839], fossor- cortical geometry. We furthermore collected simplified descriptions of ial:arboreal: r(19) = 0.635, p = 4.33e-3, 95% CI = [0.206, 0.859], species behavioural ecology in terms of sociality, activity pattern and terrestrial:arboreal: r(19) = 0.933, p < =1e-6, 95% CI = [0.797, 0.979], preferred habitat (Fig. 1; Supplementary Data 2). We obtained a cali- Supplementary Data 4a) that implied an ordering in terms of habitat brated phylogeny for the analysis from the consensus tree of 100 complexity (Figs. 4a and 6b). We performed a joint analysis of the Nature Communications | (2023)14:2252 2 Article https://doi.org/10.1038/s41467-023-37574-x Fig. 1 | Phylogenetic tree of the dataset used in this study. The dataset covers The resulting surface models are compared with endocasts (Supplementary 90 species (58 primates, 28 rodents, 2 lagomorphs, 1 scandentia, 1 dermoptera), for Data 1b) obtained from phylogenetically closely related fossil endocasts at the which 5 binary ecological and behavioural categorical indicator variables were branching points indicated by the letters A-I, highlighting the morphological collected (Large group size, diurnality, arboreality, terrestriality and fossoriality, plausibility of the results and consequently the underlying surface see Supplementary Datas 1a, 2a and 2b). Estimates of the morphology of potential correspondences. ancestral shapes were computed after iterative pairwise alignment of sister species. habitats ordered from arboreal to terrestrial to fossorial species become more attuned to specific types of sensory processing49,50. In (Fig. 4c; Supplementary Data 4b) that revealed a sequential decrease in extant species, this adaptation was successful and resulted in increases relative cortical expansion in visual, motor, frontal parietal as well as in overall body and brain size. The combination of these effects leads default mode related areas. We quantified the consistency of these to an expansion of cross-modal processing regions and a segregation effects using repeated measurements correlation47 (Supplementary of sensory processing regions51. Data 4c), which were significant in all but the ventral attention areas In order to test if the proposed common reference frame (rrm(241) = −0.048, p = 0.45, 95% CI [−0.174, 0.078]), whereas visual reflects this cortical organising principle, we projected a map of (rrm(323) = −0.802, p < 0.0001, 95% CI [−0.837, −0.759]) and dorsal cortical sensory specialisation for auditory, visual and somatosen- attention regions (rrm(243) = −0.762, p < 0.0001, 95% CI [−0.810, sory processing of individual brain regions52 onto the common −0.704]) were most strongly affected by differences in habitat com- space to estimate the functional modal specialisation of cortical plexity. Together, these results show that the morphology of the cer- areas in the rodent-primate ancestor (Fig. 5). Using the cosine dis- ebral cortex reflects ecological and behavioural factors that pose tance between sensory specialisation measures we then performed specific cognitive demands to individual species48. semivariogram modelling with respect to geodesic distance on all cortical surfaces. In spatial statistics, the largest distance at Evolution of the topography of modality-specific functional which two measurements are spatially autocorrelated is called the areas in the cerebral cortex range of the model. A decreasing range value corresponds to Evolutionary adaptation to specific environmental niches and neuro- increasing segregation into modality specific areas (Supplemen- plasticity of the cerebral cortex cause individual cortical areas to tary Fig. 9). Nature Communications | (2023)14:2252 3 Article https://doi.org/10.1038/s41467-023-37574-x Fig. 2 | Reconstructing the cortical shape of a common ancestor. Overview of dense correspondence between the cortices of the two species; (c) the shape of the the steps required to perform shape alignment and ancestral state reconstruction common ancestors is reconstructed based on these correspondences by inter- between the surface models of the cerebral cortices of two sister species. a For two polating in the space of smooth shells; (d) finally, cortical features are estimated for species, anatomy and geometric features are assigned to each point on their cor- the common ancestor, and the reconstruction is iterated until the root of the tical mesh; (b) after spectral embedding of both surfaces, alignment establishes phylogenetic tree is reached. Fig. 3 | Relative expansion along ecological dimensions reveals functionally colour-coded for each socio-ecological variable. Retaining the 99th percentile of associated areal patterns across the cortex. The changes in surface area between most strongly correlated terms (Supplementary Data 3) indicate that these the extremes of synthetically generated linear combinations of pPCA modes expansion maps encode related, semantically meaningful concepts, thus high- associated with individual ecological variables are subjected to decoding via the lighting the association of evolutionary adaptive processes in both behavioural and NeuroSynth database. Correlation of the expansion maps and individual terms are ecological variables and cortical morphology. Nature Communications | (2023)14:2252 4 Article https://doi.org/10.1038/s41467-023-37574-x Fig. 4 | The correlation structure between expansion patterns associated with the effects of habitat on the surface area of functionally defined regions in the ecological and behavioural variables reveals a putative ordering of cortical cerebral cortex revealed a putative ordering related to habitat complexity. Violin adaptation to habitat complexity. a We used two-sided Pearson correlation plots of the distribution of local expansion values in each region for each map (n = 21, the number of pPCA dimensions used in the analysis) between the matrices associated with a specific habitat are shown together with median values and first of pairwise effect measures of each observed ecological and behavioural variable to and last quartile + /− 1.5 interquartile range (IQR), as well as 95% percentiles of the highlight the ordered nature of the effect of habitat complexity on the differential output of the repeated measurements correlation model of the same values. The expansion of cortical areas. Significant correlation are indicated by *p < 0.05, dorsal attention areas for terrestrial and fossorial and the ventral attention regions **p < 0.01, ***p < 0.001. Statistical information is provided in Supplementary Data 4a. for arboreal and terrestrial species were the only regions showing no significant b Computation of relative measurement correlation requires a metric for the (two-tailed Friedman test) effects in terms of relative expansion, while the limbic explanatory variables of the individual groups. A measure for the relative habitat areas where the only ones to show significant relative expansion in association to complexity of fossorial species (−0.278) is computed from a tentative ordering of fossoriality (r = 0.628, p < 1e-6, 95% CI = [0.584, 0.671], FDR-corrected). All statistical habitat complexity and by fixing the two more complex habitats that contain information is available in Supplementary Data 4b). Repeated measurement corre- “arboreal” (1) and “terrestrial” (0) species. This scaling is applied globally to all lation analysis (shown as predicted correlation for each functionally defined region relative expansion maps and does thus not bias the analysis of relative expansion of together with 95% confidence intervals) corroborated this result by indicating non- specific cortical regions with respect to each other amongst habitats. c Analysis of significant effects only for the ventral attention network (Supplementary Data 4c). In extant species, the range parameter is consistently higher in diurnality are related with less segregation, while fossoriality coincides primates (0.567–0.998) than in rodents (0.273–0.545, Fig. 6c; Sup- with higher segregation of modality specific areas. In rodents, analysis plementary Fig. 6; Supplementary Data 6a), indicating an increase in of the relationship between ecological variables and spatial statistics of surface area attributable to multi-modal association cortex. In terms of modal integration revealed significant negative effect for fossorial socio-ecological parameters, the joint analysis of rodents and primates habitat (r(21.881) = 0.599, q = 0.011, Δ = −0.052 95% CI [−0.083, −0.021) is dominated by the inter-order differences and showed significant indicating higher segregation, whereas in primates, we observed difference in range for arboreality (r(88) = 0.397, q < 0.001, Δ = 0.338, increased modal integration in more gregarious species (r(56) = 0.374, 95% CI [0.094, 0.342]), fossoriality (r(88) = 0.488, q < 0.001, q = 0.01, Δ = 0.088, 95% CI [0.029, 0.146]) as well as in diurnal com- Δ = −0.375, 95% CI [−0.449, −0.291]) and group size (r(88) = 0.368, pared to nocturnal species (r(41.206) = 0.57, q < 0.001, Δ = 0.0.096, q < 0.001, Δ = 0.157, 95% CI [0.054, 0.233]) as well as diurnality 95% CI [0.053, 0.1460]). (r(88) = 0.62, q < 0.001, Δ = 0.392, 95% CI [0.170, 0.385]). The value of Δ Finally, we estimated the evolution of spatial dependence of is positive if the range value is higher for the present compared to the modal specificity by fitting semivariogram models on the ancestral absent socio-ecological parameter. Thus, arboreality, group size and state reconstructions of the cortical surfaces in the deep ancestral Nature Communications | (2023)14:2252 5 Article https://doi.org/10.1038/s41467-023-37574-x Fig. 5 | The ancestral state of regional functional modal specialisation of cor- dataset to estimate the distribution and evolution of the relationship between tical areas. Cortical surface shape alignment across species and to common modal specificity and surface shape. Renderings of the mapping of the ancestral ancestors allows for the estimation of the spatial distribution of primary sensory state estimate onto all surfaces of the dataset are provided in Supplementary areas (vision: blue, auditory: green, somatosensory: red) in the ancestral state of Data 11. primates and rodents. The ancestral state can be mapped to all surfaces in the human lineage. We observed a strong positive relationship (silhouette coefficient 0.574) that summarise the functional correlates (r(8) = 0.714; p = 0.002; Fig. 6b; Supplementary Data 6b) between of the morphological evolution of the human cerebrum. The maxima estimated divergence times and the spatial correlation range. of association between expansion and functional categories of the components revealed a sequence of neuromorphological stages, Localised evolutionary cortical expansion in Homo starting with the expansion of visual areas, followed by para- The fossil record does not allow for the precise reconstruction of the hippocampal regions in the late Cretaceous, auditory and sensory- cortical shape, but much information about the behaviour and ecology motor areas in the early Oligocene until the mid-Miocene and finally of ancestral species can be gathered from skeletal and artifactual expansion of association areas in the late Miocene and Pliocene up remains53. Mapping cortical surface models of all species to a common until the present day. reference frame enables approximating the progression of local mor- We then compared the evolutionary expansion patterns in the two phological changes in individual lineages during evolution. Provided distinct lineages of mice and humans (Supplementary Fig. 4) as they the proposed alignment is correct, we expect the sequence of ances- represent the two most commonly studied species of Euarchontoglires. tral states of the cerebral cortex shape in the evolutionary history of We calculated the relative area of each element in the cortical surface humans to reflect adaptations to ecological niches and to therefore models in each lineage, interpolated these values at consistent time- agree with established paleontological knowledge. At the same time, points and computed the pairwise correlation between each resulting the reconstructed sequence enables a deeper investigation beyond the vector. The resulting measure of progressive divergence between mice detail level of prior evidence. and humans during the evolution of their cortical shapes showed In order to test this hypothesis, we calculated the expansion of strong correlation with estimates of atmospheric oxygen levels55 brain regions between estimates of ancestral cortical shapes at (r(84) = 0.915, p < 0.001, 95% CI [−0.944, −0.873], Supplementary branching points of the underlying deep phylogeny and decoded these Data 8a). That is, we observed accelerated divergence during periods of maps by correlating them with human cortical activation patterns rapid decline in atmospheric oxygen levels. In order to quantify the associated with neuroscientific terms54. At each time-point, we kept the origin of this divergence, we aggregated the areas of each model using 1% most strongly correlated terms, resulting in a description of the a functional parcellation of the human cortex46 (Supplementary functional