Daphoenodon (Borocyon) neomexicanus, Hunt, 2009

Robert M. Hunt, Jr., 2009, Long-Legged Pursuit Carnivorans (Amphicyonidae, Daphoeninae) From The Early Miocene Of North America, Bulletin of the American Museum of Natural History 2009 (318), pp. 1-95 : 23-83

Daphoenodon (Borocyon) neomexicanus , new species Figures 7 , 8 , 12–16

HOLOTYPE: F: AM 49239 , a skull with associated mandibles (figs. 7D, E, 12, 13).

HOLOTYPE HORIZON AND LOCALITY: Piedra Parada Member, Zia Sand Formation , collected in 1947 by Ted Galusha and party at Standing Rock Quarry , Canyada Piedra Parada , 7 mi south and 1.5 mi west of San Ysidro , Sandoval Co., New Mexico, NW1/4, sec. 11, T14 N, R1 E.

DIAGNOSIS: Smallest species of Borocyon , with basilar skull length of, 245 mm, and c– m2 length of, 97–100 mm, averaging small- er than both B. niobrarensis and B. robustum in body size and dentition (figs. 17, 18; tables 2, 3), and in males, by a more developed bony exostosis on the distal radius. Differs from Daphoenodon superbus (basilar skull length, 215–222 mm) by its greater size, longer forelimbs, larger m1 (fig. 18A), a ‘‘folded’’ M2, and proportionately longer m2 (fig. 18B).

REFERRED SPECIMENS: All referred material was collected at Standing Rock Quarry during the 1947 field excavation by the Frick Laboratory (American Museum): (1) F:AM 49240, left mandible, complete, with i2–i3, c, p1, p3–m1, alveoli for m2–m3; (2) F:AM 49240A, right mandible with c, p2–p4; (3) F:AM 49241, juvenile left mandibular fragment with p3–m3—this is probably the same individual as a basicranium (F:AM 68240) and an MT3 with unfused epiphysis (F:AM 68240A); (4) F:AM 68241, several ribs, podials, axis vertebra, both humeri and ulnae, right scapula, various vertebrae, a canine, metacarpals 1 and 2, and metatarsal 5; (5) F:AM 68242, left femur, articulated vertebrae, phalanges, carpals, both radii (without exostoses), partial articulated tarsus in matrix, metacarpals 4 and 5, and metatarsals 2, 3, and 4; (6) F:AM 68243, right femur, complete pelvis lacking anteriormost ilial blades, atlas vertebra; (7) F:AM 68244, partial fibula, phalanges, tarsals, articulated vertebrae, metacarpals 3 and 4, metatarsal 1; (8) various postcranial elements cataloged under field numbers as follows: Jemez 7-105, radius (with exostosis); Jemez 7-106, tibia; Jemez 6-86, femur; Jemez 5-77, distal femur; Jemez 5-72, distal humerus, proximal ulna; Jemez 5-73, distal humerus; Jemez 5-74, ulna; Jemez 5-77, atlas vertebra; Jemez 7-117, atlas and two articulated vertebrae.

DISCUSSION: The Standing Rock Quarry sample of Borocyon is made up of at least three disarticulated individuals including a juvenile (fig. 7C). In skull proportions, den- tition, and the postcranial skeleton this species is very similar to both Daphoenodon (D.) superbus and the large species of Daphoenodon (Borocyon) and serves as a structural and temporal intermediate. The mammalian fauna found with B. neomexicanus indicates a latest Arikareean (Ar4) age, hence the New Mexican carnivore precedes the larger earliest Hemingfordian Borocyon niobrarensis from the Great Plains, and apparently postdates the sample of D. superbus from the carnivore dens at Agate National Monument on faunal grounds.

The hypodigm from Standing Rock Quarry includes partially articulated postcranial material requiring additional preparation and study. Only the diagnostic holotype skull and mandibles, an unworn juvenile dentition, and previously prepared limb and foot bones necessary to establish the species as the oldest member of the Borocyon lineage are included in this study.

