The full text of this article hosted at iucr.org is unavailable due to technical difficulties.

ORIGINAL ARTICLE
Free Access

EARLY EVOLUTION OF SEXUAL DIMORPHISM AND POLYGYNY IN PINNIPEDIA

Thomas M. Cullen

E-mail address:thomas.cullen@mail.utoronto.ca

Department of Earth Sciences, Carleton University, Ottawa, Ontario, Canada

Search for more papers by this author
Danielle Fraser

Department of Biology, Carleton University, Ottawa, Ontario, Canada

Search for more papers by this author
Natalia Rybczynski

Palaeobiology, Canadian Museum of Nature, Ottawa, Ontario, Canada

Search for more papers by this author
Claudia Schröder‐Adams

Department of Earth Sciences, Carleton University, Ottawa, Ontario, Canada

Search for more papers by this author
First published: 18 February 2014
Cited by: 12

All data included as supplementary information.

Abstract

Sexual selection is one of the earliest areas of interest in evolutionary biology. And yet, the evolutionary history of sexually dimorphic traits remains poorly characterized for most vertebrate lineages. Here, we report on evidence for the early evolution of dimorphism within a model mammal group, the pinnipeds. Pinnipeds show a range of sexual dimorphism and mating systems that span the extremes of modern mammals, from monomorphic taxa with isolated and dispersed mating to extreme size dimorphism with highly ordered polygynous harem systems. In addition, the degree of dimorphism in pinnipeds is closely tied to mating system, with strongly dimorphic taxa always exhibiting a polygynous system, and more monomorphic taxa possessing weakly polygynous systems. We perform a comparative morphological description, and provide evidence of extreme sexual dimorphism (similar to sea lions), in the Miocene‐aged basal pinniped taxon Enaliarctos emlongi. Using a geometric morphometric approach and combining both modern and fossil taxa we show a close correlation between mating system and sex‐related cranial dimorphism, and also reconstruct the ancestral mating system of extant pinnipeds as highly polygynous. The results suggest that sexual dimorphism and extreme polygyny in pinnipeds arose by 27 Ma, in association with changing climatic conditions.

Charles Darwin wrote extensively about the evolution of behavioral and morphological differences between the sexes or sexual dimorphism (Darwin 1871). Morphological sexual dimorphism is manifested as differences in size or shape, including ornamentation, between the sexes of a given species, some of which are preserved in the fossil record. The occurrence of sexual dimorphism (SD) has been reported numerous times in the vertebrate fossil record (e.g., sharks, ungulates, carnivorans, turtles, dinosaurs) (Janis 1982; Lund 1985; Chapman et al. 1997; Carrano et al. 1999; Mead 2000; Plavcan 2000; Deméré and Berta 2002; Joyce et al. 2012; Plavcan 2012) and yet, the evolutionary history of such sexually dimorphic traits remains generally uninvestigated for most lineages (Hendrick and Temeles 1989). Understanding the evolution of sexual dimorphism (e.g., how and when does it arise?), which is often in opposition to Darwinian natural selection, is one of the original problems posed to evolutionary biologists, and has important implications for our understanding the coevolution of behavior (e.g., mating system) and intraspecific dimorphism (Leutenegger and Kelly 1977; Hendrick and Temeles 1989). For example, studies of ungulates suggest a strong association between SD and mating system; early fossil forms were generally monogamous and monomorphic, with dimorphism and polygyny evolving later in conjunction with the expansion of open grassland environments (Janis 1982; Pérez‐Barbería et al. 2002). Within primates, sexual dimorphism in body and canine size are related to male intra‐sexual selection and can be used to infer polygyny in extinct taxa (Leutenegger and Kelly 1977; Plavcan 2000). Such correlated evolution between morphology and behavior, if present in other taxa, can be used to reconstruct and constrain inferences of the evolutionary history of sexually selected traits, and better predict their phylogenetic distribution (Plavcan 2000; Plavcan 2012). Tracing the evolution of SD along a phylogeny can allow us to reconstruct the evolutionary history of mating systems, giving us a better understanding of the timing of major morphological shifts and provide context for the origins of behavior and ecology seen in modern taxa (Weckerly 1998; Cassini 1999; Plavcan 2000).

Pinnipeds possess a wide range of reproductive strategies and sexual dimorphism, encompassing the extremes observed in modern mammals (Bartholomew 1970; Weckerly 1998; Naughton 2012). Otariids + odobenids display considerable SSD, a highly polygynous harem‐based mating system, and a preference for land‐based breeding (Bartholomew 1970; Cassini 1999; Lindenfors et al. 2002). In contrast, almost all phocids display promiscuity with weak polygyny, a preference for breeding in the water or on ice floes, and relatively little sexual size dimorphism (SSD) (Bartholomew 1970; Cassini 1999; Lindenfors et al. 2002). The exceptions to this are three derived and phylogenetically distant genera, Halichoerus and Mirounga, which appear to converge on more otariid‐like dimorphism and behavior (harem‐based land breeding), and Cystophora, which exhibits noticeable SSD, and form otariid‐like breeding colonies (though on ice floes, not shore) (Fig. 1) (Cassini 1999; Naughton 2012; Nyakutura and Bininda‐Emonds 2012). Water or ice‐based mating occurs in taxa (i.e., most phocids) living in primarily higher latitude waters with greater primary productivity, whereas colony‐forming taxa (i.e., otariids) live in comparably lower latitude waters with lower productivity (Ferguson and Higdon 2006; Jones and Goswami 2010b). It has been suggested that environmental factors, particularly ocean productivity, have been an important driver of the evolution of the form and mating system of modern pinnipeds (Ferguson and Higdon 2006).

image
Pinniped phylogeny. Taxa sampled in this study indicated with arrows. Bold font indicates the modern phylogenetic distribution of extreme sexual dimorphism and polygyny indicates extinct taxa (both fossil and recently extinct). Tree topology based on that of Nyakutura and Bininda‐Emonds (2012).

