AMNH

American Museum of Natural History, New York

Otters have often been divided into two functional groups based on dental morphology: those with bunodont carnassials and those whose carnassials form blades [33], [42] and [56]. The first group is often referred to as “crab-eating otters” and the second as “fish-eating otters”, although this is an inaccurate description of their diets [32]. In the original formulation by Pohle [55], the crab-eating otter group included the small-clawed (Amblonyx), clawless otters (Aonyx), and the sea otter (Enhydra), all of whom include crabs in their diet. “Crabotters” were considered by him to belong to a taxonomic group distinct from all other otters. As today's categories are more about dental morphology, the sea otter is the sole “crab-eating otter” as it is the only extant lutrine with bunodont teeth. All other extant lutrines are members of the second group. To reduce confusion, extant otters will be referred to as “bunodont” or “nonbunodont” in this paper.

While both bunodont and nonbunodont lutrines have been found in the fossil record, this paper focuses on bunodont forms from Africa. Confusion exists over which African material should be referred to Sivaonyx and which to Enhydriodon, reflecting the general uncertainty about these taxa. A smaller form, Vishnuonyx, is also present. The taxonomy of all bunodont forms, including those from Eurasia and North America, has been the subject of discussion for some time [8], [42], [43], [52], [65], [66] and [69] and will not be dealt with here (a discussion of valid and invalid bunodont taxa by continent can be found in Pickford's review [52]).

Although rare, postcrania presumably belonging to extinct bunodont lutrines are known [1], [21], [25], [33], [49], [50], [53] and [56]. Assignment to genus in the past has often been based on craniodental material from the same locality. As with the extant lutrines, grouping fossil lutrines solely based on dental morphology obscures the diversity of postcranial morphology present in this group. Despite the great variation in size and morphology, few studies have attempted functional analyses of this material [33] and [49]. As Enhydriodon and Sivaonyx are being found at an increasing number of sites in Africa [18], [21], [25], [26], [27], [42], [43], [49], [51], [59] and [66], understanding their behavior and ecology becomes critical for reconstructing the past. This paper, therefore, examines the most common postcranial element in the fossil record of African lutrines, the femur, to determine similarities and differences in behavior among extinct forms and among extinct and extant forms with the hope of better understanding the behavior and ecology of these forms. The first descriptions of these femora are also provided.

While taxonomic issues are beyond the scope of this paper, the various taxonomic assignments of these specimens will be discussed briefly. While two of the included specimens have not been mentioned previously in the literature, two undescribed specimens have received attention. All fossils can be found in Fig. 1, while extant lutrine femora can be found in Fig. 2 for comparison.

The four fossils clearly differ in overall morphology from extant lutrines, other mustelids, and ursids. One would predict that as a nonaquatic species evolves to a larger size, it will avoid greater stresses by changing some aspect of its design (e.g., morphology, posture, locomotor behavior) to avoid mechanical failure [3], [4] and [5]. However, in several cases (e.g., midshaft ML width, biepicondylar width), the fossil femora retained similar proportions to extant lutrines. This morphological similarity suggests that they must have changed some aspect of their locomotion and/or posture if they were terrestrial. Another possibility is that some or all are aquatic or semiaquatic, as the more aquatic the species, the fewer forces engendered by contact with a resistant substrate (e.g., river bank, tree limbs, and rocks). In that case, body size presumably affects morphology as one might need to alter morphology to maintain efficiency while propelling the larger body mass through water. Such changes would presumably have more to do with altering muscle size and/or leverage than with features associated with weight-bearing.

The fossils also differ from one another in both size and morphology. The Omo and Hadar femora are more alike than they are to the other two specimens despite the differences in size between the two. In the variables considered above, the Hadar specimen was the most extreme in condylar morphology. The West Turkana specimen has a distal end that is more similar to extant lutrines than the other fossil femora. It is also shorter relative to its overall femoral size than the other fossils. While the West Turkana and the Langebaanweg specimens were most similar overall, they differ in many respects. For a given variable, the Langebaanweg specimen was more likely to fall with generalized mustelids than with any other group.

The question then becomes, what is the meaning of these differences? It is clear that these specimens are not merely enlarged versions of extant mustelids, whether lutrine or not. While increasing body size probably plays a role in shaping the morphology of these femora, these fossils do not always follow allometric trends seen in extant taxa.

