Introduction

“The brain is the most important single mammalian organ, and knowledge of any extinct group must be considered very inadequate until it includes some data on the endocranial anatomy.” (Simpson 1933a:1).

As the quotation above from George Gaylord Simpson suggests, knowledge of the soft tissue anatomy of the central nervous system (CNS) is essential for a complete picture of the anatomy and sensory capabilities of extinct mammals, which in fossil mammals can be provided by cranial endocasts, or three-dimensional representations of the space within the endocranial cavity. Indeed, cranial endocasts reveal substantial information about the external anatomy of the brains of mammals and are useful for estimating brain volume (e.g., Jerison 1973; Radinsky 1977, 1978; Kielan-Jaworowska 1983, 1984, 1986; Rowe et al. 2011). Furthermore, cranial endocasts provide data on internal cranial anatomy that are useful for uncovering phylogenetic relationships among fossil and extant taxa (e.g., Kielan-Jaworowska 1997; Macrini et al. 2007b).

Notoungulata, an enigmatic group of extinct South American placental mammals (Simpson 1948, 1980; Patterson and Pascual 1968; Rose 2006; Croft et al. 2020), has been the subject of a number of endocranial studies based on artificial (plaster) and natural cranial endocasts (Simpson 1932, 1933a, b; Patterson 1937; Stirton 1953; Dechaseaux 1958, 1962; Radinsky 1981; Dozo 1997), and more recently of digital endocasts (Macrini et al. 2010, 2013; Dozo and Martínez 2016; Martínez et al. 2019). Many of these works provided descriptive anatomical and volumetric data on notoungulate endocasts that allowed for subsequent comparative studies on relative brain size (e.g., Jerison 1973; Radinsky 1981). Both Jerison (1973) and Radinsky (1981) determined endocranial volumes for endocasts described by previous workers to calculate body mass corrected Encephalization Quotients (EQs) and examine relative brain sizes in notoungulates. Radinsky (1981), like many previous workers (e.g., Simpson 1932, 1933a, b; Patterson 1937; Dechaseaux 1958, 1962), also compared anatomical differences between notoungulate cranial endocasts to study brain evolution in the group. However, none of those anatomical comparisons analyzed data using a phylogenetic framework, nor did they attempt to organize anatomical differences between different notoungulate taxa into characters for phylogenetic analyses.

Our study combines data from previously published anatomical descriptions, natural endocasts, and previously extracted plaster endocasts, with new data from high-resolution X-ray micro-computed tomography (µCT), to undertake the first broad comparative study of cranial endocasts of notoungulates, shedding much needed light on some aspects of the phylogenetic interrelationships of the group as well as the placement of Notoungulata within Placentalia. While molecular evidence supports a relationship with Perissodactyla and other “euungulates” (Buckley 2015; Welker et al. 2015; Westbury et al. 2017) or to Paenungulata within Afrotheria in a combined molecular and morphological analysis (O’Leary et al. 2013), phylogenetic analyses of the relationship of notoungulates to other placental mammals based solely on morphological data have yielded conflicting and generally unresolved results (Cifelli 1993; Horovitz 2004; Billet 2010; Agnolin and Chimento 2011; Billet and Martin 2011). Likewise, analyses of the interrelationships of taxa within the group have not provided a generally well-supported phylogenetic tree, with many relationships still poorly resolved (e.g., Cifelli 1993; Madden 1997; Shockey 1997; Cerdeño and Bond 1998; Nasif et al. 2000; Croft et al. 2004; Flynn et al. 2005; Croft and Anaya 2006; Hitz et al. 2006; Billet et al. 2009; Billet 2010, 2011; Shockey et al. 2012). Morphological phylogenetic analyses have relied primarily on dental, external cranial, and some postcranial characters, and none of these analyses has sampled extensively from the interior of skulls, but this also is the case for morphological phylogenies of most mammalian groups.

We generate a new phylogenetic character matrix of internal cranial anatomy features based on our current study and prior analyses by others (e.g., Gabbert 1997; Macrini et al. 2010, 2013), to more completely sample notoungulate anatomy and hopefully help resolve some of the longstanding phylogenetic uncertainties for the group. The specific aims of our study are as follows: 1) describe new cranial endocast material of notoungulates based on high resolution X-ray µCT imagery of skulls; 2) synthesize new endocast data with published data on notoungulate endocranial cavities from plaster and natural endocasts; 3) score the first extensive matrix of characters based on notoungulate endocranial cavities, and combine these data with a more extensive character matrix sampling from the entire skeleton and dentition, to conduct the most comprehensive phylogenetic analysis to date of notoungulate interrelationships; and 4) to examine relative brain size evolution in notoungulates.

Institutional abbreviations- AMNH, American Museum of Natural History, Vertebrate Paleontology Fossil Mammal Collections, New York, NY, USA; FMNH, Field Museum of Natural History, Chicago, IL, USA; MNBA, Museo Nacional de Historia Natural, Buenos Aires, Argentina; SGOPV, vertebrate paleontology collections, Museo Nacional de Historia Natural, Santiago, Chile; UCMP, University of California Museum of Paleontology, Berkeley, CA, USA.

Materials and Methods

Specimens

One natural and ten plaster brain endocasts of notoungulates were analyzed from collections at the AMNH and FMNH. In addition, two new digital endocasts were extracted from high resolution X-ray µCT imaging of notoungulate skulls. Anatomical descriptions of the new digital endocasts are provided below.

The litoptern Tetramerorhinus lucarius (formerly Proterotherium cavum; Simpson 1933b) and the astrapotheres Trigonostylops wortmani and Astrapotherium magnum were scored based on plaster endocasts and used as outgroups in the analyses. The “condylarths” Hyopsodus lepidus and Phenacodus primaevus also were used as outgroups and scored based on the descriptions of their endocasts by Orliac et al. (2012) and Simpson (1933a), respectively. A complete list of the specimens examined is provided in Table 1.

Table 1 Cranial endocasts examined in this study (*) and endocranial character data sources
Full size table

A skull of Notostylops murinus (FMNH P13319) from the Gran Barranca, Argentina (Simpson 1948), from a fauna corresponding to the Casamayoran SALMA (Cifelli 1985), was imaged using high-resolution x-ray CT, and a digital endocast was extracted (see below). This skull represents an advanced growth-stage juvenile individual with milk teeth still in place (dP2-4, M1-2; Riggs and Patterson 1935; Simpson 1948). We used this endocast to score brain endocast phylogenetic characters along with an incomplete natural endocast of an adolescent individual of Notostylops pendens (AMNH 28614; Simpson 1933a, 1948) and a description of an incomplete endocast of Notostylops ampullaceus by Simpson (1932), an apparent adult specimen (Simpson 1948). We note that the endocast characters used in our study vary little with ontogeny, as they do not deal with shape or allometric relationships (see Macrini et al. 2007c). The inner ear of the juvenile specimen of N. murinus (FMNH P13319), previously described by Macrini et al. (2010), was used to score inner ear characters for this taxon in our matrix. Ekdale (2010) demonstrated that adult sizes of several inner ear characters, such as the arcs of semicircular canals, are achieved before the inner ear is fully functional in marsupials (Monodelphis domestica). As there is overlap between the characters examined by Ekdale (2010) and those used by Macrini et al. (2010), we think that the dimensions for the inner ear of the juvenile specimen of N. murinus (FMNH P13319), a placental mammal, should be representative of those of an adult.