Given that, with few exceptions, phocids and otariids + odobenids display disparate reproductive strategies (and patterns of SSD) it is impossible to reconstruct the ancestral condition of SSD in crown pinnipeds based solely on modern taxa. There is some evidence that SSD may have deep origins within Pinnipedia; desmatophocids (e.g., Desmatophoca, Allodesmus), an extinct sister clade to the phocids, show SSD (Deméré and Berta 2002). However, the phylogenetic position of desmatophocids is after the divergence of phocids and otariids + odobenids (Boessenecker and Churchill 2013) (Fig. 1), contributing little to the study of SSD evolution among early pinnipeds; SSD was either lost in phocids (and later regained in two modern genera) or occurred in desmatophocids as a result of parallel evolution with otariids + odobenids.

Here, we use a combination of evidence from fossil and modern taxa to reconstruct the ancestral mating system of pinnipeds. We include material of the late Oligocene to early Miocene‐aged (∼27–20 Ma) fossil taxon Enaliarctos, an intermediate form regarded as an outgroup to crown pinnipeds (Berta and Adam 2001; Bebej 2009), and therefore important for reconstructing ancestral states. Unlike the freshwater, web‐footed fossil taxon Puijila darwini (Rybczynski et al. 2009), Enaliarctos is found in marine deposits, and is a flippered form resembling an otariid in general appearance (Berta 1991; Bebej 2009). While Enaliarctos is intermediate in morphology, sedimentological, and biomechanical evidence suggests it was highly aquatic and fed in water like a modern pinniped (Berta 1991; Bebej 2009). Given that sexual size dimorphism is correlated with mating systems in modern pinnipeds (and other mammals), we can use morphological evidence to reconstruct the mating system of Enaliarctos and other extinct pinnipeds (Leutenegger and Kelly 1977; Cassini 1999; Plavcan 2000). We describe a cranial specimen of Enaliarctos emlongi, report on new morphological evidence of sexual dimorphism in this taxon, and use 2D geometric morphometrics and phylogenetic comparative methods to characterize skull‑shape dimorphism in extant pinnipeds and Enaliarctos. We perform ancestral character estimation with and without Enaliarctos thereby allowing us to infer the mating system in the common ancestor of crown pinnipeds.

Methods

COMPARATIVE DESCRIPTION AND CT SCANNING

Following Rybczynski (2009), we use the stem definition of Pinnipedia, in order to include both the extant and fossil taxa (the latter occasionally referred to as pinnipedimorphs). High Resolution Computed Tomography scans of the small skull specimen of Enaliarctos emlongi (USNM 314290) were performed in order to examine the middle ear morphology (as the surface of the tympanic bulla had been previously removed from the holotype of E. emlongi, as well as in most other USNM Enaliarctos specimens in which the bullae are preserve, for examination of the internal structures, but was completely intact in this specimen), for use in the comparative description of the specimen, and analyzed using Avizo and 3D Studio Max.

EXTANT DATA COLLECTION

Skulls of adult male and female individuals, representing the morphological diversity of extant phocid and otariid lineages (Fig. 1), were selected from the collections of the Canadian Museum of Nature (CMN), in Ottawa, Canada, and the Smithsonian Institution's National Museum of Natural History (USNM/NMNH), in Washington, D.C. The degree of sutural fusion (in particular, the basisphenoid‐presphenoid, interfrontal, and maxillary sutures), and specimen collection data (i.e., sex/age at death and/or whether individual was gestating at time of collection) were used to establish the adult status of specimens (Brunner et al. 2004). Collection data is particularly valuable for taxa (i.e., phocids) where complete suture fusion is known to occur before sexual maturity (Jones and Goswami 2010a). Phocid species included Pusa hispida (n = 20, 50% male), Halichoerus grypus (n = 15, 47% male), Cystophora cristata (n = 20, 50% male), Erignathus barbatus (n = 14, 50% male), Monachus tropicalus (n = 20, 50% male), Leptonychotes weddellii (n = 8, 50% male), and Hydrurga leptonyx (n = 10, 30% male). Otariids included Callorhinus ursinus (n = 20, 50% male), Arctocephalus australis (n = 9, 33% male), and Eumetopias jubatus (n = 20, 50% male). Skulls specimens, without mandibles, were placed in a photostand and photographed at an approximate distance of 180 cm in order to reduce parallax (sensu Rybczynski et al. 2010). Skull measurements were taken using digital calipers with 0.1 mm accuracy, and repeated three times to ensure precision (Supporting Information Appendix 2a). Thirty‐one landmarks, adapted from several other mammalian (primarily pinniped) dimorphism and/or morphometric studies (Brunner et al. 2004; Sanfelice and de Freitas 2008; Jones and Goswami 2010a; Jones and Goswami 2010b; Rybczynski et al. 2010), were selected from dorsal (1–12) and ventral views (13–31) with the purpose of capturing skull shape and the positions of relevant morphological, ontogenetic, or functional features (Fig. 2, Supporting Information Appendix 2b). Landmark coordinates were obtained directly from specimen photographs using the Pointpicker plugin for ImageJ (Fig. 2, Supporting Information Appendix 2c) (Thévenaz 2012). In order to reduce potential errors in the morphometric analyses arising due to individual asymmetry, only one side of the skull was analyzed for each specimen, in both dorsal and ventral views. The use of 2D data, instead of 3D, prevents certain shape information, such as the heights of the sagittal crest and canine teeth from being taken into account. However, to allow both comparison with previously published studies (i.e., Sanfelice and de Freitas 2008), and to accommodate for the diagenetically (primarily dorsoventrally) compressed nature of the fossil material, 2D morphometrics were considered more appropriate. In general, the use of 3D data does confer certain advantages that should be considered, in addition to our 2D data, in future studies (Jones and Goswami 2010b).

image
Morphometric landmarks. Dorsal (1–12) and ventral (13–31) landmarks used in this study, placed on example specimen of Eumetopias jubatus. Landmark descriptions can be found in Supporting Information Appendix 2b.

RELATIONSHIP BETWEEN SKULL SIZE AND DIMORPHISM

Given the lack of full body size information for some specimens (in particular fossil forms), skull size was used as a surrogate for body size SSD. The average skull length for males and females of each species was measured, and compared using a two‐tailed t‐test assuming unequal variance (α = 0.05) in order to confirm if skull size was similarly dimorphic in species showing body size dimorphism. Additionally, to investigate the effect of small sample size in the fossil taxon on comparisons of male to female skull‑length ratios, in the modern taxa individual male to female skull‑lengths ratios were calculated in order to show both the mean and maximum skull length ratios of each species.