association to evolutionary changes in the cerebral cortex Data 8b). This revealed both a significant increase in relative cortical morphology (Fig. 7; Supplementary Data 7). surface area (β(105) = 0.15, p < 1e-6, 95% CI [0.1341, 0.1659]) as well as a Hierarchical clustering (Supplementary Fig. 3) of the resulting positive interaction with ancestral time (β(105) = 0.0017, p < 1e-6, 95% expansion trajectories revealed 7 distinct evolutionary components CI [0.0014, 0.0020]) for the limbic areas of the cerebral cortex in mice Nature Communications | (2023)14:2252 6 Article https://doi.org/10.1038/s41467-023-37574-x Fig. 6 | Segregation of primary sensory cortical areas during evolution. Map- social/non-social and diurnal/nocturnal primates, together with median values as ping of species-specific cytoarchitectonic atlases allows for the determination of well as upper and lower quartile + /− 1.5 interquartile range. Significance of group the functional topography of cortical regions and the relationship of the relative differences of the range parameter for primates (n = 58) and rodents (n = 28) was positioning of functional cortical areas and species ecology. a Spatial statistics assessed using two-tailed t-tests and FDR-corrected for multiple comparisons. indicate an evolutionary segregation of primary sensory cortical areas in the deep Results show that in primates, group size and diurnality are associated (q < 0.001, ancestral human lineage (green path), with a (b) significant positive relationship see Supplementary Data 6a) with a higher range parameter, indicating more wide- between the integration of sensory areas and evolutionary time (r(8) = 0.714, spread spatial correlation of measurements of areal sensory specialisation. In p = 0.002, shaded area indicates the 95% confidence interval of the linear model. contrast, in rodents, fossoriality is associated with a lower range parameter See also Supplementary Data 6b). c Violin plots of the values for range parameters (q < 0.05, Supplementary Data 6a), indicating more segregated sensory processing of the spatial statistics model for fossorial and non-fossorial rodents as well as areas in a subterranean habitat. relative to humans, whereas the human lineage showed an increase of ecological niches are associated with the shape of the cerebral cortex motor (β(105) = 0.0011, p < 1e-6, 95% CI [0.0007, 0.0014]) but also to a in extant nonhuman species. Furthermore, we showed how the com- lesser degree default mode (β(105) = 4.3230e-04, p = 0.011, 95% CI plexity of both their ecological and social environments is associated (0.0001, 0.0008)) and frontoparietal (β(105) = 4.7530e-04, p = 0.005, with the selective expansion of functionally specific cortical regions. 95% CI (0.0001, 0.0008)) areas with time, indicating a possible evolu- Using spatial statistics, we demonstrated a gradual evolutionary tionary sensory specialisation for olfaction in mice and for complex decrease in the spatial range at which the modal specificity of these motor and association capabilities in humans. cortical areas becomes independent, suggesting the emergence of areas with multiple co-located associations51. Discussion By decoding the neurological function of the estimated cortical In this study, we demonstrated how ecological and behavioural expansion in the deep ancestral human lineage, we were able to pro- adaptations during evolution are mirrored in the shape of the cerebral pose, for the first time, a sequence of both temporally and spatially cortex across extant species. Our analysis is based on a novel common circumscribed neuromorphological events that shaped the cerebral geometrical reference frame for the cortical geometry of 90 species of cortex of our ancestors over the past 77 million of years. Many of these Euarchontoglires. Using this common reference frame, we demon- events align with established results from paleoneurology such as an strated that cortical morphology contains information on the evolu- increase in the visual cortex in both primates and rodents56–58 as well as tionary processes that produced its natural present diversity. The the impact of arboreality48,59,60 and fossoriality on the complexity of phylogeny of cortical shapes can be decomposed into individual the cortex48,60,61. components in both space and deep evolutionary time, both demon- strating concise functional interpretations and relationships to ecol- Evolutionary neuroecology of group size and diurnality ogy. The spatial heterogeneity and inter-species variability of these In performing meta-analytic decoding of the cortical expansion asso- processes highlights the importance of considering the evolutionary ciated with large social group size, we recover different aspects of what origins of the structures and functions studied in neuroscience, and has been termed the social interaction network62 (Fig. 3). We find that the resulting inherent intricacies in translating experimental results group size correlates with the expansion of anatomical regions of the between species. prefrontal default mode network such as cuneus and precuneus, cin- By analysing the principal modes of variation of the dataset using gulate, medial prefrontal cortex (mpfc), posterior and anterior cingulate activation maps obtained from human functional neuroimaging stu- cortex (PCC, ACC) that are known to be related to social behaviour63,64 as dies, we showed how neuroanatomical correlates of specific socio- well as areas associated with functional terms such psychosis and Nature Communications | (2023)14:2252 7 Article https://doi.org/10.1038/s41467-023-37574-x Fig. 7 | Estimated sequence of local cortical expansion from the last common Pleistocene. The common x-axis of all subplots (a–c) denotes million years ago ancestor of rodents and primates to Homo. a Consecutive expansion maps (Mya). The colormap on the cortical surfaces (a) indicates relative area expansion between the cortical surfaces are obtained from ancestral state reconstruction of between estimated speciation points. Markers for correlations in subplot (b) are all ancestors of Homo until the LCA with Glires. b Decoding these expansion maps jittered in horizontal direction to reduce overlap. Correlations are summarised for as correlations with statistical maps of human cortical function54 and retaining the each cluster using violin plots in subplot (d) and joined at their median values 1% most highest correlated maps at each time-point allows to perform (c) hier- through time. Statistical significance for differences between distributions was archical clustering of the sequential correlation of expansion patterns in the deep assessed using repeated measurements ANOVA. Statistical information as well as ancestral human lineage (Supplementary Data 7a). d This reveals a decomposition sample numbers for each cluster and time-point are omitted for space reasons and of evolutionary cortical surface expansion that can be approximated as a sequence can be found in Supplementary Data 7c, license information for artwork used in of steps (Supplementary Data 7b), starting with the expansion of primary visual, Supplementary Data 7d. Abbreviations: P.I. Ptilocercus iowii. N.t. Notharctus auditory and motor regions in the Cretaceous up until the Oligocene, followed by tenebrosus. A.a Archicebus achilles, C. Cercopithecoidea, O.b. Oreopithecus bam- the expansion of higher order association areas in the Miocene, Pliocene and bolii, S.t. Sahelanthropus tchadensis, A.r. Ardipithecus ramidus, H.s. Homo sapiens. emotional responses. Additionally, we observe a strong correlation with relationship lie in evolutionary advantages in predator avoidance terms related to visual processing (lingual gyrus, primary visual) as well posed by congregating in larger groups in a more terrestrial habitat73. as the default mode network per se (default mode, deactivation). While more arboreal apes can rely on the safety of their habitat, more While the ACC is considered integral in understanding the inten- terrestrial species have to more actively secure their environment by tions of others and in observational learning65, both PCC and dmPFC forming large social groups74. are crucial elements of the default mode network66. The PCC has been The association between social group size and activity pattern in related to self-referential functions67, and the dmPFC is deeply involved primates is reflected in the correlation between variation in cortical in processing observed social interactions68 and judgements69. Toge- morphology attributable to social group size and expansion of visual ther with the cuneus and precuneus, these regions have been proposed areas. Partial correlation analysis of the relationship of functional to form a “social-affective default network”70 with strong links to face categories derived from cortical expansion with the evolution of processing71. sociality, controlling for diurnality (Supplementary Figs. 7 and 8; Expansion maps derived from ancestral state reconstruction of Supplementary Data 9 and 10a) revealed a significant positive corre- the cortical surfaces locate the time point of maximum correlation lation with area size associated with the terms amygdala, hippocampus between cortical expansion and terms associated with large group size and neutral faces, and significant negative correlations with area size (ACC, cingulate, cingulate cortex) at the divergence between Homi- associated with terms associated with motor function such as motor, ninae and Ponginae, some 12Mya. Compared to other great apes, premotor, finger, force (Supplementary Data 10b), highlighting ele- orangutans (Pongo) are more arboreal and less social72. These results ments of emotional control75,76 as differentiating between effects of are indicative of an indirect influence of decreasing arboreality in social behaviour and diurnality. While diurnality facilitates foraging for african apes on brain evolution. Possible explanations for this high-quality foodstuff, an actual increase in a species’ dietary efficiency Nature Communications | (2023)14:2252 8 Article https://doi.org/10.1038/s41467-023-37574-x requires the integration of more complex cognitive and social environment, and burrowing brings the animals’ gustatory system into capabilities77. Contrary to social group size, we could find no sig- closer proximity with chemical cues. The association of fossoriality nificant relationship between activity pattern and cortical morphol- with an increased expansion of the gustatory areas of the anterior ogy. However, analysis of the variability of the functional topography insula (Fig. 3), as well as a relative expansion of the limbic regions of the of the neocortex revealed a stronger multi-modal integration in both functional parcellation of the cortex we observed (Supplementary group-living as well as diurnal primates, indicating a potential evolu- Fig. 5d) can be explained by an adaptation to this environment. Fur- tionary strategy for success in cognitively more demanding thermore, fossorial rodents also exhibited less modal integration than environments. their arboreal and terrestrial counterparts (Fig. 6c), which might also Our results indicate cortical correlates of a strong evolutionary relate to the fact that perception and interpretation of a comparatively interplay78 between this complexification of visual processing, social impoverished ecological environment does not require the integration cognition79 and ecology implicating the default mode network80,81. The of multiple sensory modalities. ability of discriminating red hues might82 be a crucial component in In contrast to the processes of mapping the position and motion diurnal predator avoidance83 and potentially foraging as well as for the of objects within the extrapersonal space, determination of object interpretation of facial features84. These adaptations might be at the identity, as enabled by the ventral pathway, is less dependent on one root of a positive feedback loop in encephalization whereby mod- individual sensory modality96. Studies in congenital blindness have ifications to the functional topography of the cerebral cortex resulting repeatedly demonstrated the recruitment of these areas of the cortex from evolutionary changes to cortical morphology allowed for the for haptic processing96–99, underlining the integrative role of the integration of multimodal cues from complex ecological and social structures of ventral pathway100,101. This modal agnosticism is also environments. This potentially increased our ancestors’ capabilities for reflected in the independence of the extent of the ventral attention social behaviour and thus predator avoidance and effective foraging. areas on habitat observed in this study (Fig. 4c), which underlines the This in turn led to increases in energy85 available for supporting cross-modal relevance of the ventral processing stream for cognition. expensive brain tissue86, enabling ever more complex social and cul- An alternative or possibly complementary explanation for a rela- tural faculties87,88. tionship between the expansion of areas of the dorsolateral cortex and complexity of a species’ environment can be found in