CT Scanning and Digital Endocast Extraction

The two skulls analyzed for digital endocast reconstruction were scanned in their entirety in the coronal plane at the Center for Quantitative X-ray Imaging at Penn State University (www.cqi.psu.edu) in University Park, PA using the HD-600 industrial scanner. Scan data were reconstructed as 1024 × 1024 pixel, 16-bit Tiff slices (e.g., images). The scan of the skull of N. murinus resulted in 1599 slices, with reconstructed pixel size in the X and Y planes of 0.062 mm and interslice spacing of 0.0684 mm. The scan of the skull of Cochilius sp. (SGOPV 3774) resulted in 1066 slices, with reconstructed pixel size in the X and Y planes of 0.060 mm and interslice spacing of 0.0690. Digital endocasts were extracted using the segmentation tools of Avizo 5.0 (2008, Mercury Computer Systems, < www.mc.com >).

Phylogenetic Character Matrix and Analyses

A total of 22 endocranial characters, 11 of them new, were scored based on specimens of 20 notoungulate taxa and five outgroups (Table 1) and data in the literature (Appendices I and II). Measurements relevant to the phylogenetic characters were taken from specimens via digital measurement tools for CT reconstructions, using calipers that are accurate to the nearest 0.05 mm for physical endocasts, or using a ruler and scale bars for figured specimens in the literature. Figure 1 illustrates how these measurements were taken. All measurements were taken three times and then averaged (Table 2). Endocast length was measured by maximum anteroposterior distance from the tip of the olfactory bulbs to the foramen magnum. Right olfactory bulbs were measured unless otherwise noted. For both the olfactory bulbs and the hypophysis, length was measured as maximum anteroposterior distance, and height was measured as the maximum dorsoventral distance. Width was measured on a single olfactory bulb, as the maximum mediolateral distance perpendicular to the midline of the endocast. The width of the hypophysis also was measured perpendicular to the midline of the endocast.

Fig. 1
figure 1

Line drawings of a natural endocast of Rhynchippus equinus (FMNH P13410) in a, left lateral; b, dorsal; and c, ventral views showing how measurements reported in Table 3 were taken. Images redrawn from Fig. 75 of Patterson (1937). Abbreviations: EL, endocast length; H, height; Hyp., hypophysis; L, length; OB, olfactory bulb; W, width

Full size image
Table 2 Measurement data from endocasts used in this study. Abbreviations: EL, endocast length; H, height; Hyp., hypophysis; L, length; OB, olfactory bulb; N/A, not applicable; W, width; ?, missing data (e.g., could not be measured because of damage to the specimen). Measurements are presented in mm
Full size table

These endocranial data were integrated into a larger character matrix that was assembled from previously published phylogenies (Cifelli 1993; Gabbert 1997; Shockey 1997; Croft et al. 2003; Reguero et al. 2003; Flynn et al. 2005; Croft and Anaya 2006; Hitz et al. 2006; Billet 2010, 2011; Shockey et al. 2012), but also adding scoring for 11 original characters (Online Resources 1 and 2). To create the initial character list, publications were scanned for character descriptions and coding, which were incorporated into the matrix and crosschecked to avoid redundancy. In some cases, the characters were modified or reworded from the original source, as indicated in Online Resource 1. The resulting matrix includes 99 characters from the dentition (three of which are new), 72 from the exterior of the skull, 25 from the inner ear (characters and data from Macrini et al. 2013), and 41 from the postcranium, in addition to the 22 endocranial characters scored for the first time here across all taxa, for a total of 259 characters examined across 65 taxa (57 notoungulates, eight outgroups) (Online Resource 1). Despite some concerns about the codependence of some inner ear characters with allometric traits, which may in turn affect the influences of these characters on phylogenetic trees (Billet et al. 2015a), most studies have shown that, in general, inner ear characters provide phylogenetically significant and diagnostic information regardless of allometry (Billet and Muizon 2013; Macrini et al. 2013; Ravel and Orliac 2014; Orliac and O'Leary 2016). This issue certainly deserves further scrutiny, but some recent studies suggest that the influence of allometry may be weak and play a secondary role to phylogenetic signal (e.g. Billet et al. 2015b). Taxa were scored based on direct examination of specimens deposited in the AMNH and FMNH and previously published descriptions. Character polarity was determined by outgroup analysis. For the outgroups, taxa representing other orders of Meridiungulata (Pyrotheria, Astrapotheria, Xenungulata, and Litopterna) and basal “ungulates” (Phenacodus primaevus and Hyopsodus lepidus) were selected, as detailed above. The complete matrix is available as Project 667 in Morphobank (http://morphobank.org/permalink/?P667) and in Online Resource 2.

Parsimony analyses were conducted using TNT (Goloboff et al. 2008b), applying new technology algorithms (Goloboff 1999; Nixon 1999) to search for the shortest trees that subsequently were used to perform standard TBR searches. All characters were considered unordered, as there is no definitive a priori justification for ordering any of them. Branch support was provided by a bootstrap analysis (Felsenstein 1985) with 1000 replicates and by Bremer support/decay index analysis (Bremer 1994).

The matrix was analyzed further using an implied weighting (IW) algorithm, which weights characters based on their homoplasy (Goloboff 1993). Weighting against homoplasy reportedly improves the robustness of phylogenetic analyses of morphological data by improving stability of clades compared with analyses that weight characters equally (Goloboff et al. 2008a, 2018a, b). The value of the concavity constant (K-value) applied was the default value (K = 3), which downweights considerably against homoplastic characters. Branch support in this analysis was determined by relative Bremer support (Goloboff and Farris 2001).

Analysis of Evolution of Endocast Characters

Using parsimony ancestral state reconstruction in Mesquite (version 3.61) (Maddison and Maddison 2019), a selected subset of endocranial characters was mapped onto a pruned phylogenetic tree including only the taxa with available endocasts, to examine the evolution of the endocranium (brain) of notoungulates (Fig. 6). The topology of this pruned tree was essentially that of the equally weighted (EW) analysis but with two differences, regarding the relationships between Tetramerorhinus lucarius, Astrapotheria (Trigonostylops wortmani and Astrapotherium magnum), and Notostylops murinus, which form a polytomy, as well as Typotheriopsis internum, Pseudotypotherium pseudopachygnathum, and Mesotherium cristatum. In those two cases, we resolved the polytomies with the topology found in the IW analysis, which is fully resolved for these taxa. The ancestral state reconstruction analyses utilized DELTRAN, delayed transformations, for optimization of missing or ambiguously optimized character states (Farris 1970).

Encephalization Quotient Analysis

Encephalization Quotients (EQs) and Phylogenetic Encephalization Quotients (PEQs), body mass estimates, and endocranial volumes for selected notoungulates are presented in Table 3. Volumes of the digital endocasts examined directly in this study were measured using Avizo. Endocranial volumes of other notoungulates were obtained from the literature as reported by the original authors (Table 3). Other authors either directly measured volumes of endocasts using volume displacement in water (Jerison 1973; Radinsky 1981), or for endocasts that could not be measured directly, the endocast volume was estimated by the original author by treating the space as a cylinder (Jerison 1973). Notoungulate body mass estimates were taken from the literature (Table 3; Online Resource 3).

