New species of Ontocetus (Pinnipedia: Odobenidae) from the Lower Pleistocene of the North Atlantic shows similar feeding adaptation independent to the extant walrus (Odobenus rosmarus)

Systematic paleontology

The original specimen (NWHCM 1996.1) was photographed by the last author on June 29, 2000 at the Norwich Castle Museum (NWHCM). The cast (NMR 7472) was photographed by the senior author on March 3, 2020 at the Institut Royal des Sciences Naturelles de Belgique (IRSNB). Both mandibles (NMR 7472, IRSNB M156) were taken via a Nikon D800E with AF-Nikkor 50 mm F1.8 lens and a Panasonic Lumix G1 with Olympus 14–42 mm lens. The figures were performed in Adobe Photoshop CS6 and Adobe Illustrator CS6. The anatomical terminology used in the present study follows Deméré (1994a, 1994b), Kryukova (2012), Boessenecker & Churchill (2013), Velez-Juarbe (2017) and Magallanes et al. (2018) for the walrus and Evans & de Lahunta (2013) for the domestic dog, as a representative for the (caniform) Carnivora. Terminologies for the mandibular teeth have been abbreviated (incisors = i, canine = c, premolars = p, first premolar = p1, second premolar = p2, third premolar = p3, fourth premolar = p4, molars = m). Terminologies for the musculature and inference of muscle insertions (areas) are inspired by Valentin (1990), Kastelein, Gerrits & Dubbeldam (1991), Lavergne, Vanneuville & Santoni (1996), Naples, Martin & Babiarz (2011) and Tseng et al. (2011).

This study includes the following morphological measurements: the minimum mandible thickness (MT) and the least mandible depth (MD), following Wiig et al. (2007). The MT is the minimum transverse thickness of the mandible posterior to the last postcanine; the MD is the minimal vertical height of the mandible posterior to the last post-canine. MT and MD have been measured using a digital caliper with an accuracy of 0.01 mm. Five angular measurements have also been taken, following Mohr (1942) and Deméré (1994a) and showing in Boisville et al. (2022): angle between (a) the anterior and dorsal margin; (b) the anterior and ventral margins; (c) the ventral and dorsal margins; (d) the medial edge of the condyle and the mandibular symphysis; (e) the lateral edge of the mandibular terminus and the symphysis; f) the horizontal and vertical rami; and (g) the coronoid process and the mandibular condyle. Other mandibular and tooth measurements follow Erdbrink & van Bree (1986, 1990), Deméré (1994b), Kohno & Ray (2008) and Magallanes et al. (2018). All measurements are provided as Table S1.

Morphometry

Multiple measurements were taken capturing the overall shape and proportions of the lower dentition and mandible (a total of 31 linear measurements and seven angles) to compare the new species to Ontocetus emmonsi and Odobenus rosmarus. All morphometric analyses were run using the R statistical environment v 4.2.3 (R Development Core Team, 2023). The R script used to run the analysis is provided in the Supplemental Data 1. We applied a 50% completeness threshold, the female Ontocetus posti (RGM.St.119589) did not pass the threshold as well as two specimens of Ontocetus emmonsi (female USNM PAL 374273 and male ROM 26116) due to their preservation stage. The morphological dataset was imported using the ‘read.csv’ function and then scaled (z-transform), distance matrices (based on pairwise dissimilarities) were computed from these scaled datasets using the ‘dist’ function from the stats v4.3.0 package. We generated morphospaces using: a Principal Coordinates Analysis (PCoA), using the ‘pcoa’ function implemented in the ape package (Paradis & Schliep, 2019) and a cluster using the hclust function from the stats v4.3.0 package. The PcoA retrieved 13 axes, the first two explaining a total of 65.36% of variance (43.02% and 22.34% respectively). Finally, a permutational multivariate analysis of variance (PerMANOVA, formerly known as non-parametric/NP-MANOVA) (Anderson, 2001) was performed using the ‘adonis2’ function from the vegan package (Dixon, 2003). PerMANOVA was performed (1,000 permutations using the ‘euclidean’ method) on the distance matrix of the ratios to test for significant differences between the new species and Ontocetus emmonsi, but also between the new species and Odobenus rosmarus.

Phylogeny

Our phylogenetic analyses are based on the morphological character matrix published by Boessenecker et al. (2024) including fossil and extant Odobenidae with various outgroup taxa were also selected from other pinniped families, including two extinct desmatophocids (Desmatophoca oregonensis Condon, 1906 and Allodesmus kernensis Kellogg, 1922), two early diverging “enaliarctines” (Enaliarctos emlongi Berta, 1991 and Pteronarctos goedertae Barnes & Perry, 1989), one otariid (Callorhinus ursinus), and two extant phocids (Erignathus barbatus Erxleben, 1777 and Monachus monachus Hermann, 1779). We used Mesquite 3.81 (Maddison & Maddison, 2023) to add the new species as a new OTU (Operational Taxonomic Unit), however, considering the material described we could only score a limited number of characters (characters 77–90, 94–95, 99–106, 122–129; see Files S1 and S2).