GEOMETRIC MORPHOMETRIC ANALYSES

Morphometric analysis protocols were modified from Rybczynski et al. (2010) and Jones and Goswami (2010a,2010b). Per taxon sample sizes were increased, and landmark selection was kept similar to Jones and Goswami (2010b), with landmark selection primarily determined by homologous and/or shared functional characters. All analyses were conducted in R (R Core Team 2012), using the “Shapes” (Dryden 2013), “Abind” (Plate and Heiberger 2011), “Calibrate” (Graffelman 2012), “MASS” (Venables and Ripley 2002), and “Car” (Fox and Weisberg 2011) packages, as well as functions contained within the book “Morphometrics with R” (Claude 2008).

Generalized least‐squares Procrustes analysis (GPA) was performed for each species to minimize the effect of absolute size differences while also aligning the combined coordinate sets. These data were then analyzed using principle component analysis (PCA) in order to identify the main morphological drivers of dimorphism, and to visualize trends both within and across species. Convex hulls were applied to the resultant plots, indicating the morphospace occupied by males and females of each species. The principle component loadings were assessed to determine the landmark variables most closely associated with dimorphism. See Supporting Information Appendices 1b, 1c, 2e, and 2f for additional details. Euclidean distance was also calculated between the male and female centroids of each species, in dorsal and ventral, and compared with the distance between male and female Enaliarctos specimens, in order to assess the magnitude of shape difference present within each taxon sample.

PHYLOGENETIC SIGNAL, INDEPENDENT CONTRASTS, AND STATISTICAL ASSESSMENT OF DIMORPHISM

Mating system, dorsal PC1 and PC2 residuals, and ventral PC1 and PC2 residuals were analyzed for phylogenetic signal using the "phytools" package in R (Pérez‐Barbería and Gordon 1999; Revell 2012). Phylogenetic independent contrasts comparing morphometric data with mating system information were performed, when necessary, using the "ape" package (Paradis et al. 2004). Cranial sexual shape dimorphism, via PC residuals, was then statistically assessed using MANOVA, and additionally compared to mating systems using OLS, with mating systems (dimorphic, harem‐based vs. nondimorphic, not harem‐based) represented within the analysis through the use of "dummy" variables (Hammer and Harper 2006; Fraser and Theodor 2011).

Ancestral character estimation (ACE). R package "geiger" was used to perform ACE on mating systems in pinnipeds, using modern taxa in one analysis, and a combination of fossil (enaliarctines and desmatophocids) and modern taxa in the other (Harmon et al. 2009). The pinniped phylogenetic tree and branch lengths were from the mammalian supertree of Nyakatura and Bininda‐Emonds (2012).

Results

COMPARATIVE DESCRIPTION OF Enaliarctos emlongi

United States National Museum of Natural History (USNM/NHNH) 314290 (Fig. 3) is a small, partially deformed skull, lacking mandibles, originally referred to as a juvenile of Enaliarctos emlongi by Berta (1991). Both USNM 314290 and the E. emlongi holotype, USNM 250345, are found along the Oregon coast in the early Miocene Nye Mudstone to Astoria Formation (there is some debate in the original description of the species as to whether the strata from which they were both collected represents the Astoria Formation or the underlying Nye Mudstone) (Berta 1991). The small skull (USNM 314290) is approximately 158 mm in total length; reliable width measurements could not be obtained due to deformation, although the estimated reconstructed maximum skull width is 83.6 mm. The skull is dorsoventrally crushed, probably due to sediment compaction, with some minor shearing of the rostrum in the mediolateral plane. Despite post‐depositional deformation, most key morphological features are preserved and can be reliably characterized. With the exception of the canines, the dentition preserved in USNM 314290 is very similar to that of the E. emlongi holotype (USNM 250345) (see Fig. 4). No incisors are preserved in USNM 314290, although the alveoli show that three incisors were present on each side of the premaxilla. The left and right canines are preserved, although the left canine has been fractured from the skull itself. The canines have a smaller transverse section than those of the E. emlongi holotype (USNM 250345) or referred snout (USNM 314540). Similarly, the canines of USNM 314290 have a smaller transverse section than those of either the holotype (USNM 314295) or referred skull (USNM 335375) of E. barnesi, while they are more similar to those of Pacificotaria hadromma (USNM 167648) or Pinnarctidion rayi (USNM 314325). The canines of USNM 314290 are similar to E. emlongi specimens in being slightly recurved and possessing a vertical crest extending medially from the posterior to lingual face. Two of the four premolars (P3 and P4) are preserved, but only on the right side. Examination of the alveoli shows P1 to have been single rooted, with P2 being double rooted. There does not appear to be an embrasure pit between P2 and P3, while a shallow embrasure pit is present between P3 and P4. P3 is very similar in shape to that of E. emlongi, and unlike other Enaliarctos species, in possessing a well‐developed crest‐like metacone (Fig. 4). P4 is triple rooted, with a large conical paracone separated from a relatively short metacone. In these respects, as well as with a lingual cingulum and similar wear patterns, the P4 of this specimen displays diagnostic characteristics of E. emlongi (Fig. 4A and B). Of the molars, only M1 is preserved, although it is present on both the left and right sides. Between P4 and M1 there is a deep embrasure pit, which is unlike E. emlongi, but similar to E. barnesi. M1 is double rooted, with a paracone and metacone of nearly equal height, a large parastyle, and a slightly sloped protocone. Unfortunately, M1 is not preserved in the holotype of E. emlongi, although the morphology of M1 in this specimen is different from that of E. barnesi in the relatively smaller size of the paracone (Fig. 4A and C). M2 is not preserved in this specimen, nor can the alveoli be identified, due to poor preservation. The palate is slightly arched dorsally, and the palatine sulci are shallow, while extending from a position medial of P4 to a position medial of P2. In these respects, the specimen is similar to E. emlongi. A spur‐like process extends posteriorly from the maxilla in this specimen, which is not seen in the holotype of E. emlongi. However, a similar process is found in cf. P. rayi (Berta 1994), E. tedfordi, and E. barnesi, with the feature considered diagnostic to the latter by Berta (1991). The incisive foramina are similar to E. emlongi in being deep and separated by a narrow crest, but are unlike E. emlongi, or other Enaliarctos specimens, in being positioned much closer to the canines and incisors. This difference in position appears to be due to the narrow, rounded shape of the snout, which is in sharp contrast to the much wider and blunt, almost squared, shape seen in the holotype of E. emlongi. This difference in shape does not appear to be related to the deformation present in this skull. Another noted difference between the skull of USNM 314290 and Enaliarctos specimens are the low sagittal and lambdoidal crests, which are more pronounced in the E. emlongi holotype. The posterior end of the anteroventral process of the jugal attaches anterolaterally to M1 in USNM 314290, a feature shared with, and diagnostic of, E. emlongi (Fig. 4A and B). The middle ear and basicranial morphology of USNM 314290 was examined by X‐ray computed microtomography (Fig. 4 AII and III) in order to assess the similarities and differences between this specimen and the E. emlongi holotype. The morphology of the middle ear is very similar to that of E. emlongi, particularly in possessing a mediolaterally expanded pear‐shaped promontory, a posteriorly oriented round window, a nearly circular epitympanic recess, a deep elliptical fossa for the tensor tympani, and a larger tympanic crest that projects into the tympanic cavity. The orientation of the round window and extent of the tympanic crest are of particular significance, as within Enaliarctos they are only present to these forms in E. emlongi (with the tympanic cavity not preserved in E. barnesi, and a more laterally directed round window and less extensive tympanic crest in E. tedfordi). As noted, USNM 314290 differs from the E. emlongi holotype in the shape of the snout (being more rounded than square as in the holotype), sagittal crest (reduced compared to the holotype), and basicranium/mastoid region (relatively narrow vs. broad in the holotype). However, none of the aforementioned differences relate to characters considered diagnostic of the species (Berta 1991), though they are considered morphological indicators of sex within modern dimorphic pinnipeds (particularly otariids), with USNM 314290 showing similarity to the female condition (e.g., slender canines, low sagittal and lambdoidal crests, and narrower snout) (Howell 1929; Brunner et al. 2004; Sanfelice and de Freitas 2008). Though there are some features shared between this specimen and E. barnesi, the considerable dental similarity between USNM 314290 and E. emlongi, coupled with middle ear similarities (Fig. 4), and the nearly exclusive reliance on dental characteristics in differentiating Enaliarctos species, supports Berta's (1991) inclusion of this individual into E. emlongi.