the affordance Variability of cortical morphology and topography in relation- competition theory of cortical organisation102. Under this framework, a ship to habitat complexity selective expansion of areas of the dorsal stream in complex envir- Our findings revealed distinct patterns of cortical expansion asso- onments could lead to an increased complexity of proposed potential ciated with individual habitats (Fig. 3; Supplementary Figs. 5b–d, 6a, c). modes of action such as action maps103. Since the complexity of Arboreal species rely mostly on vision for perceiving their selecting amongst them however stays comparatively constant, no environment89, which enables them to locate important objects such expansion of the ventral stream is required. as food sources or predators at a distance and has, alongside the The evolutionary adaptations in the deep ancestral human lineage evolution of binocular colour vision and consequent depth percep- represent only one of many courses of cortical evolution. A common tion, led to the complexification of cortical visual processing. This is reference frame for the cortical shape evolution of different lineages reflected in our data by a selective expansion of occipital and dorso- enables the analysis of the rate of their divergence and its association lateral aspects of the cerebral cortex in association with arboreality. with ecological and environmental factors. This analysis (Supplemen- Expansion of these regions of the dorsal pathway90 likely reflect the tary Fig. 4) shows that periods of rapid oxygen level104,105 decline fact that the functions it supports such as processing of object loca- coincide with accelerated divergence between the cortical morphol- tion, motion and concordant planning of movements are more rele- ogy in human and mice lineages, suggesting different adaptation vant and also more elaborated in the complex 3D environment of the strategies being fostered by the corresponding evolutionary pressure. trees compared to the less intricate settings in which terrestrial and fossorial species live91. Regional cortical expansion reflects sensory segregation and In terrestrial and fossorial environments, proprioception is behavioural evolution algorithmically less challenging and perception at distance is, espe- Hypotheses about the evolutionary history of primates are manifold106 cially for small animals such as rodents, of lesser importance, as it is but have remained elusive to quantitative evaluation22. Our analysis of inherently limited by grass cover or subterraneity. Concurrently, we the joint morphological space of the neocortex of Euarchontoglires observed a graded reduction of the relative extent of visual areas in enables for the first time to study both the conserved topographical association with both terrestriality and fossoriality, related to the structure of cortical function as well as deriving a data-driven modular reduced relevance of visual information in these habitats (Supple- decomposition of its evolution. Our resulting estimate of the sequence mentary Figs. 5c–d, 6c). On the other hand, terrestrial and fossorial of evolutionary steps of cortical development in the deep ancestral species make stronger use of olfactory and gustatory cues for per- human lineage can be interpreted within the frameworks of ception. This is reflected in the selective expansion of cortical areas neuroethology107, ecomorphology106, and phylogenetic refinement15. associated with the processing of these chemical signals, in this case The gradually decreasing range of functional specificity on the particularly the insular cortex. We observed rather fine-grained dif- cortex during the evolution of the deep ancestral human lineage ferences in the expansion of the insula in association with terrestriality (Fig. 6b) indicates a relative reduction of the size of primary sensory and fossoriality, the former correlated with higher extent of its pos- areas compared to emerging larger intermediate regions without terior, the latter with higher extent of its anterior aspect (Fig. 3). Ter- unimodal assignment. In support of the tethering hypothesis51, our restriality is additionally associated with selective expansion of the results show that the modular segregation of primary cortical areas cingulate cortex. Meta-analytic decoding maps this pattern onto terms increased during human evolution. At the same time, the observed such as nociceptive, noxius, or pain. The cingulate is deeply involved in stronger modal integration appears to be favourable or even necessary fear response92 and considered as a mediator of emotional states and in cognitively demanding tasks such as social interactions. action selection93. This function might have increased relevance in By splitting local cortical evolutionary expansion into clusters terrestrial animals that rely on olfaction for survival by channelling related to vision (Fig. 7, cluster 1), orientation (Fig. 7, cluster 2), tem- otherwise functionally pervasive stress responses94. poral language areas (Fig. 7, cluster 6) and memory and retrieval (Fig. 7, A fossorial lifestyle poses physical constraints on cranial shape95 cluster 7) on the one hand and auditory perception (Fig. 7, cluster 3), and limits the distance at which animals are able to perceive their sensorimotor (Fig. 7, cluster 4) and prefrontal/dorsolateral/parietal Nature Communications | (2023)14:2252 9 Article https://doi.org/10.1038/s41467-023-37574-x (Fig. 7, cluster 5) areas on the other hand, the observed topmost and thus evolutionary fitness in increasingly harsh environments140, hierarchy of evolutionary functional expansion patterns reflects the giving rise to the Homo sapiens of today. dual stream hypothesis of cortical topography90,108. At the lowest hierarchical level in the ventral super-cluster, hierarchical clustering Limitations groups together patterns associated to parahippocampal areas (Fig. 7, There remain numerous limitations to our work. For one, the breadth cluster 2), memory and retrieval (Fig. 7, cluster 7). In the dorsal of our sample requires us to rely on single and oftentimes post- supercluster (Fig. 7, clusters 3-5), the grouping of expansion patterns mortem exemplars of each studied species showing varying degrees of related to auditory processing (Fig. 7, cluster 3), and sensorimotor degradation. This, together with the inherent geometric limitations in function (Fig. 7, cluster 4) at the lowest level reflects the close rela- mapping and averaging protruding features, might lead to an under- tionship of these functions concerning proprioception. estimation of the ancestral size of the olfactory bulb41,46. Furthermore, From a historic perspective, our results indicate that the diver- excessive evolutionary expansion of the neocortex in all extant species gence of rodents and primates started with the development of visual of Euarchontoglires may lead to an overestimation of its size38,47, a bias processing areas (Fig. 7, cluster 1), likely adaptive to the increased which is especially present in the reconstruction of Cretaceous sensory complexity posed by living in the complex 3D arboreal ancestors48,49. We also suspect that these data-dependent limitations, environment (Supplementary Fig. 5) as well as an increased visual together with a range of unobserved covariates as well as intra-species information content due to binocular and chromatic vision82. This shift variability lies at the root of the comparatively low part of extant away from a previous ground-dwelling lifestyle led to an increase in morphological cortical variability being explained by the factors stu- foraging range in early primates107, where the capability of forming and died in this paper. Notably, the ecological categories used in the ana- storing mental images of scenes such as locations of temporally lim- lyses represent stark simplifications of complex behaviours that ited food sources proved evolutionary advantageous, leading to the encompass specific modes of locomotion, perception and social expansion of parahippocampal areas109 (Fig. 7, cluster 2) specialised for interactions. We don’t propose that the results reflect a complete spatial navigation and memory110–113. The increase in mobility resulted generative model of brain shape evolution. However, the recovery of in the necessity of communicating over larger distances and thus the significant influence of socio-ecological factors on cortical shape, the development of cortical areas related to vocalisation114 and auditory qualitative compatibility of its ancestral reconstructions with fossil processing115–117, (Fig. 7, cluster 3). These evolutionary adaptations to endocasts and results obtained from quantitative analysis of its evo- arboreal locomotion and vocalisation occurred concordantly to the lutionary development serve as evidence for the validity of the pro- adoption of increasingly diurnal activity pattern and social behaviour posed common reference frame and the estimates of common (Supplementary Figure 5) and coincide with the development of ancestor cortices. We hope to see this tool being used and refined by complex motoric capabilities103, which is reflected in the ongoing other researchers using different modalities and ecological variables. expansion of somato-motor areas (Fig. 7, cluster 4). We also acknowledge the inherent anthropocentrism in the Due to the development of modality-specific processing areas, the methods used for interpreting our results. Therefore, our description primate brain around 21Mya may have been capable of performing of the neuroanatomical foundations of human perceptive, interpretive complex visual, auditory and somato-motor signal processing, and communicative capabilities can only serve as a schematic repre- increasing the necessity for mechanisms for both selective attention sentation of processes that occur at multiple spatial and temporal and action selection118 to maintain the ability to act. This may have scales, from genetic regulation to social interactions, from axonal been achieved by the expansion of prefrontal and parietal control depolarisation to evolutionary adaptation. Despite these limitations, areas in the cortex17,119–121, (Fig. 7, cluster 5). The resulting capacity of to our knowledge the data presented in this paper represents the most selectively attending to specific stimuli entails the formation of com- diverse collection of Euarchontoglires brains analysed to date and plex mental representations122,123. Importantly, previously established provides important insights into the origins and ramifications of cor- executive control over complex motor systems enables the commu- tical functional topography. nication of these internal concepts in signed124 or vocalised form125, which has important evolutionary advantages126 and leads to a com- Methods plexification of social behaviour via the establishment of primordial We performed geometric cortical shape alignment across species, and forms of language125,127,128. subsequently modelled the phylogeny of the cortical differences Based on our results, these developments were accompanied by among extant species. First, we extracted a mesh surface of each cer- a shift back to a terrestrial lifestyle starting some 10Mya and ulti- ebral hemisphere from available segmentations (Fig. 2a). Labels for mately bipedalism. This in turn potentially led to a freeing of the anatomical structures were projected onto the surface by volumetric upper limbs from requirements of locomotion, and indirectly interpolation and converted to pseudo-probabilities using a gaussian allowed for the development of greater fine muscle control129 (Fig. 7, kernel on the geodesic distances on the mesh. Surface features cluster 4), thereby enabling tool building, primarily for hunting130. describing local mesh properties were computed from the surface The resulting shifts in nutritional efficiency would have reduced the models. In parallel, the global geometry was encoded via the spectral selection pressure for enlarged masticatory apparatus and led to embedding of the Laplacian Matrix associated with the surface mesh. craniofacial changes which, together with the evolution of an These two sources - anatomy and geometry - of information were orthograde body plan and the consequent rostral shift of the fora- combined on a per-vertex basis to encode the local cortical shape. We men magnum53, would have allowed for the expansion of medial and established correspondence between the cortical surface meshes of lateral parietal regions of the cerebrum131,132(Fig. 7, cluster 7). These two species, by matching their cortical shape encodings in a high- de-novo available cortical territories assumed auto-analytic func- dimensional embedding space. Dense surface correspondences could tions of memory retrieval and episodic memory133 which form fun- then be established from the resulting alignment between these two damental faculties for establishing shared semantic concepts134. This aligned representations (Fig. 2b). To estimate the potential shape of establishment of social135 and cultural capabilities could have cre- the cortical surface of an ancestor of two sister species in the