The Evolutionary Origin(s) of the Trunk in Proboscidea (Mammalia, Afrotheria) based on Osteology Submitted to the Evolutionary Studies Institute, University of the Witwatersrand, in partial fulfilment of the requirements for the Degree Master of Science (Palaeontology) By Mpilo Nxumalo Student no. 884171 January 2022 http://orcid.org/0000-0002-3739-9655 Supervisors: Dr. Julien Benoit and Prof. Paul R. Manger Evolutionary Studies Institute, University of the Witwatersrand, Private Bag 3, WITS 2050, Johannesburg, South Africa. http://orcid.org/0000-0002-3739-9655 Contents Abstract 1 Introduction 2 Reconstruction, Evidence and Evolution of the Proboscis 3 The Evolutionary History and Diversification of Proboscidea 9 The Evidence for the Evolution of the Proboscis in Proboscidea 15 Research Problem 21 Hypothesis 21 Aims and Objectives 21 Material and Methods 22 Results 31 I- Descriptions 31 II- Statistics 58 Discussion 64 Conclusion 68 Acknowledgements 69 References 70 1 | Page Abstract To date, no quantitative assessment of the evolution of the trunk in elephant ancestors has been undertaken. This study aims to quantify the dimensions of the infraorbital canal (volume and surface area of the infraorbital foramen) and retraction of the naris, to test how these metrics correlate with the dimensions of the nose/trunk in some selected modern mammals and apply these assessments to the fossil record to determine the evolutionary origin(s) of the trunk in Proboscidea. The gross morphology of the infraorbital canal in these species is described in detail for the first time. The results indicate that there is a significant and strong correlation between the surface area of the infraorbital foramen and the dimensions of the snout. The dimensions of the infraorbital canal increase with the size of the facial appendage. There is no correlation between the degree of narial retraction and dimension of the snout in extant taxa. There is a strong taxonomic and phylogenetic signal in the morphology of the infraorbital canal. The application to fossils suggests that Numidotherium possessed a trunk shorter than that of tapirs, and Deinotherium a more tapir-like trunk length. Gomphotheres, Anancus and E. rackei possessed snout lengths shorter than proposed in the literature. The trunk did not originate as a single event at the root of Proboscidea but is a result of a long evolution that spanned the proboscidean phylogenetic tree for at least 60 My, between the last common ancestor of Proboscidea and that of Sirenia. 2 | Page Introduction Carl Illiger was the first to name the Order Proboscidea (Mammalia, Afrotheria), which contains extant elephants and their extinct relatives (Illiger, 1811). Molecular data suggest that the Afrotheria originated in the African Cretaceous, followed by an Afro-Arabian endemic evolution (from ~105 to 25 my ago) (Adaci et al., 2007; and Alberdi et al., 2011). The proboscideans are named after their single-most diagnostic feature, the trunk (Dierenfeld & Milewski, 2013; and Shoshani, 1998). The elephants are the only living proboscideans, from the 164 confirmed species that lived in the past, and that they originated in the late Paleogene of the Afro-Arabian landscape (Abraha et al., 2007; Fisher, 2018; and Gheerbrant & Tassy, 2009). Although the range of extant species is limited to Africa and Asia, extinct species were widely distributed on most continents (except Antarctica and Australia), inhabiting a variety of habitats, from deserts to mountains (Fisher, 2018; and Osborne, 1942). The proboscis, or trunk, is a flexible facial appendage that is formed by the fusion of nostrils and nasal passages with the upper lip musculature (Dierenfeld & Milewski, 2013; Shoshani, 1998; and Shoshani & Tassy, 2005). This is supported by the observation of early fetal development in elephants, where the upper labial surface and nose are clearly separated (Shoshani, 1998). Different from the true trunk, Wall (1980: pp968) defined the mobile upper-lip as “an enlarged movable lip which is adapted for seizing or grasping food but does not drastically change the position of the nostrils”. The elephant trunk is extremely sensitive and functionally used to detect subtle tactile, thermal, and pressure changes in the environment (Shoshani, 1998; and Shoshani & Tassy, 2005). It is also used for breathing, sound production, touching and navigation, olfaction, and manipulation of objects (Galton et al., 2006; Shoshani, 1998; and Shoshani & Tassy, 2005). 3 | Page Based on morphology and behavioral function, beyond the order Proboscidea, the only extant mammals observed to possess a true proboscis are the four species within the superfamily Tapiridae (Order Perissodactyla) (Dierenfield & Milowski, 2013; Galton et al., 2006; Giannini & Moyano, 2018; and Wall, 1980). The Tapiridae meets the minimum requirements of this definition in that they possess an extension formed by the narial joint and upper lip musculature, which is both flexible and tubular, and used partly for feeding and olfaction (Dierenfield & Milowski, 2013). The tapirs’ trunk differs from elephants’, in that the trunk’s anterior occlusion beyond the incisors is minimal and the nasal bones are relatively less reduced and retracted (Dierenfield & Milowski, 2013). Saiga tatarica (Clifford & Witmer, 2004a; and Frey & Hofmann, 1996), Alces alces (Clifford & Witmer, 2004b; and Linnaeus, 1758), elephant-shrews (Macroscelidea: Rhynchocyon, Elephantulus, Galegeeska, Petrosaltator, Petrodromus and Macroscelides spp.) (Kingdon, 1997), elephant seals (Mirounga spp.) (Galimberti et al., 2007; and Bryden & Ling, 1992), proboscis monkeys (Nasalis larvatus) (Yeager, 1989) and dik-diks (Rhynchotragus and Madoqua spp.) (Frey & Hofmann, 1996; and Kingdon, 1997) possess facial appendages that are referred to as a “proboscis”, but do not meet the minimum requirements of the proposed definition (Dierenfield & Milowski, 2013). None of these mammals use their trunk for grasping (Dierenfield & Milowski, 2013). Instead, they possess a prorhiscis (Dierenfield & Milowski, 2013; and Giannini & Moyano, 2018). Beyond Mammalia, no fish or reptile possesses a true proboscis or a facial appendage that meets the above-mentioned minimum requirements for a true trunk (Clifford & Witmer, 2004a; Dierenfield & Milowski, 2013; and Witmer, 1999). Reconstruction, Evidence and Evolution of the Proboscis 4 | Page The elephant trunk is composed of soft-tissue, and it is not readily preserved in the fossil record. Paleontologists make use of osteological proxies, which are believed to be associated with the possession of a trunk in extant taxa, in order to reconstruct its potential presence in the fossil record. Based on cranial modifications that are associated with the presence of a trunk, there are many extinct taxa that have been hypothesized to possess a trunk and are reconstructed with a proboscis- like facial appendage (Galton et al., 2006). According to Galton et al. (2006), the first published inference of a proboscis in extinct taxa was the historical study carried out by Cuvier (1796: pp309) on the Megatherium (Xenarthra: Megatheriidae) from South America. Shortly after, each discovery of non-proboscidean extinct mammalian taxa with similar cranial modifications (i.e. the degree of narial retraction), but with varying cranial gross morphologies, were assigned a proboscis (Galton et al., 2006). Wall (1980) studied in detail the cranial modifications of the nasal region, that are believed to be associated with the presence of a trunk or a prehensile upper lip. A detailed comparative study of Cadurcodon (Rhinocerotoidea, Amynodontidae) and Tapirus pinchaque, revealed compelling evidence that Oligocene amynodonts bore a trunk (Wall, 1980). A review of other extinct mammalian taxa, hypothesized to possess a proboscis (Palaeomastodon, Astrapotherium, and Macrauchenia) suggested that the characters which are important when reconstructing and inferring the presence of the trunk are: retraction of the external nares, enlargement of the nasal opening and the presence of large muscle attachment sites, or entheses (Wall, 1980). Osborn (1898) noted these cranial modifications as evidence for the presence of the trunk or prehensile upper lip in amynodonts. It was suggested by Troxell (1921) that Metamynodon bore a prehensile upper lip based on the caliber of the infraorbital canal, morphology of the nasal 5 | Page opening, and roughening of the supraorbital ridge. Gromova (1954) suggested that some of the advanced amynodonts likely bore a proboscis; however, she did not state which cranial modifications were associated with the proposed proboscis. The amynodont genus that has been found to possess most of the cranial modifications associated with a proboscis is Cadurcodon. The skull of AMNH 26029 is the best example for these cranial modifications (Wall, 1980). Wall (1980) conducted a detailed comparison between this skull and that of Tapirus pinchiqua, and identified twelve cranial features that are common between the two taxa and are associated with the presence of the proboscis (Table 1) (Wall, 1980). Palorchestidae, a late Oligocene family of eastern Australian marsupial megafauna, are well known for their odd ‘tapir-like’ cranial morphology (Adams et al., 2019). A detailed description of Propalorchestes shows a heavily retracted nasal morphology (Sharp & Trusler, 2016). The presence of this characteristic has been used to support the hypothesis that they could have possessed a proboscis (Adams et al., 2019; and Sharp & Trusler, 2016). They possessed a mosaic of morphological characters not seen in other mammals, including hyper retraction of the osseous naris, specialised forelimbs resembling a ‘marsupial-tapir’, and overall nasal bone morphology that superficially resembles those of tapirs (Adams et al., 2019); however, the hypothesis of the proboscis-bearing Palorchestidae has been challenged, in favor of a sensitive prehensile lip (Adams et al., 2019; and Sharp & Trusler, 2016). Like Propalorchetes, Macrauchenia patachonica possesses extremely retracted nasal bones and opening of the osseous naris (Blanco et al., 2021). Macrauchenia patachonica belongs to the Macraucheniidae, and the members of this family are characterised by their degree of narial retraction (Blanco et al., 2021). The morphology of the vomer along with the retracted nares were used as evidence for the presence of a trunk (Burmeister, 1864; and Scott, 1910). The wide muscle 6 | Page attachment surfaces were interpreted as evidence for a prolonged trunk, much longer than the tapir trunk, but not quite elephantine (Lydekker, 1903). Other authors contested this view and argued that Macraucheniidae did not bear a trunk, but a prehensile Alces-like lip (Moyano & Giannini, 2018). Other evidence such as the dimensions of the infraorbital foramen have been used to infer the presence of the trunk, in addition to the modification of the nasal region (Shoshani, 1998). The infraorbital ramus of the maxillary branch (V2) of the trigeminal nerve innervates the follicles of the sensory hairs and skin of the elephant trunk (Crumpton & Thompson, 2013; Galton et al., 2006; Muchlinski, 2008; Nabavizadeh, 2015; Nabavizadeh & Reidenberg, 2019; Osborn, 1936, 1942; Sampson et al., 1999; and Wall, 1980). It passes through the infraorbital canal, which opens caudally within the orbit (maxillary foramen) and rostrally on the lateral aspect of the maxilla (infraorbital foramen) (Benoit et al., 2019; Crumpton & Thompson, 2013; and Muchlinski, 2008) (figure 1). 7 | Page Figure 1. Procavia capensis skull in left dorso-lateral view, showing the length of the infraorbital canal (in red). 