Table 3 Encephalization Quotient (EQ) and Phylogenetic Encephalization Quotient (PEQ) calculations for notoungulates and outgroups, using different methods to assess encephalization. Abbreviations: EQ (Eisenberg), calculated using equation from Eisenberg (1981): EQ = EV / (0.055 * [BM]0.74); EQ (Jerison), calculated using equation from Jerison (1973): EQ = EV / (0.12 * [BM]0.67); PEQ, Phylogenetic Encephalization Quotient: PEQ = EXP (lnEV-0.6835 * lnBM + 3.0113); EV, endocranial volume; BM, body mass
Full size table

Encephalization Quotients (EQs) were calculated for all taxa using several approaches (Table 3). First, we used the widely applied EQ equations for mammals from Jerison (1973) and Eisenberg (1981). Encephalization values also were calculated by deriving an equation empirically from the Phylogenetic EQ (PEQ) introduced by Ni et al. (2019). This approach calculates an encephalization quotient from phylogenetic generalized least squares (PGLS), an analysis that also corrects for phylogenetic nonindependence between data points (Symonds and Blomberg 2014). The resulting equation for the taxa analyzed in this study can be expressed as PEQ = EXP (lnEV—0.6607 * lnBM + 2.5023), where EV = Endocranial Volume and BM = Body Mass, and this equation was used to calculate PEQ values for each species. These values were mapped using squared-change parsimony for continuous data (Maddison 1991) on a pruned topology of the trees generated from our phylogenetic analysis as described above, but in this case including only the taxa for which was possible to calculate EQ and PEQ values. For this analysis, we used the Jerison (1973) EQ equation, which allowed for direct comparison with other studies using the same equation (e.g., Boddy et al. 2012), in addition to our new empirically determined PEQ values that incorporate the phylogenetic context for comparative encephalization in the Notoungulata clade.

Results

Description of New Digital Endocasts of Notoungulata

Cochilius sp. (SGOPV 3774; Fig. 2) — This endocast description of Cochilius sp., from the Upeo Fauna of the Abanico Formation of Chile (likely Deseadan SALMA [late Oligocene] in age; see age discussion and analyses of the inner ear in Macrini et al. 2013), represents the first for this taxon. In lateral view (Fig. 2a, a’), the degree of cranial flexure (sensu Macrini et al. 2007a) is 20˚, and the olfactory bulb casts are dorsoventrally deep. The endocast is widest across the posterior portions of the cerebral hemisphere casts. Linear measurements and aspect ratios for this endocast are presented in Table 2. The endocranial volume of this endocast was measured as 6,839 mm3, and the olfactory bulb volume (the total from both left and right bulbs) is 292 mm3. Based on these numbers, the olfactory bulbs comprise 4.3% of the total endocast volume.

Fig. 2
figure 2

Digital cranial endocast of Cochilius sp. (SGOPV 3774) in a, left lateral; b, dorsal; and c, ventral views. Line drawings of the cranial endocast with anatomical labels in a’, left lateral; b’, dorsal; and c’, ventral views. Abbreviation: IAM, internal acoustic meatus (for cranial nerves VII and VIII). Scale bar equals 1 cm

Full size image

Forebrain— The cast of the cerebrum is moderately gyrencephalic; it displays (from medial to lateral) a lateral sulcus, Sylvian sulcus, an anterior rhinal fissure, and a rhinal fissure (Fig. 2). In dorsal view, a pronounced circular fissure separates each olfactory bulb from the isocortex of the cerebrum. The annular ridge of the frontal is the bony feature responsible for forming the circular fissure on an endocast (Macrini et al. 2007a, b). The rhinal fissure, marking the ventral edge of the isocortex and its boundary with the piriform cortex, is discontinuous, with separate anterior and posterior portions. A thin sliver of the piriform cortex is visible in dorsal view, but the cortex is more prominent in lateral and ventral views (Fig. 2). The lateral cerebral venous sinus is visible on the right side of the endocast, running adjacent to the posterior portion of the rhinal fissure; this sinus is not visible on the left side of the endocast (Fig. 2b, b’). The median sulcus, which separates the right and left cerebral hemispheres, is only visible on the anterodorsal portion of the cerebral hemisphere as a shallow, undulating sulcus (Fig. 2b, b’).

Midbrain— The midbrain is not exposed on this endocast, as it is covered dorsally by the isocortex and obscured ventrally by the piriform cortex and hindbrain.

Hindbrain— The small, rounded cerebellar hemispheres and the vermis are clearly visible in the dorsal view of the hindbrain (Fig. 2b, b’). The paired paraflocculi representing the space in the subarcuate fossae, are cone-shaped in Cochilius (Fig. 2). (Some authors, such as Lopez-Torres et al. 2020, equate the paraflocculus with the petrosal lobule, but some others, such as Sánchez-Villagra 2002, consider the petrosal lobe to be part of the paraflocculus.) The subarcuate canal (= petromastoid canal) is clearly visible on both sides of the skull (coronal CT slices C906-977), extending from the subarcuate fossa to the mastoid antrum, an air-filled cavity in the petrosal (Gray 1977). In humans, this canal, which is wide in children and narrower in adults, houses the subarcuate artery and vein (Krombach et al. 2002). Few details are visible on the ventral surface of the hindbrain, as the boundaries of the pons and medulla oblongata are not discernible (Fig. 2c, c’). These brain features were likely obscured by cisterns within the subarachnoid space and other structures within the meningeal space, thus precluding their impressions on the cranial (inner) surface of the basioccipital bone.

Cranial nerves— The endocranial surface of the cribriform plate leaves its impression on the ventral surface of the olfactory bulbs. Fibers of the olfactory nerve (cranial nerve [CN] I) passed through the various openings of the cribriform plate. Paired optic foramina accommodated the optic nerve (CN II); these foramina appear confluent. Cranial nerves III, IV, V1, V2, and VI presumably passed through the sphenorbital fissure. The mandibular branch of the trigeminal nerve (CN V3) likely passed through the foramen ovale, as in other mammals. Cranial nerves VII and VIII undoubtedly passed through the internal acoustic meatus. The jugular foramen for exit of cranial nerves IX, X, XI, and XII is not clearly visible.

Midventral endocast surface— The most prominent feature on this portion of the endocast is the hypophysis, which housed a portion of the pituitary gland. The hypophysis is oval (Fig. 2c, c’), with its long axis oriented in the coronal plane. Portions of the canals for the internal carotid arteries are visible on the posterolateral edges of the hypophysis (Fig. 2c, c’).

Notostylops murinus (FMNH P13319; Fig. 3) — This description of the digital endocast of N. murinus, from Casamayoran SALMA (middle Eocene) faunas of Argentina, is a supplement to the published descriptions of a natural endocast of another specimen of N. murinus (Simpson 1932) and a plaster endocast for its near relative, N. pendens (Simpson 1933a). The right side of this skull appears to exhibit a slight obliquely oriented dorsoventral distortion; this shear is evident when comparing the right and left sides of the digital endocast (Fig. 3). Linear measurements and aspect ratios for this endocast are presented in Table 2. The endocranial volume was measured as 9,807 mm3, and the total olfactory bulb volume (both left and right) is 629 mm3. The olfactory bulbs, which are much larger compared to overall endocast size than in Cochilius, comprise about 6.4% of the total endocast volume in Notostylops. The cranial flexure of Notostylops is 34˚, which is substantially greater than that of Cochilius.