dataset, ated a positive feedback loop for increased encephalization136 the ancestral state reconstruction was first performed on global shape whereby capacities of Theory of Mind, abstract thought and the properties. This allowed the positioning of the ancestral shape along capability of externalising and transmitting this knowledge via the geodetic connection in the space of smooth shells between the language137,138 and artefacts124,139 may have been amplified with each global shape properties of the two species. Since linear interpolation generation of hominids, increasing the species’ adaptive capabilities does not reflect the geometric properties of the surfaces, shape Nature Communications | (2023)14:2252 10 Article https://doi.org/10.1038/s41467-023-37574-x interpolation was performed in the space of smooth shells to estimate fragmented and harder to distil into viable categories. In order to the ancestral shape (Fig. 2c). In order to iterate the procedure a-c and retain statistical power while capturing the essence of these com- estimate further ancestors, the label information of the two sister plexities, we abstracted species habitat in the following way: Arboreal - species was mapped onto the estimated surface mesh of their com- spends most of its time in trees, in a cluttered environment, is able to mon ancestor. If a label was present in both sister species, information move in all directions; Terrestrial - spends most of its time on the was combined for estimating its ancestral state. However, similarly to ground, in an open-environment, does not habitually move in vertical the reconstruction of the ancestral surface model, computation in the direction; Fossorial-spends most of its time underground, affecting its Euclidean space did not account for the properties of the domain of range of movement and perception. Individual species can belong to the pseudo-probabilities and would have led to blurring and infor- multiple categories in cases where the literature is ambiguous. While mation loss. Instead, optimal transport on meshes was used to recover this classification does not do justice to the complexities of ecological an estimate of the extent and shape of the ancestral label pseudo- and behavioural diversity observed in the species under study, we probabilities (Fig. 2d). This procedure was then iterated by following found this nomenclature to be the most consistent among such a large the phylogenetic tree to its root. diversity of rodent, lagomorph, primate, dermopteran and scanden- A more detailed description of the technical aspects of the tian species. Species that were reported as either diurnal, cathemeral methods employed for phylogenetic modelling, surface reconstruc- or arrhythmic in the literature were labelled as “diurnal”. Species that tion, surface matching and estimation of spatial statistics can be found are not reported to be solitary or pair-living were labelled as living in a in the Supplementary Methods. “large social group”. Dataset description Cortical shape encoding. We performed pairwise matching of the Brain imaging. For the present study, we collected imaging data of 90 cortical surface models of sister species in the underlying phylogenetic different species from various sources (Supplementary Data 1a), con- tree using methods from computational geometry. Specifically, we sisting of 58 primates, 28 rodents, 2 lagomorphs, 1 scandentian and 1 exploit the fact that the eigenvectors of the cotangent graph dermopteran. In order to collect a maximum number of different laplacian146 of a mesh representation of a shape are closely related for species, we combined imaging data from different modalities of both isometric shapes. This property has been leveraged in numerous in-vivo and ex-vivo specimens. In total, 75 specimens were imaged geometric applications, most notably shape description147,matching148 using MRI, 7 using diffusible iodine-based contrast-enhanced Com- and alignment149. puted Tomography (DiceCT) and 8 using serial histological sections. As we are however specifically interested in deviations from For the serial histology section data, StackReg141 was used to correct isometric scaling due to localised expansion of individual cortical for rigid displacement between slices. Data for 10 species was available areas, direct application of spectral matching would lead to erro- in vivo, while 80 scans were obtained from ex vivo data. Of the 80 ex- neous alignment. Therefore, we augment the global intrinsic spec- vivo specimens, 43 were fixated ex-situ, while 37 were imaged inside tral representation of the shapes by additional extrinsic features the skull. The data for 7 species consist of atlas averages of multiple representative of cortical morphology. Since the patterns of cortical individuals. For 7 specimens, previously published segmentations of folding show strong evolutionary conservation, we use sulcal depth the cerebral structures were available. For 31 specimens, segmentation maps, their surface derivatives, gyrification indices as well as the of the cerebrum was available via the BrainCatalogue5. For the Hamilton–Jacobi Skeleton of surface curvatures150 as descriptors of remaining 52, we manually segmented the two cerebral hemispheres the sulcal geometry. Furthermore, we include manual labels for the using ITK Snap142. In cases where it remained intact after preparation, olfactory bulb, the corpus callosum and the cerebral medial wall as the olfactory bulb was manually segmented as a separate label, as well scalar values defined on each surface vertex of each surface model. as the Corpus Callosum. We further included commonly used geo- The label maps are generated by manually labelling the individual metric spaces for model animals (Mus musculus, Rattus norvegicus, structures in the volume, interpolating these binary maps onto the Macaca mulatta, Callithrix jacchus, Homo sapiens) in the dataset surface using Nearest Neighbour interpolation and consecutively (Supplementary Data 1a). Unfortunately, it was not possible to obtain computing the geodesic distance transform151 on the surface to multiple individual scans for all species. We therefore limit our dataset obtain a continuous function value for each surface vertex. The to one exemplary shape per species, as the morphological variability is geodesic distance is then subjected to an exponential kernel to limited for the cerebral cortex in most species. For humans, where localise the label information. anatomical variability is especially large, we use the “Colin” template space143. Dense surface alignment. From the spectral decomposition of the mesh Laplacian and the extrinsic surface features, we obtain a vector Surface model construction. We computed surface models indivi- representation of global and local shape information at each surface dually for each cerebral hemisphere in each specimen. In the case of vertex, which we consider elements of a combined spectral/euclidean 7 species (Supplementary Data 1a) where either only one hemi- embedding space. Matching of elements of that space is performed sphere was available or one showed defects that were too large for using Coherent Point Drift (CPD)152. CPD is a statistical formulation of manual correction, the model of the contralateral one was mirrored the problem of shape correspondence based on Gaussian Mixture along the midline to replace the missing hemisphere. Spherical Models and solved using Expectation Maximisation. In this formula- topology of the resulting surface model was ensured by respective tion, the feature vectors corresponding to the vertices of one shape functionality provided by the CAT toolbox144 as well as FreeSurfer43. represent the centers of Gaussian distributions in a high-dimensional Spherical mappings of the resulting surface models were space, whose positions are optimised such that the feature vectors of constructed145 to serve as parametric reference frames for geometry the other shape are realisations of the resulting mixture of Gaussians processing. under Maximum Likelihood. After convergence, correspondences between points can simply be obtained by selecting the center of the Ecological and behavioural data. Ecological and behavioural data Gaussian with the highest weight for each surface point. This prob- were collected from literature sources (Supplementary Data 2). We abilistic formulation relaxes the isometry assumption of the original collected data for daytime activity pattern and social group size, which spectral matching and simultaneously serves as regularisation for the we simplified into binary categories of“Diurnal” and “Large social ill-posed shape matching problem, effectively penalising strong local group”. We also collected data for preferred habitat, which are more stretching and bending149,153. Nature Communications | (2023)14:2252 11 Article https://doi.org/10.1038/s41467-023-37574-x Post-processing of surface alignment. Due to the nonlinear corre- the underlying calibrated phylogeny. However, this would imply the spondence between the original Euclidean three dimensional space in unrealistic assumption of uniform relationship between genetic simi- which the cortical surface models are observed and the spectral larity on which the phylogeny is based and the morphological differ- embedding space, additional processing of the correspondences is ences in the shapes of the cerebral cortices of sister species. We required in order to obtain continuous, invertible maps between two therefore first estimate the ancestral states of global shape variables in cortical surface models. First, we use a method proposed for the the whole tree which does not requires a-priori correspondences (cf. matching of developing fetal brain surfaces in humans154, which shows “Phylogenetic Model of Global Cerebral Morphology” for a detailed a comparable range of geometric variability as our dataset from lis- explanation). This allows for the computation of a parameter reflecting sencephalic first trimester of gestation to strongly gyrated in the third the estimated scaling factor between genetic and morphological dif- trimester. Instead of defining corresponding points in the two shapes ferences between an ancestor and its daughter states. We then gen- directly using nearest neighbours in the embedding space, this erate the shell space interpolation at these distances between the two method first performs edge-based smoothing of the embedding sister species as an approximation of the ancestral state of the cortical coordinates, resulting in smoother maps. In order to further enforce surfaces. diffeomorphic correspondences between the shapes, the resulting correspondences are used as inputs to a joint Laplacian matching Label fusion using optimal transport. After establishment of pair- step149,155. Contrary to the original procedure of computing alignments wise correspondences between sister species, available annotations between individual embeddings of the two shapes, this method is of specific homologous structures (Corpus Callosum, olfactory based on the spectral decomposition of a joint Laplacian matrix. The bulb, thalamus, hypothalamus, septum) are mapped onto the joint Laplacian of two shapes has a block structure, with the mesh ancestral morphology estimated via shell-space interpolation. In topology and geometry of the individual meshes encoded in two cases where the same label is available for both daughter shapes, this diagonal blocks, containing edges between neighbouring vertices information needs to be combined in the ancestral state. As with the weighted by their geodesic distance, and the pair-wise corre- shapes themselves, label information can only be poorly processed spondences between the two shapes encoded on the off-diagonal using euclidean methods, as linear averaging would inadvertently blocks, either as binary entries for corresponding vertices or also result in blurring of the boundaries of the labels due to both mis- weighted by their respective distances in their aligned embedding alignment of structures due to regularisation in the matching pro- space. As before, correspondences of one mesh to vertices of the other cedure but also changes in the true extent of individual structures can be established using nearest neighbours in the resulting embed- during evolution. Instead, we represent labels as pseudo- dings space of the now joint Laplacian. The third regularisation step probabilities of the extent of a structure at a specific position of that we perform in order to ensure reversibility of the resulting maps is the cerebral cortex and use optimal transport geometry to compute based on the Dirichlet energy of the resulting map. As of now, we only averages of these distributions on the estimated surface models of considered the locations of correspondences of the vertices of one ancestral states using Convolutional Wasserstein Distances158. This surface mesh in the other. However in practice, it is important to procedure results in estimates of the distribution for each label in ensure that both maps from either mesh onto the other are well- the ancestral state, but also breaks the commutativity of label cor- behaved, eg. are