8 | Page Table 1. The cranial modification characters associated with the presence of the trunk (Wall, 1980). Character state (number) Character or feature 1 Enlargement of the nasal opening and its caudal retraction to the level of the orbit 2 Process of the premaxilla and its nasal contact is lost or absent. A portion of the maxilla is incorporated along the border of the external nares 3 Nasal bone length reduced 4 The turbinal bone and cribriform plate are migrated posteriorly to a level medial to the Orbit 5 The lateral concavity on the maxilla is medially developed to the level of the orbit to accommodate the displacement of the nasal diverticulum 6 Lacrimal knobs for muscle attachment 7 The preorbital elements of the skull are reduced in overall length 8 Orbit migrated anteriorly relative to the cheek teeth 9 Infraorbital canal and foramen dimensions are increased in size 10 Premaxillae fused and vertically thickened 11 Enlargement and presence of the frontal sinus 12 Occipital region increased in width, due the presence of well developed, and strong neck Muscles Outside Mammalia, it has been noted that Diplodocus possesses a single, relatively enlarged narial orifice that is retracted to the level of the orbits (Galton et al., 2006). Based on the morphology of this character, it has been suggested that diplodocus possessed a proboscis, like an elephant (Galton et al., 2006). Hypotheses of proboscis-bearing hadrosaurs (Ornithopoda: Hadrosauridae) have 9 | Page been proposed in a series of papers by Wilfarth (1938, 1939, 1940, 1948, 1949). Wilfarth argued that the beak of the duck-billed dinosaurs is used to attach strong proboscis muscles (Wilfarth, 1939, 1940, 1948). Sternberg (1939), noted the redundancy of evolving a proboscis on top of a beak, and compellingly disproved this hypothesis. Contrary to the condition in extant elephants, the relatively small size of the maxillary branch of the trigeminal nerve in Diplodocus (based on analyses of the endocranial cast) suggests that there is no paleo-neuroanatomical evidence to support the presence of an elephantine trunk (Galton et al., 2006). Living mammals that have a trunk possess well-developed intrinsic narial muscles (Clifford & Witmer, 2002b), and as a result, they possess an enlarged maxillary branch of the trigeminal nerve (Galton et al., 2006). Loxodonta africana and Elephas maximus possess a significantly enlarged maxillary branch of the trigeminal nerve (Shoshani, 1998). In pinnipeds, the infraorbital foramen is enlarged to accommodate the enlarged infraorbital nerve (Adam & Berta, 2002). Wall (1980) also noted that Cadurcodon possesses a relatively enlarged diameter of the infraorbital canal, to accommodate the enlarged infraorbital nerve. Diplodocus possesses an infraorbital canal that has a relatively small diameter, suggesting the absence of paleo- neuroanatomical evidence for the presence of an elephant-like proboscis (Galton et al., 2006). The Evolutionary History and Diversification of Proboscidea Before tackling the question of what is currently known about the evolution of the proboscidean trunk, it is important to detail the major phases in the evolutionary history of the Proboscidea. Early proboscideans, the ‘Plesielephantiformes’ originated in the Afro-Arabian Palaeogene, (Abraha et al., 2007; Gheerbrant & Tassy, 2009; and Harzhauser et al., 2002). This suborder incorporates primitive, bilophodont proboscideans such as phosphatheres, barytherioids, 10 | Page numidotheres, moeritheres, and daouitheres (Amaghzaz et al., 2002; and Sanders et al., 2001a). The ‘Plesielephantiformes’ are a paraphyletic assemblage, formed by a group of stem proboscideans at the base of the Proboscidea clade (figure 2) (Amaghzaz et al., 2005). They share the plesiomorphic bilophodont dental condition of the primitive proboscideans (Todd, 2006). ‘Plesielephantiformes’ differ from Deinotheriidae and Elephantiformes in the size of the tusks (relatively smaller on this suborder), and dimensions of the inferred proboscis (short, tapir- like) (Sanders et al., 2001a). Deinotheres possess lophodont dentition, however, they descended from a bunodont ancestor (Harris, 1978; and Kappelman et al., 2004). The lifestyle of the ‘plesielephantiforms’ has been a centre of controversy for a century now, with different lines of evidence indicating terrestrial, semiaquatic and aquatic lifestyles (Domning & McKenna, 1986; and Osborn, 1909). Evidence from palaeontology, morphology and embryology led authors to hypothesize that elephants and sirenians share a common amphibious ancestor (Liu et al., 2008). Measurements of oxygen isotopes on the tooth enamel of the genera Barytherium and Moeritherium show an isotopic pattern that is different from terrestrial mammals, but more similar to that of aquatic and semiaquatic mammals (Liu et al., 2008). 11 | Page Figure 2: The Phylogeny of Proboscidea, along with the proposed amphibious phase and the three waves of diversification based on ecomorphological evidence, developed from Todd, (2006); Liu et al., (2008); and Fisher, (2018). The carbon-13 measurements obtained from both these genera indicate that their diet was a mixture of fresh-water vegetation from fresh-water swamps, and C3 terrestrial plants (Alberdi et al., 2011; and Liu et al., 2008). These results support the hypothesized Oligocene amphibious phase (figure 2), and aquatic ancestry of Proboscidea (Berenbrink et al., 2013; and Liu et al., 2008). Moeritherium possesses a mosaic of features in the skull that have been observed in both aquatic and semiaquatic 12 | Page mammals; its auditory region is adapted to terrestrial hearing, not underwater hearing; however, its skull is relatively telescoped, with anteriorly placed orbits, seen mostly in marine mammals (Court, 1994; Liu et al., 2008; and Matsumoto, 1923). The postcranial skeleton of Moeritherium shows evidence of hind limb reduction similar to that observed in the semiaquatic Desmostylia (i.e. small pelvic hip socket) (Simons, 1968). Elephants share many common features with sirenians, such as double-apex hearts and their lungs are not surrounded by a pleural cavity; which both groups are hypothesizes to have inherited from a common aquatic ancestor (Sukumar, 2003). The sedimentology of the Fayum region of Egypt, where Moeritherium was found, indicates that these fossils were deposited in a periodically flooded lagoon, lake or pond near the Ocean; shark and marine fish remains were recovered, with the majority of the remains constituting terrestrial vertebrates (Liu et al., 2008). Berenbrink et al. (2013) studied the evolution of the diving capacity on various afrotherian mammals using myoglobin net surface change. The results from this study support an amphibious ancestry for Proboscidea, and suggests that the ‘Plesielephantiformes’ possessed muscle oxygen storage capacity that is relatively higher than that of other early paenungulates (Berenbrink et al., 2013). The muscle oxygen storage capacity of late Eocene to early Oligocene proboscideans was likely comparable to that of extant sirenians (i.e. estimated diving capacity of Moeritherium ~ 10 min), followed by a secondary decrease in diving ability in elephantiforms (i.e. diving capacity of Elephas maximus ~ 2.5 min) (Berenbrink et al., 2013). At the beginning of the Cenozoic world (65 – 50 Mya) the warm and wet African environment had very dense forests (Guillermo & Spencer, 2010). An environment with dense vegetation favours small-bodied animals, as opposed to large ground- dwelling savannah animals (Guillermo & Spencer, 2010). This is when proboscideans originated, with forms such as Eritherium and 13 | Page Phosphatherium in the late Paleocene to early Eocene (Amaghzaz et al., 2005). They were adapted to living in swampy, wet forested environments (Liu et al., 2008). In the early Eocene, the climate was even warmer, tropical forests were widespread all the way to polar latitudes, with an evenly spread annual rainfall (Guillermo & Spencer, 2010). Christine Janis hypothesized that the more phased distribution of yearly rainfall during the peak warming triggered the spacing of forests, and thus promoted the undergrowth and evolution of grasses (Janis, 1989). From the middle to late Eocene, the global climate became cooler and arid, resulting in the retreat of tropical forests and spread of grasses (Markov et al, 2001). It has been hypothesized that the evolution of features such as body mass and brain mass in elephantids and ‘plesielephantiforms’ is better explained by environmental and climatic changes in Africa (Benoit et al., 2019; Burckle et al., 1995; and Vrba, 1993). The evolution of the Proboscidea is hypothesized to have followed environmental changes, which led to a series of adaptive shifts in the dentition and skull, as members of the Order evolved various dietary specialisations (Fisher, 2018; and Shoshani, 1998). Todd, (2006) argues that even though this view has been prevalent in the literature, there are other selective pressures, which are equally important in driving the evolution of Proboscidea, such as competition and biotic interaction. Niche partitioning in proboscideans was driven by ecomorphological innovations (Alberdi et al., 2011). Competition and biotic interaction may have led to resource partitioning; for example, Loxodonta is proposed to have migrated into open savanna niches with the demise of Elephas in Africa around 500,000 ya (Alberdi et al., 2011; and Todd, 2006). Three periods of diversification and radiation are recognized for the Order Proboscidea, based on masticatory functional morphology (figure 2): 14 | Page The first period occurred in the late Eocene, from 41 to 21 Ma, and is characterised by Barytherium, Moeritherium, Phiomia, and Palaeomastodon (Fairfield, 1934; Fisher, 2018; Gohlich, 1999; Harris, 1978; Shoshani, 1998; and Todd, 2006). Proboscideans from this period of diversification are mostly trilophodont, possessing molars with high cusps (i.e Deinotheriidae, Phiomia and Palaeomastodon), which are used in a vertical shearing motion during chewing (Moeritherium and Barytherium have bilophodont molars) (Alberdi et al., 2011; and Todd, 2006). Based on ecomorphological evidence, members from the families Moeritheriidae, Barytheriidae and Deinotheriidae are also included in this period (Todd, 2006). The transition from the first to the second radiation is marked by major trends such as an increase in body size, skull size and shape, tusk size, physiological changes, and changes in cheek dentition (Fisher, 2018; and Shoshani, 1998). All the taxa in the first radiation had the typical mammalian vertical tooth displacement mechanism (Shoshani, 1998), whereas the proboscideans in the second radiation evolved the typical horizontal, conveyor-belt type of tooth replacement that characterises proboscideans (Abraha et al., 2007). Proboscidean masticatory functional diversity changed at a constant rate during this period, with very low species diversity (Alberdi et al., 2011). The second diversification period contains the trilophodont gomphotheres, and occurred from 21 to 11 Ma (Fisher, 2018; Shoshani, 1998; and Todd, 2006). The most important event of this period is the biogeographic expansion of Proboscidea beyond Africa (Gomphotherium datum event), and they reached the Americas by 16 Ma (Alberdi et al., 2011). This expansion event led to a dramatic explosion in proboscidean eco-morphology and functional diversity, and the origin of several lineages (Alberdi et al., 2011). The proboscideans of this period are characterised by bunodont molars (with isolated conical cusps), two pairs of tusks, and a chewing motion that is a combination of predominantly horizontal (lateral and forward), and vertical motions (Alberdi 15 | Page et al., 2011; and Todd, 2006). The taxa of the second and third radiations are characterised by horizontal tooth