Two metapodials found in Blick Quarry (fig. 15), situated a few miles north of Standing Rock Quarry, can be referred to Borocyon robustum , suggesting that the transition from the smaller Standing Rock carnivore to large B. robustum is possibly documented in northern New Mexico. The failure to find B. neomexicanus in Arikareean rocks of Nebraska and adjacent states in the central Great Plains despite extensive field collecting suggests its actual absence from the fauna of that region, indicating that Borocyon may have migrated north into the midcontinent from the southwestern United States in the earliest Hemingfordian.

CRANIODENTAL OSTEOLOGY

CRANIAL: Borocyon robustum is represent- ed by three large, nearly intact crania (UNSM 25547, 25686, 26416) and B. neomexicanus by only one (F:AM 49239), crushed nearly flat. A skull assigned to B. niobrarensis (UNSM 44827, an aged individual), although without a rostrum and with extremely worn teeth, is intermediate in size. These species retain a similar skull form (see frontispiece), which derives from the short, broad skull of Daphoenodon superbus (CM 1589, 2774). The cranium, mandible and teeth are best represented for B. robustum ; this species is therefore the principal subject of this section, with B. niobrarensis and B. neomexicanus supplementing the account where appropriate.

Skull lengths of B. robustum were determined from the three crania and the sample of mandibles of B. robustum from the Runningwater Formation. An undistorted mandible (UNSM 25684) was fitted to the paratype cranium (UNSM 25686) so that upper and lower teeth occluded properly, and then, estimating from that relationship, basilar skull lengths corresponding to the remaining mandibles were calculated. A strong correlation exists between mandibular length and skull length for living Carnivora ( Christiansen and Adolfssen, 2005) . Dimorphism, presumably sexual, is reflected in the range of values for mandibles, and this is supported by ‘‘male’’ and ‘‘female’’ examples from a single quarry (UNSM 25549, 25550, UNSM locality Bx-7). The calculated values for skull lengths (using basilar length) for B. robustum , derived from seven mandibles and the three known skulls, yield a range of 258– 328 mm, a mean of 294.5 mm, with standard deviation of 20.8 and coefficient of variation of 7.0.

Basilar lengths of Borocyon robustum skulls (table 1) indicate the large size attained by these predators, and they show that the, 30–31 cm maximum value recorded for Borocyon is similar to this measurement for large New World early Miocene amphicyonine amphicyonids ( Amphicyon , Ysengrinia ), values similar to those for living large felids such as Panthera leo (, 24–34 cm), P. tigris (, 22–30 cm), and North American mainland Ursus arctos .

The skull of Borocyon robustum displays a rather short rostrum coupled to a long postorbital cranium (fig. 4, UNSM 26416; frontispiece, UNSM 25686), similar to the proportions of its contemporary Amphicyon galushai ( Hunt, 2003) . Among living carnivorans of similar size, the snout and orbital region of B. robustum are most like those of Ursus arctos , although the snout of the beardog appears somewhat shorter. It is not so swollen as in Panthera leo and P. tigris due to the absence in Borocyon of the particularly robust canines and incisors of these cats. Borocyon is also similar to U. arctos in the diameter and lateral placement of the orbits, whereas in the large felids the orbit is relatively larger and more forwardly directed. The enlarged infraorbital foramina of the large cats for sensory fibers supplying the facial vibrissae are lacking in Borocyon , which has a small foramen comparable to that of U. arctos .

Borocyon robustum retained the broad forehead of Daphoenodon superbus for expanded frontal (paranasal) sinuses, with the cranium constricted behind the orbits to enclose a relatively small braincase surmounted by impressive sagittal and occipital crests (see frontispiece). The volume of the braincase is proportionately smaller in B. robustum relative to living lions, tigers, and U. arctos of similar size; hence, the sagittal crest for attachment of temporal muscles in Borocyon is more prominent. At the base of the braincase adjacent to the narrow occiput, the squamosal root of each zygoma departs the skull at essentially a right angle, and the zygomatic arches extend farthest from the braincase at this point. The breadth of the zygomata and height of the sagittal crest demonstrate that a massive jaw musculature filled the temporal fossae, this volume augmented by the relatively small size of the braincase. The width of the palate is considerable, attaining, 9 cm in some individuals, measured between the posterior limit of the P4 carnassial blades.