reversible with as little distortion as possible. To respondence between parent and daughter states. We therefore ensure this, we optimise the correspondences obtained from the joint perform an additional matching step between the newly labelled Laplacian embedding using the method proposed in Ezuz et al.156, ancestral shape and its daughter shapes to reestablish consistency which explicitly optimises a joint energy functional consisting of terms between the shape and label correspondences. for smoothness and reversibility of the resulting deformations using block coordinate descent. Consistent topology of cortical morphology. After iterative pairwise matching of the surface models of individual cortical shapes, global Approximating ancestral shapes of cortical surfaces. We iterate the correspondence between all shapes in the dataset can be established procedure of establishing pairwise correspondences between the by propagating the correspondences throughout the phylogenetic surface models of the cerebral cortex in sister species, traversing the tree. Throughout the iterative surface mapping, we use the more phylogenetic tree from the leaves until the root, corresponding to densely sampled cortical surfaces of the two sister shapes as the last common ancestor (LCA) of rodents and primates. Enabling topology of the estimated ancestral shape. Once the LCA of rodents this iterative procedure requires the interpolation between matched and primates at the root of the tree is reached, we resample all surfaces. However, simple linear interpolation via weighted aver- shapes in the dataset to fsaverage6 topology to allow the easy aging in Euclidean space can result in both geometric and topolo- application of published results on cortical topography in humans. gical errors such as excessive thinning of structures and fold-overs The resulting surface models are available at https://github.com/ due to the intrinsic nonlinearity of shape space. We therefore per- cirmuw/EvolutionOfCorticalShape. form shape averaging in the space of discrete shells41,42. This allows the incorporation of biomedically motivated penalties to the cal- Phylogenetic modelling culation of intermediate shapes between shapes of matched topol- Estimation of phylogenetic relationships. We obtain the phyloge- ogy. The underlying biomechanical model is equivalent to a thin netic relationships for the taxa used in the present study from a time- elastic shell, which is only a rough approximation of the non-linear calibrated phylogeny of the mammalian class39. The phylogeny is material properties of the true cortical tissue157. However, a com- constructed using a Bayesian “backbone-and-patch” method from a 31 plete incorporation of the biomechanics of brain tissue in the gene Supermatrix, resulting in probabilistic estimates of the branch matching procedure is both mathematically and computationally lengths and divergence times in the tree. We draw 100 samples from challenging due to the non-linear tissue properties, and of ques- the posterior distribution of trees of the 90 species contained in our tionable validity due to the unknown evolutionary changes they dataset and compute the consensus tree to obtain a single phyloge- might be subjected to. We therefore leave these extremely inter- netic topology and divergence times. As the human brain showed esting questions for future research and limit the biomechanics of extensive expansion compared to other species, we augment the the matching procedure to linear elasticity. resulting consensus tree with that of hominin fossils obtained from We could obtain the relative weightings of the surface models of Melchionna et al.159. We merge the two trees using phylogenetic each sister species in the computation of their average directly from inference under matrix representation160. Nature Communications | (2023)14:2252 12 Article https://doi.org/10.1038/s41467-023-37574-x Phylogenetic modelling of global cerebral morphology. Estimation and the multivariate OU models used for modelling the localised of the ancestral shape of the cortical surface models of two sister evolution of the cortical shapes are not subject to the same simplifying species requires a measure of the degree of morphological changes assumption as those used in estimating the rate shifts. This approach each sister species has been subjected to. However, divergence time results in a strong link between changes in evolutionary rates of the estimates in the used phylogeny are based on genetic information, and global and local cortical morphology, but nonetheless allows for the there is no unique universal scaling factor that would relate changes to independent estimation of correlation structure for the local change in the genetic code to cortical morphology. On the other hand, estima- cortical morphology during evolution. tion of an evolutionary model for changes in cortical shape requires a We perform phylogenetic model selection based on the Gen- priori correspondences between surfaces. To avoid this circular pro- eralized Information Criterion (GIC)166, which selects an adaptive OU blem, we assume a linear relationship between the changes to the model with 2 rate shifts (BM: -460733.9, OU: -480972.5, EB: -460731.9 global shape measures of surface area (SA) and gyrification and the OU_shift: -496364.8) (Supplementary Fig. 1). We perform phylogenetic degree of local changes to cortical morphology. As a first step in our PCA165 on the low-resolution representation of the dataset in order to phylogenetic analysis, we then use the combined tree of extant and account for phylogenetic correlation in the shapes. We then use L2- extinct species to determine an appropriate evolutionary model for regularisation as proposed in Clavel et al.167 to estimate the phyloge- these global morphological properties of the cerebral cortex. netic covariance matrix, as its unregularised computation is still ill- From the available 90 species, we estimate both the cortical SA posed due to the large number of landmarks compared to the number and the area of the convex hull of each hemisphere, as their ratio of samples. The ratio of one to ten between dimensionality of the constitutes a measure of gyrification161. We then fit scalar Ornstein- samples and their number is comparable to that used in other works on Uhlenbeck (scOU) models with evolutionary rate shifts40 to the SA phylogenetic brain shape analysis with similar or smaller sample measurements of the cerebral cortex and its convex hull. Briefly, the sizes168. scalar formulation of the OU mixture models simplifies the definition of rate shifts in multivariate correlated traits in a phylogeny by shifting Analysis methods for cortical surface maps the inter-trait covariance to the Brownian motion part of the OU model Meta-analytic decoding. We use meta-analytic decoding to analyse while modelling the selection pressure jointly as a scalar parameter for the cortical surface maps obtained from phylogenetic modelling all observed traits. We assume that this simplification is acceptable for (pPCA, ancestral state reconstruction). We first map the measurements modelling the evolution of the SA of the cerebral cortex and its convex obtained on the cortical surface into the standard MNI space46. The hull due to the existence of strong scaling laws162–164 relating the folding resulting volumetric maps are then correlated with 3228 individual and scaling of mammalian cerebra. Estimation of rate shifts is then maps obtained from 14371 individual neuroscientific studies of brain performed in an Expectation-Maximisation framework, which allows activation in humans54. We retain the most strongly correlated 1% of for the automatic determination of evolutionary rate shifts. However, these terms to interpret the expansion maps in terms of neuroanato- the scalar formulation of mixture of OU models is defined only for mical localisation and neurological function (Supplementary Datas 3 ultrametric trees (eg. trees whose tips all have the same age). We and 4). We note the anthropocentrism inherent to the utilisation of therefore shift the tip ages for the fossil data to the present day. For the human reference maps for the decoding maps related to evolutionary extinct hominin species included in the augmented tree, only esti- processes. However, to the best of our knowledge there unfortunately mates of the SA of the convex hull of the brain are available from does not exist to date any comparable dataset for other species which endocasts of the fossil crania; the SA of the cerebral cortex is con- would improve on this. sidered missing in phylogenetic modelling. We assume that the slight misspecification of the tip-dates for the fossil hominin data is imma- Projecting maps onto functionally defined parcellation. Addition- terial to the relative BIC values used for model selection. We conse- ally to interpreting the maps obtained from phylogenetic modelling quently perform ancestral state estimation of the two SA values for using meta-analytic encoding, we also analyse the obtained values in each case and hemisphere and prune the tree to cover only the avail- relation to their position in a commonly used parcellation of the able extant species for further processing. The resulting estimates of human cortex from resting state functional MRI data46. While this sort the surface areas of the cerebral cortex and its convex hull are then of analysis can be criticised for its anthropocentrism, the proposed available throughout the phylogenetic tree. We use the ancestral state parcellation has been shown to be highly consistent for the analysis of reconstructions of these values to determine the interpolation factors a detailed battery of cognitive tasks169, highlighting the robustness of between two matched cortical surface models to reconstruct their the underlying topography in describing functionally distinguishable putative ancestral state without a priori matching of all shapes in the regions of the cerebral cortex. dataset as described above. Merging of cytoarchitectonic and functional atlases. Cytoarchitec- Phylogenetic geometric morphometrics of cortical morphology. tonic cortical atlases are the maps that have been the most consistently After having obtained an estimate for the rate of evolutionary change defined for different species170–172. Specifically, detailed atlases have in global morphological properties of the primate and rodent cere- been published for Mus musculus173, Rattus norvegicus173, Callithrix brum, we aim at determining the shape space of the cerebral cortices jacchus174, Macaca mulatta175 and recently Homo sapiens176. Addition- of rodentia, lagomorphs, scandentia, dermoptera and primata. How- ally, continuous work in analysing functional specificity of individual ever, the dimensionality (40962 vertices per hemisphere, as defined in cortical regions has led to the proposition of a detailed atlas of the FreeSurfer fsaverage6 space) of the dataset is too large in relation to specificity of cortical areas in terms of auditory, visual and motor the sample size (90 cortical surface models). We therefore use low- processing52 in humans. We combine these two sources of information resolution icosahedral subdivision to first reduce the dimensionality of about the structure and function of cortical regions in order to analyse the data to 162 vertices per hemisphere, and remove any samples in the the changes their organisation was subjected to during evolution. medial wall, which results in 900-dimensional representations of each Specifically, we first map the values for modal specificity into the pair of hemispheres for each taxon in the dataset. We fit both simple individual highest resolution cytoarchitectonic parcellations in each of (Brownian Motion (BM), Early Burst (EB) and Ornstein-Uhlenbeck the 5 species for which cytoarchitectonic atlases are available, yielding (OU)) as well as OU models with evolutionary shifts (OU_shift) to the 3 maps ranging from 0 to 1 at each vertex for each species. We estimate shape data165. Note that the position of these evolutionary shifts have the ancestral state of each of these maps using the phylogenetic cov- been determined a priori from the global cortical shape parameters, ariance matrix obtained from the cortical surface shape models. This Nature Communications | (2023)14:2252 13 Article https://doi.org/10.1038/s41467-023-37574-x procedure can be thought of as a constrained surface smoothing 20 non-overlapping bins. For each bin, we compute the average pair- procedure in which the blurring of functional specificity of individual wise cosine distance of the three-dimensional (propensity for visual, regions in the cortex is constrained both by the species-specific auditory and somatosensory function) vectors measured at the inclu- cytoarchitecture and the phylogenetic stability of these maps. ded surface vertices, which we again normalise to unit range. We then perform 100 independent, randomly initialised fits of linear, spherical Statistical treatment and exponential models to the resulting empirical variogram and General statistical methods. Throughout the paper, significance was retain the model best-fitting