Otters exhibit a range of swimming behaviors whose usage is determined by water depth and displacement pattern [14]. In a study of North American river otters [14], Lontra canadensis, alternate hindlimb paddling occurred at the surface when swimming in a straight line and dorsoventral undulation was used for submerged swimming. Quadrupedal or forelimb paddling was used for maneuvering. In contrast, Enhydra undulates the lumbosacral and caudal spine dorsoventrally in combination with simultaneous strokes of the hindfoot as its primary means of swimming quickly [29], [61] and [71]. At slower speeds only the hindlimbs and tail are involved. Enhydra is also more buoyant than river otters. When resting on its back, the hindlimbs may be used alternately to propel the animal in a straight line. The expanded foot is moved caudally by extension of the knee and ankle during the propelling stroke. At the end of the stroke, the limb is moved cranially by flexing these joints [61]. Tarasoff et al. do not mention femoral movement, as the femur of Enhydra is hidden within loose skin.

No otter oscillates a true hydrofoil (pectoral, pelvic or caudal), as seen in pinnipeds, sirenians, and cetaceans [15]. Hydrofoils provide the most efficient means of swimming. Most otters have long tails that are not particularly efficient during swimming. However, Enhydra has reduced its tail length [61] and Pteronura has broadened its tail [14]. Enhydra also has more of the hindlimb contained within loose body skin than most other otters [61]. Shortening and widening the tail and encasing portions of the lower limb in loose skin are presumably evolutionary steps towards usage of the caudal end of the body as a hydrofoil.

In a study using landmark analysis of mustelid humeri, ulnae, femora, and tibiae, Schutz and Guralnick [57] demonstrated that the morphology of the femur and ulna of aquatic mustelids is not as differentiated from other mustelids as is that of the humerus and tibia. They found that the femora of lutrines were shorter and the greater trochanter and gluteal tuberosity were more robust than in other mustelids, findings that are supported by the present paper.

Schutz and Guralnick grouped Enhydra, Lutra, and Lontra together in their aquatic group and did not consider them separately. However, it is clear that the femoral morphology of Enhydra is unlike that of other otters (Fig. 2). While many mustelids may swim, they are locomotor generalists. Only the sea otter, E. lutris, and to some degree the marine otter, L. felina, are truly aquatic. Therefore, examining the postcrania of other aquatic carnivorans, namely the pinnipeds, can be instructive in understanding morphology associated with aquatic behavior.

Among the three families of pinnipeds, the otariids (fur seals and sea lions) and odobenids (walruses) are most like aquatic mustelids in their terrestrial posture. Otariids and odobenids use their hindlimb for weight-bearing locomotion on land. In contrast, the phocid hindlimb extends posteriorly and cannot be brought under the body to bear weight during terrestrial locomotion, thus the orientation of the femur relative to the os coxae is quite different. Otariids rely on pectoral oscillation for locomotion, while phocids and odobenids rely on pelvic oscillation [15].

Pinniped femora (Fig. 13), in general, are short, relative to their body size, as in otters. The fovea capitis is difficult to see as this group has lost the round ligament [2]. The shaft has become compressed anteroposteriorly to a degree not seen in extant mustelids. The condyles, in particular, are very short anteroposteriorly, a feature seen to a lesser degree in the West Turkana, Omo and especially the Hadar femora.

Most significantly, pinnipeds have reoriented their head more superiorly relative to the shaft and shortened the neck. What is left of the neck is also angled more superiorly such that the distal border, which is usually inferior, is now medial. The greater and lesser trochanters are extremely enlarged.

Since lutrines use a different form of aquatic or semiaquatic locomotion, one would not expect them to be identical to pinnipeds. For truly aquatic forms, such as Enhydra, one would expect some similarities to pinnipeds due to the shared femoral posture (posterior extension) during undulation.

Enhydra and the Omo and Hadar fossils are similar in some ways to the pinniped morphology described above (note that some of these features were not quantified in the present study as angles were not included). The Omo fossil expresses this morphology to the greatest degree with the superiorly oriented articular surface of the head, shortened neck (to a lesser degree than pinnipeds), enlarged greater and lesser trochanters, and the inferior border of the neck having become more medial. Enhydra and the Hadar femur have enlarged trochanters, but the head and neck are not rotated superiorly to the same degree. Enhydra and the Omo and Hadar femora also share a feature not seen in pinnipeds: a well-excavated superior portion of the neck. The West Turkana specimen has a shortened neck, but has not reoriented the head or enlarged the lesser trochanter. The Langebaanweg specimen has none of these features.