Some modifications were made from the matrix of Boessenecker et al. (2024). First, characters 78 and 82 were fused in a new character 78 describing both the mandibular symphysis and the potential presence of a furrow: 0 = absent, 1 = present with furrow, 2 = present without furrow. Character 81 was modified to 0 = same level or anterior to p2, 1 = posterior to level of p2 (the polymorphic state was removed and polymorphic specimens were coded as 0 and 1), same for character 83. The polymorphism as an independent state was excluded because sexual dimorphism may affect the coding of character states. Male character states are generally more definitive compared to those of female individuals, because of hypermorphotic nature in males. Therefore, if possible, using male specimens is preferable for coding characters. Characters 81, 83 (now 82), 99 (now 98), 127 (now 126) were treated as unordered in order to not exclude independent evolutionary changes and possible convergent evolution. We also included two new characters: Character 144 describing the p1 as follows (0 = present, 1 = absent) as well as character 145 describing the direction of coronoid process projection (0 = vertical, 1 = posterodorsally projected).

We used TNT v 1.5 (Goloboff & Catalano, 2016) to perform the phylogenetic analysis under implied weighting using a script modified from Chatar, Michaud & Fischer (2022). We expanded the memory of TNT to a maximum of 100,000 trees. We set the search parameters at: New Technology Search, 200 ratchet iterations, 10 cycles of drifting, five hits and five replications for each hit. We then used the tree branch bisection and reconnection algorithm (TBR) to fully explore the tree islands identified by the ratchet. Nodal support was measured using symmetric resampling with 1,000 replications, each replication involving a New Technology search with a change probability of 33%. We choose symmetric resampling over bootstrapping or jack-knifing, as this measure is not affected by character weighting and is thus more appropriate to deal with implied weights (Goloboff et al., 2003) (see File S2). The best score was obtained when K was set to 12 so we computed the time calibration on the strict consensus cladogram for K = 12. To do so, we used the ‘equal’ method of the ‘timePaleoPhy’ function from the strap v1.4 package in R (Bell & Lloyd, 2015) and a table containing the biostratigraphic range of each OTU obtained from Dewaele, Lambert & Louwye (2018a), Magallanes et al. (2018), Boessenecker et al. (2024). We then generated the time tree using the ‘geoscalePhylo’ function from the paleotree v3.3.25 package (Bapst, 2012).

Generic allocation

Our phylogenetic analysis presented an odobenin with mosaic characteristics, exhibiting traits assignable to Ontocetus such as (1) a lower canine markedly larger than the cheek teeth (2) and it is rather separated from the cheek toothrow; (3) the presence of the lower first premolar (p1) with four lower postcanines considered as p1–p4; (4) a broad and short edentulous mandibular terminus with two lower incisors (i2, i3); (5) an elevated and slanted anterior margin, and (6) posterodorsally projected coronoid process. It also displayed characteristics of Odobenus, such as (1) the fusion of the mandibular symphysis; (2) an oval-shaped mandibular symphysis that is shorter in proportion to the total length of the mandible; (3) a horizontal ramus with a well-curved lateral occlusal outline; and (4) a shorter space between the canine and the cheek teeth. Morphometric results suggested that this mosaic-character combination within the odobenins positioned significantly closer to Ontocetus than to Odobenus (Fig. 7). Phylogenetic analyses suggested that our specimens are sister to the Valenictus + Odobenus clade (Fig. 8). These results indicate that On. posti might be more closely related to the later diverging species including Valenictus and Odobenus, potentially indicating an ancestral morphology for the first representative of Valenictus + Odobenus clade, absent from the North Atlantic during the Early Pleistocene. Although our result did not directly support parallel evolution of On. posti with Odobenus, it is apparent from previously proposed odobenin synapomorphies that such derived characters of On. posti shared with Valenictus + Odobenus clade could be considered to be strongly related to their suction-feeding behavior. Thus, while the present analysis recognizes morphological synapomorphies for On. posti and Valenictus + Odobenus clades, potential homoplasies for such characters directly related to their feeding behavior within the clade limit their diagnostic value.