model in terms of sums of squares error assumed at a 5% false positive level. When required, multiple com- (SSE). From the selected model, we estimate the sill (maximum value of parison correction was performed using Benjamini-Hochberg correc- spatial independence) and range (distance at which the sill is reached, tion for False Discovery Rate. Normality of distributions of analysed e.g. measurements of modal specificity at pairs of points separated data was assessed using one-sample Kolmogorov-Smirnov (KS) tests. from each other by a higher geodesic distance can be considered Dependent and independent t-tests were used for determining sig- independent), where only the latter is of interest due to the a-priori nificance of group differences when normality could be assumed for unit-normalisation of both regressor and regressed variables. both tested distributions. Wilcoxon rank sum tests were used In cases where normality could not be assumed in either of the distributions. Software. All analysis was implemented in Matlab R2014a and R2019a, When testing for equality of the means of multiple dependent mea- Python 2.7.13 and 3.7.3 as well as R 3.6.0 and 4.0.3. Additional pro- surements over time, we use repeated measurements analysis of var- cessing was performed using Convert 3D 1.1.0, FSL 6.0.4, FreeSurfer iance (RANOVA) after testing for normality using KS tests at each time- 6.0.0 and 7.1.1, ANTs 2.3.4 and ImageJ 1.49 u. point, followed by post-hoc testing using the Games-Howell method. Equal variance was tested for using Bartlett, sphericity using Mauchly Reporting summary tests. In case of violation of the sphericity assumption, the Further information on research design is available in the Nature Greenhouse-Geisser approximation was used for p-value adjustment. Portfolio Reporting Summary linked to this article. In case of unequal variances, the Games-Howell method was used for multiple comparison post-hoc testing. Data availability In cases where independence of the measurements could not be All quantitative data supporting the findings of this study are provided assumed (see “Null models for brain maps” below), significance of as Supplementary Information to the article. Sources of all imaging group effects was established by permutation testing against effect data used in the study as well as reference publications are available in distributions. Group effects were determined using Kruskall-Wallis Supplementary Data 1. Aligned surface models used to define the tests, post-hoc testing using pairwise Wilcoxon-Rank-Sum/Mann- proposed common reference frame, as well as ancestral state esti- Whitney U-Test that were corrected for multiple comparisons using mates obtained from it are publicly available at https://github.com/ the Benjamini-Yekutieli procedure for non-independent tests177. cirmuw/EvolutionOfCorticalShape and https://doi.org/10.5281/ zenodo.7713847. Expansion maps used for meta-analytic decoding Testing for the effect of species habitat on functionally defined are additionally made publicly available at https://neurovault.org/ brain regions. We use repeated measurements correlation collections/IHFSXSES/. (RMCORR)47 for testing the association of non-independent measure- ments in brain regions with species habitat, where we consider an Code availability ordering of habitat in terms of sensory complexity from arboreal to Computer code written to perform surface alignment and analysis will terrestrial to fossorial. To enable such an analysis, we first compute the be available from the corresponding author upon reasonable request. overall brain expansion computed from linear combinations of all pPCA modes significantly associated with a specific habitat. We use References summary measures for each habitat to transform the ordering of 1. Mesulam, M.-M. Principles of Behavioral and Cognitive Neurology. habitats into a tentative continuous variable (Fig. 4a). The correlation (Oxford University Press, 2000). values of each functional cortical area from RMCORR then correspond 2. Smaers, J. B., Gómez-Robles, A., Parks, A. N. & Sherwood, C. C. to a measure of relative consistency in the expansion or contraction of Exceptional evolutionary expansion of prefrontal cortex in great a specific region relative to the whole cortex. apes and humans. Curr. Biol. 27, 714–720 (2017). 3. Smaers, J. B. et al. The evolution of mammalian brain size. Sci. Adv. Null models for brain maps. Measures obtained from different loca- 7, eabe2101 (2021). tions on the cortex violate the independence assumption, which pre- 4. O’Leary, M. A. et al. The placental mammal ancestor and the post- cludes the application of most parametric and non-parametric K-Pg radiation of placentals. Science 339, 662–667 (2013). procedures for estimation of significance of results. When needed, we 5. Heuer, K. et al. Evolution of neocortical folding: a phylogenetic therefore rely on permutation testing under appropriate null hypoth- comparative analysis of MRI from 34 primate species. Cortex 118, esis. Specifically, when estimating the significance of results corre- 275–291 (2019). sponding to a specific subregion of a brain map against others, we use 6. Eichert, N. et al. Cross-species cortical alignment identifies dif- a test distribution of 1000 randomly generated brain maps showing ferent types of anatomical reorganization in the primate temporal the same spatial autocorrelation at different locations of the brain lobe. Elife 9, e53232 (2020). obtained from a generative model178. The corresponding null hypoth- 7. Mars, R. B. et al. Whole brain comparative anatomy using con- esis is that the observed effect is related to spatial autocorrelation of nectivity blueprints. Elife 7, e35237 (2018). the measurements and not any process-specific topographic feature179. 8. Xu, T. et al. Cross-species functional alignment reveals evolu- tionary hierarchy within the connectome. Neuroimage 223, Spatial statistics. We use spatial statistics to quantify the relationship 117346 (2020). between distances on the cortical surface and differences in the dis- 9. Krubitzer, L. A. & Prescott, T. J. The combinatorial creature: cortical tribution of cognitive modalities at these vertices. We approximate the phenotypes within and across lifetimes. Trends Neurosci. 41, geodesic distance between two surface vertices using the shortest 744–762 (2018). path in the graph defined from the surface mesh topology, normalise 10. Finlay, B. L. & Uchiyama, R. Developmental mechanisms chan- the resulting value to the range [0,1] and bin the resulting values into neling cortical evolution. Trends Neurosci. 38, 69–76 (2015). Nature Communications | (2023)14:2252 14 Article https://doi.org/10.1038/s41467-023-37574-x 11. Imam, N. & L Finlay, B. Self-organization of cortical areas in the relative to nonhuman primates. Proc. Natl Acad. Sci. U. S. A. 115, development and evolution of neocortex. Proc. Natl Acad. Sci. U. E5183–E5192 (2018). S. A. 117, 29212–29220 (2020). 36. Suárez, L. E. et al. A connectomics-based taxonomy of mammals. 12. Cadwell, C. R., Bhaduri, A., Mostajo-Radji, M. A., Keefe, M. G. & bioRxiv 2022.03.11.483995 https://doi.org/10.1101/2022.03.11. Nowakowski, T. J. Development and arealization of the cerebral 483995 (2022). cortex. Neuron 103, 980–1004 (2019). 37. Assaf, Y., Bouznach, A., Zomet, O., Marom, A. & Yovel, Y. Con- 13. DeCasien, A. R. & Higham, J. P. Primate mosaic brain evolution servation of brain connectivity and wiring across the mammalian reflects selection on sensory and cognitive specialization. Nat. class. Nat. Neurosci. 23, 805–808 (2020). Ecol. Evol. 3, 1483–1493 (2019). 38. Ardesch, D. J. et al. Scaling principles of white matter connectivity 14. Friedrich, P. et al. Imaging evolution of the primate brain: the next in the human and nonhuman primate brain. Cereb. Cortex 32, frontier? Neuroimage 228, 117685 (2021). 2831–2842 (2022). 15. Cisek, P. Resynthesizing behavior through phylogenetic refine- 39. Upham, N. S., Esselstyn, J. A. & Jetz, W. Inferring the mammal tree: ment. Atten. Percept. Psychophys. 81, 2265–2287 (2019). species-level sets of phylogenies for questions in ecology, evo- 16. DeCasien, A. R., Barton, R. A. & Higham, J. P. Understanding the lution, and conservation. PLoS Biol. 17, e3000494 (2019). human brain: insights from comparative biology. Trends Cogn. 40. Bastide, P., Ané, C., Robin, S. & Mariadassou, M. Inference of Sci. 26, 432–445 (2022). adaptive shifts for multivariate correlated traits. Syst. Biol. 67, 17. Kaas, J. H. The evolution of brains from early mammals to humans. 662–680 (2018). Wiley Interdiscip. Rev. Cogn. Sci. 4, 33–45 (2013). 41. Heeren, B., Rumpf, M., Schröder, P., Wardetzky, M. & Wirth, B. 18. Finlay, B. L. & Darlington, R. B. Linked regularities in the develop- Exploring the geometry of the space of shells. Comput. Graph. ment and evolution of mammalian brains. Science 268, Forum 33, 247–256 (2014). 1578–1584 (1995). 42. Fröhlich, S. & Botsch, M. Example-driven deformations 19. Barton, R. A. & Harvey, P. H. Mosaic evolution of brain structure in based on discrete shells. Comput. Graph. Forum 30, 2246–2257 mammals. Nature 405, 1055–1058 (2000). (2011). 20. Neubauer, S., Hublin, J.-J. & Gunz, P. The evolution of modern 43. Dale, A. M., Fischl, B. & Sereno, M. I. Cortical surface-based ana- human brain shape. Sci. Adv. 4, eaao5961 (2018). lysis. I. Segmentation and surface reconstruction. Neuroimage 9, 21. Ponce de León, M. S. et al. The primitive brain of early Homo. 179–194 (1999). Science 372, 165–171 (2021). 44. Krubitzer, L. & Dooley, J. C. Cortical plasticity within and across 22. Wartel, A., Lindenfors, P. & Lind, J. Whatever you want: incon- lifetimes: how can development inform us about phenotypic sistent results are the rule, not the exception, in the study of pri- transformations? Front. Hum. Neurosci. 7, 620 (2013). mate brain evolution. PLoS One 14, e0218655 (2019). 45. Revell, L. J. Size-correction and principal components for inter- 23. Silcox, M. T. & López-Torres, S. Major questions in the study of specific comparative studies. Evolution 63, 3258–3268 (2009). primate origins. Annu. Rev. Earth Planet. Sci. 45, 113–137 (2017). 46. Yeo, B. T. T. et al. The organization of the human cerebral cortex 24. Smaers, J. B. & Soligo, C. Brain reorganization, not relative brain estimated by intrinsic functional connectivity. J. Neurophysiol. size, primarily characterizes anthropoid brain evolution. Proc. Biol. 106, 1125–1165 (2011). Sci. 280, 20130269 (2013). 47. Bakdash, J. Z. & Marusich, L. R. Repeated measures correlation. 25. Grandjean, J. et al. Common functional networks in the mouse Front. Psychol. 8, 456 (2017). brain revealed by multi-centre resting-state fMRI analysis. Neuro- 48. Bertrand, O. C., Püschel, H. P., Schwab, J. A., Silcox, M. T. & image 205, 116278 (2020). Brusatte, S. L. The impact of locomotion on the brain 26. Milham, M. P. et al. An open resource for non-human primate evolution of squirrels and close relatives. Commun. Biol. 4, imaging. Neuron 100, 61–74.e2 (2018). 460 (2021). 27. Hayashi, T. et al. The nonhuman primate neuroimaging and neu- 49. Northcutt, R. G. & Kaas, J. H. The emergence and evolution of roanatomy project. Neuroimage 229, 117726 (2021). mammalian neocortex. Trends Neurosci. 18, 373–379 (1995). 28. Sudlow, C. et al. UK biobank: an open access resource for iden- 50. Kaas, J. H. The origin and evolution of neocortex: From early tifying the causes of a wide range of complex diseases of middle mammals to modern humans. Prog. Brain Res. 250, 61–81 and old age. PLoS Med. 12, e1001779 (2015). (2019). 29. Van Essen, D. C. et al. The WU-Minn human connectome project: 51. Buckner, R. L. & Krienen, F. M. The evolution of distributed asso- an overview. Neuroimage 80, 62–79 (2013). ciation networks in the human brain. Trends Cogn. Sci. 17, 30. Boyer, D. M., Gunnell, G. F., Kaufman, S. & McGeary, T. M. Mor- 648–665 (2013). phosource: archiving and sharing 3-d digital specimen data. 52. Glasser, M. F. et al. A multi-modal parcellation of human cerebral Paleontol. Soc. Pap. 22, 157–181 (2016). cortex. Nature 536, 171–178 (2016). 31. Mars, R. B., Jbabdi, S. & Rushworth, M. F. S. A common space 53. Fleagle, J. G. Primate Adaptation and Evolution. (Academic approach to comparative neuroscience. Annu. Rev. Neurosci. 44, Press, 2013). 69–86 (2021). 54. Yarkoni, T., Poldrack, R. A., Nichols, T. E., Van Essen, D. C. & Wager, 32. Barron, H. C., Mars, R. B., Dupret, D., Lerch, J. P. & Sampaio-Bap- T. D. Large-scale automated synthesis of human functional neu- tista, C. Cross-species neuroscience: closing the explanatory gap. roimaging data. Nat. Methods 8, 665–670 (2011). Philos. Trans. R. Soc. Lond. B Biol. Sci. 376, 20190633 (2021). 55. Westerhold, T. et al. An astronomically dated record of Earth’s 33. Englund, M. et al. Comparing cortex-wide gene expression pat- climate and its predictability over the last 66 million years. Science terns between species in a common reference frame. bioRxiv 369, 1383–1387 (2020). 2021.07.28.454203 https://doi.org/10.1101/2021.07.28. 56. Silcox, M. T., Benham, A. E. & Bloch, J. I. Endocasts of Microsyops 454203 (2021). (Microsyopidae, Primates) and the evolution of the brain in pri- 34. Chaplin, T. A., Yu, H.