The virtual brain endocast of Trogosus (Mammalia, Tillodontia) and its relevance in understanding the extinction of archaic placental mammals

Abstract

After successfully diversifying during the Paleocene, the descendants of the first wave of mammals that survived the end‐Cretaceous mass extinction waned throughout the Eocene. Competition with modern crown clades and intense climate fluctuations may have been part of the factors leading to the extinction of these archaic groups. Why these taxa went extinct has rarely been studied from the perspective of the nervous system. Here, we describe the first virtual endocasts for the archaic order Tillodontia. Three species from the middle Eocene of North America were analyzed: Trogosus hillsii, Trogosus grangeri, and Trogosus castoridens. We made morphological comparisons with the plaster endocast of another tillodont, Tillodon fodiens, as well as groups potentially related to Tillodontia: Pantodonta, Arctocyonidae, and Cimolesta. Trogosus shows very little inter‐specific variation with the only potential difference being related to the fusion of the optic canal and sphenorbital fissure. Many ancestral features are displayed by Trogosus, including an exposed midbrain, small neocortex, orbitotemporal canal ventral to rhinal fissure, and a broad circular fissure. Potential characteristics that could unite Tillodontia with Pantodonta, and Arctocyonidae are the posterior position of cranial nerve V₃ exit in relation to the cerebrum and the low degree of development of the subarcuate fossa. The presence of large olfactory bulbs and a relatively small neocortex are consistent with a terrestrial lifestyle. A relatively small neocortex may have put Trogosus at risk when competing with artiodactyls for potentially similar resources and avoiding predation from archaic carnivorans, both of which are known to have had larger relative brain and neocortex sizes in the Eocene. These factors may have possibly exacerbated the extinction of Tillodontia, which showed highly specialized morphologies despite the increase in climate fluctuations throughout the Eocene, before disappearing during the middle Eocene.

We show that Eocene archaic placental mammals such as Trogosus retained plesiomorphic endocranial features in comparison with contemporaneous Eocene crown clades. Several anatomical features are also identified as potential synapomorphies that might unite Tillodontia with Pantodonta and Arctocyonidae. Finally, Trogosus had a relatively small neocortex, which could have negatively impacted its survival and been linked to the extinction of this genus and of Tillodontia during the middle Eocene.

1. INTRODUCTION

Sixty‐six million years ago, the end‐Cretaceous mass extinction led to the collapse of ecosystems on land, in the air, and in the sea. Mammals survived and recovered relatively rapidly, with the appearance of a plethora of new species of placentals and close relatives during the Paleocene (Rose, 2006). Many of these species are considered archaic placentals, as they have been challenging to place in the tree of life because of their unusual morphology compared to modern groups that we know today as crown clades and which include, for example, artiodactyls, perissodactyls, carnivorans, rodents, euprimates, and bats (Rose, 2006). Remnants of the descendants of this first wave of archaic placental mammals remained throughout the Eocene, when crown clades started diversifying. However, the archaic lineages began to decrease in diversity and by the end of the Eocene, most of them were extinct. The reason for their extinction remains contentious, but a prevalent hypothesis holds that they may have been outcompeted by the crown clades (Lucas & Schoch, 1998). More recently, new evidence relying on the study of the neurosensory system has reached similar conclusions. Crown clades had higher relative brain and neocortical sizes translating into an increase in behavioral flexibility compared to archaic species, which may have given a selective advantage to the former (Bertrand et al., 2022). However, more information on the detailed neurosensory anatomy of archaic placentals is necessary to better understand the possible link between their brains and their extinction. Here, we focus on one particular genus belonging to the archaic order, Tillodontia, and describe its brain anatomy for the first time using virtual endocasts.

Tillodonts first appeared in the fossil record in the early Paleocene of Asia, which suggest that they originated and radiated from this part of the world (Rose, 2006; Rose et al., 2009). Material has also been recovered from North America (early Paleocene to middle Eocene) and Europe (early Eocene). Recently, tillodont teeth have also been found in the early Eocene of India (Rose et al., 2009, 2013). These various occurrences strongly suggest faunal exchange between the different landmasses. The last tillodonts were found in North America and Asia during the middle Eocene (Lucas & Schoch, 1998; Rose et al., 2009). Tillodonts had a relatively specialized morphology and a rather low taxonomic diversity (~15 species). They are characterized by hypertrophied gliriform incisors with restricted enamel. Additionally, the second pair of incisors is enlarged, and in younger species, these teeth are ever‐growing and hypsodont (Gazin, 1953; Rose, 2006). Postcranial remains suggest that the earliest tillodonts were terrestrial and may have been able to climb (e.g., esthonychines; Rose, 2001), while more derived members such as trogosines were specialized in digging behavior and used their moderately large and recurved claws to unearth roots and tubers (Gingerich & Gunnell, 1979).

The interordinal phylogenetic relationships of Tillodontia are still an open question and two main hypotheses have been proposed. They may have been closely related to arctocyonid ‘condylarths’ (Gazin, 1953; Szalay, 1977), or part of the larger group Cimolesta, within which they would have been closely related to pantodonts (McKenna & Bell, 1997; Rose, 2006). The relationship between pantodonts and tillodonts is supported by a uniquely shared feature: dilambdodont dentition (Chow & Wang, 1979; Gingerich & Gunnell, 1979; Lucas, 1993). Moreover, based on a study of a large number of morphological characters, pantodonts and tillodonts were found closely related in a recent higher level phylogenetic analysis of mammals (Muizon et al., 2015). Tillodonts are all placed in one family Esthonychidae, subdivided into two subfamilies mainly based on the morphology of the second incisors. Esthonychinae is the more generalized subfamily and includes two genera (Gingerich & Gunnell, 1979), while Trogosinae is composed of five genera and displays a more specialized morphology including rootless second incisors (Gazin, 1953).

One aspect of tillodont evolution that remains particularly understudied is their neurosensory system. The only brain endocast ever studied for this group is a physical endocast of Tillodon fodiens. A simple reconstruction and very brief description of the cast was first published by Marsh (1876: fig. 1). Later, Edinger (1929: fig. 119b,c) produced more detailed illustrations of the cast with a more detailed but still brief description, as she did not have the cranium at her disposal. Gazin reused Edinger's drawings, added a ventral view of the cast, and made a more detailed description of the cast. Unfortunately, the cast made for this specimen was not of high quality and displayed for example an unexpectedly dorsally enlarged cerebellum in which hardly any structures were identifiable (see Gazin, 1953: fig. 20b). This would suggest the possible presence of large canals for vessels that would have collapsed, leading to a larger endocranial cavity. The low number of studies on the brain evolution of Tillodontia is not an anomaly. Before CT scanning technology became widely accessible, very little work had focused even on groups for which high‐quality skull fossils are available. Generating brain endocasts for extinct species required destructive techniques such as making cross‐sectioned slices of the actual specimens and peeling the cranium to access the imprint of the brain (e.g., Meng et al., 2003; Novacek, 1986). Another method, only possible in crania without sediment inside, consisted of filling the endocranial cavity with latex, which can be removed once solidified (Russell & Sigogneau, 1965).

With the use of CT data to virtually extract brain endocasts, studies describing the brain evolution of mammals and other vertebrates have flourished over the past two decades (Dozo et al., 2023). In particular, a growing body of literature on Eocene crown placental mammals such as rodents (Bertrand et al., 2016, 2019b), euprimates (Harrington et al., 2016, 2020; Kirk et al., 2014; Ramdarshan & Orliac, 2016), artiodactyls (Orliac & Gilissen, 2012), afrotherians (Benoit et al., 2013), and chiropterans (Maugoust & Orliac, 2021) has been published. However, very few virtual endocasts of archaic Eocene mammals have been described; these include plesiadapiforms (Silcox et al., 2009, 2010), the ‘condylarth’ Hyopsodus (Orliac et al., 2012), the anagalid Anagale (López‐Torres et al., 2023), and notoungulates (Perini et al., 2022). Descriptions of virtual endocasts from Paleocene archaic mammals are also rare; these include Labidolemur (Silcox et al., 2011), the plesiadapiforms Plesiadapis and Niptomomys (Orliac et al., 2014; White et al., 2023), the pantodont Alcidedorbignya (Muizon et al., 2015), the ‘condylarth’ Carsioptychus (Cameron et al., 2019), the taeniodont Onychodectes (Napoli et al., 2018), and the ‘condylarth’ Chriacus (Bertrand et al., 2020). Finally, a large number of archaic mammals were virtually segmented by the study from Bertrand et al. (2022), but many of these brain endocasts have yet to be described in detail.

Here, we describe the first virtual endocasts of a tillodont, belonging to the middle Eocene Trogosus. This genus is well known from North American localities in Wyoming and Colorado, but remains have also been found in California, Utah, and British Columbia (Gazin, 1953; Miyata, 2007b; Miyata & Deméré, 2016; Robinson, 1966) and Japan (Miyata, 2007a). The phylogenetic relationships among the species within the genus Trogosus are not resolved. Several authors have suggested that the sympatric species Trogosus castoridens and Trogosus hyracoides from the Green River Basin in Wyoming could be synonymous (Lucas & Schoch, 1998; Robinson, 1966) and rostrum length disparity between both taxa might be related to sexual dimorphism (Gazin, 1953). A similar hypothesis was proposed for another sympatric pair, Trogosus hillsii and Trogosus grangeri from the Huerfano Basin in Colorado, also displaying a difference in the length of the rostrum. More recently, Miyata and Deméré (2016) described a new specimen from the Delmar Formation (San Diego, California) that they attributed to Trogosus castoridens. They also generated a new phylogenetic analysis including this specimen and new cranio‐dental and postcranial elements. Miyata and Deméré (2016) found similar groupings (i.e., sister taxon pairs of Tr. hillsii + Tr. grangeri and Tr. castoridens + Tr. hyracoides) but maintained the four taxa as separate species based on cranial differences. They hypothesized that the shortening of the rostrum may have evolved convergently in Tr. hillsii and Tr. castoridens. Ultimately, a consensus has been difficult to reach because of the small sample size and the fragmentary nature of most Trogosus fossils. A study of the neurosensory system of Trogosus, one of the most derived tillodont genera, may provide new insight into the unresolved phylogenetic relationships of Tillodontia, and raise hypotheses for whether, and how, the neurosensory system may have been related to their eventual extinction.