Fig. 3
figure 3

Digital cranial endocast of Notostylops murinus (FMNH P13319) in a, left lateral; b, dorsal; and c, ventral views. Line drawings of the cranial endocast with anatomical labels in a’, left lateral; b’, dorsal; and c’, ventral views. Abbreviation: IAM, internal acoustic meatus (for cranial nerves VII and VIII). Scale bar equals 5 cm

Full size image

Forebrain— The cerebrum is mostly lissencephalic (smooth); however, it does display a suprasylvian sulcus and rhinal fissure, both of which are more obvious on the left than the right side of the endocast (Fig. 3). The median sulcus clearly separates the right and left cerebral hemispheres along the midsagittal plane. The rhinal fissure appears continuous on the left lateral side (Fig. 3a, a’), and is best seen in dorsal view on the right side (Fig. 3b, b’). A small portion of the piriform cortex is visible in dorsal view on both the left and right sides of the endocast and is most evident on the right side (Fig. 3b, b’). A deep circular fissure separates the prominent olfactory bulbs from the cerebral hemispheres.

Midbrain— Similar to the endocast of Cochilius, the midbrain is not exposed in Notostylops. The midbrain was covered dorsally by the isocortex and obscured ventrally by the piriform cortex and the hindbrain as in many extant mammals (e.g., Monodelphis domestica; Macrini et al. 2007c).

Hindbrain— The vermis is clearly visible on the dorsoposterior portion of the cerebellum (Fig. 3a, b), but the cerebellar hemispheres are not distinct from the rest of the cerebellum. The paraflocculi are broad and rounded in Notostylops. Similar to the endocast of Cochilius, the boundaries of the pons and medulla oblongata are not discernible on the ventral surface of the hindbrain of Notostylops (Fig. 3c, c’).

Cranial nerves— The endocranial surface of the cribriform plate is not as well preserved in the Notostylops skull as in Cochilius; therefore, the olfactory nerve fibers (CN I) are not visible (Fig. 3a, c). As in Cochilius, the paired optic foramina (for passage of CN II) appear to be confluent. As there is no clear evidence for a foramen rotundum in Notostylops (Simpson 1933a), the maxillary branch of the trigeminal nerve (CN V2) presumably passed through the sphenorbital fissure, along with CNs III, IV, V1, and VI, as is typically seen in placental mammals. The paired foramina ovale (for CN V3) are oriented anteroposteriorly (Fig. 3a, c; Simpson 1933a) rather than ventrally as in Cochilius. The internal acoustic meatus clearly shows two nerve canal pathways for CNs VII and VIII (Fig. 3a, a’). The jugular foramina for exit of CNs IX, X, and XI, and the hypoglossal foramina for exit of CN XII, are not clearly visible on the digital endocast but are clear on a natural endocast of Notostylops (AMNH 28614; Simpson 1933a).

Midventral endocast surface— The hypophysis is oval (Fig. 3c, c’), with its long axis oriented anteroposteriorly. The internal carotid arteries are not visible.

Phylogenetic Analysis

The unweighted parsimony analysis of the complete phylogenetic matrix (259 characters scored for 57 notoungulates and eight outgroups) resulted in 112 most parsimonious trees, each measuring 935 steps in length (CI = 0.311; RI = 0.678) (strict consensus shown in Fig. 4a). The implied weighting (IW) analysis produced two most parsimonious trees, and their strict consensus is shown in Fig. 4b.

Fig. 4
figure 4

Strict consensus parsimony trees from analyses of complete phylogenetic matrix of 259 characters examined across 65 taxa (57 notoungulates and eight outgroups): a, unweighted; b, implied weighting. Numbers above branches in the unweighted tree (a) are Bremer support values and below branches are Bootstrap values; numbers below branches in the implied weighting tree (b) are relative Bremer support values. Color of taxon names corresponds to their traditional family assignments in the key; light blue are unassigned notoungulates of enigmatic or controversial relationships, and black are outgroups

Full size image

In the unweighted tree, N. murinus (generally considered to be a notoungulate) forms a polytomy with Pyrotherium (Pyrotheria), Carodnia (Xenungulata), and a monophyletic clade including the rest of the species traditionally placed within Notoungulata (Fig. 4a). However, the IW analysis recovered Pyrotherium and Carodnia as sister taxa, and a clearly monophyletic Notoungulata, with N. murinus as the first-diverging taxon within this clade (Fig. 4b). Within Notoungulata, our results from both analyses support the monophyly of Toxodontia and Typotheria (sensu Cifelli 1993). In the unweighted analysis, a clade formed by Anisotemnus distentus and Thomashuxleya externa (traditionally “isotemnids”) forms a polytomy with Toxodontia and Typotheria, although both species are included within Toxodontia in the IW analysis, forming a clade with several other taxa traditionally classified as Isotemnidae or Homalodotheriidae. The best known and sole henricosborniid in the analysis, Simpsonotus, forms an early diverging taxon within Notoungulata in both analyses. The only member of Homalodotheriidae in this analysis, the best-known member of the group (Homalodotherium), is among the basalmost taxa in the Toxodontia radiation in the unweighted analysis, but instead is nested within an unexpectedly monophyletic Isotemnidae in the IW analysis (e.g., Isotemnidae is termed “demonstrably nonmonophyletic” in Wyss et al. 2018). Species traditionally assigned to Isotemnidae represent a paraphyletic assemblage near the base of either Notoungulata or Toxodontia in the unweighted analysis, consistent with prior notions of paraphyly of this group. But as noted above, Isotemnidae are monophyletic in the IW analysis, albeit with relatively weak support; this clade, inclusive of the homalodotheriid Homalodotherium, forms the sister clade to all other toxodontians in the IW analysis. Leontiniidae is monophyletic in both analyses, exclusive of Colpodon, although relationships within the clade remain poorly resolved in the unweighted analysis. Both analyses surprisingly exclude Colpodon propinquus from the clade formed by other members traditionally assigned to this family, as monophyly of Leontiniidae has had substantial support in prior phylogenetic analyses (e.g., Shockey et al. 2012; see discussion in Wyss et al. 2018). Colpodon is instead positioned at the base of a toxodontian clade of Notohippidae plus Toxodontidae in both analyses. As suggested by many previous studies (e.g., Cifelli 1993; Shockey 1997; Billet 2011; Shockey et al. 2012; Wyss et al. 2018), Notohippidae is a paraphyletic assemblage at the base of a monophyletic Toxodontidae or potentially a clade of toxodontids plus leontiniids.

Members of Oldfieldthomasiidae, Archaeopithecidae, and Campanorco are the earliest-diverging members of Typotheria in the unweighted analysis, with oldfieldthomasiids paraphyletic; the oldfieldthomasiid Colbertia forms a polytomy with: 1) a clade including Oldfieldthomasia as the sister group to a clade of the archaeopithecid Acropithecus and the campanorcid Campanorco; and 2) other typotheres. In the IW analysis, oldfieldthomasiids are polyphyletic, with Oldfieldthomasia debilitata, Acropithecus rigidus, and Campanorco inauguralis forming a sequentially diverging series of branches at the base of a clade formed by Mesotheriidae, Archaeohyracidae, and Hegetotheriidae, whereas Colbertia magellanica is the sister group to Interatheriidae (including a paraphyletic assemblage of taxa historically referred to as Notopithecinae or basal interatheriids as in Hitz et al. 2000, 2006). The unweighted analysis recovered Notopithecinae as a paraphyletic radiation at the base of Typotheria rather than part of Interatheriidae; the IW analysis, however, recovered notopithecine taxa in a more conventional position (Hitz et al. 2000, 2006; Billet 2011), as a paraphyletic series of early diverging interatheriids. Other traditionally recognized groups of Typotheria are monophyletic, including Mesotheriidae, Archaeohyracidae, and Hegetotheriidae. Interatheriidae (exclusive of Notopithecinae in the unweighted analysis) is the sister group of a well-supported clade formed by Mesotheriidae and Hegetotheriidae + Archaeohyracidae, as suggested by some previous works (e.g., Cifelli 1993), but unlike the relationships recovered in some more recent phylogenies (e.g., Billet 2010, 2011).