image
Skull of Enaliarctos emlongi. At top: dorsal views of USNM 314290, with photograph (left) and interpretive drawing displaying visible sutures (right). At bottom: ventral views of USNM 314290, with photograph (left) and interpretive drawing displaying visible sutures (right). Dashed lines indicate sutures. Scale bar = 2 cm. Anatomical abbreviations—bo, basioccipital; c, canine; f, frontal; if, incisive foramen; jf, jugular foramen; ju, jugal; l, lacrimal; m1, first molar; mf, mandibular fossa; mx, maxilla; n, nasal; o, occipital; oc, occipital condyle; p, parietal; p3, third premolar; p4, fourth premolar; pl, palatine; pmx, premaxilla; pop, postorbital process; rw, round window; sc, sagittal crest; sq, squamosal; tb, tympanic bulla. Skull element labels capitalized, other features in lower case.
image
Anatomical comparison of fossil skulls. Ventral views of skulls, dentition, and middle ear structures of USNM 314290 (Enaliarctos emlongi), USNM 250345 (Enaliarctos emlongi), USNM 314295 (Enaliarctos barnesi), and USNM 314325 (Pinnarctidian rayi). (A) Ventral view of skull, dentition (I), micro‐CT internal view of left tympanic bulla (II), and 3D reconstruction of left tympanic bulla and middle ear structures as obtained from micro‐CT scans of USNM 314290. (B) Ventral view of skull, dentition (I), and right tympanic region with bulla removed, showing middle ear structures (II), of USNM 250345. (C) Ventral view of skull, and dentition (I), of USNM 314295. (D) Ventral view of skull, dentition (I), and right tympanic region with bulla removed, showing middle ear structures (II), of USNM 314325 of particular note in comparison are: (1) the shape of the cusps of the third and fourth upper premolars, and (2) the extent of the shelf of the tympanic crest and the posterolateral orientation of the promontorium, all of which share very similar morphology in both USNM 314290 and USNM 250345, further uniting them within E. emlongi, and separating them from the differing morphologies seen in USNM 314295 and USNM 314325. Anatomical abbreviations—c, cingulum; mc, metaconid; p, promontorium; P3, third upper premolar; P4, fourth upper premolar; pc, paraconid; ptc, protoconid; rw, round window; tc, tympanic crest.

The rationale for the assigned juvenile status of USNM 314290 by Berta (1991) was not specified, as the specimen was never given a full description, but was presumably related to the small size of the skull in relation to the much larger skull of the E. emlongi holotype. The sutures of USNM 314290 exhibit a closed pattern (Fig. 4B and D) consistent with those of fully adult (sensu Brunner et al. 2004) modern pinnipeds (particularly otariids), including full closure of the occipito‐parietal, squamosoparietal, interparietal, interfrontal, coronal, basisphenoid‐presphenoid, and basioccipito‐basisphenoid sutures, and probable full closures of the maxillary and premaxillary–maxillary sutures (the degree of closure in some regions is partially obscured by deformation and/or surface damage). Work by Brunner et al. (2004) showed that among modern pinnipeds, the basisphenoid‐presphenoid, interfrontal, maxillary, and premaxillary–maxillary sutures were the last to close, signalling physical maturity. Of note, fractures that appear on USNM 314290 do not appear to follow sutures preferentially, suggesting that the sutures are fully fused and thus did not offer obvious zones of weakness for fracture propagation. Given that the skull sutures of USNM 314290 are closed (particularly those that fuse later in life), the status of USNM 314290 as a juvenile can be rejected. This specimen represents an adult individual of E. emlongi, though it is both considerably smaller, and differing from the holotype in the shape of several cranial characters often associated with females in extant pinnipeds.