2. INSTITUTIONAL ABBREVIATIONS

AMNH, American Museum of Natural History, New York, NY, USA; CR, Cernay Les Reims, France; CRL, Conglomérat de Cernay of Lemoine collection and quarry of Mont de Berru, Paris, France; MCZ, Museum of Comparative Zoology, Cambridge, MA, USA; MHNC, Museo de Historia Natural ‘Alcide d'Orbigny’, Cochabamba, Bolivia; MNHN, Muséum Nationale d'Histoire Naturelle, Paris, France; SDSM, South Dakota School of Mines, SD, USA; SDSNH, San Diego Natural History Museum, San Diego, California, USA.; UM, University of Michigan, Michigan, USA; USNM, United States National Museum of Natural History, Washington, D.C., USA; YPM, Peabody Museum of Natural History, Yale University, New Haven, Connecticut, USA.

3. MATERIALS AND METHODS

This study includes skulls of three different species belonging to the genus Trogosus: Tr. hillsii, Tr. grangeri, and Tr. castoridens that are all from the middle Eocene. The skulls of Tr. hillsii and Tr. grangeri were both described by Gazin (1953) and these two specimens are the types of both species. The majority of the skull is preserved in Tr. hillsii (USNM 17157) but lacks parts of the rostrum and is from the upper Huerfano B (Huerfano Basin, Colorado; no details on the site). Trogosus grangeri (AMNH 17008) includes a complete skull and some postcranial elements. It is also from Huerfano B (Huerfano‐Muddy divide, 3 miles west of Gardner, Huerfano Basin, Colorado). Photographs of both crania were published by Miyata and Deméré (2016). Trogosus castoridens was first described by Leidy in 1871 based on a partial mandible with left and right dentaries and incomplete dentition (illustrated in Gazin, 1953), which was the only material known for the species until a new specimen was described by Miyata and Deméré (2016). Trogosus castoridens (SDSNH 40819) is a nearly complete cranium with an almost complete dentition and some postcranial elements (Miyata & Deméré, 2016). This specimen is from the upper portion of the Delmar Formation (San Diego County, California).

3.1. Comparative sample

We compared Trogosus brain endocasts with potentially closely related groups. As a framework, we considered the phylogenetic analysis of Muizon et al. (2015), which placed Tillodontia within Laurasiatheria as part of a cluster of early placental mammals, and not particularly closely related to any crown clades in the most parsimonious trees from their analyses. Instead, Tillodontia was placed close to the archaic groups Pantodonta and Arctocyonidae (Muizon et al., 2015: figs. 121, 122). However, the tree topologies are relatively unstable due to the presence of a high‐level of homoplasy and these results should be interpreted with caution (Muizon et al., 2015). We made comparisons with the tillodont Tillodon fodiens (YPM 11087; Gazin, 1953), the pantodont Alcidedorbignya inopinata (MHNC 8372; Muizon et al., 2015), the arctocyonid ‘condylarths’ Chriacus baldwini (MCZ 20676; Bertrand et al., 2020), Arctocyon primaevus (MNHN CR 700), and Arctocyonides arenae (MNHN CR 733; Russell & Sigogneau, 1965), the ‘condylarth’ Hyopsodus lepidus (AMNH 143783; Orliac et al., 2012), the ‘condylarth’ Pleuraspidotherium aumonieri (MNHN CRL 252; Russell & Sigogneau, 1965), and the periptychid ‘condylarth’ Carsioptychus coarctatus (AMNH 27601; Cameron et al., 2019). Because of possible close relationships with the eutherian, and possibly basal placental, mammal group Cimolesta (Rose, 2006), we compared Trogosus with the taeniodont Onychodectes tisonensis (AMNH 785; Napoli et al., 2018), the leptictid Leptictis (AMNH FM 96730; Novacek, 1982, 1986), and the palaeoryctid Eoryctes melanus (UM 68074; Thewissen & Gingerich, 1989; Wible, 2022). For natural or handmade endocasts, we only used well‐described and illustrated specimens. Therefore, we did not make comparisons with drawn endocasts from the publications of Cope (1877), Marsh (1876), and Schoch (1983).

3.2. Virtual endocast acquisition

Two of the three specimens were CT scanned for this project: Tr. grangeri (AMNH FM 17008) was scanned at the Microscopy and Imagery Facility (MIF) at the AMNH and Tr. hillsii (USNM 17157) was scanned at the University of Texas, in the high‐resolution X‐ray CT facility (UTST). They were scanned at a resolution (voxel size), respectively, of 0.08918041 and 0.145 mm. The specimen of Tr. castoridens (SDSNH 40819) was kindly shared with us by K. Miyata and T. Deméré. The voxel size for this specimen is 0.1 mm. This specimen was scanned by K. Miyata using a high‐resolution X‐ray CT scanner, TXS320‐ACTIS (TESCO Co., Yokohama) at the Fukui Prefectural Dinosaur Museum.

The three specimens were segmented in Avizo® 9.7.0 software (Visualization Sciences Group, 1995‐2019). New labelfield modules were created to segment the endocranial cavity of each specimen in order to visualize the brain endocast separately from the cranium. The pen tool was used during the segmentation process because of the low contrast between the density of the bone and the endocranial cavity that was filled with matrix. Trogosus castoridens (SDSNH 40819) was the most challenging to segment because the density of the bone and the matrix contained inside the endocranial cavity were very similar. In certain regions, the bone was not preserved, and a straight line was used to link the two nearest pieces of bone. We followed Bertrand et al. (2020), for the brain endocranial anatomical descriptions. For the cranial vascular system description, we followed Muizon et al. (2015) and Wible (2022). A surface rendering was generated for the three virtual endocasts using unconstrained smoothing. To estimate the volume of the endocast, the module ‘generate surface’ was used directly onto the labelfield module of Tr. hillsii (USNM 17157). Originally, we reconstructed the olfactory bulbs in Tr. grangeri (AMNH FM 17008) and included them in the measurement dataset of Bertrand et al. (2022). We re‐checked the CT data and were not confident in the original reconstruction; the cribriform plate surrounding the olfactory bulbs could not be identified with certainty because of limited preservation and low contrast of the CT scan. Because of this uncertainty, we decided not to report the endocranial volume for Tr. grangeri (AMNH FM 17008). The virtual brain endocast of Tr. castoridens (SDSNH 40819) appears crushed dorsoventrally, and therefore no volumetric measurements were performed on this specimen.

3.3. Palaeobiological calculations

3.3.1. Body mass estimation

We used dental measurements to determine the body mass of Tr. hillsii (USNM 17157) using data reported by Miyata (2007a) for this specimen. The body masses for Tr. grangeri (AMNH FM 17008) and Tr. castoridens (SDSNH 40819) were not estimated because the brain endocasts were too incomplete for providing an accurate brain endocranial volume. We used the following dental equations to estimate the body mass of Tr. hillsii (USNM 17157): the ‘All mammal curve’ M₁ area equation from Legendre (1989), and the ‘non selenodont taxa’ M₁ length and area equations from Damuth (1990). The equations and body mass estimates are presented in Table 1. We used the three values for our quantifications below to take into account the uncertainty of the body mass. Because the cranium of Tr. hillsii (USNM 17157) is incomplete rostrally, skull length could not be used to determine body mass. We could not use postcranial elements to determine the body mass of Tr. hillsii (USNM 17157). An almost complete humerus AMNH 17011 from the Huerfano has been attributed to the genus Trogosus, but cannot be attributed to a species (Gazin, 1953: fig. 29; Miyata, 2007b) and therefore would not represent a good candidate. Femora have been recovered for Tr. hillsii, but linear measurements for these bones have not been generated yet (Gazin, 1953: fig. 35).

TABLE 1.

Equations used to calculate the body masses of Trogosus hillsii (USNM 17157).

Formula	r 2	References	Body mass (kg)
10^(3.17 × (LOG10 (mL length)) + 1.04)	0.98	Damuth (1990)—Non selenodont taxa	66.41
10^(1.51 × (LOG10 (mL area)) + 1.44)	0.97	Damuth (1990)—Non selenodont taxa	97.38
EXP (1.7054 × (LN (mL area)) + 2.247)	0.97	Legendre (1989)—Mammals	96.27

Note: Specimen used for estimating body mass: USNM 17157 (Miyata, 2007a).

Abbreviations: BM, body mass; CL, cranial length; EXP, exponential; LN, natural logarithm; LOG10, logarithm with base 10; m1, lower molar 1.

3.3.2. Neurobiology

Linear measurements of the three virtual brain endocasts were taken and ratios of these dimensions were generated following Bertrand and Silcox (2016; Table 2). The olfactory bulb volume for Tr. hillsii (USNM 17157) was estimated using the module ‘volume edit’ tool in Avizo® 9.7.0 software (Visualization Sciences Group, 1995‐2019). For Tr. grangeri (AMNH 17008), the value is not reported here because of poor preservation that may bias an accurate measurement. We also want to avoid overestimation of the volume of these structures and of the overall endocranial volume. The volume was also not generated for Tr. castoridens (SDSNH 40819) because of poor preservation. The neocortical surface area of Tr. hillsii (USNM 17157) was quantified by virtually isolating it from the rest of the brain endocast using the pen tool in Avizo® 9.7.0 software (Visualization Sciences Group, 1995‐2019). The surface of the neocortex was not estimated in Tr. grangeri (AMNH 17008) and Tr. castoridens (SDSNH 40819) because the boundaries of the neocortex were hardly distinguishable.