A pruned phylogeny of taxa that are represented by endocasts, based on the unweighted analysis (Fig. 4a), is shown in Fig. 5 (see above). This phylogeny was also used as the framework for optimizations of endocranial traits (Table 4 and Fig. 6). This tree was further pruned for the tree-based PEQ analyses (see above; Fig. 7), as not all available casts provided enough information for an endocranial volume calculation. Therefore, Oldfieldthomasia debilitata, Pachyrukhos moyani, Paedotherium bonaerense, Mesotherium cristatum, and Trachytherus alloxus were excluded from the EQ analyses. One exception is Interatherium robustum, which was added to and included in the EQ analysis but was not included in the original pruned tree as we were not able to examine its endocast firsthand; Radinsky’s (1981) description is not detailed enough to score endocranial characters in the phylogenetic matrix, but it does provide an estimate of endocranial volume for this species.

Fig. 5
figure 5

Pruned tree (including only taxa with endocast data) of the Notoungulata phylogenies obtained here that was used in endocast character optimization analyses. Topology determined as explained in the text

Full size image
Table 4 Parsimony ancestral state reconstructions for endocast characters listed in Appendices 1 and 2, and Online Resources 1 and 2. Reconstructions and clade names are based on the pruned topology shown in Fig. 5
Full size table
Fig. 6
figure 6

Character transformations of six selected endocast characters (Appendix 1) based on DELTRAN optimization on the topology shown in Fig. 5. A striped branch leading to a terminal taxon indicates an unknown character state for that particular taxon

Full size image
Fig. 7
figure 7

Encephalization quotients (EQ) of notoungulate species with character transformations reconstructed using parsimony. The tree used is pruned from the trees obtained in Fig. 4. a, EQ calculated using Jerison’s (1973) equation; b, EQ calculated using a Phylogenetic Encephalization Quotient (PEQ) corrected equation. Colors of branches (to the left of taxon names) correspond to binned ranges of EQs noted in the heat map explanation key for a-b. Colors of EQ bars (to the right of taxon names) correspond to families, with notoungulates of enigmatic or controversial relationships indicated by light blue; colors: light green, Leontiniidae; orange, “Notohippidae”; grey, Toxodontidae; purple, Interatheriidae; dark green, Hegetotheriidae; dark blue, Mesotheriidae

Full size image

EQ Analysis

Notoungulates show relatively low EQs compared with modern mammals (Boddy et al. 2012), with the last common ancestor (LCA) of Notoungulata estimated as having an EQ of 0.5198 based on Jerison’s equation (Fig. 7a). This low EQ is maintained in the early diverging notoungulate N. murinus, with slightly higher EQs in basal Toxodontia (Rhyphodon and Homalodotherium), although the relatively low EQs for this clade optimize as a secondary reduction because there is an ancestral increase in EQ at the Toxodontia + Typotheria node. EQ further increases in the toxodont lineage that includes Leontinia plus the clade of Rhynchippus + Toxodontidae. There are independent reversals to lower EQs in Adinotherium + Nesodon and in Toxodon. The highest EQ among notoungulates is found within Toxodontia, in Leontinia, with Proadinotherium also standing out among toxodontids due to its high EQ compared to other members of the family (Table 3 and Fig. 7a). At the base of Typotheria, there is an increase in EQ for the Hegetotheriidae + Mesotheriidae clade, with Typotheriopsis showing the highest EQ among typotheres (Fig. 7a). Interatheriids, on the other hand, show a reduction in EQ relative to the LCA of Typotheria + Toxodontia, with Cochilius showing a reduction to even lower EQ within the family and Interatherium optimizing as a slight increase relative to more basal interatheriid nodes.

PEQ results (Fig. 7b) are similar to EQ results using Jerison’s (1973) equation, with an increase in PEQ in the LCA for Toxodontia + Typotheria, but noticeable reversals to lower PEQ values in Rhyphodon, Homalodotherium, and Cochilius (Fig. 7b). Proadinotherium, Leontinia, and Rhynchippus again stand out due to their exceptionally high PEQs. Interatheriids show a slight reduction in PEQ relative to the LCA of typotheres, with Interatherium reversing to a higher PEQ, while there is a tendency for a general increase in PEQ in the Hegetotheriidae + Mesotheriidae clade.

Discussion

Comparison of New Digital Endocasts of Notoungulata With Published Data From Near Relatives

Cochilius

Comparisons of the new digital endocasts of Cochilius sp. are made with previously described endocasts of other interatheriids— Miocochilius anomopodus (UCMP 39651 [Stirton 1953; Radinsky 1981]); species of Protypotherium (AMNH 9246 [Simpson 1933b; Protypotherium sp.]; FMNH P13002 [Radinsky 1981; Pr. australe]), and Interatherium robustum (FMNH P130547 [Radinsky 1981]).

A natural endocast of a Miocene interathere from Colombia, Miocochilius anomopodus, was described by Stirton (1953) with additional observations by Radinsky (1981). Distinctive features of this endocast that are similar to those of Cochilius include small olfactory bulbs relative to those of Notostylops, prominent endocast flexure, presence of a prominent Sylvian sulcus, exposure of only a shallow dimple marking the rostral portion of the median sulcus (i.e., longitudinal sulcus, medial longitudinal sulcus, central sulcus) that divides the right and left cerebral hemispheres, and no dorsal exposure of the midbrain (Stirton 1953; Radinsky 1981). Miocochilius differs from Cochilius in possessing a well-developed osseous tentorium, rounded paraflocculi, prominent anterior lobe of the cerebellum, and preservation of the pons on the ventral surface (Stirton 1953).

A partial artificial endocast extracted from a skull of Protypotherium sp. was described by Simpson (1933b); the olfactory bulbs, ventral surface, and occipital surface are partially or completely missing. This description was supplemented by Radinsky (1981) based on an additional endocast of P. australe. The dorsal surface of the cerebral hemisphere casts of Protypotherium is triangular in shape and shows that the lateral sulcus, Sylvian sulcus, only the rostral portion of the median sulcus, and the lateral cerebral venous sinus in the cerebral region all are similar to Cochilius (Simpson 1933b; Radinsky 1981). The paraflocculi are cone-shaped in both Protypotherium and Cochilius. According to Simpson (1933b), the endocast of Protypotherium possesses a suprasylvian sulcus, unlike that of Cochilius.

An artificial endocast of another interathere, Interatherium robustum, was prepared by Radinsky (1981) to augment his description of the endocast of Protypotherium. Similar to Protypotherium and Cochilius, the dorsal surface of the cerebral hemisphere casts of Interatherium is triangular in shape, possesses a prominent Sylvian sulcus, and only the rostral portion of the median sulcus is visible (Radinsky 1981).

Notostylops

The digital endocast of the juvenile specimen of N. murinus described here closely compares to the endocast figures and descriptions of ontogenetically more advanced specimens of Notostylops described by Simpson (1932, 1933a). In particular, the overall shape and proportions of the cerebral hemispheres, olfactory bulbs, and cerebellum of the digital endocast and older specimens are similar in dorsal and profile views (Fig. 3; Simpson 1932, 1933a). Even so, there are some differences that are discussed below.