By contrast, the oldest records of the genus Odobenus correspond to an isolated tusk from the Pliocene-Pleistocene Shiranuka Formation of Japan (Kohno et al., 1995) and a nearly complete Late Pliocene skull from the bottom of the Sea of Okhotsk (Miyazaki, Kimura & Ishiguri, 1992). Mandibles assigned to Odobenus mandanoensis (ca. 0.5 Ma) (Tomida, 1989), Odobenus rosmarus (ca. 0.7 Ma) (Kimura et al., 1983; Ishiguri & Kimura, 1993), and a partial skull from the Yabu Formation (Chiba Prefecture) from the Middle Pleistocene strata (Miyazaki et al., 1995) have also been recorded from the North Pacific realm. Although some cervical vertebrae of Odobenus sp. have been mentioned by Brigham (1985) from the Gubik Formation (Alaska), it is unclear to estimate the origin of these remains, as the author seems to mention several sections for Odobenus remains, from Aminozone 2 (ca. 0.5 Ma) and/or Aminozone 3 (1.7–0.7 Ma), and does not provide any illustration of vertebrae. The oldest known specimen assigned to Odobenus rosmarus outside the North Pacific comes from Northern European Russia (Pechora River), dated to the end of the Middle Pleistocene (Ponomarev et al., 2023), potentially suggesting a dispersion outside the North Pacific at least during the latter half of the Middle Pleistocene. In contrast, Ontocetus was already present along the western coasts of the North Atlantic during the Early Pliocene (Berry & Gregory, 1906; Kohno & Ray, 2008), with occurrences in the North Sea in the early Late Pliocene (e.g., Lankester, 1865; du Bus, 1867; van Beneden, 1871; Hasse, 1909) until the Early Pleistocene (Erdbrink & van Bree, 1990). Remains assigned to Ontocetus have been found in the late Late Pliocene Moroccan deposits of Ahl al Oughlam (Geraads, 1997), and a final population may have survived on the east coast of the United States until about 1.1 Ma (Boessenecker, Boessenecker & Geisler, 2018). Considering these elements, and pending more field exploration of Pliocene–Middle Pleistocene fossil localities within the Arctic Circle, particularly the Gubik Formation (Alaska), but also more odobenin material from the Early Pleistocene of the North Sea, it is preferable to assign these specimens to Ontocetus, presenting an independent evolutionary convergence of suction-feeder with Odobenus.

Functional interpretations toward suction-feeding specialization

While the first odobenids were piscivorous, they rapidly evolved toward molluscivory (Adam & Berta, 2002), with some taxa, like Valenictus chulavistensis, exhibiting an extreme reduction of dentition except for the upper canines just as the extant walrus (Deméré, 1994a). Ontocetus, previously considered a molluscivore (Deméré, 1994a, 1994b; Adam & Berta, 2002; Boessenecker, Boessenecker & Geisler, 2018), however, exhibits features suggesting less specialization, making its diet a bit more complex to assess. Odobenus rosmarus exhibits a unique combination of characters that optimize its ability to engage in suction-feeding (Kastelein & Gerrits, 1990; Kastelein, Gerrits & Dubbeldam, 1991). NWHCM 1996.1 and IRSNB M156 exhibit characteristics specific to Ontocetus, such as a significantly larger lower canine compared to the cheek teeth, four post-canine teeth, and incisors (i2). However, some features of those specimens are also found in Odobenus including mandibular fusion and shortening, oval mandibular arch, less developed and flatter mandibular condyle, and thin septa between the various teeth. The robustness of the mandibular symphysis is directly associated with its condition, subjected to torsional forces experienced during feeding (Adam & Berta, 2002). Mandibular fusion, concomitant with symphyseal size reduction, serves as a preventive measure against substantial mandibular deformation, particularly during tongue retraction in suction-feeding scenarios (Gordon, 1984; Kastelein & Gerrits, 1990; Deméré, 1994a; Adam & Berta, 2002). The fusion of the mandible and the pachyosteosclerotic reinforcement of the anterior part of the mandible further contributes to enhanced resistance, by supporting the large tusks. Suction-feeding specialization is further evidenced by the dorsally concave vaulting of the hard palate in the Odobenini, evident in both transverse and longitudinal planes. This palatal configuration, as proposed by Kastelein & Gerrits (1990), results in an enlarged intraoral space, facilitating a more extensive tongue protraction before retraction during suction-feeding. The curvature and oval shape of the mandible in occlusal view correspond to the arched structure of the hard palate vaulting. Additionally, the shortening of the skull, and consequently of the mandible, enhances the containment of negative intraoral pressure, facilitating the suction of mollusks when water reaches the palate (Werth, 2006a, 2006b).

Concurrently, the reduction in dental series among the Odobenini, as documented by Boessenecker & Churchill (2013), manifests in Odobenus with narrower spacing between teeth compared to Ontocetus. This dental morphology is likely a consequence of Odobenus specialization in suction-feeding, accompanied by diminished reliance on teeth and a relatively free of obstruction feature into the oral cavity that would enhance the efficiency of suction-feeding (Adam & Berta, 2002). Additionally, the reduced and flattened mandibular condyle in Odobenus is indicative of an adaptation to suction-feeding, paralleling the morphological similarity observed in Valenictus, identified as the most specialized suction-feeder among the Odobenini (Deméré, 1994a). This suggests that Ontocetus posti was able to use suction to catch its prey and may have once occupied an ecological niche similar to the extant Odobenus rosmarus in the Pliocene of the eastern North Atlantic before the latter species came into the North Atlantic (Fig. 9).