TABLE 2.

Linear measurements, volumes, and surface areas for the different endocasts of Trogosus described in this study.

	Trogosus hillsii	Trogosus grangeri	Trogosus castoridens
	USNM 17157	AMNH FM 17008	SDSNH 40819
Measurements (mm)
Total endocast length (TL)	82.78	—	—
Olfactory bulb length (OL)	18.44	—	—
Olfactory bulb width (OW)	32.17	—	—
Olfactory bulb height (OH)	20.01	—	—
Cerebrum maximal length (CRML)	34.52*	38.1*	26.84*
Cerebrum maximal width (CRMW)	49.24	48.28	25.6*
Cerebrum maximal height (CRMH)	32.76	34.43	—
Cerebellum width (CLW)	47.67	43.60	25.47*
Cerebellum maximal length (CLML)	17.85*	17.11*	9.2*
Ratios linear measurements (%)
OL/TL	22.28	—	—
CRML/TL	41.70	41.65	—
CLML/TL	21.56	18.71	—
CLW/CRMW	96.81	90.31	99.49
OW/CRMW	65.33	—	—
OW/CLW	67.48	—	—
Surface areas (mm2) and volumes (mm3)
Total endocast surface area	12676.30	18170.63*	—
Neocortical surface area	1313.58	—	—
Total endocast volume	61021.20	—	—
Olfactory bulb volume	4428.85	—	—
Ratios surface areas and volumes (%)
NS/TS	10.36	—	—
OV/TV	7.26	—	—

Note: Values with the asterisk “*” are estimations.

3.3.3. Statistical analyses

We performed a series of statistical analyses on middle Eocene taxa (Bertrand et al., 2023: Tables S1–S7). The age of each specimen was based on the locality where the specimen was found or on the species age when the locality was unknown. Brain endocasts of species for which the age could not be attributed to a single epoch subdivision were categorized using the oldest subdivision (e.g., middle to late Eocene was categorized as middle Eocene). For more information about the exact age in millions of years of the specimens, see Bertrand et al. (2022: Table S19). The decision not to include early and late Eocene taxa is based on the presence of a temporal effect on brain size in mammals (Bertrand et al., 2022; Jerison, 1973), which could lead to misleading interpretations. Indeed, there is a clear increase in relative brain size from the Paleocene to the present, which occurred independently in different mammalian lineages (Bertrand et al., 2022; Jerison, 1961). We investigated the relationship between the (1) brain volume and body mass, (2) olfactory bulb volume and brain volume, (3) olfactory bulb volume and body mass, and (4) neocortical surface area and brain surface area. We elected to perform an OLS regression without taking the effect of phylogeny into account because of the uncertainty of the relationships among archaic and crown taxa, and among the three tillodonts themselves. Additionally, the sensitivity analyses performed by Bertrand et al. (2022) showed that phylogeny did not have an impact on the results. We plotted the size of the olfactory bulbs against the endocranial volume to compare the proportional differences in the olfactory bulbs and neocortical sizes. The olfactory bulbs were plotted against body mass to estimate the actual size variation of these structures.

All analyses were performed in R v3.6.2 (R Core Team, 2019) and R studio v2022.07.2 (RStudioTeam, 2022). For the OLS linear regressions, we used the function “ggplot” in the package ggplot2 (Wickham, 2016) for visualization and the function “gls” in the package nlme (Pinheiro et al., 2018) to create three regression lines for the four endocranial relationships: (1) all taxa, (2) archaic taxa, and (3) crown clades (Bertrand et al., 2023: Table S2). We generated the residuals for all taxa (1) using the function “residuals” in the package stats (Chambers & Hastie, 1992; see Bertrand et al., 2023: Table S3). Then, we produced boxplots of these residuals to compare individual crown clades and archaic taxa with Tr. hillsii using the function “ggboxplot” in the package ggplot2 (Wickham, 2016). Normality of the data was assessed using a Shapiro–Wilk test (data normally distributed: p‐value >0.05). Then, homogeneity of the variances (equality of the variance: p‐value >0.05) was verified using Levene's test (data not normally distributed) or Bartlett's test (data normally distributed). When the variances were equal, we used Fisher–Pitman permutation tests with the functions “oneway‐test” in the package coin (Hothorn et al., 2006) and “pairwisePermutationTest” in the package rcompanion (Mangiafico, 2017). Otherwise, we used the Welch test (Welch, 1951) and the functions “welch. test” and “paircomp” in the package onewaytests (Dag et al., 2018). These tests allowed us to assess whether groups had significant differences in endocranial residuals but were only performed with more than two individuals per category (Bertrand et al., 2023: Table S4). Additionally, we performed ANOVAs on the four endocranial relationships to evaluate the difference in slope and intercept between crown clades and archaic taxa (Bertrand et al., 2023: Table S5).

Finally, the data were also organized by dietary groups: herbivores and carnivores–omnivores for the ungulate and carnivoran species only based on the morphology of their teeth and previous work (Bertrand et al., 2023: Tables S1–S7). In these analyses, we also made the distinction between Eocene archaic taxa and crown clades. We analyzed the endocranial residuals (Bertrand et al., 2023: Table S6) using this classification to investigate whether there was any significant difference in the neurosensory variables between dietary guilds. We visualized these analyses using the function “ggplot” in the package ggplot2 (Wickham, 2016). Permutation tests were also conducted for this set of analyses using similar methods as described above (Bertrand et al., 2023: Table S6).

We produced encephalization quotient values and residuals from the OLS linear regression between brain volume and body mass of our middle Eocene sample (Bertrand et al., 2023: Table S2). The encephalization quotient (EQ) was first proposed by Jerison (1973; Ei/Ec) and corresponds to the ratio between the actual brain size (Ei) of a given species (i) and the brain size expected for a hypothetical ‘typical’ mammal of the same body mass (Ec; Martin, 1990). Here, we generated our own EQ equation based on the OLS regression for our sample of middle Eocene taxa (see EQ values in Bertrand et al., 2023: Table S3). We elected to not use the equations from Jerison (1973) or Eisenberg (1981) to generate EQ values because of biases due to the sample size used to generate these EQ equations.

4. DESCRIPTIONS AND COMPARISONS

The brain endocast of Tr. hillsii is complete (Figures 1, 2 and 3a,b) with a brain volume of 61021.2 mm³ (Table 2). The brain endocast of Tr. grangeri (Figures 2b,e and 3c,d) is relatively complete; however, fewer details of the anatomy are visible. Trogosus castoridens (SDSNH 40819; Figures 2c,f and 3e,f) is flattened dorsoventrally compared to the two other specimens; however, the cranium appears undeformed (see Miyata & Deméré, 2016). Possible endocranial deformation may have occurred as the bone appears very thick in the cross‐section of the CT data (Bertrand et al., 2023: fig. S1).

FIGURE 1.

Virtual endocasts inside the cranium of Trogosus hillsii (USNM 17157). (a) Dorsal, (b) ventral, and (c) lateral views. Scale = 10 mm.

FIGURE 2.

Virtual reconstructions in dorsal and ventral views of the brain endocast of (a,d) Trogosus hillsii (USNM 17157), (b, e) Trogosus grangeri (AMNH FM 17008), and (c, f) Trogosus castoridens (SDSNH 40819). Scale = 10 mm.

FIGURE 3.

Virtual reconstructions in lateral views of the brain endocast of (a, b) Trogosus hillsii (USNM 17157), (c, d) Trogosus grangeri (AMNH FM 17008), and (e, f) Trogosus castoridens (SDSNH 40819). Scale = 10 mm.

4.1. Olfactory bulbs

The olfactory bulbs are well preserved in Tr. hillsii. In dorsal view, the posterior width of the olfactory bulbs is wider than the anterior width. The bulbs are also medially separated by an antero‐posterior gap (Figure 2a). Although not complete, the gap is visible at the base of the olfactory bulbs in Tr. grangeri, which suggests a similar olfactory bulb morphology (Figure 2b). Leptictis also displays a wider posterior width for the olfactory bulbs, but no gap in between (Novacek, 1986: fig. 30). Similar to Trogosus, Al. inopinata has a gap running medially along the olfactory bulbs, but their shape is more ovoid (Figure 4d). The bulbs are ovoid and conjoined in H. lepidus (Figure 4c). In Ar. primaevus, there is a small gap between both bulbs anteriorly; however, their shape is difficult to describe because of limited preservation (Russell & Sigogneau, 1965). Still in dorsal view, the olfactory bulbs appear to be ovoid in Ch. baldwini, O. tisonensis (Figure 4a,b), Ca. coarctatus (Cameron et al., 2019: fig. 3c), and Arctocyonides arenae (Russell & Sigogneau, 1965). It is not possible to comment on the condition in Ti. fodiens, Pleuraspidotherium, and E. melanus because of lack of preservation. The olfactory bulbs are positioned posteriorly to the molars in Tr. hillsii (Figure 1c), which is similar to the condition observed in O. tisonensis (Napoli et al., 2018: fig. 1a), Ch. baldwini (Bertrand et al., 2020: fig. 3b), and Ca. coarctatus (Cameron et al., 2019: fig. 2d). The olfactory bulbs are positioned above the teeth in H. lepidus (Orliac et al., 2012: fig. 2a). This condition can only be determined while visualizing the endocast and the translucent cranium together. Therefore, it could not be determined for Ti. fodiens, Al. inopinata, Leptictis, Ar. primaevus, Arctocyonides arenae, Pleuraspidotherium, and E. melanus. The circular fissure separates the olfactory bulbs from the cerebrum and is greatly expanded in Tr. hillsii (Figures 2a and 3a) and Tr. grangeri (Figures 2b and 3c). A similar pattern is visible in the Paleocene taxa Ch. baldwini, Al. inopinata, O. tisonensis (Figure 4), Ar. primaevus, and Arctocyonides arenae (Russell & Sigogneau, 1965). In dorsal view, the circular fissure appears narrower in Leptictis (Novacek, 1986: fig. 30) and E. melanus (Thewissen & Gingerich, 1989: fig. 7). The condition is not clear in Ti. fodiens, H. lepidus, Ca. coarctatus, and Pleuraspidotherium.