replacement (Shoshani, 1998). Due to the modification in the skull morphology, the mandible became too short to house all the molars and premolars simultaneously (Nabavizadeh, 2015; Nabavizadeh & Reidenberg, 2019; and Shoshani, 1998). The last period of radiation and diversification occurred from 7 to 1 Ma, and it includes Elephantidae (Fisher, 2018; Gohlich, 1999; Shoshani, 1998; and Todd, 2006). During this period, the first major proboscidean extinction event is recorded; driven by palaeoclimatic dynamics, and a high ecomorphological specialisation (Alberdi et al., 2011). The proboscideans of this period are characterised by a strictly proal chewing motion, and hypsodont molars (Todd, 2006). Based on masticatory functional morphology, the elephants of this period are further subdivided into two groups: the first are the horizontal shearing browsers with cusp-like occlusal surfaces (i.e.) Primelephas and Stegotetrabelodon, and the second are the horizontal shearing grazers and they were hypsodont, i.e., the Elephantidae (Alberdi et al., 2011; and Todd, 2006). The Evidence for the Evolution of the Proboscis in Proboscidea Early proboscideans from the late Palaeocene to early Oligocene, such as Phosphatherium, have a relatively small nasal opening that is positioned rostral on the snout (Shoshani, 1998). This is a primitive condition, which many authors have used to argue for the absence of a trunk or the presence of a fleshy, mobile upper lip (Shoshani & Tassy, 2005; and Sukumar, 2003). Most authors agree that the slightly more phylogenetically crownward Numidotherium koholense possessed a short, tapir-like proboscis, based on the retracted external nares and the elevated skull (Shoshani, 1998; Shoshani & Tassy, 2005; and Sukumar, 2003). Even though these authors have 16 | Page qualitatively used the degree of the narial retraction in their arguments and restorations, none have tested if there is a real correlation between the trunk and the degree of narial retraction. Moving crown-ward on the phylogenetic tree (figure 2), Larramendi inferred the presence or absence of the trunk in various proboscideans, on restorations based on femoral measurements and estimated body sizes (Larramendi, 2016). Moeritherium possesses a long neck with an elongated mandible, but lacks the cranial modifications (or degree of modification) associated with the presence of a short trunk (Larramendi, 2016). According to the reconstructions of Larramendi, (2016), Moeritherium seemingly did not possess a trunk, nor even the mobile upper lip similar to rhinos, as proposed by other authors (i.e. Shoshani, 1996). The dimensions of the infraorbital foramen are correlated to the number of nerve fibres passing through the infraorbital canal in mammals (Muchlinski 2008). As the proboscis developed during proboscidean evolution, it is thus inferred that the size of the infraorbital foramen on fossilised skulls would reflect the increasing innervation of the “growing” trunk (Andrews, 1904; and Osborn , 1936, 1942). To the best of our knowledge, no quantitative approach to trace the evolution of the dimensions of the proboscidean infraorbital foramen has been undertaken, and only qualitative accounts are available. It is noteworthy that even the basal-most “Plesielephantiformes”, such as Eritherium, Phosphatherium, and Numidotherium (Amaghzaz et al., 2005; Ameur et al., 1984; and Gheerbrant 2009), already present with a relatively large infraorbital foramen, surrounded by a deep infraorbital fossa (or canine fossa) for the attachment of a presumably well- developed levator alae nasi muscle (Boas, 1908; and Shoshani & Eisenberg, 1982). This strongly suggests that a mobile and prehensile upper lip was already present in the basal-most proboscideans and is likely a plesiomorphic feature of the Tethytheria (Amaghzaz et al., 2005). 17 | Page Deinotheriidae and Elephantiformes, including the basal elephantiform Palaeomastodon, possess a very large infraorbital foramen, comparable to that of modern elephants (Andrews, 1904; Osborn, 1936, 1942; and Sanders et al., 2010), although some variations exist and remain to be fully explored, like in Gomphotherium angustidens, which exhibits a condition where the infraorbital canal is divided into a small dorsal foramen and a relatively larger ventral one (Tassy, 2013). In general, the infraorbital canal is long and runs horizontally in basal “plesielephantiforms” but becomes relatively short and more obliquely oriented in deinotheres and elephantiforms as the rostrum shortens and the external nares are retracted (Andrews, 1904; Osborn, 1936, 1942; and Sanders et al., 2010). The proboscis, molars, and tusks combined weigh hundreds of kilograms (Larramendi, 2016; and Shoshani & Eisenberg, 1982) contributing 5 to 10% of total body mass in extant elephants. The proboscideans have evolved a highly pneumatised skull with deep insertions for the nuchal ligaments, seemingly to compensate for the increased cranial mass (Andrews, 1904; Bezuidenhout et al., 1995; Osborn, 1936, 1942; Sanders et al., 2010; and Shoshani & Tassy, 1996). Eritherium, Phosphatherium, and Moeritherium show few signs of cranial pneumatisation, whereas Numidotherium and Barytherium do reveal aspects of cranial pneumatisation (Amaghzaz et al., 2005; Ameur et al., 1984; Delmer, 2005b; Gheerbrant, 2009; and Gheerbrant & Tassy, 2009). This makes it difficult to determine the exact origin of a pneumatised skull among “plesielephantiforms”. It is nevertheless likely that Moeritherium secondarily lost its cranial pneumaticity as an adaptation to a semi-aquatic lifestyle (Matsumoto, 1923; and Tassy, 1985). The deinotheres, Palaeomastodon, and more derived elephantiformes all share the presence of cranial pneumaticity and deep nuchal fossae for ligamentous attachment (Andrews, 1904; Osborn, 1936, 1942; Sanders et al., 2010; and Shoshani & Tassy, 1996). 18 | Page Due to the evolution of the proboscis, the proboscidean skull changed in overall gross morphology to accommodate attachments of the heavy labial and nasal musculature needed to operate the massive trunk, i.e. the nares became increasingly large and retracted, the snout shortened and the premaxilla became wider (Andrews, 1904; Gheerbrant & Tassy, 2009; Osborn, 1936, 1942; and Shoshani 1998). The earliest proboscidean to display an enlarged narial opening is Numidotherium koholense, whereas the first hints of narial retraction appear with Barytherium and Moeritherium (Ameur et al., 1984; Andrews, 1906; Delmer, 2005, 2005c; and Sanders et al., 2010). These anatomical changes are consistent with a gradual increase in size of the pre-existing mobile upper lip. Larremendi (2016) argues that Deinotherium had a short neck, columnar limbs, and a long femur; therefore, it must have possessed a long proboscis to aid during feeding (Larramendi, 2016). Since their discovery 50 years ago, there has been a lot of debate and controversy about the facial morphology of deinotheres (Markov et al., 2001). Abel (1922) reconstructed deinotheres with a more elephantine trunk, along with downturned lower tusks, based on the presence of a large nasal opening. Harris (1975), Markorv et al. (2001), Svistun (1974), Tassy (1998), and Tarabukin (1974), all argued that this reconstruction is improbable both for evolutionary and anatomical reasons. They opposed this reconstruction despite the presence of the large nasal opening, which has been used by previous authors to support the presence of the trunk (Markorv et al., 2001). The skull does not have enough surface area for muscle insertion of an elephantine-sized trunk, and the orientation of the premaxilla would have hindered its movement (Markov et al., 2001). Markorv et al. (2001) attempted to reconstruct the facial morphology, and then elucidated the feeding behaviour of Deinotherium. They concluded that the gross cranial morphology of Deinotherium suggests that they possessed a giant tapir-like trunk. Proboscideans from the 2nd wave (Oligocene to Miocene) are marked by enlarged and heavily retracted nares (Shoshani, 1998). These advanced features are observed as early as the Miocene “gomphotheres”, and are well established on derived taxa such as mammoths (Shoshani, 1998). 19 | Page Phiomia and Palaeomastodon are usually reconstructed as bearing a trunk that is slightly longer than the tapir-like condition, but not as long as the elephantine trunk (Sukumar, 2003). On the restorations of Phiomia and Palaeomastodon from Sukumar (2003), the osteological proxies used were not specified. Wall (1980) noted that Palaeomastodon possesses cranial modifications that are associated with the presence of the trunk, including retracted external nares, enlarged nasal opening, and large muscle attachment scars. Andrews (1904) and subsequent authors (e.g. Nabavizadeh, 2015; and Nabavizadeh & Reidenberg, 2019) hypothesized that the onset of a very long mandibular symphysis in basal elephantiforms (i.e. Palaeomastodon, Mammutida, Gomphotheriinae, Choerolophodontinae, Amebelodontinae and other “gomphotheres”) and deinotherids accompanied the evolution of the proboscis. The proboscis would occlude with the symphysis to enhance trophic activities and food processing, and as such, the growth of the trunk would parallel the lengthening of the symphysis throughout phylogeny (Nabavizadeh, 2015). This initial lengthening is coupled with the formation of tusk-like upper and shovel-shaped lower incisors (Andrews, 1904; Hautier et al., 2008; and Nabavizadeh, 2015). The maximum length of the mandibular symphysis is reached in Choerolophodontinae, and Amebelodontinae indicating that a trunk comparable to that in modern elephants was present as early as the middle Miocene, and is followed by the convergent, secondary reduction of the symphysis in the late Miocene and Pliocene in the Mammutida and Stegodontidae (modern elephant ancestors) whilst the proboscis remained stable (Andrews, 1904; Bezuidenhout, 2010; Nabavizadeh, 2015; Osborn, 1936, 1942; and Tassy, 2013). The convergent loss of lower tusks may be correlated to the decrease of global temperature and humidity in the upper Miocene and Pliocene, as the presence of four tusks would enhance heat loss (Guillermo & Spencer, 2010). Based on the retraction of the narial opening, length of the mandibular symphysis, enlargement of the infraorbital foramen, and other cranial adaptations, it is most likely 20 | Page that basal “plesielephantiforms'' had a prehensile upper lip (Nabavizadeh, 2015). The facial and narial musculature eventually evolved into a large and prehensile proboscis in the last common ancestor of the Deinotheriidae and Elephantiformes in the late Eocene (Andrews, 1904; Nabavizadeh, 2015; and Osborn, 1936, 1942). The presence of a prehensile upper lip would account for the relatively large cerebellum of Moeritherium, which makes up about one-third of the total length of the endocast in dorsal view (Benoit et al., 2019). The endocast of all known Elephantiformes displays an enlarged cerebellum comparable to that in modern elephants (Benoit, 2016; and Benoit et al., 2019). In the rare occasion when it is preserved and depicted, the cast of the trigeminal nerve is correspondingly large on the endocranial cast of elephantiforms (Andrews, 1906; Dechaseaux, 1958; Giovinazzo & Palombo, 2005; and Kupsky et al., 2006). Though the cerebellar morphology of deinotherids is unknown, the size of the foramen rotundum indicates that the trigeminal nerve was relatively large (Harris, 1975). Finally, there are special cases where we have direct evidence of a proboscis and do not need to rely on proxies (i.e., the mummified remains of mastodons and mammoths) (Fisher, 2018). In the absence of direct anatomical evidence, we can use information from ichnology and trace remains to