Suction-feeding is observed independently in various groups of marine mammals, starting with pinnipeds, as in the bearded seal Erignathus barbatus. Nevertheless, its adaptations differ somewhat. The rostrum of the bearded seal is flatter, highly mobile, and capable of diverse shape changes (Marshall, Kovacs & Lydersen, 2008; Marshall, 2016). The bearded seal retains incisors and possesses a relatively flat palate, although the maxillary alveolar processes are expanded, resulting in a concave ventral rostral surface in transverse section. Adam & Berta (2002) propose that Erignathus barbatus may employ suction-feeding, possibly to a lesser extent than Odobenus rosmarus, although this has yet to be confirmed with experimental data. Other pinnipeds, such as the leopard seal Hydrurga leptonyx and the harbor seal Phoca vitulina, employ suction for prey capture (Hocking, Evans & Fitzgerald, 2013; Marshall et al., 2014; Churchill & Clementz, 2016; Kienle & Berta, 2016; Hocking et al., 2017), albeit without undergoing modifications as significant as those observed in the mandible and skull of Odobenus rosmarus. Suction-feeding is also present independently in contemporary toothed whales and dolphins (beaked whales, monodontids and sperm whales), involving the loss of elongated mandibles and skulls, accompanied by a reduction in teeth (Heyning & Mead, 1996; Werth, 2000; Kane & Marshall, 2009). This specific feeding technique is also surprisingly present in baleen whales, in the sole gray whale Eschrichtius robustus, predominantly employing this capability for aspirating prey-laden sediment residing on the seabed. This behavior aligns more closely with suction-filter feeding than a complete suction-feeding specialization (Ray & Schevill, 1974).

In the fossil record, suction-feeding has been proposed with the cetacean Odobenocetops (de Muizon & Domning, 2002), possessing an arched and vaulted palate, distinctive of this feeding specialization. Furthermore, de Muizon & Domning (2002) suggested that tusks also facilitate suction-feeding, acting as a “sled” in order to maintain a useful distance between the gape and the benthic substrate, so that clams or other mollusks can be more efficiently sucked up and harvested from the seafloor. In Odobenus rosmarus, tusks have often been considered as a function of hauling out of water or digging in ice for holes (Kastelein & Gerrits, 1990), however it would appear that tusks were used mainly as secondary sexual characteristics, with diversity of shape and size throughout the evolutionary history of the Neodobenia, linked to mating behaviour (lekking and/or female-defense polygyny) and used as a display and in rarer cases combat (Deméré, 1994a; Boessenecker & Churchill, 2021). Although Fay (1982) demonstrated that extant walruses did not use their tusks for benthic feeding, recent observations of males linked abrasive wear on the anterior surface of tusks with the tusks being dragged along the sediment during feeding in a sledge-like manner (Levermann et al., 2003), while anterior surface spanning the length of the tusks was not observed in any of the known female specimens. This observation corroborates sexually dimorphic feeding behaviors (McLaughlin et al., 2022). According to Born et al. (2003), male individuals predate larger bivalves than females. It may possible that behavioral and ecological sexual dimorphism exist in Od. rosmarus, with females potentially foraging in shallower water, outside same beaching areas than males (Lydersen, 2018). It is quite difficult to estimate the exact function of the tusks in Ontocetus, given that they are shorter and more procumbent than those of Od. rosmarus. In the absence of clear evidence, it is preferable to consider the primary function of the tusks of Ontocetus as secondary sexual characteristics, mainly used by the males for display or combat during female-defense polygyny. However, further studies are necessary to assess whether Ontocetus would have had sexually dimorphic feeding behavior like Od. rosmarus (McLaughlin et al., 2022), especially in view of the potential geographical sexual segregation that would have existed in Ontocetus (Kohno & Ray, 2008). NWHCM 1996.1 demonstrates distinctive muscular insertions, particularly evident on both the horizontal and vertical ramus. Observable variations are discernible among Odobenus rosmarus, Ontocetus posti, and Ontocetus emmonsi (Fig. 10). On the horizontal ramus, the hyoid muscle insertion zones can be identified, bounded ventrally by the mylohyoid line (anteriorly) and the mylohyoid groove (posteriorly), proximate to the mandibular foramen. To some extent, the insertion zone for the mylohyoideus muscle appears broader in Ontocetus compared to Odobenus rosmarus. Specimen preservation enables the observation of inferior mental spines, corresponding to the geniohyoideus muscle insertion, and, to some extent, superior mental spines, corresponding to the genioglossus muscle insertion. Ontocetus appears to possess proportionally more developed genioglossus and geniohyoideus muscles. The digastricus muscle can be estimated, characterized by the presence of a digastric prominence. However, Ontocetus posti exhibits a less pronounced digastric prominence, resembling that of Odobenus rosmarus. On the vertical ramus, the masseteric fossa is notably more pronounced in Ontocetus compared to Odobenus rosmarus. This fossa corresponds to the muscular insertion area of the masseter profundus muscle. Ventral to the condyloid crest, the muscular insertion area of the masseter superficialis appears distinguishable, with Ontocetus emmonsi having a proportionally larger muscular insertion area. Consequently, it is plausible that Ontocetus emmonsi possesses a more developed masseteric muscle compared to Ontocetus posti and Odobenus rosmarus. The pterygoid fossa on the medial face of the coronoid process also appears more pronounced in Ontocetus compared to Odobenus rosmarus. This fossa corresponds to the insertion zone of the temporalis muscle, differentiating into the temporalis pars medialis in the medial zone of the coronoid process and the temporalis pars lateralis located at the apex of the process. Ontocetus also exhibits a proportionally larger zone and, consequently, a more developed temporal muscle than Odobenus rosmarus. Lastly, the insertion zone of the pterygoideus muscle, specifically its medial part (pterygoideus medialis), can be observed. This zone appears slightly smaller in size in Ontocetus emmonsi compared to Odobenus rosmarus and Ontocetus posti. In general, the comparison and interpretation of muscle functions between the fossil record and the present pose challenges. Considering the proportions of muscular insertion zones found on the mandibles of Ontocetus and Odobenus rosmarus, it appears that fossil taxa have slightly larger muscular insertion areas in proportion, potentially indicative of more developed muscles. This suggests a possible moderately lesser specialization toward suction-feeding, given the different proportional changes in muscular insertion zones related to this mechanism.