-H., Soares, J. G. M., Gattass, R. & Rosa, M. G. P. mitive primates. J. Hum. Evol. 58, 505–521 (2010). A conserved pattern of differential expansion of cortical areas in 57. Bertrand, O. C., Amador-Mughal, F. & Silcox, M. T. Virtual endo- simian primates. J. Neurosci. 33, 15120–15125 (2013). casts of Eocene Paramys (Paramyinae): oldest endocranial record 35. Donahue, C. J., Glasser, M. F., Preuss, T. M., Rilling, J. K. & Van Essen, for Rodentia and early brain evolution in Euarchontoglires. Proc. D. C. Quantitative assessment of prefrontal cortex in humans Biol. Sci. 283, 20152316 (2016). Nature Communications | (2023)14:2252 15 Article https://doi.org/10.1038/s41467-023-37574-x 58. Bertrand, O. C. & Silcox, M. T. First virtual endocasts of a fossil 81. Li, W., Mai, X. & Liu, C. The default mode network and social rodent: Ischyromys typus (Ischyromyidae, Oligocene) and brain understanding of others: what do brain connectivity studies tell evolution in rodents. J. Vert. Paleontol. 36, e1095762 (2016). us. Front. Hum. Neurosci. 8, 74 (2014). 59. Bertrand, O. C., Amador-Mughal, F. & Silcox, M. T. Virtual endocast 82. Kaas, J. H. Evolutionary Neuroscience. (Academic Press, 2020). of the early Oligocene Cedromus wilsoni (Cedromurinae) and 83. Pessoa, D. M. A. et al. The adaptive value of primate color vision for brain evolution in squirrels. J. Anat. 230, 128–151 (2017). predator detection. Am. J. Primatol. 76, 721–729 (2014). 60. Bhagat, R., Bertrand, O. C. & Silcox, M. T. Evolution of arboreality 84. Carvalho, L. S., Pessoa, D. M. A., Mountford, J. K., Davies, W. I. L. & and fossoriality in squirrels and aplodontid rodents: Insights from Hunt, D. M. The genetic and evolutionary drives behind primate the semicircular canals of fossil rodents. J. Anat. 238, color vision. Front. Ecol. Evol. 5, (2017). 96–112 (2021). 85. Simmen, B. et al. Primate energy input and the evolutionary 61. Bertrand, O. C., Amador-Mughal, F., Lang, M. M. & Silcox, M. T. transition to energy-dense diets in humans. Proc. Biol. Sci. 284, Virtual endocasts of fossil Sciuroidea: brain size reduction in the 20170577 (2017). evolution of fossoriality. Palaeontology 61, 919–948 (2018). 86. Aiello, L. C. & Wheeler, P. The expensive-tissue hypothesis: the 62. Sliwa, J. & Freiwald, W. A. A dedicated network for social inter- brain and the digestive system in human and primate evolution. action processing in the primate brain. Science 356, Curr. Anthropol. 36, 199–221 (1995). 745–749 (2017). 87. Humphrey, N. K. The social function of intellect. Grow. points 63. Putnam, P. T. & Chang, S. W. C. Social processing by the primate Ethol. 37, 303–317 (1976). medial frontal cortex. Int. Rev. Neurobiol. 158, 213–248 (2021). 88. van Schaik, C. P., Isler, K. & Burkart, J. M. Explaining brain size 64. Finlayson-Short, L., Davey, C. G. & Harrison, B. J. Neural correlates variation: from social to cultural brain. Trends Cogn. Sci. 16, of integrated self and social processing. Soc. Cogn. Affect. Neu- 277–284 (2012). rosci. 15, 941–949 (2020). 89. Heesy, C. P. Seeing in stereo: the ecology and evolution of primate 65. Burgos-Robles, A., Gothard, K. M., Monfils, M. H., Morozov, A. & binocular vision and stereopsis. Evol. Anthropol. 18, 21–35 (2009). Vicentic, A. Conserved features of anterior cingulate networks 90. Goodale, M. A. & Milner, A. D. Separate visual pathways for per- support observational learning across species. Neurosci. Biobe- ception and action. Trends Neurosci. 15, 20–25 (1992). hav. Rev. 107, 215–228 (2019). 91. Cartmill, M. Arboreal adaptations and the origin of the order pri- 66. Buckner, R. L. & DiNicola, L. M. The brain’s default network: mates. In The Functional and Evolutionary Biology of Primates updated anatomy, physiology and evolving insights. Nat. Rev. 97–122 (Routledge, 2017). Neurosci. 20, 593–608 (2019). 92. Soudry, Y., Lemogne, C., Malinvaud, D., Consoli, S.-M. & Bonfils, P. 67. Brewer, J. A., Garrison, K. A. & Whitfield-Gabrieli, S. What about the Olfactory system and emotion: common substrates. Eur. Ann. ‘Self’ is processed in the posterior cingulate cortex? Front. Hum. Otorhinolaryngol. Head. Neck Dis. 128, 18–23 (2011). Neurosci. 7, 647 (2013). 93. Rolls, E. T. The cingulate cortex and limbic systems for emotion, 68. Wagner, D. D., Kelley, W. M., Haxby, J. V. & Heatherton, T. F. The action, and memory. Brain Struct. Funct. 224, 3001–3018 (2019). dorsal medial prefrontal cortex responds preferentially to social 94. Bombail, V. Perception and emotions: On the relationships interactions during natural viewing. J. Neurosci. 36, between stress and olfaction. Appl. Anim. Behav. Sci. 212, 6917–6925 (2016). 98–108 (2019). 69. Bzdok, D. et al. The modular neuroarchitecture of social judg- 95. Agrawal, V. C. Skull adaptations in fossorial rodents. Mammalia 31, ments on faces. Cereb. Cortex 22, 951–961 (2012). 300–312 (1967). 70. Amft, M. et al. Definition and characterization of an extended 96. Pietrini, P. et al. Beyond sensory images: object-based repre- social-affective default network. Brain Struct. Funct. 220, sentation in the human ventral pathway. Proc. Natl Acad. Sci. U. S. 1031–1049 (2015). A. 101, 5658–5663 (2004). 71. Schwiedrzik, C. M., Zarco, W., Everling, S. & Freiwald, W. A. Face 97. Snow, J. C., Strother, L. & Humphreys, G. W. Haptic shape pro- patch resting state networks link face processing to social cog- cessing in visual cortex. J. Cogn. Neurosci. 26, 1154–1167 (2014). nition. PLoS Biol. 13, e1002245 (2015). 98. Sathian, K. Analysis of haptic information in the cerebral cortex. J. 72. Van Schaik, C. P. & van Hooff, J. A. R. A. M. Toward an under- Neurophysiol. 116, 1795–1806 (2016). standing of the orangutan’s social system. in Great ape societies, 99. Amedi, A., Raz, N., Azulay, H., Malach, R. & Zohary, E. Cortical (pp (ed. McGrew, W. C.) vol. 328 3–15 (Cambridge University activity during tactile exploration of objects in blind and sighted Press, xx, 1996). humans. Restor. Neurol. Neurosci. 28, 143–156 (2010). 73. Shultz, S., Opie, C. & Atkinson, Q. D. Stepwise evolution of stable 100. Weiller, C. et al. The ventral pathway of the human brain: a con- sociality in primates. Nature 479, 219–222 (2011). tinuous association tract system. Neuroimage 234, 117977 (2021). 74. Van Schaik, C. P. Why are diurnal primates living in groups? 101. Ptito, M. et al. Crossmodal recruitment of the ventral visual stream Behaviour 87, 120–144 (1983). in congenital blindness. Neural Plast. 2012, 304045 (2012). 75. Fusar-Poli, P. et al. Functional atlas of emotional faces processing: 102. Cisek, P. Cortical mechanisms of action selection: the affordance a voxel-based meta-analysis of 105 functional magnetic reso- competition hypothesis. Philos. Trans. R. Soc. Lond. B Biol. Sci. nance imaging studies. J. Psychiatry Neurosci. 34, 418–432 (2009). 362, 1585–1599 (2007). 76. Adolphs, R. What does the amygdala contribute to social cogni- 103. Graziano, M. S. A. Ethological action maps: a paradigm shift for the tion? Ann. N. Y. Acad. Sci. 1191, 42–61 (2010). motor cortex. Trends Cogn. Sci. 20, 121–132 (2016). 77. Zuberbühler, K. & Janmaat, K. Foraging cognition in non-human 104. Falkowski, P. G. et al. The rise of oxygen over the past 205 million primates. Primate Neuroethology 64–83 (2010). years and the evolution of large placental mammals. Science 309, 78. Rosati, A. G. Foraging cognition: reviving the ecological intelli- 2202–2204 (2005). gence hypothesis. Trends Cogn. Sci. 21, 691–702 (2017). 105. National Research Council, Division on Earth and Life Studies, 79. Dunbar, R. I. M. The social brain hypothesis. Evol. Anthropol. 6, Board on Earth Sciences and Resources & Committee on the Earth 178–190 (1998). System Context for Hominin Evolution. Understanding Climate’s 80. Mars, R. B. et al. On the relationship between the ‘default mode Influence on Human Evolution. (National Academies Press, 2010). network’ and the ‘social brain’. Front. Hum. Neurosci. 6, 189 106. Soligo, C. & Smaers, J. B. Contextualising primate origins - an (2012). ecomorphological framework. J. Anat. 228, 608–629 (2016). Nature Communications | (2023)14:2252 16 Article https://doi.org/10.1038/s41467-023-37574-x 107. Ghazanfar, A. A. Primate Neuroethology. (OUP USA, 2012). 130. Isler, K. & Schaik, C. P.Van How humans evolved large brains: 108. Goodale, M. A., Króliczak, G. & Westwood, D. A. Dual routes to comparative evidence. Evol. Anthropol. 23, 65–75 (2014). action: contributions of the dorsal and ventral streams to adaptive 131. Bruner, E. Human paleoneurology: shaping cortical evolution in behavior. In Progress in Brain Research vol. 149 269–283 (Else- fossil hominids. J. Comp. Neurol. 527, 1753–1765 (2019). vier, 2005). 132. Cavanna, A. E. & Trimble, M. R. The precuneus: a review of its 109. Schilder, B. M., Petry, H. M. & Hof, P. R. Evolutionary shifts dra- functional anatomy and behavioural correlates. Brain 129, matically reorganized the human hippocampal complex. J. Comp. 564–583 (2006). Neurol. 528, 3143–3170 (2020). 133. Schurz, M. et al. Toward a hierarchical model of social cognition: a 110. Murray, E. A., Wise, S. P. & Graham, K. S. Representational spe- neuroimaging meta-analysis and integrative review of empathy cializations of the hippocampus in phylogenetic perspective. and theory of mind. Psychol. Bull. 147, 293–327 (2021). Neurosci. Lett. 680, 4–12 (2018). 134. Whiten, A. & Erdal, D. The human socio-cognitive niche and its 111. Herweg, N. A. & Kahana, M. J. Spatial representations in the human evolutionary origins. Philos. Trans. R. Soc. Lond. B Biol. Sci. 367, brain. Front. Hum. Neurosci. 12, 297 (2018). 2119–2129 (2012). 112. Solomon, E. A., Lega, B. C., Sperling, M. R. & Kahana, M. J. Hip- 135. Son, J.-Y., Bhandari, A. & FeldmanHall, O. Cognitive maps of social pocampal theta codes for distances in semantic and temporal features enable flexible inference in social networks. Proc. Natl spaces. Proc. Natl Acad. Sci. U. S. A. 116, 24343–24352 (2019). Acad. Sci. U. S. A. 118, e2021699118 (2021). 113. Naya, Y. & Suzuki, W. A. Integrating what and when across the 136. Muthukrishna, M., Doebeli, M., Chudek, M. & Henrich, J. The cul- primate medial temporal lobe. Science 333, 773–776 (2011). tural brain hypothesis: how culture drives brain expansion, soci- 114. Gavrilov, N. & Nieder, A. Distinct neural networks for the volitional ality, and life history. PLoS Comput. Biol. 14, e1006504 (2018). control of vocal and manual actions in the monkey homologue of 137. Seyfarth, R. & Cheney, D. The Social Origins of Language. (Prin- Broca’s area. Elife 10, e62797 (2021). ceton University Press, 2017). 115. Song, X., Osmanski, M. S., Guo, Y. & Wang, X. Complex pitch 138. Veissière, S. P. L., Constant, A., Ramstead, M. J. D., Friston, K. J. & perception mechanisms are shared by humans and a New World Kirmayer, L. J. Thinking through other minds: a variational monkey. Proc. Natl Acad. Sci. U. S. A. 113, 781–786 (2016). approach to cognition and culture. Behav. Brain Sci. 43, 116. Hackett, T. A. Information flow in the auditory cortical network. e90 (2019). Hear. Res. 271, 133–146 (2011). 139. Stout, D. & Hecht, E. E. Evolutionary neuroscience of cumulative 117. Aboitiz, F. A brain for speech. Evolutionary continuity in primate culture. Proc. Natl Acad. Sci. U. S. A. 114, 7861–7868 (2017). and human auditory-vocal processing. Front. Neurosci. 12, 140. Maslin, M. A., Shultz, S. & Trauth, M. H. A synthesis of the theories 174 (2018). and concepts of early human evolution. Philos. Trans. R. Soc. 118. Haladjian, H. H. & Montemayor, C. On the evolution of conscious Lond. B Biol. Sci. 370, 20140064 (2015). attention. Psychon. Bull. Rev. 22, 595–613 (2015). 141. Thevenaz, P., Ruttimann, U. E. & Unser, M. A pyramid approach to 119. Brass, M., Ullsperger, M., Knoesche, T. R., von Cramon, D. Y. & subpixel registration based on intensity. IEEE Trans. Image Pro- Phillips, N. A. Who comes first? The role of the prefrontal and cess. 7, 27–41 (1998). parietal cortex in cognitive control. J. Cogn. Neurosci. 17, 142. Yushkevich, P. A. et al. User-guided 3D active contour segmen- 1367–1375 (2005). tation of anatomical structures: significantly improved efficiency 120. Duncan, J. The multiple-demand (MD) system of the primate brain: and reliability. Neuroimage 31, 1116–1128 (2006). mental programs for intelligent behaviour. Trends Cogn. Sci. 14, 143. Aubert-Broche, B., Evans, A. C. & Collins, L. A new improved ver- 172–179 (2010). sion of the realistic digital brain phantom. Neuroimage 32, 121. Cole, M. W. & Schneider, W. The cognitive control network: inte- 138–145 (2006). grated cortical regions with dissociable functions. Neuroimage 37, 144. Yotter, R. A., Dahnke, R. & Gaser, C. Topological correction of brain 343–360 (2007). surface meshes using spherical harmonics. In Medical Image 122. Bor, D. & Seth, A. K. Consciousness and the prefrontal parietal Computing and Computer-Assisted Intervention – MICCAI 2009 network: insights from attention, working memory, and chunking. 125–132 (Springer Berlin Heidelberg, 2009). Front. Psychol. 3, 63 (2012). 145. Crane, K., Pinkall, U. & Schröder, P. Robust fairing via conformal 123. Marchetti, G., Benedetti, G., & Alharbi, A. (eds.) Attention and curvature flow. ACM Trans. Graph. 