SKULL SIZE DIMORPHISM IN PINNIPEDS

All three otariid taxa show significant differences (Arctocephalus, p = 3.65 × 10−4; Callorhinus, p = 2.44 × 10−10; Eumetopias, p = 2.62 × 10−16) between male and female skull length with adult males being larger (see also Supporting Information Appendix 2d). The phocids show no statistical differences between male and females, with the exceptions of Cystophora and Halichoerus, which both show significant male–female differences, though of considerably less magnitude than in otariids. In Cystophora (p = 1.29 × 10−4) and Halichoerus (p = 2.82 × 10−5), males have significantly greater average skull lengths than females. Comparisons of male to female skull length ratios show that otariids possess large mean length ratios (Arctocephalus, 1.30; Callorhinus, 1.29; Eumetopias, 1.25), with much higher maximum ratios (Arctocephalus, 1.46; Callorhinus, 1.40; Eumetopias, 1.30) (see also Supporting Information Appendix 2k). Phocid male to female skull length mean ratios show little to no dimorphism (0.95–1.06), with Cystophora (1.13), as above, being the exception. Halichoerus shows a different signal of male to female skull length, with low mean ratio (0.95), but high maximum (1.32), and very high standard variation (0.95, with no other taxa showing standard deviation above 0.09). The male to female skull length ratio of Enaliarctos emlongi is 1.45. Skull size dimorphism is consistent with body size dimorphism in taxonomic distribution within pinnipeds, and although the ratio of male to female skull length in E. emlongi is quite large, it is not outside the dimorphism seen in modern pinnipeds.

SKULL SHAPE DIMORPHISM IN PINNIPEDS

Procrustes‐superimposed principle component (PC) analyses were run on fossil and extant pinniped skull landmarks (Fig. 2), producing both dorsal and ventral results (Figs. 5 and 6; see also Supporting Information Appendices 2e and 2f). Dorsal PC loadings (see also Supporting Information Appendix 1b) show that most variation was concentrated around landmarks relating to the position and extent of the nasal and rostrum, both laterally and antero‐posteriorly. Analysis of dorsal landmarks (Fig. 5) identified three distinct groups. The first group comprises the three otariid taxa (Arctocephalus, Callorhinus, and Eumetopias), which show some separation between male and female morphospaces. In Arctocephalus and Eumetopias, this separation is along the PC2 axis, relating to increased relative length of the nasal compared to the parietal. In Callorhinus, the separation is along the PC1 axis, relating to the lateral extent of the nasal and jugal. The second morphospace group is the largest and includes the phocids Pusa, Monachus, Erignathus, Halichoerus, Hydrurga, and Leptonychotes, although the latter is separated from the majority of the group. Within this phocid group, Halichoerus shows some evidence of separation between males and females, although it is less pronounced than among the otariids, and is distributed across both PC axes, while Pusa, Monachus, Hydrurga, Erignathus, and Leptonychotes show no clear separation. The third distinct group is positioned a considerable distance from all other morphospaces and contains only one taxon, Cystophora, which shows some male–female separation along the PC1 axis representative of lateral nasal and jugal extent. The otariids and phocids are primarily separated in morphospace along the PC1 axis, with otariids positioned negatively and phocids positively. Along the PC2 axis, the otariid taxa cluster close to the zero position, while the phocids are distributed across negative and positive positions (with most species in mid to low positive positions, and Cystophora positioned negatively). The male and female individuals of the fossil taxon Enaliarctos emlongi plot near otariids, though the female skull plots further from the main group.

image
Dorsal morphometric results. PCA of extant and fossil pinnipeds, based on dorsal landmarks (Supporting Information Appendices 2e and 2f). Three primary groups are present: the otariids and fossils, with more negative PC1 values; the phocids (not counting Cystophora), with positive PC2 values; and Cystophora, with positive PC1 and negative PC2 values. Within species, the otariids, Cystophora, and Halichoerus appear to show male/female separation in morphospace. At bottom, legend explaining above plots. Individuals of modern species indicated by filled shapes, fossils by unfilled shapes, females represented by circles, males by triangles, male/female groups by convex hulls, and species by color.

Ventral PC loadings (See also Supporting Information Appendix 1c) showed that most variation was concentrated around landmarks relating to the lateral extent of the basicranium and mastoid, with additional variation centered on the lateral extent and relative length of the rostrum. Ventral landmark analysis (Fig. 6) shows two distinct groups. The first is composed of Arctocephalus, Callorhinus, Eumetopias, and Hydrurga, all showing apparent separation between male and female morphospaces. The second group contains Pusa, Monachus, Erignathus, Halichoerus, Leptonychotes, and Cystophora. Within this group, only Halichoerus and Cystophora show evidence of male–female separation, although to a lesser extent than among otariids. Ventrally, all dimorphic taxa, with the exception of Hydrurga and Leptonychotes, show a similar trend; male morphospaces are displaced from those of females through a positive shift along both PC1 and PC2 axes, representing increases in lateral extent of basicranium to palate, and relative length of palate to basicranium, respectively. In Hydrurga and Leptonychotes the pattern is reversed. The ventral analysis also places E. emlongi within the otariid morphospace.

image
Ventral morphometric results. PCA of extant and fossil pinnipeds, based on ventral landmarks (Supporting Information Appendices 2e and 2f). Two groups are present: the otariids, fossils, and Hydrurga, with generally positive PC1 values and negative to neutral PC2 values; and the phocids, with positive PC1 values and a range of PC2 values. Within species, the otariids, Cystophora, and Halichoerus appear to show male/female separation in morphospace, although otariids show a more consistent and similar pattern between male and female differences. The fossils plot within otariids. At bottom, legend explaining above plots. Individuals of modern species indicated by filled shapes, fossils by unfilled shapes, females represented by circles, males by triangles, male/female groups by convex hulls, and species by color.