Although we have attempted to describe these insertion zones, Kienle, Cuthbertson & Reidenberg (2021) conducted a comparative examination of craniofacial musculature in pinnipeds and its role in aquatic feeding. Contrary to their initial hypothesis, approximately half of the biting species in this study (i.e., harbor seals and ringed seals) do not exhibit specific musculoskeletal adaptations for biting or suction-feeding. Moreover, several pinniped species without morphological adaptations for suction have been shown to be extremely capable suction-feeders (e.g., Hocking, Evans & Fitzgerald, 2013; Marshall et al., 2014; Kienle et al., 2018). Thus, it seems that most pinnipeds are opportunistic predators, and their non-specialized craniofacial musculature allows flexible feeding behavior. In light of this information, it is more likely that osteological changes (mandible fusion, reduction of teeth, skull shortening, or vaulted and arched palate) are specific to the walrus lineage in the specialization for suction-feeding, rather than representing a change in craniofacial musculature.

In this regard, Ontocetus emmonsi is inferred to be moderately less specialized in suction-feeding based on observations of a lower canine wider than the cheek teeth, the presence of incisors and the absence of mandibular fusion. On the other hand, some features observed in both Ontocetus posti and the extant walrus such as the presence of a fused and reduced mandibular symphysis and a well-curved mandibular arch, attesting of a shorter skull, may imply that this taxon might have been more specialized in suction-feeding than previously thought for the genus Ontocetus. Indeed, the absence of Odobenus rosmarus left this specialized molluscivore/suction-feeder niche open for Ontocetus in the Early Pleistocene of the North Sea. We hereby suggest that Ontocetus posti might have occupied a similar niche to Odobenus rosmarus. It is still interesting to note that despite those adaptations to suction-feeding more marked than in Ontocetus emmonsi, the mandibular and dental proportions of Ontoceus posti are still much more similar to Ontocetus emmonsi than Odobenus rosmarus (Fig. 7).

Sexual dimorphism in Ontocetus posti

Sexual dimorphism is a key morphological characteristic in pinnipeds, usually associated with their polygynous behavior (Bartholomew, 1970; Kovacs & Lavigne, 1992; Garlich-Miller & Stewart, 1998; Weckerly, 1998; Lindenfors, Tullberg & Biuw, 2002; Ralls & Mesnick, 2009; Jones & Goswami, 2010; Velez-Juarbe, 2017; Mesnick & Ralls, 2018). In Odobenus rosmarus, adult males exhibit conspicuous differences in size and weight compared to adult females, and several studies highlight sex-dependent morphological variations in their crania and mandibles (Mohr, 1942; Fay, 1982; Kovacs & Lavigne, 1992; Garlich-Miller & Stewart, 1998; Boisville et al., 2022). Among the odobenins, Ontocetus emmonsi has been hypothesized to display extreme sexual dimorphism, even more marked than in Od. rosmarus (Kohno & Ray, 2008). Larger individuals (males) are approximately 20–40 percent larger than smaller individuals (females) according to cranial, mandibular and postcranial sizes, based on the sample of the Yorktown Formation (Lee Creek Mine, North Carolina, USA; Kohno & Ray, 2008). Similarly, a clear difference in upper canines indicates distinctions between males (larger and stocky tusks) and females (smaller and slender tusks).