supplement the morphological fossil record (Cawthra et al., 2021). Along the coastline of the South African Cape, thirty-five Pleistocene proboscidean track remains have been identified (Cawthra et al., 2021). Elongated grove impressions found along track marks from these sites have been interpreted as elephant proboscis drag marks (Cawthra et al., 2021). These drag impressions were found in association with a tusk and at least five other occurrences of elephant skeletal material all dated to no later than the Pleistocene (likely Loxodonta sp.) (Cawthra et al., 2021). No one so far has attempted to use such trunk drag impressions to reconstruct the dimensions of the trunk in the fossil record; however, the very existence of drag marks suggests that 21 | Page the proboscis was long enough in these Pleistocene proboscideans to reach the ground. Most of the osteological proxies used by authors to argue for the presence or absence of the trunk on this literature review can be easily measured (i.e. degree of narial retraction, surface area of the infraorbital foramen and volume of the infraorbital canal); however, to date the correlation of the trunk to these proxies have not been examined in a statistical framework. Research Problem To date, no quantitative assessment of the evolution of the trunk in elephant ancestors has been undertaken. We do not know exactly when the trunk evolved, how it grew (assuming it started small), and we do not have a quantifiable metric or systematic approach for determining whether a trunk was present or not in fossil species, and for determining how large such a trunk may have been. This project aims to develop a better understanding of the evolution of the trunk in elephants, which has implications regarding our understanding of the evolutionary success of elephants, their feeding and drinking habits, and intra- and interspecific behaviours (e.g. tactile and scent recognition). This project will thus shed new light on a hitherto understudied research question. Hypotheses 1. The evolutionary origin(s) of the proboscis in Proboscidea occurred gradually during the course of their evolution. 2. The presence and dimensions of a proboscis can be reconstructed using simple measurements on dry and fossil skulls. Aims and Objectives 1. To quantify the dimensions of the infraorbital canal (i.e. volume and surface area of the infraorbital foramen) and retraction of the nares on the dry skulls of a selected sample of modern “ungulates” and afrotherians. 2. To test how these metrics correlate with the dimensions of the nose/trunk in the corresponding 22 | Page live animals (as measured in lateral view). 3. Apply these assessments to fossil proboscideans, from the basal-most ‘Plesielephantiformes’ to more derived Elephantidae. Material and Methods CT scan samples: This project is based on observations and measurements made on an existing dataset of 126 CT- scanned dry skulls representing 126 modern species of Mammalia (113 Artiodactyla, 8 Perissodactyla and 5 species of Paenungulata). The CT scans of dry skulls of mammals that are proposed to possess a trunk or “trunk-like” structures (the saiga antelope, Asian and American tapirs, elephant shrew and African elephants) are compared to the skulls of outgroup taxa with no trunk. These medical quality CT scans were made from the collections of the American Natural Museum (ANMH), Ditsong National Museum of Natural History (AZ and TM), Natural History Museum of Basel (NMB), Evolutionary Studies Institute of the University of the Witwatersrand (BP), School of Anatomical Science of the University of the Witwatersrand (MS and ZA), Wits Life Science Museum (WLSM), Yale Peabody Museum of Natural History (YPM), and Zoological Museum of the University of Zurich (ZM). The list of the specimens (and their CT scanning parameters) is available in the appendix (supplementary Table 1 and 2), and in Benoit et al., (2020). The elephant shrew (AMNH 161535) was Laser Scanned by Leif Tapanila (published 06/16/2020), at Idaho Virtualization Lab (Idaho State University) available on morphosource; (www.morphosource.org/concern/media/000119087?locale=en). Segmentation and Measurements of the CT data: All the measurements on the CT scans were undertaken using the software AVIZO 9 (FEI VSG, Hillsboro OR, USA) at the virtual imaging lab of the Evolutionary Studies Institute, University of the Witwatersrand. The volume of both the left and right canals (if available) was measured on the http://www.morphosource.org/concern/media/000119087?locale=en 23 | Page same software, and the average between the two was used for statistical analyses. If only one side was preserved, only one canal was used. The measured volume includes all the branches of the infraorbital canal, from the maxillary foramen to the rostral alveolar branch. The retraction of the naris (e.g. 2D measurement, figure 3), skull length (i.e. 2D measurements, figure 4), and the surface area of the infraorbital foramen (i.e. 3D measurements of the diameter, of both the long-Y and short-X axis, figure 5) were measured on the same software. The retraction of the osseous naris on the skull was quantified as the distance between the anterior margin of the skull and the posterior-most margin of the narial opening in lateral view (figure 3). The length of the skull was defined as the distance between the anterior-most margin of the premaxilla to the tip of the condyle (figure 4). The surface area was calculated using the formula A = π(a.b), where a is half the height of the infraorbital foramen (long axis-Y) and b is half the width of the infraorbital foramen (short axis-X). These measurements co-vary with body size, so this effect is accounted for by taking the ratio of the infraorbital canal volume to log body mass, surface area of the infraorbital foramen over the head length, and the ratio of snout length to head length (body masses are available in the literature; Benett et al., 2012; Benoit et al., 2020 and, Holling & Lambert, 1998). The internal gross morphology of the infraorbital canal (branching patterns) across these mammalian taxa was documented and described. 24 | Page Figure 3. Skull of Bos taurus in lateral view, showing the measurement of the narial retraction. Figure 4. Skull of Bos taurus in ventral view, showing the measurement of the skull. 25 | Page Figure 5. Diameter of the infraorbital foramen in Loxodonta africana, in oblique anterior view. Measurements on pictures: After taking measurements of the dimensions of the infraorbital canal and naris on dry skulls, measurements of the snout and lower jaw length (head length) were taken on living species using the software Image J (figure 6). The measurements on living species were taken in order to test the correlation of the snout to the proposed osteological proxies (i.e. calibre of the infraorbital canal). Pictures of zoo animals in lateral view were taken using a camera equipped with a spirit level. The images were taken in the years 2018 and 2019 at the Zoologischer Garten (Germany), Zoopare of Beauval (France), Ménagerie du Jardin des Plantes (France), Lory Park and Owl Sanctuary, Johannesburg Zoo (South Africa), National Zoological Garden (South Africa), Montecasino Bird Garden (South Africa), and Chester Zoo (United Kingdom). K. H. Vogel provided the images of the saiga, and this dataset represents about 10,000 pictures documenting the lateral view of 129 species and is available at https://osf.io/4vpnj/?view_only=3dc98 7012f cd44a 6a64a d7d8949ec0 1f (https://doi.org/10.17605 /OSF.IO/4VPNJ (Benoit et al., 2020). The linear distance between the eyeball and the distal-most extremity of the 26 | Page snout (or of the trunk when applicable) was measured in order to quantify the length of the nose (or trunk) (figure 6). Figure 6.The heads of A, Tapirus indicus and B, Equus zebra, in both lateral view, showing the measurements of the snout (Sl) and lower jaw (Hl) lengths on the images. Regression models The aim of this study is to test if there is a significant correlation between the dimensions of the trunk and the osteological proxies listed above using linear regressions. Simple linear regression tests were conducted using the software PAST (see Hammer, 1999-2021). Three regression models were generated and tested on the sample (n = 43) used for the current study. These models include: (1) Snout length/Head length (photos) vs Surface area of Infraorbital foramen/Skull length (dry skull); (2) Snout length/Head length (photos) vs Narial retraction; and (3) Snout length/Head length (photos) vs Average Volume Infraorbital canal/log (body mass). If one or more of these osteological measurements prove to be strongly correlated to the dimensions of the trunk (which 27 | Page is the expected result based on the available literature), the regression formula will then be applied to the following extinct proboscidean taxa. Application to fossils Dr J. Benoit provided the CT scan of the basal-most Numidotherium koholense (UOK5 from the early Eocene of Algeria, 48 million years old, scanned at 81.66 μm voxel size, at the AST-RX platform of the MNHN in Paris). The Pleistocene Mammut americanum (ANSP VP 13307) was scanned using surface scanning, converted into image stack using the Scan Surface to Volume function of Aviso, at Greenfield 3-D Imaging Lab of the Academy of Natural Sciences (Drexel University), and uploaded by Kyle Luchenbill (01/09/2919) (acquired from morphosource.org: https://www.morphosource.org/concern/media/000066352?locale=en). The volume of the canals and surface area of the cross-section of the infraorbital foramen, and retraction of the naris of Numidotherium and Mammut were measured using Avizo 9 (FEI VSG, Hillsboro OR, USA) (Mammut is represented by partial maxilla; its narial retraction can be measured from images in the literature). The body mass of Numidotherium used here is 250 -300 kg (Larramendi, 2016), and Numidotherium skull length is 37.0 cm (Ameur et al., 1984). The body mass of Mammut americanum is 5761 kg (Larramendi, 2016). The skull length of Mammut americanum is 97.7 cm (Branstrator & Woodman, 2008). Dr J. Benoit also provided measurements of the infraorbital foramen in fossil proboscidean taxa, from the Nairobi National Museum, Kenya (see table 2): Table 2. The measurements of the infraorbital foramen (IOF) from the museum of Nairobi, Kenya. Specimen no. ID IOF height (cm) IOF width (cm) Surface area (cm²) KNM-LU975 Anancus kenyansis 5 3.2 12.57 KNM-WS12874 Gomphotheriidae indet 1.9 1.9 2.84 KNM-KP30204 Loxodonta sp. 7 8 43.98 http://www.morphosource.org/concern/media/000066352?locale=en) http://www.morphosource.org/concern/media/000066352?locale=en) http://www.morphosource.org/concern/media/000066352?locale=en) http://www.morphosource.org/concern/media/000066352?locale=en) 28 | Page KNM-ER5711 Elephas recki 5 12 47.12 KNM-ER1087 Deinotherium bozasi 6.5 4 20.42 KNM-WS12660 Gomphotheriidae indet 3 5.5 12.96 The skull lengths of these fossil taxa were calculated from the literature: Anancus kenyansis = 70 cm (Brunet et al., 2009); Elephas recki = 112 cm (Larramendi, 2016); Deinotherium bozasi = 115 cm (Harris, 1975). For Gomphotheriidae, the species is indeterminate. As such, an average skull length was calculated and used for the gomphothere specimens: G. tassyi = 76 cm (Jaroon et al., 2017); G. productum = 106.25 cm (Larramendi, 2016); G. steinheimense = 93.75 cm (Larramendi, 2016); Notiomastodon, N. platensis MECN82 = 81.25 cm (Larramendi, 2016); Stegomastodon, S. mirificus NMNH 10707 = 87.5 cm (Larramendi, 2016); G. hondurensis = 57.5 cm (Guillermo & Spencer, 2010); Cuvieronius = 90 cm (Guillermo & Spencer, 2010); and the average skull length = 90 cm. Sampling: This data is heterogeneously sampled, with a bias towards ruminants, especially bovids which are over-represented. As this will yield biased results towards ruminants, taxonomic re-sampling was undertaken, to diminish the confounding effects of phylogeny and create a more homogenous data set for statistical analyses. It was sampled so that the species within the three Orders Afrotheria, Perissodactyla and Artiodactyla are more evenly represented (n = 43). In addition, an in-group sub- sample containing all the taxa that possess a trunk or proboscis-like snouts in the study (n = 8; 2 Loxodonta, 2 tapirs, 1 elephant shrew, 2 saiga and 1 dik-dik) were isolated for some of the analyses. This dataset does not include proboscis monkeys, elephant seals or other carnivores because ungulates are the closest ecological relatives of proboscideans, and afrotherians are the closest phylogenetic relatives of proboscideans. 