We tested for phylogenetic signal among our principal component scores using Pagel's λ. Dorsally, neither PC1 (λ = 7.33 × 10−5) nor PC2 (λ = 7.33 × 10−5) show evidence of phylogenetic signal, whereas in ventral, PC1 (λ = 0.509) shows a small degree of signal, and PC2 (λ = 0.869) shows a relatively large influence due to phylogeny (see also Supporting Information Appendix 2g). MANOVAs were then performed on ventral PC1, and dorsal PCs 1 and 2. The ventral MANOVAs for PCs 1 and 2 confirmed the qualitative relationships displayed in morphospace, showing significant differences between male and female results in Arctocephalus (PC1 p = 0.04; PC2 p = 0.01), Callorhinus (PC1 p = 0.0005; PC2 p = 0.00001), Eumetopias (PC1 p = 0.00005; PC2 p = 0.001), Cystophora (PC1 p = 0.001; PC2 p = 0.002), and Halichoerus (PC1 p = 0.002; PC2 p = 0.01). Dorsal PC1 showed significant male–female differences in Arctocephalus (p = 0.04), Callorhinus (p = 0.00001), Cystophora (p = 0.003), and Halichoerus (p = 0.05), while dorsal PC2 showed significant male–female differences only in Eumetopias (p = 0.002) (see also Supporting Information Appendix 2h). Ordinary least‐squares regression was then used to compare the previous results to mating system (harem‐based or nonharem‐based, via “dummy” variables), with ventral PC1 values showing a significant correlation with mating system (R2 = 0.652, p = 0.001), while neither dorsal PC1 (R2 = –0.03, p = 0.44) nor PC2 (R2 = 0.11, p = 0.16) showed any relationship to mating system (see also Supporting Information Appendix 2i). Ventral PC2 was compared to mating system through phylogenetically independent contrasts, and showed a significant correlation (R2 = 0.38, p = 0.03) (see also Supporting Information Appendix 2j).

In addition, Euclidean distance was calculated between the male and female centroids of each species, in dorsal and ventral, and compared with the distance between male and female Enaliarctos specimens (see also Supporting Information Appendix 2l). In dorsal, phocid distances ranged from low in Pusa (29.95), Monachus (24.76), Leptonychotes (30.27), and Erignathus (36.75), to high in Hydrurga (67.58), Halichoerus (93.52), and Cystophora (109.32). Dorsal otariid distances were generally high, as in Arctocephalus (67.52), Eumetopias (60.15), and Callorhinus (75.67). The distance in Enaliarctos in dorsal was very high, 212.75, though this is likely due to the potentially anomalous position of the female. In ventral, phocid distances were generally low, such as in Pusa (10.35), Monachus (23.05), Hydrurga (39.50), and Erignathus (23.67), with higher values in Leptonychotes (66.35), Halichoerus (78.01), and Cystophora (112.83). Otariid values were high, as in Arctocephalus (105.06), Eumetopias (84.09), and Callorhinus (108.88). The distance for Enaliarctos in ventral was high, 114.14, and comparable to the centroid distance of highly dimorphic taxa.

ANCESTRAL CHARACTER ESTIMATION

We performed ancestral character estimation for the pinniped mating system (i.e., high polygyny, harem‐based vs. low polygyny, nonharem‐based) using an entirely extant pinniped phylogeny (Fig. 7). ACE suggests that a highly polygynous system (gray) is plesiomorphic for otariids, and that a nonharem‐based, low polygyny system (white) is plesiomorphic for phocids, with two more recent appearances of high polygyny. However, the estimated ancestral condition for Pinnipedia remains inconclusive, with both mating systems estimated at near equal probability (low polygyny, no harem: 48.8% vs. high polygyny, harem: 51.2%).

image
Ancestral character estimation. (A) ACE of extant pinnipeds, based on the mammal supertree of Nyakutura and Bininda‐Emonds (2012). This analysis resolves otariids as ancestrally dimorphic/harem based (gray), and phocids as ancestrally nonharem based (white). The base of Pinnipedia remains unresolved, with nearly equal probability of either mating system. (B) ACE of pinnipeds, with fossils (Enaliarctos, Desmatophoca) included. This analysis resolves pinnipeds as ancestrally dimorphic/harem based (gray), with a reversion to a nonharem, less polygynous system (white) at or just after the divergence of phocids from desmatophocids.

We repeated the ancestral character estimation including the fossil taxa (i.e., Enaliarctos, Desmatophoca), with fossil mating systems inferred from the degree of sexual dimorphism, following near exclusive association between polygynous mating system and dimorphism in extant relatives (results of this study; Berta 1991; Deméré and Berta 2002; Pörschmann et al. 2010). Inclusion of Enaliarctos and Desmatophoca increased the statistical support of a polygynous and harem‐based system as plesiomorphic for crown Pinnipedia (99.6%) (Fig. 7). However, the ancestral state of phocids was less confidently reconstructed (low polygyny, no harem: 48.7% vs. high polygyny, harem: 51.3%).

Discussion

Ancestral reconstruction of pinniped reproductive strategy (i.e., highly polygynous, harem‐based or weakly polygynous, not harem‐based) performed without fossil taxa is equivocal (Fig. 7). We use a morphometric approach to characterize sexual dimorphism in extant pinnipeds and use these results to identify the presence of dimorphism in a fossil taxon, Enaliarctos emlongi. Also, we present evidence of correlated evolution between mating system and sexual shape dimorphism in modern pinnipeds, suggesting that E. emlongi was likely highly polygynous. Given the phylogenetic placement of E. emlongi among the basal pinnipeds, phylogenetic ancestral reconstruction suggests the early evolution of polygyny among pinnipeds and reversion to a weakly polygynous mating system among phocids (Fig. 7). Our results show that the inclusion of even a small number of fossil taxa can have a transformative effect on inferences about the evolution of behavior and strongly supports a combined paleontological/neontological approach.

Cranial dimorphism in pinnipeds has been previously discussed only qualitatively (Howell 1929), or treated quantitatively with a particular focus on otariid (Brunner et al. 2004; de Oliveira et al. 2005; Sanfelice and de Freitas 2008) or phocid species (Mo et al. 2009), or with a broader taxonomic sampling using smaller sample sizes and/or fluctuating landmark selection (Jones and Goswami 2010a, 2010b). This study builds on previous studies to show which traits tend to be most dimorphic, and would be most reliable for identifying dimorphism in fossil taxa. All of the highly polygynous taxa examined in this study (otariid taxa; Arctocephalus, Callorhinus, and Eumetopias, and phocids; Cystophora and Halichoerus) show significant sexual dimorphism in skull shape. Males are separated from females by the lateral expansion of the mastoid, basicranium, and rostrum, as well as the length of the palate relative to the basicranium (for additional details see Supporting Information Appendix 2f), all of which are found as landmarks in ventral view. Ecologically, expansion of the rostrum, as well as the basicranium and mastoid, in males may reflect larger canine teeth and larger attachment areas for neck musculature, respectively, presumably associated with land‐based male–male combat (Howell 1929; Cassini 1999; Brunner et al. 2004). The only exception, given its water‐based mating and ice‐based breeding, is Cystophora, which shows less dimorphism in the basicranium and a greatly altered rostrum, a pattern that may relate to an inflatable display structure used by males to attract mates (Cassini 1999). A similar morphometric pattern exists in both dimorphic otariids and phocids, with a single exception, suggesting it is primitive for the group, and can be used to test for sexual dimorphism, and infer mating system, in basal (i.e., fossil) taxa. Dorsal landmarks do not show phylogenetic signal or a significant correlation with mating system, suggesting other factors, such as ecology, and geographic or individual variation, may play a stronger role in influencing dorsal morphology. In contrast, dimorphism in ventral skull shape shows significant phylogenetic signal and is significantly correlated with mating system. This study highlights, and provides new quantitative evidence for, the cranial morphological traits (rostrum width, relative palate length, basicranium width, lateral extent of mastoid), in addition to qualitative observations of sagittal and lamdoidal crest size, and canine length, that vary most with sexual dimorphism and mating system in pinnipeds.