Among other extants, only the giant otter, Pteronura, shows any similarity to Enhydra. Pteronura has enlarged the greater trochanter and has a relatively short, mediolaterally wide shaft. The lesser trochanter is not enlarged to the degree seen in Enhydra and is within the size range of other extant otters. The giant otter, as the name implies, is quite large and is the longest otter, reaching up to 2 m in length [32]. However, despite its length, it is not as heavy as Enhydra. These morphological similarities may be due not only to Enhydra and Pteronura being the two largest extant otters, but also to similarities in the structure of their habitat. Enhydra swims in deep ocean water, while Pteronura prefers the deeper regions of lakes [6] and [32]. All other nonbunodont otters prefer shallower areas along the margins of bodies of water [32]. While no laboratory analyses have been done of the locomotion of Pteronura as they have for other otters, one would predict that they might include more undulation and Enhydra-like movements on a daily basis than the typical river otter.

Having a superiorly oriented head and neck, as seen in the Omo fossil and to a lesser degree in Enhydra, changes the orientation of the entire femur with respect to the os coxae. This reorientation changes both rotatory ability and muscle leverages. Rather than the femur being relatively valgus and moving parasagittally, as in terrestrial cursorial mammals, or more varus, as in ambulatory mammals, the femur can be rotated posteriorly more easily. In fact, Enhydra has been shown to bring its hindlimb further superiorly past the tail when swimming for increased thrust, while other otters only bring their hindlimb to the level of the tail [61].

Enhydra has a large lip on the ilial portion of the acetabulum (arrow in Fig. 14). When Enhydra swims ventrum down or when it is terrestrial, this lip would be anterior and slightly superior. This lip prevents hyperflexion and excessive medial rotation. There is little to no lipping of the ilial portion of the acetabulum in other otters.

Enhydra also has widened, laterally flaring iliac blades in the ossa coxae such that the proximal half of the ilium projects laterally at roughly a 135° angle from the main axis of the rest of the os coxae (Fig. 14). This flaring not only provides a wider surface area of attachment for muscles, but also a radical change in the orientation of those muscles. Several muscles attach to the iliac blade including the iliocostalis and longissimus lumborum on the internal surface and the gluteus medius on the external surface. The longissimus lumborum is an important extensor of the lumbosacral region of the vertebral column. Presumably, having broader, everted ilia increases the leverage for this group muscle and would create more powerful dorsoventral undulation in this region of the spine. The iliocostalis lumborum is important in lateral flexion, which is important for turning, but not a component of rectilinear movement in sea otters. A somewhat similar ilial morphology in phocine seals has been suggested to be due to the need to increase the surface area of attachment for the iliocostalis lumborum, a muscle that is important for much of the lateral body movements used in swimming in this subfamily [2] and [30].

In terrestrial mammals, the gluteus medius is the strongest extensor of the hip and an abductor [44]. The insertion point for the gluteus medius, the greater trochanter, is also enlarged in Enhydra. The greater trochanter also serves as the insertion point for the piriformis and the gluteus profundus, which support the actions of the gluteus medius. Perhaps Enhydra ilial morphology also helps to increase the powerstroke from the hindlimbs during undulation or while changing direction. Extension is apparently a very important part of Enhydra locomotion as the gluteal tuberosity (= third trochanter) is larger in Enhydra than in any other extant otter (it is also enlarged in the Omo femur, although it projects posteriorly and not laterally). This tuberosity is the site of attachment for the gluteus superificialis, another hip extensor. One should note that the lesser trochanter is also quite enlarged in Enhydra. This is the insertion point for the iliopsoas, an important flexor of the hip used during the recovery phase of locomotion.

In contrast to Enhydra, other extant otters have nonflaring ilia that are very similar to nonlutrine mustelids. The size of the greater trochanter varies and the lesser trochanter is prominent, but not as enlarged. The gluteal tuberosity is discernible, but not enlarged. These differences may reflect the fact that undulation is not a dominant form of locomotion in these otters.

Unfortunately, no os coxae were found with the femora included in this study. One might predict that the Hadar and Omo femora would be associated with ossa coxae more similar to those of Enhydra based on the morphology of the head, neck, and greater trochanter. The West Turkana and Langebaanweg specimens would probably have been associated with ossa coxae like other lutrines. In fact, S. beyi, a species that has been reconstructed as not being particularly aquatic (see Figure 7 in Peigné et al. [49]), from the published photographs appears to have had the nonflaring ilia like river otters given what remains of the fragmentary os coxae.

Of the four fossil femora included in this study, it is clear that the Langebaanweg specimen is least likely to be aquatic in the manner of Enhydra. In some features, it appears more like nonlutrine mustelids. As a good generalist, it is possible that it was semiaquatic, like extant river otters, or occasionally aquatic, like many nonlutrine mustelids.