Consequently, it is justified to investigate the sexual dimorphism in Ontocetus posti. The size of mandible as well as the circumference of the lower canines in NWHCM 1996.1 corresponds to the dimensions of male specimens of Ontocetus emmonsi (IRSNB M168, ROM 26116, USNM PAL 475482). Although incomplete, IRSNB M156 also exhibits proportions consistent with the male morphotype of Ontocetus emmonsi, whereas RGM.St.119589 displays proportions much similar to the female morphotype (USNM PAL 9343). It is noteworthy that IRSNB M156 presents measurements larger than any documented Ontocetus. Males and females can be easily distinguished using the symphysis height/symphysis length ratio, with females exhibiting a lower ratio indicative of a slender mandible (Kohno & Ray, 2008). IRSNB M156 exhibits values similar to the male Ontocetus emmonsi, while NWHCM 1996.1 and RGM.St.119589 exhibit values closer to females. Mandibular height (see MD in Table S1) and mandibular width (see MT in Table S1) after the last cheek tooth, indicate that NWHCM1996.1 aligns with male Ontocetus emmonsi values, and IRSNB M156 surpasses these values, suggesting an increased mandibular thickness. The length of the cheek toothrow and the length between the last cheek tooth and ascending ramus (see Table S1) is also sexually dimorphic. NWHCM 1996.1 appears closer to female Ontocetus emmonsi values (USNM PAL 9343), whereas IRSNB M156 exhibits broader proportions, particularly with a longer cheek toothrow than the largest males. Furthermore, transverse width of the mandible at the lower canine is narrower in NWHCM 1996.1 and RGM.St.119589, closer to female Ontocetus emmonsi values (USNM PAL 9343, USNM PAL 374273), while IRSNB M156 reaches male Ontocetus emmonsi values. A narrow width between the lower canines may reflect a narrower space between the upper canines and thus a relatively narrow rostrum, corresponding to a female morphotype in walruses (e.g., Boisville et al., 2022; Boessenecker et al., 2024). Regarding ontogenetic age, NWHCM 1996.1 and IRSNB M156 should be regarded as adult individuals, based on their overall size, complete tooth eruption, tooth wear facets and complete fusion of the symphysis, while RGM.St.119589 should be considered as subadult based on the smaller relative dimensions of its teeth (i.e., thinner cementum of tooth roots) and the presence of a distinctive symphyseal furrow still present on the mandible, indicating that the fusion was not entirely complete.

Based on these comparisons, we identify NWHCM 1996.1 as an adult female, RGM.St.119589 as a subadult female, and due to its extreme proportions, IRSNB M156 is recognized as an old adult male, likely considered the largest Ontocetus individual ever found. Ontocetus posti may have had a significant sexual dimorphism, like Ontocetus emmonsi. However, intraspecific variation with respect to the mandibles of Ontocetus posti appears to be less than Ontocetus emmonsi, approaching that found in Odobenus rosmarus.

Biochronological dispersal and evolution of Ontocetus in the North Atlantic

The youngest record of Ontocetus in the North Atlantic is dated to the Early Pleistocene (1.8–1.1 Ma) in the Upper Waccamaw Formation deposits (Austin Sand Pit, South Carolina, USA; Boessenecker, Boessenecker & Geisler, 2018). Although the Easton Bavents specimen is slightly older (2.2–1.7 Ma), it represents the final occurrence of this genus in the North Sea. The biochronology and dispersal of Ontocetus is well-established, with its initial appearance traced back to the Upper Miocene/Lower Pliocene boundary age Tatsunokuchi Formation (Okamoto & Kohno, 2019) and Lower Pliocene Joshita Formation in Nagano Prefecture, Japan (Kohno, Narita & Koike, 1998). Additional specimens have been uncovered in contemporaneous deposits in Japan (Yasuno, 1988; Kohno et al., 1995). This suggests that Ontocetus originated in the North Pacific; those more primitive forms also seem distinct from those discovered in Lower Pliocene deposits on the United States east coast or in the North Sea (Kohno et al., in prep.).