FIGURE 4.

Virtual endocasts of compared archaic mammals. Chriacus baldwini (MCZ 20676; Bertrand et al., 2020) in (a) dorsal, (e) ventral, and (i) lateral views. Onychodectes tisonensis (AMNH 785; Napoli et al., 2018) in (b) dorsal, (f) ventral, and (j) lateral views. Hyopsodus lepidus (AMNH FM 143783) in (c) dorsal, (g) ventral, and (k) lateral views. Alcidedorbignya inopinata (MHNC 8372; Muizon et al., 2015) in (d) dorsal, (h) lateral, and (l) ventral views. cif, circular fissure; jf, internal jugular vein and cranial nerves IX, X, XI; nc, neocortex; ob, olfactory bulb; otc, orbitotemporal canal; II, optic nerve; pl, petrosal lobule; rf, rhinal fissure; sf₁, ophthalmic veins and cranial nerves III, IV, V₁, and VI; sf₂, ophthalmic veins and cranial nerves III, IV, V₁, V₂, and VI; sss, superior sagittal sulcus; ve, vermis; V₂, maxillary nerve; V₃, mandibular nerve; VII, facial nerve; VIII, vestibulocochlear nerve; XII, hypoglossal nerve. Scale = 10 mm.

4.2. Cerebrum and midbrain

In dorsal view, the cerebrum has an ovoid shape anteriorly and lacks the preservation of the superior sagittal sulcus defining both hemispheres antero‐medially in the well‐preserved Tr. hillsi (Figure 2a). This pattern is unlikely to be due to preservation; however, it cannot be ruled out for the two other Trogosus specimens because they are less well‐preserved (Figure 2b,c). In contrast, the cerebral hemispheres have a straight outline anteriorly in Leptictis (Novacek, 1986: fig. 30) and E. melanus (Thewissen & Gingerich, 1989: fig. 7). This feature cannot be identified for other taxa because of lack of preservation. The cerebral hemispheres are interrupted medially by the superior sagittal sulcus running between the frontal lobes of the neocortex in Al. inopinata, Ch. baldwini, H. lepidus (Figure 4), Ar. primaevus, Arctocyonides arenae (Russell & Sigogneau, 1965), Leptictis (Novacek, 1986: fig. 30), and E. melanus (Thewissen & Gingerich, 1989: fig. 7). Because of preservation, this feature is not easily identifiable in Ti. fodiens, O. tisonensis, Ca. coarctatus, and Pleuraspidotherium. Trogosus appears to have had a completely lissencephalic brain (Figure 2a–c) similar to H. lepidus (Figure 4c). Possible sulci have been identified in Ar. primaevus, Arctocyonides arenae (Russell & Sigogneau, 1965), Al. inopinata (Muizon et al., 2015: fig. 54), Leptictis (Novacek, 1986: fig. 30), E. melanus (Thewissen & Gingerich, 1989: fig. 7), and O. tisonensis (Napoli et al., 2018: fig. 2). Because of preservation, we cannot verify the presence of sulci in Ti. fodiens, Ch. baldwini, Ca. coarctatus, and Pleuraspidotherium.

The cerebrum is composed of two regions in mammals known as the neocortex and the paleocortex (= piriform lobe). The separation between both structures corresponds to the rhinal fissure (Martin, 1990). This key landmark can be used to provide information on the degree of expansion of the neocortex (Jerison, 2012; Long et al., 2015). The rhinal fissure is visible in dorsal view, in mammals with less expanded neocortices or in lateral view, in taxa with more expansive neocortices. In Euarchontoglires, there is an overlap between the rhinal fissure and the orbitotemporal canal (e.g., Bertrand & Silcox, 2016; Martin, 1990; Silcox et al., 2010). However, this relationship is not consistent across Mammalia, and in some groups, the rhinal fissure may occupy a more dorsal position compared to the orbitotemporal canal due to a lesser expansion of the neocortex. Furthermore, the ramus supraorbitalis of the stapedial artery (Wible, 1987) may not always be enclosed inside an orbitotemporal canal, as observed in the anagalid Anagale gobiensis (AMNH 26079; López‐Torres et al., 2023) where no canal could be identified. The orbitotemporal canal is preserved and visible in Tr. hillsii and it overlaps with the rhinal fissure posteriorly but then runs ventrally from the latter anteriorly. After separating from the course of the orbitotemporal canal, the rhinal fissure slopes up and points to the direction of the dorsal aspect of the olfactory bulbs (Figure 3a). Both structures are challenging to identify in Tr. castoridens and Tr. grangeri because of poor preservation of the area (Figure 3c,f). The orbitotemporal canal runs ventrally to the rhinal fissure in Al. inopinata, Ch. baldwini (Figure 4i,l), Pleuraspidotherium, Ar. primaevus, and Arctocyonides arenae (Russell & Sigogneau, 1965). The rhinal fissure and the orbitotemporal canal have been described as running together anteriorly in Leptictis (Novacek, 1982) and E. melanus (Thewissen & Gingerich, 1989). The orbitotemporal canal is not visible in H. lepidus. The relationship between these two structures cannot be reliably identified in Ti. fodiens, O. tisonensis, and Ca. coarctatus.

In the dorsal view of Tr. hillsii, the midbrain is identified as a small patch with boundaries challenging to trace and without clearly defined colliculi (Figure 2a). The lack of good preservation of this region in Tr. grangeri and Tr. castoridens does not allow us to confidently identify this structure in these two specimens. A similar condition is observed in Al. inopinata, H. lepidus (Figure 4c,d), E. melanus (Thewissen & Gingerich, 1989: fig. 7), Pleuraspidotherium, Ar. primaevus, and Arctocyonides arenae (Russell & Sigogneau, 1965) in which the midbrain is exposed and not covered by the cerebrum. In contrast to other taxa, E. melanus displays a more complex midbrain with the presence of caudal colliculi (Thewissen & Gingerich, 1989: fig. 7). The midbrain was described as not being exposed to Leptictis (Novacek, 1982, 1986). This condition cannot be identified in Ti. fodiens, Ch. baldwini, O. tisonensis, and Ca. coarctatus because of a lack of preservation.

4.3. Cerebellum

The cerebellum of Tr. hillsii is complete but detailed structures, including the boundaries between the vermis and the lobes of the cerebellum, are hard to identify (Figure 2a). The width of the cerebellum and cerebrum are similar in dorsal view for the three Trogosus specimens (Table 2; Figure 2a–c) as well as in Ti. fodiens. In contrast, the cerebellum appears narrower in Ch. baldwini, Al. inopinata, H. lepidus (Figure 4a,c,d), Leptictis (Novacek, 1986: fig. 30), E. melanus (Thewissen & Gingerich, 1989: fig. 7), and Ca. coarctatus (Cameron et al., 2019: fig. 3c). The remaining specimens are not complete enough to make comparisons (i.e., O. tisonensis, Pleuraspidotherium, Ar. primaevus, and Arctocyonides arenae).

The petrosal lobules, as defined in previous studies as regions of the cerebellum filling the subarcuate fossa (e.g., Bertrand et al., 2020; Lang et al., 2022), are very subtle in Tr. hillsii, and this is unlikely due to poor preservation of the area (Figure 2a). The two other specimens, Tr. grangeri and Tr. castoridens, display even fewer details of the anatomy of the cerebellum (Figure 2b,c). Trogosus may have a subarcuate fossa where the petrosal lobules are usually present such as in rodents (Bertrand et al., 2017), but because we could not reconstruct the semicircular canals of the inner ear, the petrosal lobules could not be correctly identified. In any case, the subarcuate fossa of Trogosus would have not been very deep and similar to Al. inopinata (Muizon et al., 2015: fig. 54a), and Pantolambda bathmodon (NMMNHS 14538; isolated petrosal), which has a very small petrosal lobule volume (0.01%; Bertrand et al., 2022). Muizon et al. (2015) also mentioned the tillodont Azygonyx (CT data of UM 68511) as not having a large subarcuate fossa. However, we could not access the CT data of this taxon to see how similar the condition was to Trogosus. The petrosal lobules are also very small in O. tisonensis (0.06%; Figure 4j) and in Ar. primaevus (0.13%; IRSNB M 2332). They are slightly larger in Ch. baldwini (0.26%; Figure 4i) and H. Lepidus (0.24%; Figure 4g). They are larger in Pleuraspidotherium aumonieri (0.53%; UCMP 61488) and Leptictis sp. (1.63%; SDSM 62369; AMNH 62369 in Bertrand et al., 2022). The quantitative measurements are from Bertrand et al. (2022). The petrosal lobules could not be identified in Ti. fodiens and were not estimated in Arctocyonides arenae, Ca. coarctatus, and E. melanus.