Using morphometric analysis and comparative descriptions, we show that a small skull specimen of Enaliarctos emlongi (USNM 314290), originally thought to be juvenile, is most likely an adult female, and that the holotype of E. emlongi (USNM 250345) most likely represents a male. This interpretation implies that E. emlongi shows considerable size dimorphism (female skull approximately 70% male length) and that the differences in snout and canine morphology between male and female was similar to that of a modern otariid (Howell 1929; Cassini 1999; Sanfelice and de Freitas 2008). The dorsal skull morphology of USNM 314290 shows morphological features (low sagittal and lambdoidal crests, reduced rostrum width) characteristic of female pinnipeds (Howell 1929; Brunner et al. 2004; Sanfelice and de Freitas 2008; this study) (Fig. 8). Dorsally, the position of the female E. emlongi skull in morphospace is unexpected, when compared with extant taxa, because it plots outside of otariids, and far from the male skull (Supporting Information Appendix 2l). The position of the female skull is likely due to postdepositional/diagenetic deformation of the female skull, distorting the dorsal landmarks, particularly in the rostrum (landmarks 1–6, and focused around landmarks 2–4 describing the structure of the nasal). The distortion of the rostrum does not appear to have had such a great effect on the configuration of the ventral morphology. Ventrally, both the male and female skulls of E. emlongi plot near otariid taxa, and show strong dimorphism along the PC1 axis (Fig. 6), with the male showing greater lateral expansion of the basicranium/mastoid and palate (in agreement with qualitative observations noted in the comparative description) (Fig. 8). However, along the PC2 the relative positions of the male and female are opposite that of the general trend seen within the otariids (with males plotting more positively along PC2), and similar to that of Hydrurga and Leptonychotes. It is difficult to draw firm conclusions on the meaning of the precise positions of the male and female relative to one another, as the apparent reversal of position along PC2 may represent a genuine difference, indicative of other influences (such as ecology) on the morphology in conjunction with sexual dimorphism, or may simply be the result of small sample size (as when there is an overlap in male and female morphospaces, or when taxa show a wide range of morphospace, the centroids of the male and female morphospaces can run opposite to the position of randomly sampled individual pairs). Overall, E. emlongi falls within the PC morphospace of modern otariids (particularly in ventral), and shows a similar magnitude of difference in morphospace, and in shape difference relating to the expansion of the basicranium and mastoid, supporting our reconstruction of E. emlongi as strongly sexually dimorphic and suggesting that otariids may represent a good approximation for the ecology of basal pinnipeds. We therefore provide the first strong evidence that Enaliarctos was highly sexually dimorphic, and additional evidence for the basal ecology of the pinniped lineage.

image
Comparative skull shape dimorphism. (A) Enaliarctos emlongi (USMM 314290) female skull (left) compared to E. emlongi (USNM 250345) male skull (right), in dorsal view. (B) E. emlongi (USNM 314290) female skull (left) compared to E. emlongi (USNM 250345) male skull (right), in ventral view. (C) Arctocephalus australis (UNSM 504898) female skull (left) compared to A. australis (UNSM 239140) male skull (right), in dorsal view. (D) A. australis (UNSM 504898) female skull (left) compared to A. australis (UNSM 239140) male skull (right), in ventral view.

Multiple issues still exist concerning the taxonomy of stem‐pinnipeds. The genus Enaliarctos is considered a "meta‐taxon," and as such lacks unambiguous characters to support monophyly or paraphyly (Gauthier 1986; Berta 1991). A number of Enaliarctos species share some supposedly autapomorphic characters, with the dental morphology seeming to provide the only traits that separate the species, although in some cases even those traits vary to produce some overlap between species. Moreover, a number of the Enaliarctos species, such as E. emlongi, E. barnesi, and E. tedfordi, have overlapping or adjacent stratigraphic distributions within the late Oligocene to early Miocene Yaquina Formation of Oregon (Berta 1991; Prothero et al. 2001) (Supporting Information Appendix 1a). Also, the ranges of both E. emlongi and the related stem‐pinniped Pinnarctidion rayi overlap in the Nye mudstone, which overlies the Yaquina formation (Berta 1991; Berta 1994). The "meta‐taxon" status of Enaliarctos, combined with the number of similarities with other time equivalent stem‐pinnipeds, such as Pinnarctidion, suggests that a major revision of the broader taxonomy of the group is required. With such a revision it might be possible to determine whether other stem‐pinniped taxa are dimorphic, further testing the new hypothesis that sexual dimorphism is primitive for the pinniped lineage.

Identifying sexual dimorphism at the base of the pinniped phylogeny has a number of implications for understanding pinniped evolution and for constraining the probable mating system of the crown‐group ancestor. Given the high degree of dimorphism in E. emlongi, and the close correlation shown between modern pinniped dimorphism and mating system, we suggest Enaliarctos exhibited a land‐based, highly polygynous (harem) breeding system similar to modern otariids. The fossil taxon Desmatophoca, currently classified as a sister clade to Phocidae, also displays evidence of SSD (Deméré and Berta 2002), further constraining the ancestral breeding system of crown Pinnipedia to have been a highly polygynous, harem‐based system. This reconstruction implies that the appearance of a nonharem system characteristic for most phocids was a reversal, assuming the pre‐enaliarctine ancestor of the pinniped lineage did not form harems. This assumption seems reasonable considering that the earliest members of the stem‐pinniped lineage were otter‐like in their body size and ecomorphology (e.g., Puijila, Potamotherium) (Savage 1957; Tedford 1976; Rybczynski et al. 2009), having been derived from a terrestrial musteloid common ancestor (Nyakutura and Bininda‐Emonds 2012). Modern musteloids range from generally monogamous to weakly polygynous (and varying degrees of promiscuity), with higher dimorphism in small‐bodied mustelid species, and none are known to form harems (Weckerly 1998; Johnson et al. 2000; McPherson and Chenoweth 2012).