32, 1–10 (2013). meaning: the attentional basis of meaning. Nova Science 146. Reuter, M., Wolter, F. E. & Peinecke, N. Laplace-spectra as fin- Publishers (2015). gerprints for shape matching. Proc. of the 2005 ACM symposium 124. Stout, D. & Chaminade, T. Stone tools, language and the brain in on Solid and physical modeling 101–106 (2005). human evolution. Philos. Trans. R. Soc. Lond. B Biol. Sci. 367, 147. Reuter, M., Wolter, F.-E., Shenton, M. & Niethammer, M. 75–87 (2012). Laplace–Beltrami eigenvalues and topological features of eigen- 125. Graham, K. E., Furuichi, T. & Byrne, R. W. The gestural repertoire of functions for statistical shape analysis. Comput-Aided Des. 41, the wild bonobo (Pan paniscus): a mutually understood commu- 739–755 (2009). nication system. Anim. Cogn. 20, 171–177 (2017). 148. Isaacs, J. C. Laplace-Beltrami eigenfunctions for 3D shape 126. Tomasello, M., Melis, A. P., Tennie, C., Wyman, E. & Herrmann, E. matching. Detection and sensing of mines, explosive objects, and Two key steps in the evolution of human cooperation: the inter- obscured targets XVI https://doi.org/10.1117/12.885642 (2011). dependence hypothesis. Curr. Anthropol. 53, 673–692 (2012). 149. Lombaert, H., Sporring, J. & Siddiqi, K. Diffeomorphic spectral 127. Hobaiter, C. & Byrne, R. W. The meanings of chimpanzee gestures. matching of cortical surfaces. Inf. Process. Med. Imaging 23, Curr. Biol. 24, 1596–1600 (2014). 376–389 (2013). 128. Fröhlich, M. et al. Unpeeling the layers of language: bonobos and 150. Shi, Y., Thompson, P. M., Dinov, I. & Toga, A. W. Hamilton–Jacobi chimpanzees engage in cooperative turn-taking sequences. Sci. skeleton on cortical surfaces. IEEE Trans. Med. Imaging 27, Rep. 6, 25887 (2016). 664–673 (2008). 129. Huber, L. et al. Sub-millimeter fMRI reveals multiple topographical 151. Sethian, J. A. Fast marching methods. SIAM Rev. 41, 199–235 (1999). digit representations that form action maps in human motor cor- 152. Myronenko, A. & Song, X. Point set registration: coherent point tex. Neuroimage 208, 116463 (2020). drift. IEEE Trans. Pattern Anal. Mach. Intell. 32, 2262–2275 (2010). Nature Communications | (2023)14:2252 17 Article https://doi.org/10.1038/s41467-023-37574-x 153. Lombaert, H., Grady, L., Polimeni, J. R. & Cheriet, F. Fast brain 174. Paxinos, G., Watson, C., Petrides, M., Rosa, M. & Tokuno, H. The matching with spectral correspondence. Inf. Process. Med. Ima- Marmoset Brain in Stereotaxic Coordinates. (Elsevier Sci- ging 22, 660–673 (2011). ence, 2011). 154. Wright, R. et al. Construction of a fetal spatio-temporal cortical 175. Paxinos, G., Petrides, M. & Evrard, H. C. The Rhesus Monkey Brain in surface atlas from in utero MRI: application of spectral surface Stereotaxic Coordinates. (Elsevier Science, 2021). matching. NeuroImage 120, 467–480 (2015). 176. Amunts, K., Mohlberg, H., Bludau, S. & Zilles, K. Julich-Brain: A 3D 155. Nenning, K.-H. et al. Joint embedding: a scalable alignment to probabilistic atlas of the human brain’s cytoarchitecture. Science compare individuals in a connectivity space. Neuroimage 222, 369, 988–992 (2020). 117232 (2020). 177. Benjamini, Y. & Yekutieli, D. The control of the false discovery rate 156. Ezuz, D., Solomon, J. & Ben-Chen, M. Reversible harmonic maps in multiple testing under dependency. Ann Stat 29, between discrete surfaces. ACM Trans. Graph. 38, 1–12 (2019). 1165–1188 (2001). 157. Budday, S., Steinmann, P., Goriely, A. & Kuhl, E. Size and curvature 178. Burt, J. B., Helmer, M., Shinn, M., Anticevic, A. & Murray, J. D. regulate pattern selection in the mammalian brain. Extreme Mech. Generative modeling of brain maps with spatial autocorrelation. Lett. (2015). Neuroimage 220, 117038 (2020). 158. Solomon, J. et al. Convolutional wasserstein distances. ACM 179. Weinstein, S. M. et al. A simple permutation-based test of inter- Trans. Graph. 34, 1–11 (2015). modal correspondence. Hum. Brain Mapp. 42, 5175–5187 (2021). 159. Melchionna, M. et al. From smart apes to human brain boxes. A uniquely derived brain shape in late hominins clade. Front. Earth Sci. 8, 273 (2020). Acknowledgements 160. Ragan, M. A. Phylogenetic inference based on matrix repre- We gratefully acknowledge everyone who contributed data to the pro- sentation of trees. Mol. Phylogenet. Evol. 1, 53–58 (1992). ject. In alphabetical order, Eike Budinger of the Leibniz Institute for 161. Rodriguez-Carranza, C. E., Mukherjee, P., Vigneron, D., Barkovich, Neurobiology for contributing data of Meriones unguiculatus, Marc J. & Studholme, C. Comparing 3D Gyrification Index and area- Dhenain of the Institut de biologie Francois Jacob for contributing data independent curvature-based measures in quantifying neonatal of Microcebus murinus, Hao Lei of Huashan Hospital for contributing brain folding. In Medical Imaging 2007: Image Processing vol. data of Tupaia belangeri, Kyoko Miura of Kumamoto University for con- 6512 65120N (International Society for Optics and Photo- tributing data of Heterocephalus glaber, Thomas Morrison for con- nics, 2007). tributing data of Mesocricetus auratus, Brian Tainor and Vanessa A. 162. Mota, B. & Herculano-Houzel, S. Cortical folding scales universally Minnie (with support from NIH R01 MH121829) for contributing data of with surface area and thickness, not number of neurons. Science Peromyscus californicus, Cody W. Thompson of the University of 349, 74–77 (2015). Michigan Museum of Zoology for contributing data of Tamias striatus 163. Herculano-Houzel, S. & Manger, P. R. Brain scaling in mammalian and Sarael Alcauter Solorzano of the Universidad Nacional Autonoma de evolution as a consequence of concerted and mosaic changes in Mexico and Jason R. Yee of the University of Veterinary Medicine Vienna numbers of neurons and average neuronal cell size. Front. Neu- for contributing data of Microtus ochrogaster. We especially want to roanat. 8, 77 (2014). acknowledge the effort of making neuroanatomical data publicly avail- 164. Herculano-Houzel, S. The remarkable, yet not extraordinary, able made by the Digital Morphology Museum of Kyoto University, the human brain as a scaled-up primate brain and its associated cost. Brain Catalogue from Institut Pasteur, the National Natural History Proc. Natl Acad. Sci. U. S. A. 109, 10661–10668 (2012). Museum Paris and the Institute for Brain and Spinal Cord, the NIN Pri- 165. Clavel, J., Escarguel, G. & Merceron, G. Mv morph: an r package mate Brain Bank of Utrecht University, the Institute of Neuroscience and for fitting multivariate evolutionary models to morphometric data. Medicine, Brain and Behaviour (INM-7) at Forschungszentrum Jülich, the Methods Ecol. Evol. 6, 1311–1319 (2015). Comparative Mammalian Brain Collections of the Brain Museum from 166. Konishi, S. & Kitagawa, G. Generalised information criteria in the University of Michigan, Michigan State University and the National model selection. Biometrika 83, 875–890 (1996). Museum of Health and Medicine, the Montreal Neurological Institute 167. Clavel, J., Aristide, L. & Morlon, H. A penalized likelihood frame- and COBRA Laboratory at McGill University, the Laboratory of Brain and work for high-dimensional phylogenetic comparative methods Cognition at the National Institute of Mental Health, the Laboratory for and an application to new-world monkeys brain evolution. Syst. Symbolic Cognitive Development at RIKEN Brain Science Institute, the Biol. 68, 93–116 (2019). Center of Advanced European Studies and Research in Bonn, the 168. Aristide, L. et al. Brain shape convergence in the adaptive radiation National Institutes of Mental Health, the Duke Center for In Vivo Micro- of New World monkeys. Proc. Natl Acad. Sci. U. S. A. 113, scopy at Duke University, the Cambridge Biotomography Centre at 2158–2163 (2016). Cambridge University, the Museum of Vertebrate Zoology at Berkeley, 169. Zhi, D., King, M., Hernandez‐Castillo, C. R., & Diedrichsen, J. the Marseille Functional MRI Center, the Institut de biologie Francois Evaluating brain parcellations using the distance‐controlled Jacob, the Institute of Basic Medical Sciences at the University of Oslo boundary coefficient. Human Brain Mapping 43, and the Museum Nationale d’Histoire Naturelle de Paris. Furthermore, 3706–3720 (2022). the authors gratefully acknowledge the collection efforts by the Amer- 170. de Sousa, A. A. et al. Comparative cytoarchitectural analyses of ican Museum of Natural History (AMNH FM 12506, AMNH 4756, AMNH striate and extrastriate areas in hominoids. Cereb. Cortex 20, FM 128167, AMNH FAM 65511, AMNH 12561), the United States National 966–981 (2010). Museum (Smithsonian Institution) (USNM V 17996, USNM 256584, USNM 171. Butti, C., Raghanti, M. A., Sherwood, C. C. & Hof, P. R. The neo- 17161), the Natural History Museum London (NHM PV M 1345), the Geo- cortex of cetaceans: cytoarchitecture and comparison with other logical Museum of the University of Wyoming (UW 12362), the Field aquatic and terrestrial species. Ann. N. Y. Acad. Sci. 1225, Museum of Natural History (FMNH UC 1642), the Musée d’Histoire Nat- 47–58 (2011). urelle de Montauban, Victor Brun (MA PHQ 289), the Jackson School of 172. Amunts, K., Schleicher, A. & Zilles, K. Cytoarchitecture of the Geosciences at the University of Texas, Austin (TMM 40688-7), the Yale cerebral cortex–more than localization. NeuroImage 37, Peabody Museum (YPM 14737) and the Museo Argentino de Ciencias 1061–5 (2007). Naturales, Buenos Aires (MACN 14128) as well as Doug Broyer, Mathew 173. Paxinos, G. & Franklin, K. B. J. Paxinos and Franklin’s the Mouse Colbert, Rachel Ives, Ruth O’Leary and Mary Silcox for providing access Brain in Stereotaxic Coordinates. (Academic Press, 2019). to endocast data. ES has received financial support for this project by the Nature Communications | (2023)14:2252 18 Article https://doi.org/10.1038/s41467-023-37574-x Austrian Research Fund (FWF) grant I 3925-B27 in collaboration with the Additional information French National Research Agency (ANR). RT and KH were funded by the Supplementary information The online version contains French National Research Agency and National Science Foundation supplementary material available at grant agreement project NeuroWebLab (ANR-19-DATA-0025), DMOBE https://doi.org/10.1038/s41467-023-37574-x. (ANR-21-CE45-0016), the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agree- Correspondence and requests for materials should be addressed to ment No101033485 (KH Individual Fellowship), and the STIC-AmSud Georg Langs. programme project STIC-AMSUD + CLANN 22-STIC-03. GL has received funding by the Austrian Research Fund (FWF) grant P 35189, the Vienna Peer review information Nature Communications thanks Leandro Aris- Science and Technology Fund (WWTF) grant 10.47379/LS20065, and tide and the other, anonymous, reviewer(s) for their contribution to the the European Commission grant 765148 TRABIT. OCB has received peer review of this work. Peer reviewer reports are available. financial support from the Marie Skłodowska-Curie Actions: Individual Fellowship (H2020-MSCA-IF-2018-2020; No. 792611). Reprints and permissions information is available at http://www.nature.com/reprints Author contributions E.S. designed the study and the experiments, created the software used Publisher’s note Springer Nature remains neutral with regard to jur- in the analysis, performed the analysis and parts of the cortical seg- isdictional claims in published maps and institutional affiliations. mentations and wrote the article. K.H.N. co-designed the study and the experiments, supported the analysis and co-wrote the article. K.H. Open Access This article is licensed under a Creative Commons initiated the project, performed parts of the cortical segmentations, Attribution 4.0 International License, which permits use, sharing, interpreted the results and commented on and edited the article. N.J. adaptation, distribution and reproduction in any medium or format, as performed scanning of specimens, interpreted the results and com- long as you give appropriate credit to the original author(s) and the mented on and edited the article, O.C.B., provided fossil endocast data, source, provide a link to the Creative Commons license, and indicate if interpreted the results and commented on and edited the article, R.T. changes were made. The images or other third party material in this initialised the project, interpreted the results and commented on and article are included in the article’s Creative Commons license, unless edited the article, G.K. provided funding for the project, interpreted the indicated otherwise in a credit line to the material. If material is not results and commented on and edited the article, D.P. provided included in the article’s Creative Commons license and your intended resources for the project and edited the article. G.L. supervised the use is not permitted by statutory regulation or exceeds the permitted project, co-designed the study, interpreted the results and co-wrote use, you will need to obtain permission directly from the copyright the article. holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/. Competing interests The authors declare no competing interests © The Author(s) 2023 Nature Communications | (2023)14:2252 19