4.4. Cranial nerves and blood vessels

On the ventral surface of the endocast, the casts for some cranial nerves and vessels can be distinguished. The canals for the optic nerves (cranial nerve II) are positioned below the posterior aspect of the circular fissure in Tr. hillsii (Figures 2d and 3a). In the natural endocast of Ti. fodiens (Gazin, 1953: fig. 20), the optic nerve canals, and the sphenorbital fissure both hold a similar position to Tr. hillsii (Figure 5b). In contrast, only one foramen and canal are present on each side in Tr. grangeri (Figures 2e and 3d) and in Tr. castoridens (Figures 2f and 3f), which would suggest that the optic nerve and content of the sphenorbital fissure exited through a common foramen in these two taxa. All compared taxa have an optic foramen and canal separated from the sphenorbital fissure (Figure 4). The ancestral condition for eutherians and placental mammals corresponds to having the ophthalmic veins and cranial nerves III (oculomotor), IV (trochlear), V₁ (ophthalmic), V₂ (maxillary), and VI (abducens), likely exiting through the sphenorbital fissure. This is exhibited by many other placental mammalian orders (e.g., dermopterans, chiropterans, carnivorans; Novacek, 1986; O'Leary et al., 2013). This appears to be the condition for Tr. hillsii, Ti. fodiens, Leptictis (Novacek, 1986: fig. 30), E. melanus (Thewissen & Gingerich, 1989: fig. 7), H. lepidus (Figure 4g), and possibly O. tisonensis (Figure 4f). The two arctocyonids, Arctocyonides arenae and Ar. primaevus (Russell & Sigogneau, 1965), display a different configuration where two separated foramina are present: posterior to the optic canal, the foramen rotundum would have contained V₂ and the sphenorbital fissure would have been for the rest of the nerves and vessels. In Al. inopinata (Figure 4h) and Ca. coarctatus (Cameron et al., 2019: fig. 3f), posterior to the optic canal, two separate canals appear to join anteriorly in a common fossa, suggesting an anterior fusion between the foramen rotundum and sphenorbital fissure. We could not identify the condition with certainty for Ch. baldwini and Pleuraspidotherium.

FIGURE 5.

Cranium of Trogosus hillsii (USNM 17157). (a) lateral view of the cranium. (b) CT cross‐sectional view of the braincase. The line in A denotes the plane of the cross‐sectional view (Slice 1287).

Posterior to the sphenorbital fissure, the mandibular nerve (V₃) would have passed through the foramen ovale and is visible in Tr. hillsii (Figures 2d and 3a), Tr. grangeri (Figures 2d and 3c), and Tr. castoridens (Figures 2f and 3e). This would have been the case for all other compared specimens. In lateral view, the exit for V₃ (= foramen ovale) is positioned posteriorly to the cerebrum and not level with it in Trogosus, Ti. fodiens (Gazin, 1953: fig. 20), Ar. primaevus, and Arctocyonides arenae (Russell & Sigogneau, 1965). In Al. inopinata, the exit for V₃ is located on the posterior aspect of the cerebrum (Figure 4l), while in O. tisonensis, Ch. baldwini, H. lepidus (Figure 4e–g), Ca. coarctatus (Cameron et al., 2019: fig. 3f), Leptictis (Novacek, 1986: fig. 30), E. melanus (Thewissen & Gingerich, 1989: fig. 7), and Pleuraspidotherium (Orliac et al., 2012: fig. 4b), it is positioned more anteriorly on the lateral aspect of the cerebrum. The casts for cranial nerves VII (facial) and VIII (vestibulocochlear) could not be reconstructed in any of the Trogosus specimens because of lack of preservation.

The cast of the jugular foramen, which would have corresponded to the passageway of the internal jugular vein and cranial nerves IX (glossopharyngeal), X (vagus), and XI (accessory) in life, is visible in all three specimens, Tr. hillsii (Figures 2d and 3b), Tr. grangeri (Figures 2e and 3c), and Tr. castoridens (Figures 2f and 3f). The internal jugular vein cast is positioned anterior to the hypoglossal foramen for cranial nerve XII (hypoglossal). These two features have a similar position in other taxa for which they are both preserved (Figure 4).

Details of the venous drainage system are visible in all three specimens. In dorsal view, the superior sagittal sulcus that would typically run antero‐posteriorly between the cerebral hemispheres cannot be identified in any of the three specimens, Tr. hillsii (Figure 2a), Tr. grangeri (Figure 2b), and Tr. castoridens (Figure 2c). It is also not visible in Ti. fodiens but it could be due to the way the endocast was prepared. Compared to other taxa, the superior sagittal sulcus is very subtle in O. tisonensis and Ch. baldwini (Figure 4a,b), but this could be due to limited preservation. The other taxa Al. inopinata, H. lepidus (Figure 4c,d), Ar. primaevus, Arctocyonides arenae (Russell & Sigogneau, 1965), Leptictis (Novacek, 1986: fig. 30), and E. melanus (Thewissen & Gingerich, 1989: fig. 7) all display a clear superior sagittal sulcus and clear anterior separation between both cerebral hemispheres. The preservation is too limited to make any statement about Ca. coarctatus and Pleuraspidotherium.

The confluence of sinuses, which would have connected with the superior sagittal and the transverse sinuses, is partially visible in Tr. hillsii on the surface of the endocast (Figure 2a), but not preserved in the two other Trogosus specimens. This pattern is typical and present in all fossil taxa preserving this region. The transverse sinus would have abutted the midbrain medially in H. lepidus (Figure 4c), Leptictis (Novacek, 1986: fig. 30), and E. melanus (Thewissen & Gingerich, 1989: fig. 7), but the condition in other taxa is not clear because of limited preservation. Laterally, the transverse sinus would have fallen against the lobe of the cerebellum in all taxa. In Tr. hillsii (Figure 2a), it remains unclear how much of the transverse sinus would have covered the midbrain because of limited preservation of the area; however, the region where the midbrain lies between the cerebral hemispheres and the cerebellum is relatively long antero‐posteriorly.

In the posterior aspect of the endocast, a complex set of large canals surrounds the cerebellum in Tr. hillsii (Figure 6). Some but not all of these canals are also preserved in Tr. grangeri (Figures 2e and 3d) and Tr. castoridens (Figures 2f and 3e,f). Below, we describe what was likely circulating in these canals based on our observations. Comparisons with Ti. fodiens, O. tisonensis, and Ch. baldwini were not possible because of lack of preservation. Concerning the venous system, the transverse sinus divides into three vessels: the capsuloparietal emissary vein, the sigmoid sinus, and the superior petrosal sinus in eutherians (e.g., Muizon et al., 2015; Novacek, 1986; Wible, 1984). The capsuloparietal emissary vein appears to have connected to the endocranial cavity via the orbitotemporal vein contained in the orbitotemporal canal, visible on the surface of the brain endocast in Tr. hillsii (Figure 6a). This would be similar to the morphology in the pantodont Al. inopinata (Muizon et al., 2015: fig. 47), the ‘condylarths’ H. lepidus, Pleuraspidotherium, Ar. primaevus (Orliac et al., 2012: fig. 4a,b,e), Arctocyonides arenae (Russell & Sigogneau, 1965), and the palaeoryctid E. melanus (Wible, 2022: fig. 4). The capsuloparietal emissary vein would have exited the cranium through the postglenoid foramen as the postglenoid vein in Tr. hillsii. This configuration is similar to the condition in Ca. coarctatus (Cameron et al., 2019: fig. 3), Al. inopinata (Muizon et al., 2015: fig. 47), in the ‘condylarths’ H. lepidus, Pleuraspidotherium, Ar. primaevus (Orliac et al., 2012: fig. 4a,b,e), A arenae, the palaeoryctid E. melanus (Wible, 2022: fig. 4), and the leptictid Leptictis (Novacek, 1986: fig. 28). Part of the capsuloparietal emissary vein cast is preserved in Tr. grangeri (Figures 2e and 3d) and Tr. castoridens (Figures 2f and 3e). The cast for a branch of the capsuloparietal emissary vein, that we interpret as the supraglenoid vein, is also visible and would have exited the cranium through the supraglenoid foramen in Tr. hillsii (Figure 6; see terminology from Cope, 1880). The foramen and the vein cast cannot be identified in the two other Trogosus specimens. The cast for the supraglenoid vein is also visible in Ca. coarctatus (=sinus drain for the temporomandibular muscles in Cameron et al., 2019: fig. 3) but could not be identified for the remaining of the compared specimens.

FIGURE 6.

Virtual reconstruction of the brain endocast and canals for blood vessels in Trogosus hillsii (USNM 17157). (a) lateral, (b) posterior, (c) ventral, and (d) dorsal views. Scale = 10 mm.

Another division of the transverse sinus, the cast of the sigmoid sinus, is visible in Tr. hillsii (Figure 6) and in Tr. castoridens (Figure 3e). The sinus would have run anteroposteriorly on the lateral side of the lateral lobes of the cerebellum and then ventroposteriorly along the cerebellum and would have exited the cranium via the jugular foramen as the internal jugular vein (Wible, 1990). This condition is similar to the pantodont Al. inopinata (Muizon et al., 2015: fig. 58), the ‘condylarths’ H. lepidus, Pleuraspidotherium, Ar. primaevus (Orliac et al., 2012: fig. 4a,b,e), Arctocyonides arenae (Russell & Sigogneau, 1965), the palaeoryctid E. melanus (Wible, 2022: fig. 4), and the leptictid Leptictis (Novacek, 1986: fig. 28). The sigmoid sinus would have also connected to the condyloid vein in Tr. hillsii (Figure 6), Tr. grangeri (Figures 2e and 3d), and Tr. castoridens (Figures 2f and 3f). The content of the condyloid canal would have exited through the foramen magnum and would have also been connected to the hypoglossal foramen in all three specimens and in the compared taxa that preserve this region. The cast for the sigmoid sinus is not preserved in Tr. grangeri. There is no evidence of the superior and inferior petrosal sinuses in any of the Trogosus specimens.