The West Turkana specimen is also relatively generalized. Among the fossils, the West Turkana specimen has the shortest functional length relative to overall bone size. In that manner, it is most similar to extant lutrines. The condyles are also relatively short anteroposteriorly and the neck is short. Overall, this specimen could easily have belonged to a semiaquatic form.

The largest fossil, Omo, does have a few features that may be indicative of more aquatic behavior: the superiorly oriented head, more sharply angled neck, and enlarged trochanters. While the Hadar femur has not reoriented the head, it does have enlarged trochanters. The Omo and especially the Hadar femora also had relatively anteroposteriorly short condyles (also seen in the West Turkana femur), although not to the degree seen in otariids. However, this is not a feature expressed in Enhydra or other less fully aquatic otters.

Could the rotation of the head and angling of the neck in the Omo specimen be due to the need to transmit increased forces due to their much larger body weight than extant mustelids? This feature is present to a lesser degree in the much smaller Enhydra. Among carnivorans, it is not seen in any extant ursid (including pandas), nor is it seen in felids, including large, extinct machairodont forms such as Homotherium. Although not recognized until now, among extinct bunodont otters, it is present in Enhydritherium terraenovae (see Figure 7 in Lambert [33]), but not in the larger S. beyi (see Figure 7 in Peigné et al. [49]). While I have not surveyed all extinct large mustelids, this does not seem to be a feature found in any specimen except some of those potentially belonging to bunodont lutrines. Granted, not many mustelids approach the size of the Hadar and Omo specimens. Even the large mustelid from Lothagam, Ekorus ekakeran (which also lacks this feature), has a femur that is only slightly longer than the West Turkana femur and is more gracile. Thus, it appears that this feature is more likely to be due to locomotor adaptation than body size within this group.

The Omo and Hadar femora do have some features that would argue against aquatic behavior along the lines of Enhydra. Both fossils have the relatively long functional lengths for their overall femoral size. Otters, with the exception of Amblonyx cinereus, tend to have short, robust femora, with the largest otters Enhydra and Pteronura taking this trend to the extreme. Among nonlutrine mustelids, the most scansorial taxa have the longest femora relative to overall bone size. The two fossils fall in the middle with the majority of lutrines and nonlutrine mustelids. Whether their femora are long relative to their body size for a lutrine or whether their femora are simply gracile relative to their overall body size for a lutrine cannot be determined as body size is unknown. In either case, this does not preclude aquatic behavior as they are certainly within the range of extant otters.

In sum, the two smallest femora from Langebaanweg and West Turkana, while large, can be reconstructed as locomotor generalists. The West Turkana form appears more like lutrines than the Langebaanweg form and was probably semiaquatic. In the Langebaanweg form, this behavior cannot be ruled out. The two largest specimens (Omo and Hadar) are unusual both in their size and in their morphology. They are clearly engaging in locomotion unlike any extant mustelid. While the morphology of the proximal end points to more aquatic behavior than in most lutrines, the relative length to overall size is more similar to that of generalist mustelids, including semiaquatic lutrines. Future study of the less common postcranial elements of bunodont otters from Omo and Hadar may help to resolve this mystery.

While it is not reasonable to reconstruct the behavior and ecology of a taxon solely on femoral morphology, some remarks can still be made. If the larger taxa were aquatic to the degree of Enhydra, they may have preferred the deepwater lakes searching for a variety of foods. While narrow streams and river banks may be favored habitats of extant otters such as Lutra [32], a gigantic, more aquatic otter might not be able to maneuver as well in such an area. Today, the largest “fish otter”, Pteronura (giant otter), prefers clear, black water which is usually the larger, fish-rich areas [6] and [32]. Moderately large size would not preclude the use of river banks in species that were more semiaquatic (e.g., the West Turkana species).

Of course, as mustelids, these species probably came out of the water as even Enhydra does. Among otters, A. cinereus, the small-clawed otter, spends the most time away from the river. This relatively long-limbed otter often forages for invertebrates at night away from the river bank in damp vegetation, while other sympatric species of otter remain in the river [32]. Amblonyx has not been reported to dive or fish in deeper waters [32], and so represents the opposite extreme from Enhydra in being the least aquatic otter.

It should be noted that while Enhydra is often thought of being predominantly durophagous, other types of foods such as fish and cephalopods can be important components of the diet. Kenyon [29] found that 50% of the stomach contents of over 300 sea otters were fish, with only 37% of the contents being mollusk and 11% sea urchin. Freshwater otters, despite being referred to as “fish otters”, also consume a wide range of foods. The giant otter, Pteronura, has been known to eat tapir dung, invertebrates extracted from the mud, and even small vertebrates (bird, small mammal, snake, small caimans, and turtles) [32]. Amblonyx and Aonyx prefer crabs and other invertebrates even without bunodont teeth [32].