The origination of a highly polygynous mating system in pinnipeds may have occurred through sexual selection for larger males following the evolution of female aggregation; females selected males possessing more elaborate musculature for territorial display and male–male competition. This was likely followed by the evolution of harems and to the concentration of reproductive success among a few individual males, with breeding pressure from satellite males on females functioning to maintain the harems (Trillmich and Trillmich 1984), while also potentially allowing some genetic variability through occasional copulations (de Bruyn et al. 2011). In modern pinnipeds there is a close association between the location of breeding colonies and ocean upwelling sites (Trillmich and Trillmich 1984; Black et al. 2011), implying that climate change, as it relates to the location and intensity of upwelling sites, may have played a role in providing the original selective regime for polygyny. The most conservative estimate for divergence of enaliarctines from the crown pinniped lineage, based on the age of the oldest Enaliarctos and its phylogenetic position, is 27 Ma, at the end of the Oligocene (Berta 1991; Nyakutura and Bininda‐Emonds 2012). In the late Oligocene and throughout the early to middle Miocene there were major changes in climate and reorganization of thermohaline circulation (Flower and Kennett 1994; Zachos et al. 2001; Pälike et al. 2006), suggesting that climatic factors, such as glaciations, an increase in ocean stratification, and development of new coastal upwelling sites (Deméré et al. 2003; Jones and Goswami 2010b), may have played a role in the origination and/or maintenance of female aggregations and, later, polygynous mating systems. The reversion to a weakly polygynous, nonharem system in phocids may have followed a shift to a more aquatic lifestyle in areas of higher marine productivity, such as the generally colder, higher latitude waters encompassing much of their current distribution (Naughton 2012). If food sources are more dispersed, the formation and maintenance of harems could be more difficult, leading to a shift toward mating in water and weak polygyny.

This study is the first to statistically confirm a link between cranial dimorphism (in both size and shape) and reproductive strategy as well as to find evidence for sexual skull shape dimorphism in fossil stem‐pinnipeds. We have shown a deep origin of skull shape dimorphism in pinnipeds and thus of highly polygynous, likely harem‐forming land‐based mating systems. These combined results provide greater insight into the evolution of pinnipeds, the modern phylogenetic distribution of their mating systems, and our understanding of the coevolution of morphology and behavior of sexually selected traits in mammalian systems.

ACKNOWLEDGMENTS

This manuscript was improved through comments by Katrina Jones and an anonymous reviewer. We thank the staff of the Smithsonian Institution's Natural History Museum for access to the modern pinniped collections and fossil material, and for loaning the small skull specimen for use in this study, the staff of the Penn State Centre for Quantitative X‐Ray Imaging for use of their microCT scanning equipment, Alex Tirabasso for his assistance in manipulating the CT scan data and creating of the 3D model of the middle ear structures, Alan McDonald for assistance with skull line drawings, the R Foundation and the various package authors, and for discussions on the methods we would like to thank J. Claude, T. Hossie, S. Hinic‐Frlog, L. Schmitz, N. Campione, C. Van Buren, and C. Brown. Funding support for this research was provided by the Natural Sciences and Engineering Research Council of Canada (NSERC), which provided a Canada Graduate Scholarship (CGS M 410157‐2011) to T.M.C., Postgraduate Scholarship (PGSD3‐391699‐2010) to D.F., and Discovery Grant (RGPIN 312193‐2005) to N.R., as well as the Canadian Museum of Nature, and Carleton University. The authors declare no competing financial interests.

    DATA ARCHIVING

    The doi for our data is 10.5061/dryad.24r7q.

    Notes :

    • All data included as supplementary information.

    Number of times cited: 12

    • , The tympanic region of Otaria byronia (Otariidae, Carnivora) – morphology, ontogeny, age classes and dimorphism, Journal of Anatomy, 232, 1, (54-71), (2017).
    • , Sex‐related differences in the postmolt distribution of Weddell seals (Leptonychotes weddellii) in the southern Weddell Sea, Marine Mammal Science, 34, 2, (403-419), (2017).
    • , Sexual Dimorphism, Encyclopedia of Marine Mammals, 10.1016/B978-0-12-804327-1.00226-0, (848-853), (2018).
    • , Pinniped Evolution, Encyclopedia of Marine Mammals, 10.1016/B978-0-12-804327-1.00196-5, (712-722), (2018).
    • , Territorial Behavior, Encyclopedia of Marine Mammals, 10.1016/B978-0-12-804327-1.00018-2, (983-990), (2018).
    • , Eotaria citrica, sp. nov., a new stem otariid from the “Topanga” formation of Southern California, PeerJ, 5, (e3022), (2017).
    • , Sexual selection on skeletal shape in Carnivora, Evolution, 70, 4, (767-780), (2016).
    • , Combining geometric morphometrics and finite element analysis with evolutionary modeling: towards a synthesis, Journal of Vertebrate Paleontology, 10.1080/02724634.2016.1111225, 36, 4, (e1111225), (2016).
    • , Cope's rule and the evolution of body size in Pinnipedimorpha (Mammalia: Carnivora), Evolution, 69, 1, (201-215), (2014).
    • , Reproductive Structures, Strategies, and Patterns, Marine Mammals, 10.1016/B978-0-12-397002-2.00013-2, (465-532), (2015).
    • , Elephant seal (Miroungasp.) from the Pleistocene of the Antofagasta Region, northern Chile, Journal of Vertebrate Paleontology, 35, 3, (e918883), (2015).
    • , New data on the early odobenid Neotherium mirum Kellogg, 1931, and other pinniped remains from the Sharktooth Hill Bonebed, California , Journal of Vertebrate Paleontology, 10.1080/02724634.2018.1481080, (1-14), (2018).