The basicranium (fig. 13) is preserved only in the holotype cranium of B. neomexicanus (F:AM 49239) and, although badly damaged, in the referred skull of B. niobrarensis (UNSM 44827). The auditory bulla and surrounding basicranial structure are essentially the same as previously described for the holotype cranium of Daphoenodon superbus (CM 1589, Hunt, 2002a, fig. 23; 2002b, fig. 18). This plesiomorphic bulla is distinguished by its simple, flask-shaped form, by thin medial and posterior bulla walls, and by lack of posterior expansion of the middle ear cavity; this bulla morphology is characteristic of both subgenera of Daphoenodon .

DENTAL: Dimensions of the upper and lower teeth of the species of Borocyon are here compared with those of other Late Arikareean–early Hemingfordian amphicyonids (tables 2, 3). From the small species Borocyon neomexicanus to large B. robustum , the teeth increase in size over time, in keeping with the increase in body size. However, the dentition is well represented only in B. robustum .

Skulls of Borocyon robustum are most easily distinguished from those of the large contemporary amphicyonine Amphicyon by the teeth. Nine mandibles and three skulls from the Hemingford Quarries in the upper Runningwater Formation of Nebraska comprise a population sample that best defines Borocyon robustum . The teeth of these individuals show a remarkably uniform morphology, principally varying in size, which in North American amphicyonids is attributable to sexual dimorphism. The population from the Hemingford Quarries (primarily Quarries 7A–7B, 12D, 22, 23) includes the largest carnivores assigned to the genus. The population sample from the Bridgeport Quarries is made up of numerous isolated teeth that correspond in size to those from the Hemingford sites. Because the molars of Borocyon robustum are diagnostic, the Bridgeport sample can be confidently allocated to the species, based not only on teeth but also isolated limb and foot elements.

Skulls and mandibles of Borocyon robustum are identified by the following defining dental traits (figs. 4–8, 17, 18): (1) the premolars are not reduced as in Amphicyon but are well-developed grasping teeth; (2) M3 is lost, a characteristic known elsewhere in North American amphicyonids only in temnocyonines (M3 is reduced in size in D. superbus and B. neomexicanus ; it is minute but still present in the only known skull of B. niobrarensis ); (3) the shearing P4 is longer and narrower than that of contemporary Amphicyon and retains a more prominent protocone; (4) M 1 in occlusal view is a large triangular tooth with tall paracone and metacone and a conspicuous notch (fig. 5) indenting the anterolingual cingulum; (5) M2 is particularly diagnostic in its ‘‘folded’’ shape in B. robustum , B. niobrarensis , and B. neomexicanus ; that is, in posterior view the lingual half forms a prominent angle (105 ° – 130 °) relative to the labial half of the tooth. Dentally plesiomorphic daphoenines ( D. superbus , D. falkenbachi , D. skinneri ) have an M2 with a much flatter occlusal surface; (6) the posterior cingulum of p4 is ‘‘squared’’ (fig. 8A), not rounded as in Amphicyon —a few individuals show some rounding of this posterior margin, but p4 always remains more laterally compressed than in Amphicyon ; (7) m1 is a massive, robust carnassial that retains a prominent metaconid and a relatively narrow talonid, not a broadened talonid as in Amphicyon (because the B. robustum M1 paracone is a tall cusp, it occludes far ventrad against the labial face of the m1 talonid, cutting a broad facet, which is particularly evident in the Hem- ingford Quarries population); (8) m2 is elongate with a broad occlusal surface for crushing against the lingual half of the ‘‘folded’’ M2—it is not as broad as m2 of Amphicyon ; (9) m3 is present, placed high on the ascending ramus, and tilted forward in an attempt to occlude with M2, due to loss of M3 (only the m3 trigonid occludes with M2, creating a small wear facet on the posterior cingulum).