Pickford [52] has suggested that Enhydriodon was a molluscivore based on the reinforced anterior snout dentition, powerful masticatory musculature, and the possession of bunodont teeth with thick enamel. In his analysis, the breadth–length relationship of m1 was found to discriminate between otters whose primary prey was shellfish and those whose primary prey was fish. Both Sivaonyx and Enhydriodon have a breadth–length relationship suggesting shellfish predation. However, Pickford suggests that those taxa with thick enamel (in his analysis, all Enhydriodon and African Sivaonyx) would have preyed on harder more abrasive food like freshwater mollusks. Pickford, therefore, suggests that Enhydriodon preyed extensively, although not exclusively, on freshwater mollusks.

Larger, extinct bunodont otters may have foraged for a wide variety of foods in addition to their primary prey. Bunodont teeth may have allowed them to deal with larger fish with harder external coverings (e.g., catfish) or even other vertebrates as seen in Pteronura. Enhydritherium, a bunodont otter from North America, has been shown to have extremely powerful neck muscles, which Lambert [33] suggested were used to hold on to extremely large fish. Bunodont dentition would not have prevented the intake of a large amount of softer foods, as seen in sea otters today, and may have been useful for dealing with relatively large fish.

The four fossil femora exhibit very different morphologies. The differences in morphology and extreme differences in size among the four specimens suggest that they each belong to a different species. While one might speculate that the Omo femur is male and the Hadar femur is female, this is extremely unlikely. In the Omo femur, the head is angled more superiorly, the neck is shorter, and the distance between the greater and lesser trochanter is much shorter relative to overall size than in the Hadar femur. The condyles of the Hadar femur are unique.

The two largest specimens, Omo L183-14 and AL 166-10, have a morphology that is most consistent with an aquatic lifestyle. Given their respective ages, it is possible that the Omo species evolved from the Hadar form. This would have required an increase in size along with an increase in the relative length of the bone. The head would become angled more sharply superiorly and the relative size of the intertrochanteric length would shrink. The condyles, however, would have had to enlarge somewhat.

The West Turkana specimen looks more like a much larger version of a river (“fish”) otter, but with a higher greater trochanter, shorter neck and an increased ridge running from the greater trochanter to the gluteal tuberosity. This femur could certainly have belonged to a lutrine, whether bunodont or not. If not a bunodont otter, this individual is much larger than any known nonbunodont lutrine in Africa such as Torolutra. The dating of this specimen overlaps with both the Hadar and Omo specimens. While it could represent the ancestral form relative to the more derived Hadar and Omo specimens, it may also simply be another large otter living in eastern Africa during the Pliocene. To evolve from the West Turkana form into the other two would require an increase in overall size, a reorientation of the head and neck more superiorly, a slight increase in neck length, and an increase in size and medial projection of the lesser trochanter.

Whether any of the above specimens belong to Enhydriodon or to Sivaonyx or to something different is unclear. Association with dental material is unclear, although it is most parsimonious to assume that they are the same taxon as any similarly-sized lutrine dental material found from the same locality and geological age. Full description of these new taxa will be left for a later paper [68].

The oldest specimen, PQ-L 41523, from Langebaanweg is more like a generalized mustelid than any of the other taxa. It is certainly possible that it is a primitive bunodont otter. The morphology is generalized enough that it could be ancestral to the others.

In sum, there are multiple species of large mustelid present in Africa during the Neogene. The four femora presented here are not firmly associated with dentition, but are definitely mustelid. While there is no clear picture of how these species might be related, it is certain that they were not engaging in exactly the same forms of locomotion. If they do belong to bunodont lutrines, then they appear to be more diverse in postcranial morphology than in dental morphology. The oldest (Langebaanweg) appears the most generalized, the West Turkana form is more like extant river otters, and the two largest forms (Omo and Hadar) are unique in both their extreme size and morphology, with features that point to aquatic behavior. The increasing number of identifications of large mustelids, both lutrine and nonlutrine, in the fossil record points to the importance of these taxa for understanding Neogene paleoecology. How might an otter (or a nonlutrine mustelid) with a femur within the size range of extant lions interact with other taxa? One of the limits of the present study was the lack of information on specific locomotor behavior across all extant lutrine and nonlutrine mustelid taxa. Only further functional studies of extant morphology and of the morphology of other craniodental and postcranial specimens of extinct mustelids will help determine whether these enormous mustelids were feared predators or gentle giants, rarely seen outside of the lake.