The cast of the occipital emissary vein can be traced in Tr. hillsii (Figure 6a) and would have exited the endocranial cavity through the mastoid foramen as in the dog (Evans & de Lahunta, 2012; Wible, 2008). This condition would have been similar to Al. inopinata for which a mastoid foramen has been identified (Muizon et al., 2015: fig. 53). Finally, the cast of the supraoccipital emissary vein can also be traced in Tr. hillsii (Figure 6a) and would have exited through a foramen that we identify as the supraoccipital foramen. The vein is not preserved in Tr. grangeri but may have been present in Tr. castoridens; however, it could not be reconstructed with confidence in this specimen. In Ca. coarctatus, the vena diploëtica magna, content of the posttemporal canal (= occipital emissary vein in Cameron et al., 2019: fig. 3d), would have exited through the posttemporal foramen, as in Al. inopinata (Muizon et al., 2015: fig. 53). There is no evidence of a posttemporal canal in Trogosus. In Leptictis, the cast of the occipital emissary vein also appears present (mastoid vein in Novacek, 1986) and would have exited posteriorly through the mastoid foramen located in the suture between the supraoccipital, parietal, and petromastoid (Novacek, 1986: fig. 29). The mastoid foramen could not be identified in E. melanus (Wible, 2022). The cast of the occipital emissary vein and supraoccipital emissary vein cannot be identified in the other compared specimens.

The common carotid artery divides into two branches: the external and internal carotid arteries. The internal carotid artery cast is visible on the surface of the endocast of Tr. hillsii, and would have entered the endocranial cavity anterior to the promontorium (Figure 2d). The stapedial artery, a branch of the internal carotid artery (Figure 6), would have entered laterally to the promontorium and given rise to the ramus superior, which is similar to the situation in E. melanus (Wible, 2022: fig. 5). There is no evidence of a ramus inferior in Trogosus. The ramus superior would have divided into the orbitotemporal artery (=ramus supraorbitalis of Wible, 2022) anteriorly and into a multitude of rami temporales dorsally that can be observed in Tr. hillsii (Figure 6) and in the other specimens of Trogosus (Figure 3d,f). This pattern is similar to that in E. melanus (Wible, 2022: figs. 5 and 6). The situation is similar in Al. inopinata but the cast of the internal carotid artery is not visible on the surface of the brain endocast (Muizon et al., 2015: fig. 57). The majority of the rami temporales would have exited through foramina in the parietal in Tr. hillsii and in E. melanus (Wible, 2022: fig. 6), while in Al. inopinata, they would have exited through the foramina in the squamosal/parietal suture (Muizon et al., 2015: fig. 47). The overall pattern is the same in Leptictis (Novacek, 1986: fig. 27), but the rami temporales (sinus canals in Novacek, 1986: fig. 17) would have all exited through the squamosal. Rami temporales are also present in H. lepidus but are less numerous.

4.5. Relative brain size

Regression lines of the relationship between brain volume and body mass for archaic taxa and crown clades are significantly different (p‐value = 0.001; Bertrand et al., 2023: Tables S2 and S5). The residual values for Tr. hillsii estimated using the Legendre M₁ area (Le) and Damuth M₁ area (DA) equations are slightly below the regression line for the middle Eocene archaic taxa, well below the one for middle Eocene crown clades, and for all taxa (Figure 7a). The same is true for the residuals generated from the Damuth M₁ length (DL) equation except that it is slightly higher than the regression line for archaic taxa (Figure 7a). All three residual values for Tr. hillsii are close to the early perissodactyl Hyrachyus modestus and the mesonychid Mesonyx obtusidens (Figure 7a). In the boxplot of the residuals for the relationship between brain volume and body mass, the three values of Tr. hillsii (Le, DA, and DL) make this taxon overlap with Hyopsodus, the plesiadapiform M. annectens, dinoceratans, and be in the low range of variation for creodonts and carnivoramorphans (Figure 7b; Bertrand et al., 2023: Table S3). Compared to crown clades, the residual value range of Tr. hillsi is lower than all middle Eocene crown clades except the primate Notharctus tenebrosus with which it overlaps (Figure 7c; Bertrand et al., 2023: Table S3). No groups are significantly different in terms of relative brain size for the middle Eocene crown and archaic groups (Figure 7c; Bertrand et al., 2023: Table S4).

FIGURE 7.

Relative size of the brain of middle Eocene mammals. (a) Linear regression of log10 (endocranial volume area vs. body mass) for archaic groups and crown clades, (b) Boxplot of the residuals from the equation in (a) for Trogosus and other archaic groups, (c) Boxplot of the residuals from the equation in (a). for Trogosus and crown clades. See Bertrand et al. (2023: Table S2) for residual values. Volumetric measurements are in cubic centimeters and body mass was measured in grams. BM, body mass; BV, brain volume.

4.6. Olfactory bulb size

The olfactory bulb volume percentage for Tr. hillsii is 7.3% (Table 2; Bertrand et al., 2023: Tables S1–S7) and is very close to the perissodactyl Hyrachyus modestus (7.9%). Compared to other middle Eocene taxa, Trogosus has a higher percentage ratio for the olfactory bulbs compared to all Euarchontoglires (1.0%–6.0%), Artiodactyla (4.5%–7.0%), the palaeanodont Metacheiromys marshi (6.4%), and the dinoceratan Tetheopsis ingens (3%). Trogosus is on the upper range of variation for Pan‐Carnivora (3.8%–8.6%). Finally, Tr. hillsii has a lower value compared to the ‘condylarth’ Hyopsodus (8.3%–8.6%) and the dinoceratan Uintatherium anceps (13.7%).

Regression lines of the relationship between olfactory bulb volume and brain volume for archaic taxa and crown clades are significantly different (p‐value = 0.003; Bertrand et al., 2023: Tables S2 and S5). Trogosus hillsii is above the regression lines for middle Eocene archaic taxa and all taxa, but below the one for crown clades (Figure 8a). The residuals from this relationship show that Tr. hillsii is very close to the value for the creodont Cynohaenodon cayluxi and it overlaps with Carnivoramorpha (Figure 8b; Bertrand et al., 2023: Table S3). Compared to other archaic taxa, Tr. hillsii has a higher residual value compared to Microsyops annectens, Sinopa lania, and Tetheopsis ingens, but lower compared to U. anceps, Hyopsodus, Leptictis, and M. marshi (Figure 8b). Compared to middle Eocene crown clades, Tr. hillsii has a higher value compared to all taxa except Hyrachyus modestus and Cebochoerus (Figure 8c). Trogosus has a slightly higher residual value compared to the rodent Pseudotomus oweni (Figure 8c; Bertrand et al., 2023: Table S3). Euprimates have significantly lower residuals compared with rodents (p‐value = 0.015; Bertrand et al., 2023: Table S4) and artiodactyls (p‐value = 0.046; Bertrand et al., 2023: Table S4).

FIGURE 8.

Relative and absolute size of the olfactory bulbs of middle Eocene mammals. (a) Linear regression of log10 (Olfactory bulb volume area vs. Endocranial volume) for archaic groups and crown clades, (b) Boxplot of the residuals from the equation in (a) for Trogosus and other archaic groups, (c) Boxplot of the residuals from the equation in (a) for Trogosus and crown clades. (d) Linear regression of log10 (olfactory bulb volume area vs. body mass) for archaic groups and crown clades, (e) Boxplot of the residuals from the equation in (d) for Trogosus and other archaic groups, (f) Boxplot of the residuals from the equation in (d) for Trogosus and crown clades. See Bertrand et al. (2023: Table S2) for residual values. Volumetric measurements are in cubic millimeters and body mass was measured in milligrams. BM, body mass; BV, brain volume; OBV, olfactory bulb volume.

Regression lines of the relationship between olfactory bulb volume and body mass for archaic taxa and crown clades are not significantly different (p‐value = 0.259; Bertrand et al., 2023: Tables S2 and S5). The residual values for Tr. hillsii estimated using the Legendre M₁ area and Damuth M₁ area equations are slightly above the regression line for archaic taxa and on the line for all taxa (Figure 8d). The residual value for Tr. hillsii estimated using the Damuth M₁ Length equations is above both archaic and all taxa regression lines. All residuals for Tr. hillsii are below the line for crown clades (Figure 8d; Bertrand et al., 2023: Table S3). Compared to archaic groups, the range of Tr. hillsii values overlaps with Hyopsodus, dinoceratans, creodonts, and the lower range of carnivoramorphans. Trogosus hillsii range of values is higher compared to M. annectens, but lower than the palaeanodont M. marshi (Figure 8e; Bertrand et al., 2023; Table S3). Lastly, compared to crown clades, Tr. hillsii overlaps with the lower range of artiodactyls and mid to higher range of rodents. It has a lower value compared to Hyrachyus modestus, but higher than euprimates (Figure 8f; Bertrand et al., 2023: Table S3). Trogosus hillsii has significantly higher residuals than euprimates (p‐value = 0.045; Bertrand et al., 2023: Table S4). Euprimates have significantly lower residuals compared to rodents (p‐value = 0.027; Bertrand et al., 2023: Table S4) and artiodactyls (p‐value = 0.045; Bertrand et al., 2023: Table S4).