Identification of Borocyon from a sinkhole or fissure in the Suwannee River, Florida, rests on a reconstructed palate with complete upper dentition (fig. 9), a few isolated teeth, edentulous mandibles, and numerous isolat- ed, diagnostic limb and foot elements. Additional craniodental material of Borocyon from the site is in private collections and not available. Although the size of the teeth in the palate suggests a large B. robustum male (particularly the canines), the isolated m1s and m2–m3 indicate small individuals, presumably females the size of B. niobrarensis . The p4s fall in the size range of B. robustum yet are not as large as p4s in the largest of the Hemingford Quarry individuals. Three edentulous mandibles do not equal the depth and robust quality of the Hemingford Quarry specimens and are similar in size to B. niobrarensis mandibles. The Florida sample is provisionally referred here to Borocyon cf. B. robustum ; postcranials indicate individuals of the population are somewhat smaller on average than the upper Runningwater hypodigm from the Hemingford Quarries— whether this is a smaller southern population coeval with B. robustum or possibly animals from a geologically older interval is uncertain. It is also possible that these individuals constitute a diachronous sample that accumulated over a significant interval in the sinkhole(s) in the Suwannee River.

Figures 17 and 18 present dental dimensions of B. robustum for m1–m2 and M1–M2 relative to other early Miocene daphoenines. The Hemingford Quarries and Bridgeport Quarries populations are plotted as B. robustum , and the few individuals from Oregon and Florida, although included in the B. robustum hypodigm, are separately indicated on the graphs. M1–M2 and m1–m2 are useful in distinguishing B. robustum from B. niobrarensis —the latter species is consis-

TABLE 2 Dental Measurements (in mm) of Lower Teeth of North American Early Miocene Amphicyonids

TABLE 2 (Continued)

TABLE 2 (Continued)

tently smaller, plotting in the lower part of the range of the former. When teeth of Borocyon neomexicanus are available, they are smaller than those of both B. robustum and B. niobrarensis .

Dimensions and the form of M1 and M2 distinguish Borocyon robustum from its contemporary Amphicyon galushai . These molars are particularly reliable in identification of the two species. M1–M2 of Borocyon are mesiodistally shorter relative to the upper molars of A. galushai , and Borocyon robustum has lost M3 whereas Amphicyon retains that tooth ( B. niobrarensis and B. neomexicanus have reduced M3s). The carnassials of B. robustum are more sectorial than those of A. galushai and the premolars better developed for grasping, together indicating a more carnivorous diet in Borocyon . Although B. robustum can often be identified solely by dental and cranial characters, the species is more confidently recognized using a combination of dental and postcranial traits. When craniodental traits of Borocyon robustum are combined with its postcranial skeletal features, the species is anatomically unique among Miocene amphicyonids of both the Old and New World.

POSTCRANIAL OSTEOLOGY

Although no complete associated skeleton of Borocyon is known, postcranial bones of B. robustum from the Hemingford Quarries, supplemented by isolated elements from the Bridgeport Quarries, provide a relatively comprehensive overview of the postcranial skeleton, particularly the limbs and feet. Additionally, the associated forelimb of the holotype of Daphoenodon (Borocyon) niobrarensis Loomis from the lower Runningwater Formation serves as a useful reference in reconstruction of the forequarters. This forelimb (ACM 3452), complete from the scapula to the ungual phalanges (fig. 11), indicates a slightly less evolved antebrachium and forepaw relative to B. robustum . With regard to B. neomexicanus , only limb proportions, the radius, and metapodials are considered here. Additional preparation of the material from Standing Rock Quarry is necessary to obtain a complete description of its postcranial skeleton.

An estimate of body size for Borocyon robustum is speculative, given the lack of associated material. However, mid-shaft diameters of femora have often been used as predictors of body mass ( Gingerich, 1990; Anyonge, 1993; see also Christiansen, 1999). The length and midshaft diameters of a complete femur (UNSM 26435: length, 378 mm; A-P and M-L diameters, 30.4 and 33.1 mm, respectively) of the species, probably representing a large adult male, exceed average values for the femora of Panthera leo (,324, 26, 28 mm) and P. tigris (,352, 27, 28 mm) presented by Bertram and Biewener (1990), suggesting at least a similar body mass perhaps slightly exceeding the mean for these large felids. On average, P. leo males range from 150 to 250 kg, and males of the Indian and smaller Sumatran P. tigris range from 100 to 258 kg ( Nowak, 1991). Based on dimensions of the femur, B. robustum surely reached 100 kg, and probably some males at, 150 kg.