29 | Page Results I – Descriptions The nomenclature used for the description of the maxillary canal branches is that of Benoit et al. (2016). The infraorbital canal is never perfectly cylindrical, but is divided into the canal for the infraorbital nerve per se (which leads to the infraorbital foramen), the rostral or incisivomaxillary alveolar canal (the longest and anterior-most branch, abbreviated RAC) which is oriented towards the canine root, and the middle (MAC) and caudal (CAC) alveolar ramus which are both oriented towards cheek teeth (Wilber, 2008, 2011; Evans & Lahunta, 2012; Rodella et al., 2012; German et al., 2015; and Benoit et al., 2016). The infraorbital foramen is more-or-less elliptical, with varying degrees of lateral and dorso-ventral compressions amongst the taxa. Afrotheria Orycteropus afer: The two aardvarks sampled here possesses a short and conical canal, with a long and well-defined RAC (figure 7). The RAC is twice as long as the infraorbital canal and gives off three to four smaller branches that diverge ventrally and medially towards the tooth row. The CAC and MAC are not present or clearly defined. 30 | Page Figure 7. The canal in Orycteropus afer: A; right lateral view, B; left lateral view, and C; dorsal view (all skulls are transparent) (scale: 20 mm). IOC; infraorbital canal, IOF; infraorbital foramen, RAC; rostral alveolar canal. The purple structure is the brain. Procavia capensis The canal of Procavia is conical caudally and becomes more tubular and relatively elongated rostrally. Specimen N291 possesses a MAC and RAC that diverge from the middle of the main canal, at the same level (figure 8 A). The MAC diverges laterally towards the check teeth, and the RAC branches anterior-ventrally towards the root of the upper tusk-like incisor. There are four Procavia capensis specimens in this study; specimen UNM462 has a gross morphology of the canal similar to that of UMN76: i.e. MAC is absent, but possesses a RAC that bifurcates to two or three branches of equal size, towards the root of the tusk-like incisor (figure 8 C). Specimen UNM305 is similar to UNM276; i.e. possesses a MAC and along with a simple RAC (figure 8 A and B). 31 | Page Figure 8. The infraorbital canal in Procavia capensis; A; UNM291 in right lateral view (scale: 10 mm, B; UNM291 in dorsal view, C; UNM76 in left lateral view. IOC; infraorbital canal, IOF; infraorbital foramen, MAC; middle alveolar canal, RAC; rostral alveolar canal. The purple structure is the brain. Macroscelides proboscideus: The elephant shrew possesses an infraorbital canal that is relatively short, conical and voluminous (3.2819 cubic centimeters). It also possesses an elongated, simple RAC without visible side branches (figure 9); however, there is a small and short branch caudal to the RAC, which appears to run parallel to the rostral alveolar canal. This condition is symmetrical as it is present on both the left and right canals. 32 | Page Figure 9. The infraorbital canal in Macroscilides proboscideus; A; right lateral view (scale: 5 mm), B; left lateral view (scale: 6mm), C; dorsal view, all skulls transparent (scale: 7 mm). IOC; infraorbital canal, IOF; infraorbital foramen; RAC; rostral alveolar canal. Proboscidea The proboscideans possess a relatively voluminous, short and cylindrical infraorbital canal with reduced side branches. The ‘Plesielephantiform’ proboscidean Numidotherium koholense possesses a relatively long and tubular canal (figure 10). Mammut americanum possesses a short and cylindrical canal, with a relatively more circular infraorbital foramen (figure 11), similar to that of Loxodonta africana (figure 10) and the Loxodonta africana with no specimen number (figure 11). Both the Loxodonta africana specimens possess a MAC that branches off ventro- medially towards the cheek teeth, with no clear RAC. Similar to Macroscelides proboscideus, Loxodonta africana possesses a small branch, which is oriented caudo-dorsal to the main canal; however, it differs from that of Macroscelides proboscideus in that it is positioned caudal to the 33 | Page main canal, instead of the RAC. The specimen on figure 13 does not seem to possess this caudo- dorsal branch. Figure 10. The infraorbital canals in Numidotherium koholense; A, opaque skull in in right lateral view, scale bar is 350 mm; B, infraorbital canal in left lateral view missing the transparent, Scale bar is 35 mm. IOC; infraorbital canal; IOF; infraorbital foramen. Figure 11. The infraorbital canal (in red) in Mammut americanum. A; right lateral view, B; caudal view; C, right lateral view, transparent skull. IOF; infraorbital foramen. Scale bar is 13 mm. 34 | Page Figure 12. Infraorbital canal in Loxodonta africana. A; right lateral view (scale: 30 mm), B; dorsal view, C; left lateral view (scale: 40 mm). IOC; infraorbital canal, IOF; infraorbital foramen; MAC; middle alveolar canal. Scale bar is 30 mm A, 77 mm B, 40 mm C. Figure 13. The infraorbital canal in Loxodonta africana, no-specimen number. A; right lateral view (scale: 41 mm), B; left view, C; left dorsall view. IOC; infraorbital canal, IOF; infraorbital canal; MAC; middle alveolar canal. Scale bar is 40 mm A, 50 mm B, 77 mm C. The purple structure in A is the brain. 35 | Page Artiodactyla Tylopoda Tylopods possess an infraorbital canal with a gross morphology that is mostly tubular, medio- laterally compressed in the middle, and much thicker rostrally in lateral view, unlike the condition observed in most ungulate taxa. Camelus bactrianus (figure 14 A and B) possesses a dorsoventrally thicker infraorbital canal than that of Camelus dromedarius (figure 14 C and D); however, its RAC is relatively shorter. The RAC of Camelus dromedarius is almost equal to that of the main canal in length, but much thinner in thickness compared to that of Camelus bactrianus. Both taxa seemingly do not possess a visible MAC nor CAC. Figure 14. Infraorbital canal in; A; Camelus bactrianus in left lateral view (scale: 75.95 mm), B; dorsal view (scale: 93.29 mm), C; Camelus dromedarius left lateral view (scale: 50 mm), D; dorsal view, all skulls transparent. IOC; infraorbital canal, IOF; infraorbital canal; RAC; rostral alveolar canal. 36 | Page Suoidea The gross morphology of the infraorbital canal in the suids is generally thicker when compared to other artiodactyls such as bovids (figures 15 - 18). Potamochoerus porcus (figure 15), has an outline that is jagged in lateral view, and is laterally compressed. The rest of the suid specimens possess a more cylindrical canal. It is conical caudally (except Phacochoerus africana, figure 17), and then elongates and tapers rostrally in dorsal view. All the suid specimens possess a well- defined, elongated RAC, with the exception of Phacochoerus africanus (figure 17), which possesses a short, dorsoventrally thickened RAC. Both the MAC and CAC are seemingly absent in suids except in Potamochoerus porcus (figure 15) which possesses a visible MAC. The RAC terminates at the root of the tusk-like canines, on the taxa with an elongated canal (figure 18). The taxa who possess a canal with a smooth outline seem to have fewer branches (i.e. figure 16), compared to the irregularly shaped types (figure 15). Figure 15. Infraorbital canal in Potamochoerus porcus; A; right lateral view (scale: 50 mm), B; left lateral view (scale: 50 mm), C; dorsal view (scale: 50 mm), all skulls transparent. IOC; infraorbital canal, IOF; infraorbital canal; MAC; middle alveolar canal, RAC; rostral alveolar canal. The purple structure on figure C is the brain. 37 | Page Figure 16. The infraorbital canal in Sus scrofa domesticus; A; left lateral view (scale: 75.81 mm, B; right lateral view (scale: 57.4 mm), C; dorsal view (scale: 57.4 mm), all skulls transparent. IOC; infraorbital canal, IOF; infraorbital canal; RAC; rostral alveolar canal. The purple structure is the brain. Figure 17. The infraorbital canal in Phacochoerus africanus; A; right lateral view (scale: 65.14 mm), B; left lateral view (scale: 65.14 mm), C; dorsal view (scale: 57.02 mm), all skulls transparent. IOC; infraorbital canal, IOF; infraorbital canal; RAC; rostral alveolar canal. 38 | Page Figure 18. The infraorbital canal in Sus scrofa (wild boar); A; right dorsal view (scale: 30.4 mm), B; left lateral view (scale: 30 mm), all skulls transparent. IOC; infraorbital canal, IOF; infraorbital foramen; RAC; rostral alveolar canal. The purple structure in both figures is the brain. Hippopotamidae Hippopotamus amphibius possesses a very complex, relatively large infraorbital canal, which is narrow caudally and thickens rostrally, when viewed laterally (figure 19). The canal is medio- laterally compressed when viewed dorsally, and contains multiple recesses on its lateral external surface. The RAC is complex, being divided into two main branches, each ramifying into multiple smaller branches. The dorsal-most branches are relatively small and go above the root of the tusk- like incisor. The ventral-most branches are larger and go below the root, along the palate (figure 19). The condition in Hexaprotodon is similar to that of H. amphibius, however, relatively simpler as it bears fewer branches (figure 20). The RAC in Hexaprotodon bifurcates into two branches 39 | Page only (with no further subdivisions), but with a similar orientation to those of H. amphibius. The MAC and CAC seem absent in both taxa. Figure 19. The infraorbital canal in Hippopotamus amphibius; A; right lateral view (scale: 65 mm), B; left lateral view (scale: 76.5 mm), C; dorsal view (scale: 93 mm), all transparent. IOC; Infraorbital canal, IOF; Infraorbital canal; RAC; rostral alveolar canal. 