The dispersal of Ontocetus from the North Pacific to the North Atlantic has traditionally been proposed to have occurred via the Central American Seaway around the Mio-Pliocene boundary (Repenning, Ray & Grigorescu, 1979; Kohno et al., 1995; Kohno & Ray, 2008). Another dispersal route via the Arctic through Bering Strait has also been hypothesized (Deméré, Berta & Adam, 2003; Boessenecker, Boessenecker & Geisler, 2018; Boessenecker & Churchill, 2021). This perspective suggests that Ontocetus evolved during the Late Miocene in the North Pacific before migrating to the North Atlantic. The earliest occurrence of Ontocetus in the North Atlantic comes from the Upper Bone Valley Formation in Florida (e.g., Kohno & Ray, 2008; Boessenecker, Boessenecker & Geisler, 2018). Subsequent deposits presenting Ontocetus appear in a geochronostratigraphic succession (Yorktown Formation in North Carolina and Virginia; Lower Tamiami Formation in Florida; Goose Creek Limestone and/or Raysor Formation in South Carolina; Pinecrest Beds of the Tamiami Formation in Florida) at similar or more northern latitudes, potentially indicative of a distribution that expanded from south to north (Kohno & Ray, 2008). Considering the corresponding latitudinal occurrences in Early Pliocene deposits (Florida, South Carolina, North Carolina), Ontocetus may appear to be adapted to warmer temperate water than the extant species, especially given the absence of Ontocetus occurrences at high latitudes, although the absence of fossiliferous rocks at these latitudes may represent a bias rather than a signal.

Additionally, an extinct baleen whale Herpetocetus has very often been found in Upper Miocene to Pliocene similar deposits associated with odobenins (i.e., Purisima Formation, California, USA; Tatsunokuchi Formation, Sendai, Japan; Yorktown Formation, North Carolina and Virginia, USA; Lillo Formation, Belgium; San Diego Formation, California, USA). However, Herpetocetus already appears to be present at least in Upper Miocene deposits of Belgium although the origin of the specimen remains undecided (El Adli, Deméré & Boessenecker, 2014). The potential presence of Herpetocetus as early as the Late Miocene in the North Sea also argues for a later arrival, probably via the southern western North Atlantic, rather than the Arctic, for Ontocetus.

Although there is a lack of fossils of Ontocetus from the entire Pacific coast of North America to date, which seems to have had its own faunal province with little migration events (Boessenecker, 2013), the presence of an another odobenin Valenictus (Late Miocene up to Late Pliocene) on the west coast of the USA (California) supports a more southerly migration of Pliocene tusked walruses via the Central American Seaway rather than through the Arctic (Repenning & Tedford, 1977; Deméré, 1994a, 1994b). The gregarious behavior, the dependence on shallow embayment along the coasts, their oldest occurrence in the western North Atlantic located in Florida, the absence of Pliocene and Lower Pleistocene fossiliferous deposits at high latitudes in the western North Atlantic and the fact that this does not significantly predate the closure of the Central American Seaway (Grossman et al., 2019), may suggest that Ontocetus may have migrated from the North Pacific to the North Atlantic via the still-open Central American Seaway at the time (Naafs et al., 2010; O’Dea et al., 2016) (Fig. 11A). More information on the understudied Pliocene-Middle Pleistocene fossil localities within the Arctic Circle (e.g., Gubik Formation, Alaska) could help clarify the migration pattern of Ontocetus (Brigham, 1985; Boessenecker, Boessenecker & Geisler, 2018).

Ontocetus then dispersed along the western North Atlantic during the Early Pliocene, populating fossil-rich formations such as the Bone Valley Formation, Raysor Formation, and Yorktown Formation (Berry & Gregory, 1906; Kohno & Ray, 2008). However, around 4.6–4.2 Ma, the closure of the Panama Isthmus disrupted equatorial currents between the Pacific and Atlantic Oceans, leading to the cooling of the Atlantic Ocean and eventual ice cap formation (O’Dea et al., 2016). Between these glaciations, during the warm periods around the late Early Pliocene (3.8 Ma), a population from the western North Atlantic may have migrated towards the North Sea. This migration was facilitated by the intense North Atlantic current transporting warm waters northward and maintaining the higher latitudes’ warmth (Naafs et al., 2010) (Fig. 11B). However, during the Late Pliocene (3.1 Ma), sea-level oscillations, likely acting in conjunction with other oceanographic alterations such as changes in productivity, ocean circulation, and biotic drivers such as prey availability, impacted neritic areas and led to the extinction of the Pliocene marine megafauna (Pimiento et al., 2017). The loss of abundant shallow embayments and seaways, coupled with faunal turnover in the Late Pliocene, likely contributed to the disappearance of mid-latitude walruses, which were highly specialized in feeding adaptations (Boessenecker, 2013, Boessenecker & Churchill, 2013; Pimiento et al., 2017).