FORELIMB

The proportions of the forelimbs of the species of Borocyon were compared with those of living ursids, felids, and canids (fig. 19A; table 4; appendix 1). A proportional increase in length of the radius relative to the humerus is evident as one progresses from ursids and felids to canine canids. Most New World amphicyonids for which an associated forelimb is known have a short lower forelimb, unlike living canids. Borocyon , on the other hand, has an elongated forelimb similar to proportions found in the wolf and cheetah (fig. 19A). An elongated forelimb (figs. 14, 15) is already present in the earliest known species of Borocyon , B. neomexicanus , and the lengthened forelimb is a principal synapomorphy of the species of the subgenus.

SCAPULA- HUMERUS

Poor representation of the scapula in Borocyon is likely due to targeted scavenging of the muscle mass concentrated at the shoulder—this bone survives for B. niobrarensis (fig. 20) and B. neomexicanus but not for B. robustum . Scapula and humerus are preserved in the associated forelimb of the holotype of Borocyon niobrarensis (ACM 3452), with the scapula nearly complete except for a small portion of the vertebral border (fig. 20A). The shape of this scapula is a composite of features of living ursids and large felids, and this shape was already characteristic of Daphoenodon superbus and Borocyon neomexicanus . No clavicle was present (it was already lost in D. superbus ); the large acromnion process is configured as in Ursus americanus (breakage precludes determination of a metacromnion).

The cranial border of the scapula and the relative volumes of the supraspinatus and infraspinatus fossae are nearly identical to those of Panthera leo and P. tigris ; in fact the overall form of the scapula is quite close to those of the large living felids. The principal osteological distinction of the Borocyon scapula from the scapulae of these cats is a broad teres process appended to the posterior axillary border, a process also found in living ursids (fig. 20). In the bears the teres process takes the form of a projecting bony plate whose lateral concave surface, termed the postscapular fossa, houses the subscapularis minor muscle. The postscapular fossa continues as a deep channel or groove (subscapular fossa of Davis, 1964) along the ventral axillary border, which houses the belly of the subscapularis minor muscle.

According to Davis (1949), fibers of the subscapularis minor of Ursus americanus maintain a separate identity from those of the subscapularis, both muscles inserting close together on the medial side of the humeral head, where they would effect medial rotation of the humerus. Canis and living felids lack the teres process and the subscapularis minor is not distinct from the subscapularis itself. In Ursus americanus the teres major muscle has been displaced to the medial side of the scapula by the subscapularis minor, where it arises from the surface of that muscle and from the thin adjacent edge of the scapula ( Davis, 1949). The form of the teres process of Borocyon niobrarensis and the postscapular fossa on its lateral face are developed as in living ursids and in amphicyonids where the scapula is known, and certain procyonids have a scapular anatomy presaging the ursid condition.

In order to confirm the anatomy of the postscapular fossa and subscapularis minor, R.M. Joeckel and I dissected the shoulder of the Malayan Sun Bear ( Helarctos malayanus ). The subscapularis minor, here a fusiform muscle belly that occupies the postscapular fossa and its continuation forward into the subscapular fossa or groove, is a ventral derivative of the subscapularis muscle. Fibers of subscapularis minor run subparallel to the bony margins of the groove. The belly of this muscle lies loosely in the postscapular fossa and along the axis

TABLE 3 Dental Measurements (in mm) of Upper Teeth of North American Early Miocene Amphicyonids

TABLE 3 (Continued)

of the groove and is separated from the bone by a thin fascial sheet; fibers do not attach strongly to the bone of the fossa or groove. As the subscapularis minor is followed craniad to the forward edge of the groove, its fibers merge with fibers at the ventral margin of subscapularis. Of interest is that subscapularis minor does not twist around the ventral border of the scapula as described for Ursus by Davis (1949), but rather joins the large subscapularis and loses its identity before reaching the medial face of the scapular blade. The subscapularis, considered a major internal rotator of the arm, had 6–7 major pinnate bellies that join to form a thick, tendinous aponeurosis covering the medial head of the humerus. The ventral edge of this muscle extends 1–2 cm ventral to the scapular border as a fleshy muscle mass, which sends a slip caudad into the postscapular fossa. In Helarctos the subscapularis minor cannot function differently than the main body of the subscapularis. Consequently, the distinctive anatomy associated with the teres process of ursids and amphicyonids probably derives from an ancestral pattern retained in these lineages but does not necessarily confer the functional significance attributed to this anatomy by Davis (1949) in which these muscles formed part of a muscular package of the shoulder that was thought to favor climbing behavior. Considering the total limb anatomy of Borocyon , a climbing facility is extremely unlikely, whereas an ability to retract the limb in digging might seem a better possibility.