4.7. Neocortical size

Trogosus hillsii has a neocortical surface area ratio of 11.46% (Table 2), which is similar to the palaeanodont M. marshi (11.47%). Trogosus hillsii has a smaller neocortex compared to middle Eocene Euarchontoglires (16.3%–43.8%), Pan‐Carnivora (16%–25.5%), the perissodactyl Hyrachyus modestus (20%), and Hyopsodus paulus (13.5%). In terms of the relationship between neocortical surface area and brain surface area, Tr. hillsii is below the regression lines for middle Eocene crown clades and archaic taxa (Figure 9a). Trogosus hillsii has a higher residual value compared to the ‘condylarth’ Hyopsodus paulus, the carnivoramorph Viverravus minutus, and the palaeanodont M. marshi, but lower than the plesiadapiform M. annectens, the creodont S. lania, other carnivoramorphans, and all crown clades (Figure 9b,c; Bertrand et al., 2023: Table S3). Rodents and euprimates are significantly different in terms of the relative size of the neocortex (p‐value = 0.017; Bertrand et al., 2023: Table S4).

FIGURE 9.

Relative size of the neocortex of middle Eocene mammals. (a) Linear regression of log10 (neocortical surface area vs. endocranial surface area) for archaic groups and crown clades, (b) Boxplot of the residuals from the equation in (a) for Trogosus and other archaic groups, (c) Boxplot of the residuals from the equation in (a) for Trogosus and crown clades. See Bertrand et al. (2023: Table S2) for residual values. Surface area measurements are in square millimeters. BS, brain surface; NS, neocortical surface.

4.8. Endocranial variables and dietary guilds

Herbivorous crown clades have significantly relatively larger brains compared to herbivorous (p‐value = 0.003; Bertrand et al., 2023: Table S7) and omnivorous–carnivorous archaic groups (p‐value = 0.038; Bertrand et al., 2023: Table S7). Omnivorous–carnivorous archaic taxa have higher residual average compared to herbivorous archaic groups, but the difference is not significant (p‐value = 0.072; Figure 10a; Bertrand et al., 2023: Tables S6 and S7). Herbivorous crown clades have significantly more expanded neocortices compared to herbivorous archaic groups (p‐value = 0.015; Bertrand et al., 2023: Table S7) but are in the range of omnivorous–carnivorous archaic groups (p‐value = 0.443; Bertrand et al., 2023: Table S7). Omnivorous–carnivorous archaic taxa have significantly larger neocortices compared to herbivorous archaic groups (p‐value = 0.005; Figure 10b; Bertrand et al., 2023: Table S7). Herbivorous crown clades are in the range of omnivorous–carnivorous archaic groups for the residuals of the olfactory bulb volume against brain volume (p‐value = 0.943; Bertrand et al., 2023: Table S7). The majority of herbivorous archaic groups overlap with the upper range of herbivore crown clades (p‐value = 0.391) and omnivorous–carnivorous archaic groups (p‐value = 0.285; Figure 10c; Bertrand et al., 2023: Table S7). When comparing olfactory bulb volume to body mass, herbivorous crown clades overlap with the upper range of both herbivorous (p‐value = 0.528) and omnivorous–carnivorous archaic taxa (p‐value = 0.213; Figure 10d; Bertrand et al., 2023: Table S7). The majority of archaic herbivores is in the mid‐range of omnivorous–carnivorous archaic groups for the olfactory bulbs against body mass plot (p‐value = 0.845; Figure 10d; Bertrand et al., 2023: Table S7). The two extreme datapoints in Figure 10c,d are dinoceratans and do not overlap with any other taxa.

FIGURE 10.

Relative size of the brain, neocortex, olfactory bulbs, and absolute size of the olfactory bulbs organized by dietary guilds and grade. (a) Boxplot of the residuals log10 (endocranial volume area vs. body mass), (b) Boxplot of the residuals log10 (neocortical surface area vs. endocranial surface area), (c) Boxplot of the residuals log10 (olfactory bulb volume area vs. endocranial volume) (d) Boxplot of the residuals log10 (olfactory bulb volume area vs. body mass). See Bertrand et al. (2023: Table S3) for residual values. BM, body mass; BS, brain surface; BV, brain volume; H, herbivorous mammals; NS, neocortical surface; OBV, olfactory bulb volume; O‐C, omnivorous–carnivorous mammals.

5. DISCUSSION

5.1. Endocranial anatomy and phylogenetic relationships

The phylogenetic relationships within Trogosinae remain contentious in part because some taxa are poorly preserved, but also because of a relatively specialized and conserved dentition within the group (e.g., Chow et al., 1996; Miyata, 2007b; Miyata & Deméré, 2016). The monophyly of Trogosinae is not disputed (Rose, 2006), but because of the rarity of well‐preserved material, questions remain about the relationships of the seven trogosine species (Miyata & Deméré, 2016). In the most recent phylogenetic analysis, Tr. hyracoides and Tr. castoridens from Wyoming are sister taxa, while Tr. hillsii and Tr. grangeri from Colorado form a clade. However, because both members of each sister taxon pair co‐occur in the same sedimentary basin, some authors have argued that the main morphological difference (i.e., rostrum length) could be attributed to sexual dimorphism (Robinson, 1966). Based on the information obtained from the brain virtual endocasts, we only note one major difference between the species. Gazin (1953) observed only one foramen for both the optic nerve and the content of the sphenorbital fissure in Tr. grangeri and Tr. hillsii. Miyata and Deméré (2016) noted a similar pattern in Tr. castoridens. However, after analyzing the CT data, Tr. hillsii appears to have two distinct foramina (i.e., optic foramen and sphenorbital fissure) as previously observed for Tillodon fodiens (Gazin, 1953). Therefore, the fusion of both the sphenorbital fissure and the optic foramen could represent a synapomorphy for Tr. grangeri and Tr. castoridens. However, it is possible that because of limited preservation of this area, the optic foramen and canal are not preserved in Tr. grangeri and Tr. castoridens. The condition could not be determined in the CT data of Tr. hyracoides because of the lack of preservation in this area. Trogosus grangeri and Tr. hillsii have overall endocranial shapes that are relatively similar but because of the limited preservation of Tr. castoridens, we cannot make any firm conclusion about the usefulness of this observation.

Placing Tillodontia on the higher level family tree of mammals has been challenging, because of a lack of clear synapomorphies uniting the members of this clade with other mammals (Rose, 2006). Tillodonts might have been related to other archaic groups such as Arctocyonidae (Gazin, 1953; Rose, 1972; Van Valen, 1963), but could also potentially be part of Cimolesta (Rose, 2006). However, material from early tillodonts discovered in Asia and their resemblance to Pantodonta led some paleontologists to consider Pantodonta as a more probable sister clade to Tillodontia (Chow & Wang, 1979; Wang & Jin, 2004). Recently, Muizon et al. (2015) produced a phylogenetic analysis that recovered them as sister clades. In this analysis, arctocyonids are also relatively closely related to the sister group of Pantodonta and Tillodontia.

Trogosus displays morphological features that could be interpreted as ancestral for mammals based on what is known of derived conditions present in crown clades. The posterior position of the olfactory bulbs behind the upper molars in Trogosus, Taeniodonta, and the ‘condylarths’ Arctocyonidae and Periptychidae, could represent a plesiomorphic state resulting from the limited development of the brain in these archaic placental mammals. In Eocene crown clades such as Rodentia and Euprimates (Bertrand et al., 2019a; Harrington et al., 2016), the olfactory bulbs are above the tooth row. The circular fissure is also relatively expanded in Trogosus, Pantodonta, Taeniodonta, and Arctocyonidae, which could also be considered ancestral. The Eocene crown clades within Euarchontoglires and Artiodactyla have a narrower circular fissure due to the expansion of the frontal lobes of the neocortex onto the circular fissure (Bertrand et al., 2019a; Harrington et al., 2016; Orliac & Gilissen, 2012). The absence of sulci on the surface of the neocortex may also represent an ancestral feature related to the limited expansion of the neocortex in Trogosus as in Hyopsodus, which also displays a lissencephalic brain. Sulci might be present in some archaic taxa (e.g., some plesiadapiforms), but they become more common in crown clades such as Rodentia, Euprimates, and Artiodactyla (Bertrand et al., 2019a; Harrington et al., 2016; Orliac et al., 2023). Also related to the development of the neocortex, the more ventral position of the orbitotemporal canal in relation to the rhinal fissure represents an ancestral feature displayed by Trogosus, Pantodonta, and Arctocyonidae. The derived condition is present in all crown and archaic Euarchontoglires that preserve both structures and display a complete alignment of the orbitotemporal canal and the rhinal fissure (Bertrand et al., 2018, 2019b; Harrington et al., 2016; Silcox et al., 2010). The covering of the midbrain has been extensively studied in many mammalian groups (Edinger, 1964). Trogosus displays an exposed midbrain, which represents the ancestral condition for mammals. Whether the midbrain is exposed, greatly depends on the development of the neocortex (Dozo et al., 2023), which occurs independently in many different clades (Bertrand et al., 2022).

In terms of possible shared derived endocranial features, Trogosus, Tillodon, and arctocyonids share a common position for the exit of V₃, which is located posteriorly to the cerebrum, and in Alcidedorbignya it is on the posterior aspect of the cerebrum compared to other taxa where it is more anteriorly positioned. This feature would not appear to be related to the development of the neocortex. Another characteristic corresponds to the low development of the subarcuate fossa in Trogosus, Alcidedorbignya, and possibly in the tillodont Azygonyx (CT data of UM 68511, see Muizon et al., 2015). Pantolambda has identifiable petrosal lobules, but they are smaller than in arctocyonids and taeniodonts (Bertrand et al., 2022). As Muizon et al. (2015) suggested for Pantodonta, relatively reduced subarcuate fossae are probably a derived condition, because they are well developed in Cretaceous taxa such as in early eutherians (i.e., Kennalestes, Zalambdalestes; Kielan‐Jaworowska, 1984), and in early Laurasiatheria (i.e., Acmeodon; Bertrand et al., 2022).