40 | Page Figure 20. The infraorbital canal in (Hexaprotodon liberiensis; A; right lateral view (scale: 26.63 mm), B; left lateral view (scale: 28 mm), C; dorsal view (scale: 31.17 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital canal; Rac; rostral alveolar canal. Moschidae and Tragulidae Though the Moschids and Tragulidae are not closely related, they share a simple infraorbital canal morphology, with no side branches. Both canals are thin, elongated and tubular; however, the canal in Tragulidae is medio-laterally compressed so that the infraorbital canal is more elliptical than that of the Moschidae (figures 21 and 22). The MAC and CAC are absent in both taxa. 41 | Page Figure 21. The infraorbital canal in Tragulus javanicus; A; right lateral view (scale: 11.87 mm), B; left lateral view (scale: 14.23 mm), C; dorsal view (scale: 11.58 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen. Figure 22. The infraorbital canal in Moschus A; dorsal view (scale: 30 mm), B; left dorso-lateral (scale: 30 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen. All skulls are transparent. Giraffidae Unlike other ruminants such as bovids, giraffids have a relatively thicker infraorbital canal , with more branches (especially Okapia johnstoni, figure 24). The infraorbital canal in the giraffe is 42 | Page thicker caudally, and then elongates and tapers rostrally (figure 23). The infraorbital canal in Okapia bears more branches than in the giraffe. The left canal bifurcates in two separate canals rostrally, of equal thickness and length, and both terminate externally at the level of the infraorbital foramen (IOF 1 & 2, figure 24). The right infraorbital canal divides into three branches rostrally, with IOC 1 and 2 being thicker, and IOC 3 being relatively thin. They also terminate externally on the surface of the maxilla, at the level of the infraorbital foramen. Both taxa do not seem to possess the MAC and CAC branches. Figure 23. The infraorbital canal Giraffa camelopardalis. A; left lateral view (scale: 141.8 mm), B; right lateral view (scale: 131.8 mm), C; left dorsal view (scale: 141.8 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen. 43 | Page Figure 24. The infraorbital canal in Okapia johnstoni; A; right lateral view (scale: 31.01 mm), B; left lateral view (scale: 36.3 mm), C; left dorsal view (scale: 33.55 mm), D; oblique ventro-lateral view of the infraorbital canal. Skulls, A to C are transparent, D is opaque. IOC; infraorbital canal, IOF; infraorbital foramen. Bovidae (non-proboscis bearing) The bovids possess a simple, thin, tubular and relatively long infraorbital canal (figure 25 to 27). They bear fewer branches compared to other artiodactyls, except for a simple RAC, when present (figure 27A). The only bovid that seem to have a relatively complex infraorbital canal morphology, with multiple branches is Raphicerus campestris (figure 28). Its infraorbital canal morphology is similar to that of the saiga. Raphicerus campestris also has a well-defined, right CAC. 44 | Page Figure 25. The infraorbital canal in bovids without a proboscis; A; Syncerus caffer, dorsal view (scale: 55.55 mm), B; Tragelaphus strepsiceros, dorsal view (scale: 60 mm), C; Connochaetes taurinus, dorsal view (scale: 34. 19 mm), D; Redunca fulvorufula, left dorso-lateral view, all transparent. IOC; infraorbital canal, IOF; infraorbital foramen; RAC; rostral alveolar canal. The purple structure in C is the brain. Figure 26. The infraorbital canal in bovids without a proboscis; A; Bos taurus, right lateral view (scale: 23.15 mm), B; Capra aegagrus, right lateral view (scale: 22.93 mm), C; Kobus ellipsiprymnus, right lateral view (scale: 78.88 mm), D; Kobus leche, right lateral view (scale: 23.01 mm) all skulls transparent. IOC; infraorbital canal, IOF; infraorbital foramen. The purple structure in B to D is the brain. 45 | Page Figure 27. The infraorbital canal in Raphicerus campestris; A; right lateral view (scale: 14.41 mm), B; dorsal view (scale: 19.03 mm), C; left lateral view (scale: 20.47 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen; MAC, middle alveolar canal, RAC; rostral alveolar canal. The purple structure in C and D is the brain. Bovid taxa who possess proboscis-like snouts All bovids possess a narrow, elongated and tubular canal; however, the proboscis-bearing bovids bear a slightly more complex canal, with multiple MAC and RAC branches. Saiga tatarica (figure 34 and 35) possesses only one clearly visible MAC, which branches at the cheek teeth level and terminates externally at the level of the infraorbital foramen. This condition is only observed on the right canal in specimen YPM 85301 (figure 34), and on the left canal in YPM 1510 (figure 35). The CACs in both specimens are not well defined. 46 | Page Figure 34. The infraorbital canal in Saiga tatarica (YPM 85301); A; left lateral view (scale: 19.92 mm), B; right lateral view (scale: 16.42 mm), C; dorsal view (scale: 22.32 mm) all skulls transparent. IOC; infraorbital canal, IOF; infraorbital foramen; MAC; middle alveolar canal, RAC; rostral alveolar canal. Figure 35. The infraorbital canal in Saiga tatarica (YPM 1510); A; right lateral view (scale: 30.49 mm), B; left lateral view (scale: 30.49 mm), all skulls transparent. IOC; infraorbital canal, IOF; infraorbital canal; RAC; rostral alveolar canal. 47 | Page Madoqua kirkii possesses an infraorbital canal with a gross morphology that is similar to that of the saiga; however, its MAC and RAC branches are relatively well defined. Its infraorbital canal morphology ranges from a canal with fewer branches (the left canal only bears a single side branch, figure 37), to one with well-defined MAC branches, in both canals (figure 36). These branches also terminate externally, at the level of the infraorbital foramen. The two canals are symmetrical in specimen YPM 3985 (figure 36), whereas in specimen YPM 5462, the condition is asymmetrical, with the right canal having more branches (figure 37). The CAC in both specimens is not clearly visible or present. Figure 36. The infraorbital canal in Madoqua kirkii (YPM 3985); A; left lateral view (scale: 18.04 mm), B; right lateral view (scale: 18.04 mm), C; dorsal view (scale: 18.04 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen; MAC; middle alveolar canal, RAC; rostral alveolar canal. The purple structure in C is the brain. 48 | Page Figure 37. The infraorbital canal in Madoqua kirkii (YPM 5462); A; dorsal view (scale: 5.33 mm), B; right lateral view (scale: 4.61 mm), C; left lateral view (scale: 5.33 mm), all transparent. IOC; infraorbital canal, IOF; Infraorbital foramen; MAC; middle alveolar canal, RAC; rostral alveolar canal. Cervidae The gross morphology of the infraorbital canal in the cervids is similar to that of non-proboscis bearing bovids. Its overall shape is tubular and elongated. However, more cervid specimens were observed to possess a RAC than in none proboscis-bearing bovids. Alces (figure 28) possesses a visible MAC, while Dama, Hippocamelus and Mazama (Figures 29 to 32) all bear a well-defined RAC. In Hippocamelus the RAC terminates externally on the surface of the maxilla, at the level of the infraorbital foramen, instead of internally within a tooth socket as is usual for mammals (Benoit et al., 2016). Similar to the majority of the bovids, the CAC is seemingly absent in all these taxa. 49 | Page Figure 28. The infraorbital canal in Alces alces; A; dorsal view (scale: 69.09 mm), B; left lateral view (scale: 55.57 mm), C; right lateral view (scale: 49.24 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen; MAC, middle alveolar canal, RAC; rostral alveolar canal. Figure 29. The infraorbital canal in Cervus elaphus; A; right lateral view (scale: 64.74 mm), B; left lateral view (scale: 64.74 mm), C; left dorsal view (scale: 69.96 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen. 50 | Page Figure 30. The infraorbital canal in Dama dama; A; right lateral view (scale: 23.51 mm), B; left lateral view (scale: 35.9 mm), C; dorsal view (scale: 38.14 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen; RAC; rostral alveolar canal. The purple structure in C is the brain. Figure 31. The infraorbital canal in Hippocamelus sp.; A; left lateral view (scale: 28.83 mm), B; right lateral view (scale: 28.14 mm), C; dorsal view (scale: 46.78 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen; rostral alveolar canal. The purple structure in C is the brain. 51 | Page Figure 32. The infraorbital canal; A; Muntjac, dorsal view (scale: 29.18 mm), B; right lateral view (scale: 29.18 mm), C; Mazama americana, left dorso-lateral view (scale: 49.96 mm), A and B are transparent. IOC; infraorbital canal, IOF; infraorbital foramen, Rac; rostral alveolar canal. The blue structure in A is the brain. Figure 33. The infraorbital canal in Cervus elaphus; A; right lateral view (scale: 35 mm), B; left lateral view (scale: 30 mm), C; dorsal view (scale: 30 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen. 52 | Page Perissodactyla Equidae The infraorbital canal morphology of the equids is similar to that of ruminants, but differs in the way its slopes – the slope is very shallow caudally and steepens rostrally. The canal is concave- down bent resulting in an upside-down U-shape gross morphology. This condition is much more visible in zebras, and almost absent in Equus asinus (figure 38). The infraorbital canal in E. asinus appears U-shaped when viewed dorsally, and this condition is absent in other Equidae. In equids, the infraorbital foramen is positioned more dorsally, unlike most ruminants where it is positioned more ventrally. All the Equidae taxa appear to bear a simple RAC (figures 38, 40 and 41). E. caballus possesses the longest RAC when compared to the other Equus taxa (figure 40). E. quagga does not possess a RAC (figure 39). None of the Equidae taxa appears to possess the MAC and CAC branches. 53 | Page Figure 38. The infraorbital canal Equus asinus; A; right lateral view (scale: 70.26 mm), B; left lateral view (scale: 65.34 mm), C; dorsal view (scale: 81.23 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen; RAC; rostral alveolar canal. The purple structure in the figures is the brain. Figure 39. The infraorbital canal Equus quagga; A; right lateral view (scale: 60 mm), B; left lateral view (scale: 70 mm), C; left dorsal view (scale: 90 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen. The purple structure in all the figures is the brain. 54 | Page Figure 40. The infraorbital canal in Equus caballus; A; right lateral view (scale: 57.56 mm), B; left lateral view (scale: 70.75 mm), C; dorsal view (scale: 57.6 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen; RAC; rostral alveolar canal. The purple structure in all the figures is the brain. Figure 41. The infraorbital canal Equus zebra; A; right lateral view (scale: 60.39 mm), B; left lateral view (scale: 52.86 mm), C; dorsal view (scale: 60.39 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen; MAC, middle alveolar canal, RAC; rostral alveolar canal. The purple structure in all the figures is the brain. 55 | Page Rhinocerotidae Rhinoceroses bear a canal morphology with multiple branches, being more complex compared to equids. The infraorbital canal in Ceratotherium (figure 42) is relatively thicker compared to that of Diceros (figure 43). It is thicker caudally and narrows rostrally in Ceratotherium simum, whereas Diceros bears a uniformly tubular canal. Ceratotherium has a MAC and RAC that are further divided into smaller single or double branches. Diceros exhibits a special condition, in which the canal divides into three smaller branches that terminate on the surface of the maxilla, at the level of the infraorbital foramen (figure 43). Both taxa do not appear to possess the CAC. Figure 42. Infraorbital canal of Ceratotherium simum; A; right lateral view (scale: 35.23 mm), B; left lateral view (scale: 51.88 mm), C; dorsal view (scale: 49.43 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen; MAC, middle alveolar canal, RAC; rostral alveolar canal. 56 | Page Figure 43. The infraorbital canal of Diceros bicornis; A; dorsal view (scale: 99.31 mm), B; left lateral view (scale: 19.03 mm), C; left lateral view (scale: 20.47 mm), all transparent except for C. IOC; infraorbital canal, IOF; infraorbital foramen; MAC, middle alveolar canal, RAC; rostral alveolar canal. C; right anterior-lateral view showing the infraorbital foramen in an opaque skull, D; dorso-ventral view showing the infraorbital foramen in a transparent skull (scale: 81 mm). Tapiridae Of all the proboscis-bearing taxa, the tapirs possess the most complex infraorbital canal condition, having more branches (especially T. indicus, figure 44). The gross infraorbital canal in tapirs is thicker caudally than rostrally, and narrows and increases in complexity rostrally as the number of branches increases. The infraorbital canal in T. indicus is much thicker than that of T. terrestris (figure 45), and bears more branches. Its overall shape is conical and broader caudally, and then 57 | Page narrows rostrally. Tapirus terrestris bears a relatively simple canal morphology, with fewer branches. It possesses a thick, short RAC, with a uniform tubular shape compared to T. indicus (figure 45). Both taxa do not appear to possess the MAC and CAC branches. Figure 44. The infraorbital canal in Tapirus indicus; A; right lateral view (scale: 14.41 mm), B; dorsal view (scale: 19.03 mm), C; left lateral view (scale: 20.47 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital canal; MAC, middle alveolar canal, RAC; rostral alveolar canal. 