Despite what Radinsky (1981:171) described as the “great exposure of the midbrain” on the endocast of N. pendens (originally described by Simpson 1933a), there are no impressions of midbrain structures on the skull roof bones and, consequently, no cast of any midbrain features on the dorsal surface of the digital endocast that we observed. Other mammalian endocasts displaying dorsal exposure of the midbrain preserve clearly delimited impressions of the colliculi (Bauchot and Stephan 1967). Meninges and the dural sinus system presumably filled in the space between the skull bones and the midbrain, as they do in many extant mammals, resulting in the smooth surface on the inferior surface of this portion of the parietals.

Simpson (1933a) described and figured an artificial endocast of N. pendens with a rhinal fissure, but Radinsky (1981) later questioned that interpretation. The new digital endocast of the specimen of N. murinus described here appears to show a rhinal fissure, supporting Simpson’s original interpretation of rhinal fissure presence for Notostylops.

Endocast Character Optimizations

Ancestral character state reconstructions are presented in Table 4 for all 22 endocast characters listed in Appendix 1, and the optimizations of select characters are mapped onto the pruned phylogeny in Fig. 6. Character state 6.1 (presence of suprasylvian sulcus on cerebrum) optimizes as a synapomorphy of Notoungulata. Character state 2.1 (olfactory bulb wider than deep) and character state 3.1 (olfactory bulbs diverge from each other near their posterior junction with the rest of the cerebrum, just anterior to the circular fissure) are synapomorphies of Toxodontidae and Toxodontia, respectively. Two additional character states are equivocal synapomorphies for other clades of notoungulates: character states 10.1 (hypophyseal fossa width and length approximately equal) for Interatheriidae and 15.0 (paraflocculus cone-shaped) for Hegetotheriidae. Character 8 (separation of the anterior portion of the cavum epiptericum leading to the sphenorbital fissure) is invariant across our taxon sample but varies in other mammals (Macrini et al. 2007b) and may be phylogenetically informative at more inclusive taxonomic levels in future studies.

Multiple character states are autapomorphic within our taxonomic sample. Character state 4.1 (presence of a vascular foramen entering the olfactory bulb fossa) is only found in Rhynchippus, whereas character state 7.1 (presence of midbrain exposure on the dorsal surface of the endocast) is only found in Hyopsodus. Character states 16.1 (subarcuate fossa posterior to the internal auditory meatus) and 20.1 (absence of a separate optic fissure from cranial nerve II) are only found in Rhyphodon, and character state 18.1 (large recess dorsal to, and confluent with, canal leading to jugular foramen) is only exhibited in Nesodon. Character state 19.1 (foramen rotundum present; separate opening for V2) is only seen in Notostylops.

Nine endocast characters examined in this analysis do not vary for our sample and are not phylogenetically informative for the available taxonomic sample but may be useful at higher taxonomic levels and/or in larger taxonomic samples of notoungulates. The synapomorphies noted above and other character states showing phylogenetically informative and/or noteworthy distributions are described in more detail below.

Character state 2.1 (olfactory bulbs that are wider than deep) is a synapomorphy for Toxodontidae, as represented by Proadinotherium muensteri, Nesodon imbricatus, and Adinotherium ovinum (Fig. 6a). This character could not be scored for Toxodon sp. based on the endocast figures in Dechaseaux (1962). It also occurs convergently in the outgroup taxon Phenacodus primaevus.

Character state 3.1 (divergence of olfactory bulbs near their posterior junction with the rest of the cerebrum [just anterior to circular fissure]) is identified as a synapomorphy for Toxodontia as represented on endocasts of Rhyphodon sp., Homalodotherium cunninghami, Leontinia gaudryi, Rhynchippus equinus, Toxodon sp., Proadinotherium muensteri, and Nesodon imbricatus (Fig. 6b). The character state also is present independently in the outgroup Astrapotherium magnum.

Character state 10.1 (hypophyseal fossa that is longer than wide) is inferred to be a synapomorphy for Interatheriidae based on its presence in Cochilius and Miocochilius; the condition in Protypotherium is unknown (Fig. 6e). This character state also is observed independently in the toxodontids Adinotherium and Toxodon, while other toxodontids (Proadinotherium, Nesodon) retain the ancestral condition for toxodontians, Notoungulata, and nearest outgroups (Character state 10.2).

Character state 15.0 (cone-shaped parafloccular lobes of the cerebellum) represents a synapomorphy for hegetotheriids in our analysis, a feature convergent in some interatheriids (Fig. 6f). The hegetotheriids Hegetotherium and Pachyrukhos show this condition on their endocasts, but the condition is unknown in Paedotherium. This state is observed in the interatheriids Cochilius and Protypotherium, but the more basal interatheriid Miocochilius anomopodus exhibits character state 15.1 (broad and rounded parafloccular lobes), retaining the ancestral condition for notoungulates and nearest outgroups.

Of note, character state 5.1 (piriform lobes of the cerebrum small and not visible in dorsal view, being obscured by the overlying isocortex) is present in all notoungulates in this analysis for which this feature could be scored except for Notostylops, Rhyphodon, Cochilius, and Mesotherium. The condition is unknown in Leontinia, Adinotherium, and Oldfieldthomasia. This character state also is observed in the outgroups Astrapotherium, Trigonostylops, and Tetramerorhinus, and optimizes as ancestral for meridiungulates (Fig. 6c). The character state is convergent with the condition seen in members of several other placental groups (e.g., primates, cetaceans, proboscideans), some marsupials, and monotremes (Macrini 2006; Macrini et al. 2006; Rowe et al. 2011).

Also of note is character state 6.1 (suprasylvian sulcus present on cerebrum), which is mostly distributed in typotheres (Miocochilius, Protypotherium, Hegetotherium, Paedotherium, Mesotherium, and Typotheriopsis), but also is present in the toxodontians Rhyphodon and Adinotherium as well as in the basal notoungulate Notostylops; it optimizes as a synapomorphy of Notoungulata (Fig. 6d). However, most other toxodontians lack this sulcus, and its absence (a reversal from the ancestral condition for Notoungulata) optimizes as a synapomorphy of toxodontians exclusive of the earliest diverging Rhyphodon, with an independent reacquisition of the sulcus in Adinotherium. The suprasylvian sulcus also is absent independently in the typothere Cochilius (character state 6.0). The suprasylvian sulcus is also known on endocasts of other mammals such as artiodactyls (Macrini 2009).

Notoungulate Phylogenetic Relationships

In our analyses, there is support for a monophyletic Notoungulata (Fig. 4). The position of N. murinus is problematic and reminiscent of the results obtained by Billet (2010) that suggest that Pyrotheria belongs within Notoungulata (Billet 2010). Despite very different dentitions, there is some phylogenetic evidence that Pyrotherium romeroi may represent a highly divergent notoungulate, being particularly close to Notostylopidae in some analyses (Billet 2010, 2011). Our unweighted analysis (Fig. 4a) recovers Py. romeroi and N. murinus in a polytomy at the base of Notoungulata, together with Carodnia vieirai. The IW analysis (Fig. 4b), on the other hand, suggests that Py. romeroi (Pyrotheria) and Ca. vieirai (Xenungulata) are sister taxa, and that N. murinus is an early diverging taxon within a monophyletic Notoungulata. The dentition of Ca. vieirai does share some similarities with Py. romeroi, but the former taxon is still incompletely known, and the position of both xenungulates and pyrotheres has been highly contentious (Cifelli 1993; Agnolin and Chimento 2011; Kramarz and Bond 2014; Muizon et al. 2015; Croft et al. 2020). Similarities between the auditory region of Py. romeroi and notoungulates were noted by Patterson (1977), and the phylogeny of Billet (2010) gives some support for a close relationship. However, our results suggest that the phylogenetic position of Pyrotheria remains unresolved. Notostylops murinus, on the other hand, although sharing a basic cranial structure typical of notoungulates, is otherwise highly divergent in many features, as already noted by Simpson (1948), aggregating a great number of autapomorphies. This distinctiveness of Notostylopidae from other notoungulates and other meridiungulates, even after inclusion of new endocranial data in our phylogenetic analyses, thus may continue to hinder disentanglement of its relationships.