No occurrences of otariids or phocids were found in the North Sea during the Pliocene, except a monachinae indet. (Dewaele, Lambert & Louwye, 2018b), and two undescribed occurrences (cf. Pagophilus sp., cf. Pusa sp.; Post & Bosselaers, 2005). Given the robust morphology of its forelimb, giving highly moving abilities by robust limbs for taking up breeding territories by example, as well as its craniomandibular anatomy (relative elongation of the skull, characterized by a more slanted cranial roof, non-fused and longer mandible, more complete dentition, particularly with functional incisors, and a more developed mandibular condyle) Ontocetus emmonsi has been proposed to have occupied a close ecological niche of the current sea lion who was absent from the North Pacific until the end of the Pliocene (Ontocetus sp., Okamoto & Kohno, 2019).

The global cooling during the Plio-Pleistocene transition had a profound impact on the warm North Atlantic Current, causing the ice cap to extend to lower latitudes (Naafs et al., 2010; Batchelor et al., 2019). During the Late Pliocene, a North Sea population of Ontocetus may have evolved and specialized in suction-feeding techniques, particularly in response to changes in mollusk fauna dominated by immigrants from the Pacific Ocean (Macoma balthica) and other bivalves (Riches, 2010). In the cold periods of the Early Pleistocene, the warm North Atlantic Current did not flow northwards but rather west-eastwards, directly towards Morocco, where the Arctic ice cap likely reached Spain (Naafs et al., 2010). Ontocetus has also been identified in the Ahl al Oughlam deposits of Morocco through the Plio-Pleistocene boundary (3.0–2.2 Ma) (Alachtherium africanum Geraads, 1997) (Fig. 11C). This raises the possibility that the Moroccan population of Ontocetus became endemic due to the altered direction of the North Atlantic Current and ice cap expansion. A population of Ontocetus from the western North Atlantic may have migrated toward Morocco, or that part of the eastern North Atlantic population may have migrated southwards through the paleo-Bay of Biscay toward Morocco, although there is no fossil evidence at present (Fig. 11C). Later, as the Early Pleistocene unfolded, warm periods became increasingly rare, contributing to an overall colder climate with drastic falls in sea levels due to ice cap expansion.

Extinction of Ontocetus in relation to global cooling through Early Pleistocene

Despite its suction-feeding adaptation to the exclusive molluscivorous diet in the North Sea, Ontocetus posti became extinct around 1.7 Ma. This extinction may also be explained by the confinement of the taxa in the North Sea and unable to escape northwards to the Atlantic Ocean due to ice sheet expansion (Van Vliet-Lanoë et al., 2002; Gibbard & Cohen, 2015; Gibbard & Lewin, 2016; Rea et al., 2018; Lein et al., 2022) (Fig. 11D). Moreover, the Dover Strait was closed and would not have reopened until the middle of the Chibanian or Holsteinian of about 400 Ka (Meijer & Preece, 1995; Westerhoff et al., 2020). Ontocetus posti may have been overly specialized in suction-feeding for molluscivorous diet, rendering it more vulnerable to rapid changes in sea level oscillation and the general global cooling during this period. Additionally, during the Late Tiglian (1.8 Ma), there was a significant drop in mollusk species, with few "warm" mollusks (Meijer & Preece, 1995). The Late Pliocene marine mammal assemblage in the southern North Sea also includes other species coexisting with Ontocetus, some of which using a similar suction-feeding technique, such as piscivorous Mesoplodon sp. (Ziphiidae), Delphinapterus sp. (Monodontidae), and Globicephala sp. (Delphinidae) (de Vos, Mol & Reumer, 1998; Post & Bosselaers, 2005). The only other pinnipeds found are two Phocidae (cf. Pagophilus sp. and cf. Pusa sp.) and a monachinae indet., which, based on their contemporary representatives, did not occupy the same niche as Ontocetus (Post & Bosselaers, 2005; Dewaele, Lambert & Louwye, 2018b).

Interestingly, one final population of Ontocetus emmonsi managed to survive at much lower and warmer latitudes in the Upper Waccamaw Formation, South Carolina, USA, until 1.1 Ma (Boessenecker, Boessenecker & Geisler, 2018). Although this is at present evidenced only by the fossil record of a relatively short and strongly curved tusk, it may suggest that the oral cavity or the rostrum of this animal had still retained a relatively generalized one. As a result, it might have been facilitated much broader dietary preferences by a less specialized feeding apparatus and suffered less quickly or significantly suffered the environmental changes during the Pliocene and Early Pleistocene. More stable and favorable climatic conditions, at more southerly latitudes compared to the North Sea, also enabled Ontocetus to survive until 1.1 Ma.

Further findings on this final population are still needed to better recognize its species-level assignment. The temporal separation of Odobenus and Ontocetus in the North Atlantic for approximately one million years suggests that they did not compete for resources and reflects separate Early Pliocene (Ontocetus) and Middle–Late Pleistocene (Odobenus) dispersals into the North Atlantic from the North Pacific. These dispersals occurred most likely through the Central American Seaway for the warmer-tolerant Ontocetus and the Arctic for the colder-tolerant Odobenus (Kohno et al., 1995; Kohno & Ray, 2008).