58 | Page Figure 45. The infraorbital canal in Tapirus terrestris; A; right lateral view (scale: 14.41 mm), B; dorsal view (scale: 19.03 mm), C; left lateral view (scale: 20.47 mm), all transparent. IOC; infraorbital canal, IOF; infraorbital foramen; MAC, middle alveolar canal, RAC; rostral alveolar canal. The purple structure in C is the brain. II - Statistics The three proposed osteological proxies (i.e. volume of the infraorbital canal, surface area of the infraorbital foramen, and degree of narial retraction) were quantified and simple linear regressions were generated on the “all taxa” sample (Table 3). The first generated model (model 1, SL/Hl vs Surface area IOF) has the most significant P–value of the three models tested on the “all taxa” sample (P–value = 9.8E-05); however, its r² is low = 0.339. The degree of narial retraction is the second most significant model with a P-value = 0.018. Interestingly, while its P-value is significant, its r² is very low (0.141). The third generated model (model 3, Sl/Hl vs Volume IOC) has a non- significant P-value = 0.57 and a low r² = 0.094. Overall, these generated models are only weakly predictive, and thus application to the fossil samples was not attempted. 59 | Page Simple linear regressions were generated and tested on taxa that possess a trunk or proboscis-like snouts (Table 4). The 4th generated model (model 4, SL/Hl vs Surface area IOF) was found to be the most significant, with a P-value = 1.73E-05. The r² was the highest of all the generated models, with a value = 0.98 (best correlated, figure 46). The 5th model (Narial retraction) was again found to be significant with a P-value = 0.006 although its r² value was very low = 0.233. The last generated model (narial retraction, model 6, Volume IOF) is non-significant with a P-value = 0.182. The r² was very low = 0.25, but higher than that of the narial retraction model for the taxa that possess proboscis-like snouts. Only the model using IOF surface area in taxa bearing a proboscis-like snout proved to be both significant and has a high determination quotient. It was thus applied to the proboscidean fossils because it has the best predictive power (highest r² and most significant P-value). The estimated ratio of the Snout length to Head length for the proboscidean fossil taxa are shown below (Table 5). The estimated ratios were obtained from model 3, and they are as follows: Numidotherium koholense = 0.8477; Deinotherium bozasi = 0.98; Mammut americanum = 1.622; G. indet (KNM- WS12874) = 0.69; G. indet (KNM-WS12660) = 0.78; Anancus kenyansis = 0.8; Elephas recki = 0.98; Loxodonta africana (KNM-Juvenile 1) = 0.76; Loxodonta africana (KNM-Juvenile 2) = 0.86; Loxodonta sp. = 1.13. The estimated snout lengths are as follows: Numidotherium koholense = 31.36 cm; Deinotherium bozasi = 113 cm; Mammut americanum = 158.47 cm; G. indet (KNM- WS12874) = 62 cm; G. indet (KNM-WS12660) = 70 cm; Anancus kenyansis = 56 cm; Elephas recki = 76.88 cm; Loxodonta africana (KNM-Juvenile 1) = 55 cm; Loxodonta africana (KNM- Juvenile 2) = 62.12 cm; Loxodonta sp. = 81.144 cm. 60 | Page Figure 46. The Snout length/Head length (pictures) vs Surface area of the infraorbital foramen/ Skull length (CT scan) in proboscis-bearing taxa, n = 8. 61 | Page Table 3. Summary of the statistics results in all taxa, n = 43. Regression Model Slope standard error a/b r² r P(slope) Intercept b 1) Snout length/Head length (image) vs Surface area of IOF/ Skull length (CT scan) 0.963 0.221 / 0.21 0.339 0.582 9.98E-05 -0.132 2) Snout length/Head length vs Narial retraction 0.003 0.001/0.122 0.141 0.376 0.018 0.563 3) Snout length/Head length vs Average IOC Volume 2.44E-05 1.24E-05/ 0.097 0.094 0.306 0.057 0.689 Table 4. Summary of the statistics results in the taxa bearing a proboscis-like snout, n = 8. Regression Model Slope standard error a/b r² R P(slope) Intercept b 4) Snout length/Head length (image) vs Surface area of IOF/ Skull length (CT scan) 0.003019 0.0012229/ 0.12191 0.98089 0.9904 1.73E-05 0.66446 5 Snout length/Head length vs Narial retraction 3.75E-08 0.0046737/ 0.67225 0.23255 0.48223 0.005753 0.71657 62 | Page 6) Snout length/Head length vs Average IOC Volume 7.87E-05 5.07E-05/ 0.32494 0.25 0.57004 0.182 0.8265 Table 5. Fossil application results Regression Model Estimated Snout length/Skull length, Numidotherium koholense Estimated Snout length/Skull length, Deinotherium bozasi 4) Snout length/Head length (image) vs Surface area of IOF/ Skull length (CT scan) 0.8477 0.98 Estimated length of the trunk in cm 31.36 113 Regression Model Estimated Snout length/Skull length, Mammut americanum Estimated Snout length/Skull length, Gomphotheriidae indet (KNM-WS12874) 4) Snout length/Head length (image) vs Surface area of IOF/ Skull length (CT scan) 1.622 0.69 Estimated length of the trunk in cm 158.47 62 Regression Model Estimated Snout length/Skull length, Gomphotheriidae indet (KNM-WS12660) Estimated Snout length/Skull length, Anancus kenyansis 63 | Page 4) Snout length/Head length (image) vs Surface area of IOF/ Skull length (CT scan) 0.774 0.8 Estimated length of the trunk in cm 70 56 Regression Model Estimated Snout length/Skull length, Elephas recki Estimated Snout length/Skull length, Loxodonta sp. 4) Snout length/Head length (image) vs Surface area of IOF/ Skull length (CT scan) 0.98 1.127 Estimated length of the trunk in cm 76.88 81.144 64 | Page Discussion The general morphology of the infraorbital canal in the taxa studied here is consistent with the general mammalian pattern described by Benoit et al., (2016). The morphology of the canal in afrotherians is relatively short, and the interspecific variation amongst these taxa ranges from short and conical, to short and cylindrical. In contrast to the afrotherians, the artiodactyls possess a thin, elongated and tubular infraorbital canal (i.e. Bovidae). The results show that Hippopotamidae possesses the most voluminous (relative to head size) and complex canal morphology (many branches) within the Ruminanta, followed by Giraffidae, Suoidea and Tylopoda. The gross morphology of the canal in Perissodactyla varies greatly between taxa. The overall shape of the infraorbital canal in equids is similar to that in bovids, but the morphoogy becomes more complex, in terms of the number of branches, in the Rhinocerotidae and Tapiridae. The comparative description of these mammalian taxa reveals that there is a strong taxonomic and phylogenetic signal in the morphological complexity of the infraorbital canal. The internal branching morphology appears to increase in complexity in the taxa with an enlarged and mobile upper lip. The infraorbital canal morphology appears to vary across taxa, and there are notable differences between the infraorbital canal of the proboscis-bearing taxa when compared to those without. The taxa that possess a true proboscis and those that bear proboscis-like structure have a larger, cylindrical canal, with a more spherical infraorbital foramen (i.e. elephants, saigas, elephant shrew’s, dik-dik and tapirs). In addition, the canal of these taxa has fewer branches in comparison with taxa that have a short proboscis and fleshy mobile upper lip since those taxa with a mobile upper lip (e.g. Rhinocerotidae, Hippopotamidae) have a highly ramified canal. It appears that there is a trade-off between the volume and length of the canal, such that the canal is long in taxa that 65 | Page have a narrow canal, whereas taxa with a large canal, such as Mammut americanum and Loxodonta africana, have a shorter canal. The taxa that have a mobile upper lip such as rhinoceroses and hippopotami have more branches (especially the RAC), or multiple infraorbital foramina in comparison to species with a proboscis-like structure, which tend to have a simpler, more cylindrical and shorter canal. Noticeably, tapirs have a relatively voluminous canal, but more complex than that of elephants (many branches), and appear thus intermediate in this respect. These qualitative observations are consistent with the literature that hypothesized that a large dimensions of the infraorbital canal and foramen are expected to be associated with the presence of a proboscis-like structure or mobile upper lip (Adam & Berta, 2002; Galton et al., 2006; Shoshani, 1998; Troxell, 1921; and Wall, 1980). The statistical results also support this observation, as there is a significant correlation between the length of the snout and the surface area of the infraorbital canal (table 3). This correlation is even more significant and the determination quotient higher when considering taxa with a proboscis or proboscis-like snouts alone (table 4). Many authors, who have attempted to study, infer and reconstruct the presence of the trunk in extinct taxa, used the degree of narial retraction (Adams et al., 2019; Blanco et al., 2021; Burmeister, 1864; Osborn, 1898; Scott, 1910; Sharp & Trusler, 2016; Shoshani & Tassy, 2005; Troxell, 1921; and Wall, 1980) to support their reconstructions. The statistical results provided above (tables 3 and 4) challenge this long-standing view, as the regression lines between snout length and narial retraction are found to be statistically significant, but the coefficients of determination are very low (tables 3 and 4). This implies that the dispersion around the regression lines is too low, both in all taxa and taxa bearing a trunk-like structure only, to enable a reliable prediction of the length of the snout. This result contradicts more than a century of untested assumptions (e.g. Burmeister, 1864; Osborn, 1898; Scott, 1910; and Wall, 1980). 