Simpsonotus praecursor is one of the earliest well-known notoungulates (Pascual et al. 1978) and is frequently considered as one of the earliest diverging members of the clade (Billet 2011). This is supported by our IW analysis (Fig. 4b), although our unweighted analysis recovers it more deeply nested within Notoungulata, as an early offshoot of Toxodontia (Fig. 4a). The two traditionally recognized major subgroups of Notoungulata, Toxodontia and Typotheria (Cifelli 1993; Billet 2011), each are monophyletic and are sister taxa. Within Toxodontia, the results differ somewhat among our analyses. In the unweighted parsimony analysis (Fig. 4a), isotemnids form a paraphyletic assemblage at the base of a monophyletic clade composed of Homalodotheriidae, Leontiniidae, Notohippidae, and Toxodontidae, as suggested by Billet (2011). The IW analysis (Fig. 4b), however, recovers a monophyletic group composed of Isotemnidae. Anisotemnus distentus and Thomashuxleya externa are sister taxa in a polytomy with Toxodontia (inclusive of the henricosborniid Simpsonotus as its earliest diverging member) and Typotheria in the unweighted analysis (Fig. 4a) but are grouped with other Isotemnidae in the IW analysis (Fig. 4b). Isotemnidae is a group of early Cenozoic notoungulates almost universally thought to be non-monophyletic (Simpson 1948; Cifelli 1993; Billet 2011; see also Wyss et al. 2018). In contrast to that widespread presumption, and despite sharing many plesiomorphic characters, the results of our IW analysis suggest that at least some isotemnids may form a monophyletic clade (Fig. 4b). Homalodotherium cunninghami is sister group to a clade including Leontiniidae, Notohippidae, and Toxodontidae in the unweighted analysis (Fig. 4a), although Billet (2011), reaching the same results, acknowledged that only highly homoplastic characters support the position of Homalodotheriidae. In our analysis, five characters support this relationship, including an endocast character (Character 6; Fig. 6), but not one of them is unambiguous. The inclusion of Homalodotheriidae within Isotemnidae in our IW tree (Fig. 4b) is similar to the results obtained by Shockey et al. (2012) and is consistent with Cifelli’s (1993) suggestion that homalodotheriids might be highly specialized isotemnids. Potential synapomorphies for a monophyletic Isotemnidae, including H. cunninghami, include a reduced-size i1 (character 94.2), which also appears independently in Notostylops, and at least two postcranial features: a well-developed supinator crest in the humerus (character 203.0), which is also a feature of some late-diverging typotheres such as mesotheriids, and a superior foramen in the astragalus (character 229.0), which also occurs in Colbertia. However, as this latter character state is the primitive condition for therians (Meng et al. 2003) and has been lost independently in many groups of placental mammals, including many notoungulates (Horovitz 2000; Shockey and Flynn 2007; Vera 2012; Croft and Lorente 2021), it is more likely a symplesiomorphy despite its optimization in our analysis. As many notoungulates and other Paleogene South American native ungulates still lack adequately known postcranial remains, the status of these character states as synapomorphies should be considered provisional. One endocast character, a sphenorbital fissure even with or just behind the circular fissure (character 9.1), may also be a synapomorphy of this clade.

The grouping of Leontiniidae, Toxodontidae, and Notohippidae in a monophyletic clade has long been recognized (e.g., Cifelli 1993; Billet 2011; Shockey et al. 2012). Leontiniidae (exclusive of Colpodon; see below) and Toxodontidae are monophyletic, but notohippids, as seems to be well established (Cifelli 1993; Shockey 1997; Cerdeño and Vera 2017; Billet 2011; Shockey et al. 2012; see also Wyss et al. 2018), form a paraphyletic array of taxa at the base of Toxodontidae in both analyses (Fig. 4a, b). Of particular interest is the position of Colpodon propinquus, which is traditionally considered a member of Leontiniidae but also shares many characteristics with Toxodontidae and Notohippidae (Cifelli 1993). In our analyses, Co. propinquus was consistently recovered at the base of the clade of Notohippidae + Toxodontidae. At least one synapomorphy supports this relationship unambiguously, a crista intermedia running lingually from the ectoloph between the protoloph and the crochet on upper cheek teeth (character 47.0), but the discovery of new characters and additional analyses may be necessary to test among these competing hypotheses of relationships for Colpodon.

Members of three families in our analyses, Oldfieldthomasiidae (represented by Colbertia and Oldfieldthomasia), Archaeopithecidae (Acropithecus), and “Campanorcidae” (Campanorco inauguralis) (García-López and Babot 2017), are poorly known taxa that usually have unstable positions in phylogenetic analyses (Cifelli 1993); they also have differing interrelationships in our two analyses, although they are members of Typotheria in both (Fig. 4a, b), and are usually assumed to be paraphyletic (e.g., Billet 2011; García-López and Babot 2015). Oldfieldthomasiidae is not clearly monophyletic in our analyses, with its two species either divided between interatheriids and remaining typotheres (Fig. 4b) or in a polytomy at the base of Typotheria (Fig. 4a, with Oldfieldthomasia linked to a clade of Acropithecus + Campanorco). Campanorco inauguralis remains an enigmatic taxon, frequently recovered in a polytomy at the base of Typotheria (Billet 2011), as in our unweighted analysis (Fig. 4a), but also sometimes considered an early member of the clade including Mesotheriidae, Archaeohyracidae, and Hegetotheriidae (Reguero and Prevosti 2010), a position supported by our IW analysis (Fig. 4b).

Results for Typotheria suggest the monophyly of most families, including Archaeohyracidae, Hegetotheriidae, Mesotheriidae, and at least interatheriine Interatheriidae (unweighted analysis; Fig. 4a), but also probably all of Interatheriidae (IW analysis; Fig. 4b). The position of Mesotheriidae as sister group of Hegetotheriidae + Archaeohyracidae was recovered in both analyses (Fig. 4a, b) and agrees with most prior phylogenies (Cifelli 1993; Croft et al. 2003; Billet et al. 2009; Reguero and Prevosti 2010), although monophyly of Archaeohyracidae has been contested in the past (Croft et al. 2003; Billet et al. 2009). Phylogenetic relationships among species of ‘Notopithecinae’ or basal interatheriids are rarely addressed, with many phylogenies including only a single member of the subfamily (Reguero et al. 2003; Billet 2011). In most studies analyzing numerous ‘notopithecines’, however, they are considered a paraphyletic assemblage of early diverging Interatheriidae (Hitz et al. 2006), although other studies recover some ‘notopithecines’ as a monophyletic group of basal interatheriids (Vera 2016). Our IW analysis supports a monophyletic Interatheriidae, with Interatheriinae nested within the paraphyletic ‘Notopithecinae’ (Fig. 4b), a commonly shared view (Hitz et al. 2006). Our unweighted analysis (Fig. 4a), however, recovers ‘notopithecine’ taxa as not closely related to Interatheriinae (and thus not interatheriids), instead forming a paraphyletic radiation at the base of Typotheria. Additional studies with even more comprehensive taxonomic and character sampling will permit testing of these alternative hypotheses of ‘notopithecine’ relationships as well as other points of uncertainty in our phylogeny, such as some unresolved relationships within Toxodontia. The inclusion of early diverging taxa within each group, in particular, may alter the relationships presented here.