In summary, the endocranial data gathered here show that Trogosus displays many ancestral features that are present in other archaic mammals: exposed midbrain, small neocortex, orbitotemporal canal ventral to rhinal fissure, and a broad circular fissure. However, we could identify some potential derived characteristics. Within Trogosinae, we observe variation in the number of foramina and canals for the optic nerve and the content of the sphenorbital fissure. Additionally, the position of the exit of cranial nerve V₃ exit (foramen ovale) and the degree of development of the subarcuate fossa could be characters to add into future phylogenetic analyses aimed at disentangling the intra‐ and interspecific relationships of Tillodontia.

5.2. Endocranial regions and ecological inferences

Trogosus has a very distinctive dentition, including gliriform ever‐growing second incisors with restricted enamel similar to rodents (Coombs, 1983). In Esthonyx, an early member of the group, the teeth are heavily worn down which would indicate the consumption of abrasive elements such as soil and grit ingested with vegetation pulled out of the ground (Gingerich & Gunnell, 1979). The curved phalanges displayed by tillodonts, including Trogosus, are consistent with digging adaptations (Gazin, 1953; Rose, 1990), suggesting that they might have been using their forelimbs to unearth roots and tubers from underground. Trogosus also bears resemblance with clawed herbivores that put more weight on their hindlimbs rather than their forelimbs (Coombs, 1983) by having a relatively shorter tibia compared to the femur (Rose, 2006).

Based on the size of the olfactory bulbs and neocortex, Trogosus probably relied more on its sense of smell compared to other senses such as vision. Eocene mammals such as early squirrels and euprimates that rely more heavily on vision generally show a posterior expansion of the neocortex (covering of the midbrain) where the occipital cortex is located compared to less derived euarchontoglirans such as plesiadapiforms and ischyromyid rodents (Bertrand et al., 2017, 2019a; Silcox et al., 2010). The petrosal lobules are also not greatly developed in Trogosus, and these structures are known to play a role in maintaining the eye position during movement (Rambold et al., 2002). In contrast, they are greatly expanded in early squirrels, which may have relied on vision while navigating among the complex 3D environment of the trees (Bertrand et al., 2021). A small neocortex has been associated with fossoriality in mammals and more specifically in aplodontiid rodents (Bertrand et al., 2018, 2021) and in the anagalid An. gobiensis (López‐Torres et al., 2023). In the case of Trogosus, because of the lack of data for earlier trogosines, a possible decrease in neocortical surface area related to ecological specialization cannot be tested. Alternatively, as a small neocortex is also the ancestral condition for mammals, the pattern seen in Trogosus may simply reflect the ancestral state. Therefore, teasing apart the effect of phylogeny and ecology will require additional data. Nevertheless, endocranial data do not contradict the idea that Trogosus was a ground dweller and engaged in digging behavior to unearth roots and tubers. It is challenging to compare the brain of Trogosus with modern mammals displaying the same type of feeding behavior, such as Suidae, because they evolved from early artiodactyls, which had already developed bigger neocortices during the Eocene, probably before any ecological specializations for ground feeding (Orliac & Gilissen, 2012). Overall, this suggests that a complex neocortex was not necessary for the survival of Trogosus in its particular ecological niche.

5.3. Extinction of Tillodontia and other archaic placental mammals

The extinction of Tillodontia in the Eocene has long been shrouded in mystery. During the Eocene, a time of significant environmental fluctuations, the genus Esthonyx shows very limited changes in its lower dental morphology compared to earlier tillodonts, suggesting that this genus and Tillodontia in general may have been more vulnerable to extinction if they were not adapting their dentitions (Luongo et al., 2019). However, this pattern could be interpreted as a morphological specialization which would have potentially been a selective advantage for this genus to allow its survival for more than 7 million years from the late Paleocene to middle Eocene (time range from Lucas & Schoch, 1998). Possible competition for similar resources, both with other archaic placentals and crown species, could have exacerbated their extinction. The early Eocene taeniodont Ectoganus copei appears to have had similar morphological adaptations for digging and hard‐object feeding to tillodonts (Coombs, 1983), and the two groups may have overlapped in terms of their ecological niches, putting them in competition for similar resources (Lucas & Schoch, 1998). Tillodontia may also have been in competition with the crown clade Artiodactyla and more specifically with taxa like Achaenodon, which displayed similar morphological adaptations (Lucas & Schoch, 1998). This genus was one of the first large artiodactyls that weighed more than 200 kg (Foss, 2001, 2007). No endocranial data have been produced for this taxon but a relatively complete cranium is known (YPM VPPU 010033) and could be CT scanned in the future. Based on published brain virtual endocasts, the oldest artiodactyl, the early Eocene Diacodexis (Orliac & Gilissen, 2012), already had a relatively expanded neocortex that was double the neocortical size of Trogosus. Relatively larger brains have been associated with greater behavioral flexibility and may represent a crucial advantage for adapting to new and/or changing environments (Ratcliffe et al., 2006; Sol, 2009; Sol et al., 2008; van Woerden et al., 2012). In turn, species with relatively larger brains tend to also have relatively expanded neocortices (Kaas, 2006). Therefore, a bigger neocortex could have provided early artiodactyls with more computational power to access the same resources and adapt more rapidly in response to environmental change.

The extinction of tillodonts may have also been related to predator–prey interaction. Previous studies have hypothesized a possible arms race between mammalian predators and prey in the Paleogene, with an increase in relative brain size from the Paleocene to the Eocene occurring in parallel in these two guilds (Bertrand et al., 2022; Jerison, 1970, 1973). This hypothesis is supported here by our finding that herbivorous middle Eocene crown clades not only had a relatively larger brain but also more expanded neocortex compared to herbivorous archaic groups. Furthermore, omnivorous/carnivorous archaic taxa have much larger neocortices compared to herbivorous archaic groups. In light of these findings, it is remarkable that tillodonts retained their relatively small brains and neocortices—inherited from the first placental mammals that radiated after the end‐Cretaceous extinction—so deep into the Eocene, when so many other mammals were changing their neurobiology. Therefore, it is possible that tillodonts did not develop the cognitive tools to keep up with the evolution of ever larger‐brained and more intelligent predators over evolutionary time, and ever‐larger‐brained competitor herbivorous species too. Ultimately, the retention of their primitively small brains and neocortices could have been a factor in the extinction of Tillodontia during the middle Eocene, when competing against artiodactyls for similar resources and avoiding/escaping predators.

6. CONCLUSION

The study of the brain endocranial anatomy of Trogosus reveals very few anatomical differences among the three species for which we have data. The only potential synapomorphy of a subset of species corresponds to the fusion of both the optic canal and sphenorbital fissure in Tr. grangeri and Tr. castoridens. Both Tr. grangeri and Tr. hillsii, which are the best‐preserved specimens, are very similar in overall endocranial shape, which reinforces the results from previous work that has struggled to resolve the phylogenetic relationships within Trogosus (Chow et al., 1996; Miyata & Deméré, 2016). Trogosus displays many ancestral features also present in other archaic taxa: exposed midbrain, small neocortex, orbitotemporal canal ventral to rhinal fissure, and a broad circular fissure. We note two potential characteristics that might help unite Trogosus with Pantodonta and Arctocyonidae: the position of the exit of cranial nerve V₃ exit and the low degree of development of the subarcuate fossa. The relatively large olfactory bulbs and small neocortex of Trogosus are consistent with a terrestrial lifestyle and a diet of roots and tubers. The reduced size of the neocortex shows that Trogosus did not require an expanded neocortex to survive. However, in addition to dental morphological specializations, a relatively small brain, particularly neocortex, could have been one of the reasons for the extinction of tillodonts during the middle Eocene. Indeed, competition with contemporary artiodactyls displaying more developed neocortices and increased predation pressure from archaic carnivorans with more developed neocortices could have exacerbated the extinction of tillodonts.

AUTHOR CONTRIBUTIONS

Ornella C. Bertrand, Marina Jiménez Lao, and Steven L. Brusatte conceived and designed the study. Ornella C. Bertrand, Jin Meng (AMNH specimen), and John R. Wible (USNM specimen) acquired the CT data. Ornella C. Bertrand and Marina Jiménez Lao did all segmentations and drafted the manuscript, tables, and figures. Analyses and interpretations were performed by Ornella C. Bertrand, and Marina Jiménez Lao and critically reviewed by Sarah L. Shelley, Stephen L. Brusatte, John R. Wible, Thomas E. Williamson, and Jin Meng. All authors revised the manuscript and provided final approval before submission.

Supporting information

Figure S1.

Tables S1–S7.

Bertrand, O.C. , Jiménez Lao, M. , Shelley, S.L. , Wible, J.R. , Williamson, T.E. , Meng, J. et al. (2024) The virtual brain endocast of Trogosus (Mammalia, Tillodontia) and its relevance in understanding the extinction of archaic placental mammals. Journal of Anatomy, 244, 1–21. Available from: 10.1111/joa.13951

DATA AVAILABILITY STATEMENT

The code to reproduce the analyses is on the Github Repository: https://github.com/Bertrand‐Ornella/Trogosus‐Brain‐evolution‐Archaic‐Mammals, and the brain virtual endocasts for the different specimens of Trogosus are accessible in MorphoSource: https://www.morphosource.org/projects/000519068?locale=en.