66 | Page It was found that the length of the snout is strongly correlated to the surface area of the infraorbital foramen. Compared to the other two proposed proxies, this proxy shows the best p-values and determination quotient, particularly if considering the analysis performed with taxa bearing proboscis-like structures only. Using the equation of this regression line, the estimated snout length for Numidotherium is 31 cm, which is short relative to estimated shoulder height and within the tapir average range (29 - 42 cm). Numidotherium has a body mass and an estimated shoulder height = 90 - 100 cm, which are both similar to that of tapirs. This finding supports the literature that proposes a tapir-like proboscis for Numidotherium (Larramendi, 2016; Shoshani, 1998; Shoshani & Tassy, 2005; and Sukumar, 2003). Contrary to what the literature above has stated (Harris, 1975, Markorv et al., 2001, Svistun, 1974, Tassy, 1998; and Tarabukin, 1974), Deinotherium was estimated to possess a trunk-length of 113 cm. This trunk condition is longer than that of tapirs, but not in the range of extant elephants (167 - 243 cm). This finding supports the reconstruction proposed by Abel (1922), with a more elephantine length for the trunk. Note: I disagree with the sentence above. Using Larramendi’s figure, I estimate that the trunk length of a 360 cm tall Deinotherium bozasi (Larramendi, 2016 pp: 556) would have had a 130 cm trunk according to the usual “tapir-like hypothesis”. As such, a 113 cm long trunk is in fact quite short, and perfectly in line with Markhov’s hypothesis. Mammut was estimated to possess a snout length of 158.47 cm, which is almost within the range of extant elephants, and closer to that of Deinotherium than to the estimated gomphothere length. The estimated shoulder height for M. americanum is 289 cm (Larramendi, 2016 pp: 571) and this estimated trunk length is quit long as it makes up more than half the limb length (and almost within the elephantine range). M. pacificus has an estimated shoulder height of (182 – 244 cm) 67 | Page (Yeagar, 1989) and this trunk length is almost as long as the lower boundary of this limb length range. The gomphotheres have estimated shoulder height 230 - 320 cm (Larramendi, 2016) and an estimated snout length of 62 – 70 cm. When using the shoulder height 230 cm, both estimated snout lengths do not make up half the limb length. Anancus has an estimated shoulder height = 310 – 350 cm (Larramendi, 2016) and the estimated trunk length is very short (the ratio of limb to trunk length is higher than that of tapirs and Numidotherium). Elephas recki has an estimated shoulder height = 540 cm (Larramendi, 2016), and the estimated trunk length is also very short compared to shoulder height (possesses the highest limb to trunk length ratio, of all the fossils in the study). These findings support the proposed snout lengths for basal elephantiforms such as Palaeomastodon (shoulder height = 220 cm; Larramendi, 2016), and challenge that of more derived taxa (i.e. gomphotheres, usually more elephantine) (Andrews, 1904; Nabavizadeh, 2015; Nabavizadeh & Reidenberg, 2019; Osborn, 1936, 1942; and Sanders et al., 2010). Results show that this model has strong predictive power, and its findings are consistent with the previous qualitative estimates for Plesielephantiformes. However, this model challenges the qualitative estimates for the trunk lengths of elephantiforms, especially that of Anancus and Elephas recki. Based on the findings of this study, we now have a quantitative metric protocol that can be used to infer the presence, and estimate the dimensions of the trunk in fossil taxa (i.e. Model 4). We now know that ‘Plesielephatiforms’ such as Numidotherium possessed a snout length much shorter than that of tapirs, but longer than a mobile upper lip for a similar overall body shape and size. The most derived “plesielephantiforms”, the deinotheres, had a tapir-like trunk. The estimated trunk length of gomphotheres does not make half the limb length, however, still longer than that 68 | Page of tapirs. The estimated snout lengths for Anancus and Elephas recki are very short compared when compared to their shoulder heights (shortet than that of tapirs and Numidotherium). The trunk did not originate as a single event at the root of the proboscidean tree, but is the result of a long evolution that spanned the proboscidean phylogenetic tree for at least 60 million years between the last common ancestor of Proboscidea and that of Sirenia. More proboscidean fossils are needed to investigate this research question, such as Eritherium, Phosphatherium, Daouitherium and Moeritherium for us to gain insight as to when exactly the trunk originated and elongated. For the gomphotheres, the actual head lengths must be used instead of the average skull length for the family in order to increase the reliability of these results. In addition, more extant taxa such as proboscis monkeys and elephant seals are needed to increase the sample size of the taxa who possess proboscis-like snouts, and strengthen these regression models. Conclusion This research aimed to investigate the evolutionary origins of the trunk in Proboscidea, quantify and test proxies that have been hypothesized to be associated with the presence of the trunk in a statistical framework, and to create a quantitative protocol that can be used to reconstruct the evolution of the trunk in the fossils record. The results indicate that there is a significant and strong correlation between the surface area of the infraorbital foramen with the dimensions of the snout (in taxa with proboscis-like structures, particularly). The results also challenge the longstanding hypothesis that narial retraction is a reliable proxy for reconstructing the presence of a trunk. It can be concluded that there is a strong taxonomic and phylogenetic signal in the condition and morphology of the infraorbital canal. 69 | Page Acknowledgements I thank the DSI-NRF Centre of Excellence in Palaeosciences (CoE-Pal) and The Palaeontological Scientific Trust (PAST), Johannesburg South Africa, for financial support, my supervisors Dr. Julien Benoit and Prof. Paul Manger for academic support, the Evolutionary Studies Institute (ESI) at the University of the Witwatersrand and the National Museum of Nairobi, Kenya. 70 | Page References Abel, O. (1922). Lebensbilder aus der Tierwelt der Vorzeit. Jena. Abraha, M., Berhe, S., Ghirmai, T., Libsekal, Y., Marchant, G. H., Sanders, J. W., ... Zinner, D. …..(2007). A Proboscidean From the Late Oligocene of Eritrea, a ‘‘Missing link’’ Between Early …..Elephantiformes and Elephantimorpha, and Biogeographic Implications. Proceedings of the …..National Academy of Sciences, Vol. 103, 46: 17296–17301. Adaci, M., Bensalah, M., Hartenberger, J. L., Jaegar, J. J., Mahboubi, M., Marivaux, L., … …..Tafforeau. (2007). Early Tertiary Mammals From North Africa Reinforce the Molecular …..Afrotheria Clade. Proceedings of the Royal Society of Britain, 274: 1159–166. Adam, P. J., and Berta, A. (2002). Evolution of Prey Capture Strategies and Diet in the …..Pinnipedimorpha (Mammalia, Carnivora). Oryctos, 4: 83–107. Adams, J. W., Evans, A. R., Fitzgerald, G. M., Richards, H. L. and Wells, R.T. (2019). The …..Extraordinary Osteology and Functional Morphology of the Limbs in Palorchestidae, a …..Family of Strange Extinct Marsupial Giants. Public Library of Scince ONE, 14 (9): e0221824. Alberdi, M.T., Donato, M., Ortiz-Jaureguizar, E., Prado, J.L., and Posadas, P. (2011). Paleo- …..Biogeography of Trilophodont Gompotheres (Mammalia: Proboscidea). Areconstruction …..Applying DIVA (Dispersion-Vicariance Analysis). Revision of Mexican Ciencias Geology. …..Vol. 28, 2: 235e245. Amaghzaz, M., Bouya, B., Cappetta, H., Gheerbrant, E., Iarochène, M., and Sudre, J. (2002). A …..New Large Mammal From the Ypresian of Morocco: Evidence of Surprising Diversity of …..Early Proboscideans. Acta Palaeontologica Polonica, 47: 493–506. Amaghzaz, M., Bourdon, E., Bouya, B., Cappetta, H., Gheerbrant, E., Iarochène, M., …Sudre, 71 | Page …..J. (2003). Les Localités à Mammifères des Carrières de Grand Daoui, Bassin des Ouled …..Abdoun, Maroc, Yprésien: Premier état des lieux. Bulletin de la Société Géologiquede …..France, 174: 279–293. Amaghzaz, M., Bouya, B., Gheerbrant, E., Iarochène, M., Tassy, P. and Sudre, J. (2005). …..Nouvelles Données sur Phosphatherium escuilliei (Mammalia, Proboscidea) de l’Éocène …..inférieur du Maroc, Apports à la Phylogénie des Proboscidea et des Ongulés Lophodontes. …..Geodiversitas 27: 239–333. Ameur, R., Crochet, J. Y., Jaeger, J. J., and Mahboubi, M., (1984). El Kohol (Saharan Atlas, …..Algeria): A New Eocene Mammal Locality in Northwestern Africa. Palaeontographica …..Africa, 192: 15–49. Andrews C. W. (1904IV). On the Evolution of the Proboscidea. Philosophical Transition of the …..Royal Society of London, B196: 99–118. Benoit, J. Manger, P. R. and Rubidge, B. S. (2016). Paleoneurological Clues to the Evolution of …..Defining Mammalian Soft Tissue Traits. Cenozoic. Scientific Reports, 6: 25604. Benoit, J., Legendre, J. L., Manger, P., Mararescul, V., Obada, T. and Tabuce, R. (2019). Brain …..Evolution in Proboscidea (Mammalia, Afrotheria) Across the Cenozoic. Scientific Reports, 9: …..9323. Benoit, J., Cousteur, S., Farke, A., Legendre, J. L., A., Mennecart, B., Merigeaud, S., …Neenan, …..J. .M. (2020). A Test of the Lateral Semi-circular Canal Correlation to the Head Posture, Diet …..and Other Biological Traits in “Ungulate” Mammals. Scientific Reports, 10: 19602. Bennett, C. N., Chiminba, C. T. and Medger, K. (2012). Seosonal Reproduction in the Eastern …..Rock Elephant–Shrew: Influenced by Rainfall and Ambient Temperature? Journal of Zoology, ….288:.283-293. Berenbrink, M., Burns, J. M., Campbell, L. K., Cossins, A. R., Mirceta, S. and Signore, A. V. ….. 72 | Page ….. (2013). Evolution of Mammalian Diving Capacity Traced by Myoglobin Net Surface Charge. …..Science, 340: 1234192. Blanco, E. R., Jone, W. W., Rinderknecht, A. and Yorio, L. (2021). Macrauchenia patachonica …..Owen, 1838: Limb Bones Morphology, Locomotory Biomechanics, and Paleobiological …..Inferences. Geobios, 30 (30): 30. Boas, J. E. V., and Paulli, S. (1908). The Elephant’s Head. Studies in the Comparative Anatomy …..of the Organs of the Head of the Indian Elephant and Other Mammals. Part I. Gustav Fischer, …..Jena. Branstrator, J., W. and Woodman, N. (2008). The Overmyer Mastodont (Mammut ameticanum) …..from Fulton County, Indiana. The American Midland Naturalist, Vol. 159 (1): 125-146. Brunet, M., Hautier, L., Lihoreau, F., Mackaye, H. T., Tassy, P. and Vignaud, P. (2009). New …..Material of Anancus kenyansis (Proboscidea, Mammalia) From Toros-Menalla (Late Miocene, …..Chad): Contribution to the Systematics of African Anancines. Journal of African Earth …..Sciences, 53: 171–176. Bryden, M. M. and Ling, J. K. (1992). Mirounga leonina. Mammalian Species, 391: 1–8. Burckle, L. H., Denton, G. H., Partridge, C. T. and Vrba, E. S. (1995). Paleoclimate and Evolution, …..with Emphasis on Human Origins, Yale University Press, New Haven, CT. Burmeister, H. (1864). Descripción de la Macrauchenia patachonica. Anales del Museo …..Público de Buenos Aires, 1: 32–66. Cawthra, C. H., Dixon, G. M., Helm, C. W., Lockley, G. M., Moolman, L., Stear, W., …Vyncka, …..C. D. J. (2021). Morphology of Pleistocene Elephant Tracks on South Africa’s Cape South …..Coast and Probable Elephant Trunk-drag Impressions. Quaternary Research, 1–15. Clifford, A. B. and Witmer, L. M. (2002b). Not All Noses are Hoses: An Appraisal of 73 | Page …..Proboscis Evolution in Mammals. Annual Meeting of the Society of Vertebrate …..Paleontology, Norman, Oklahoma. Journal of Vertebrate Paleontology, 22 (Supplimentary to …..3): 66A. Clifford A. B, and Witmer, L. M. (2004a). Case Studies in Novel Narial anatomy: 3 Sturcture and …..Function of the Nasal Cavity of saiga (Artiodactyla: Bovidae: Saiga tatarica). Journal of …..Zoology, Vol. 264 (3): 217-230. Clifford A. B, and Witmer, L. M. (2004b). Case studies in Novel Narial Anatomy: 2. The …..Enigmatic Nose of Moose (Artiodactyla: Cervidae: Alces alces). Journal of Zoology 262: …..339-60. Court, N. (1994). Limb Posture and Gait in Numidotherium koholense a Primitive Proboscidean …..From the Eocene of Algeria. Zoological Journal of the Linnean Society, 111: 297-338. Crumpton, N. and Thompson, R. S. (2013). The Holes of Moles: Osteological Correlates of the …..Trigeminal Nerve in Talpidae. Journal of Mammal Evolution, 20: 213–225. Cuvier, G. (1796). Sur le Squelette d’une Très-grande Espèce de Quadrupède Inconnu Jusqu’à …..pPésent, Trouvé au Paraguay, et Déposé au Cabinet D’histoire Naturelle de Madrid. Magasin …..Encyclopédique, 1: 303–310. Dechaseaux, C. (1958). L’encephale d’Elephas meridionalis. Annales de Paleontologie, 54: …..269–278. Delmer, C. (2005). Les Premières Phases de Différenciation des Proboscidiens (Tethytheria, Mammalia): …..Le Rôle du Barytherium Grave de Libye. Unpublished PhD Dissertation, Muséum National d’Histoire …..Naturelle, Paris, 470 pp. Delmer, C. (2005b). Phylogenetic Relationships Between Paleogene Probosci− Deans: New Data …..and Impact on Tethytherian Systematics. Journal of Vertebrate Paleontology, Supplement 74 | Page …..25: 50A. Delmer, C. (2005c). Early Proboscideans Phylogeny and Systematics: The Role of Barytherium …..Grave (Proboscidea; Mammalia). Abstracts of the Plenary Symposium, Poster and Oral …..Papers Presente