Evolution of Encephalization in Notoungulates

The mean Encephalization Quotient (EQ) among notoungulates evaluated in this study is low compared to living mammals with presumably similar habits (Fig. 7a and Table 3), being 0.55 (using Eisenberg’s (1981) equation) or 0.52 (using Jerison’s (1973) equation). Mean EQ values for notoungulates are well below extant ungulates, such as perissodactyls (1.0) and artiodactyls (minus Cetacea; 1.42), and even lower than dominantly smaller bodied and herbivorous rodents (0.6; Boddy et al. 2012). Body mass estimates, especially in extinct taxa like notoungulates without close living analogs, can be challenging. Prior authors (e.g., Jerison 1973; Radinsky 1981; Cassini et al. 2012a, b; Elissamburu 2012) have provided widely varying estimates, and this can result in markedly divergent EQs. We provide justifications of our body mass estimates in Online Resource 3 but acknowledge that using different estimates could alter some of the conclusions reached here and that this issue should always be carefully considered in studies that reach conclusions based on indirect data in extinct species.

Many factors are thought to influence EQ in mammals. Relatively small brains may reflect maintenance of the ancestral condition because early mammals tended to have relatively low EQs in comparison with later diverging forms (Jerison 1973). However, as divergence or phylogenesis continues, patterns of encephalization within groups tend also to become more complex (Schultz and Dunbar 2010), and EQ starts to vary widely in response to other factors. At this point, purely directional explanations are not sufficient to explain variation in EQ within some groups, as other factors come into play.

One of the factors that may influence brain size is the large energetic expenditure of this organ, which may be especially metabolically costly for large bodied mammals since larger bodies also imply an increase in energetic requirements (Marquet and Taper 1998). This relationship may be particularly relevant for large bodied notoungulate groups, such as the Toxodontia. Without evolutionary pressures to maintain large brains, brain reduction may occur in some large bodied mammals, especially those with other metabolically expensive organs and tissues, such as the large and complex guts of herbivores (Striedter 2005; Boddy et al. 2012).

However, one of the most common explanations for the low EQ of notoungulates is brain reduction due to insularism. It is well known that some mammal lineages may undergo brain size reduction when evolving under isolated conditions (van der Geer et al. 2010), which is usually attributed to reduction of predation pressure and low competition in island ecosystems. Low levels of predation also have been proposed to drive decline in EQ under other circumstances, such as reduced selection due to the evolution of antipredator defenses (Stankowich and Romero 2017).

Because South America was an isolated continent during most of the Cenozoic, notoungulates were predated upon by relatively few, generally small brained metatherians and non-mammalian predators (Croft 2001, 2006), which may account for relatively small brains of notoungulates compared to mammals on other continents (Jerison 1973; Fernández-Monescillo et al. 2019). The phenomenon of brain reduction in large bodied South American herbivorous mammals is well documented. Some neoepiblemid rodents, which achieved large body sizes and may have shared some of the large bodied herbivorous mammal ecological niches with various larger bodied notoungulates, show remarkably low EQs when compared with other rodents and similarly sized mammals, and this also has been attributed to lower predation pressures (Ferreira et al. 2020). However, it has been argued that predator guilds in Cenozoic South America were not particularly impoverished compared to other continents based on studies of an early Miocene ecological community of Patagonia (Kay et al. 2012; Rodríguez-Gómez et al. 2020). In either case, metatherian predators may be subject to intrinsic ecological and evolutionary limitations (Croft et al. 2018), which could include less developed brain and cognitive abilities, implying reduced pressure on their prey.

In spite of the overall generally low EQ of notoungulates relative to ecologically comparable herbivores on other continents, substantial variations in EQ can be detected along branches of the Notoungulata tree, including both clade-specific changes and independent increases or decreases (Fig. 7). Both traditional EQ analyses and PEQ yield similar results, with a very low EQ and PEQ for the LCA of Notoungulata, and relatively low EQs/PEQs ancestrally in early diverging toxodontians, most toxodontids, and interatheriid typotheres, with Cochilius in particular showing a substantial secondary reduction in EQ/PEQ within interatheriids. Notable increases in EQ and PEQ occur in the LCA of leontiniids + notohippids + toxodontids, with a further independent increase in Leontinia, Rhynchippus, and Proadinotherium among toxodontians, and in the LCA of hegetotheriids + mesotheriids among typotheres.

Fernández-Monescillo et al. (2019) found that mesotheriids displayed relatively low EQ values compared with other notoungulates, which was attributed to their semifossorial habits. In contrast, the two mesotheriids evaluated in our study (Pseudotypotherium and Typotheriopsis) show relatively high EQs and PEQs compared to other notoungulates (Fig. 7 and Table 3). The notoungulate EQ values found by Fernández-Monescillo et al. (2019), as well as the EQ values of notohippids and leontiniids of Dozo and Martínez (2016) and Martínez et al. (2019), generally are lower than those found here, likely due to differences in estimation of body mass (see above).

On the other hand, there are some instances of marked increases in EQ/PEQ among notoungulates (Fig. 7). Proadinotherium is one of the smallest toxodontids, with an estimated weight around 40 kg (Radinsky 1981), but it has the highest EQ/PEQ in the family and one of the highest among notoungulates. The notohippid Rhynchippus and the leontiniid Leontinia also show higher EQs/PEQs than most notoungulates, with Leontinia in particular having the highest EQ/PEQ among all notoungulates sampled. Toxodontians and typotheres together show an increase in EQ/PEQ relative to the ancestral notoungulate condition. The sampled species of typotheres maintain relatively homogeneous EQs/PEQs overall but with a substantial secondary decrease in EQ/PEQ in the interatheriid Cochilius and an increase in the mesotheriid + hegetotheriid clade, with the mesotheriid Typotheriopsis showing a particularly noticeable increase in EQ/PEQ (Fig. 7). Among toxodontians, higher EQs/PEQs seem to be associated with smaller bodied taxa, even though these measures already standardize for body size, with the notable exception of the large-bodied Leontinia. Increased brain size is more commonly associated with more complex social behavior (Dunbar 1995; Striedter 2005; Schultz and Dunbar 2010), and the results herein are consistent with inferences of gregarious behavior in these species. However, many additional factors may be involved in brain size evolution in mammals, especially in species with such distinctive histories as the endemic South American ‘ungulates’, and it remains to be further tested to determine if the broad conclusions of this study remain supported when additional taxa are sampled.

Little is still known about social behavior in notoungulates and possible consequences of brain structure for specific cognitive abilities among extinct South American ungulates. This may represent an area ripe for future investigation of patterns in evolution of particular brain regions that clearly are associated with particular cognitive functions in extant mammals. There still is but a small sample notoungulate endocasts available relative to the rich species diversity in the group, which can be seen by comparing our pruned tree with the sampling in the full phylogenetic analyses. As our knowledge of the brain morphology and encephalization of notoungulates increases, primarily as CT scanning makes comparative data more readily available, we expect to continue to expand knowledge of the brain anatomy, sensory capabilities and cognition of notoungulates.