Whether the anatomy of the shoulder in Borocyon relative to that of large living felids is of significance is questionable, given that the role of the subscapularis minor and its postscapular process in Helarctos and that of the subscapularis in felids probably differs little in terms of the efficient function of the shoulder joint. This view also finds support from electromyographic investigations of muscle actions involving the shoulder joint.

The head of the humerus in B. niobrarensis (ACM 3452) and in the Suwanee River Borocyon (KU 113751) displays well-defined scars for the insertion of rotator cuff muscles (supraspinatus, infraspinatus, subscapularis) that stabilize the shoulder and aid in extension of the forelimb; their placement is the same as these scars in much smaller D. superbus (CM 1589) and similar to the wolf, particularly the flat insertion scar for infraspinatus situated far forward beneath the greater tuberosity. In felids and ursids the infraspinatus scar is more posteriorly situat- ed. Not only do rotator cuff muscles stabilize the shoulder joint, but their simultaneous contractions participate in maintaining the balanced anatomical alignment of the parasagittal digitigrade limb: electromyographic studies of the rotator cuff muscles of domestic dogs and cats ( Tokuriki, 1973a, b; English, 1978; Goslow et al., 1981) show that they act in concert not only during the stance phase when the forefoot contacts the ground but also during extension of the shoulder joint at limb protraction. These cuff muscles appear to act together to control movements of the shoulder joint brought about by loading on substrate contact as well as accurately positioning the forelimb during protraction prior to placement of the forepaw on the ground. A controlled rotation and translation of the shoulder apparatus during

TABLE 4 Length Ratios Demonstrating Fore- and Hindlimb Proportions of Daphoenine and Amphicyonine Amphicyonids, Living Ursids, Felids, and Canids

the gait of carnivores is known to contribute to stride length. English (1978), based on electomyographic studies, thought that ‘‘the patterns of activity of the muscles of the shoulder girdle during stepping in most [living] carnivores … may be qualitatively similar.’’

The scapular spine for deltoid musculature and the prominent deltopectoral crest of the humerus extending distad for two-thirds the length of the diaphysis suggest developed deltoid, cephalohumeralis, pectoral, and brachialis muscles supporting the forequarters and fixing the shoulder, with cephalohumeralis acting in forelimb protraction, much as in large living ursids ( Davis, 1964); the deltopectoral crest does not extend as far distad in large felids such as Panthera leo but is a common trait of amphicyonids and ursids regardless of locomotor style. It seems probable that the deltopectoral crest of the humerus and the scapula’s teres process with its postscapular fossa represent skeletal traits of Borocyon rooted more in phylogenetic considerations than in adaptive differences bearing on forelimb function.

Although we lack a complete humerus of the large B. robustum from the Hemingford Quarries, distal ends from these quarries and complete humeri and distal ends of B. robustum from the Bridgeport Quarries show a narrow distal humerus (fig. 21), differing from all New World amphicyonines. However, the distal humerus in Daphoenodon superbus is broader, essentially the same form as in amphicyonines, and in fact this type of

(ZM 1870); (C) Panthera tigris (ZM 14602): ps, postscapular fossa; tp, teres process.

humerus is the plesiomorphic amphicyonid state. The mediolateral breadth of the distal humerus of D. superbus is due to a developed medial epicondyle for attachment of the flexor muscle group; also, in posterior view, the outline of the olecranon fossa is asymmetric because of the ulnar articulation, involving a slight outward angulation of the elbow joint. This type of distal humerus is also found in felids and ursine ursids. In running carnivorans such as the wolf, the distal humerus is much narrower, and the medial epicondyle reduced, thereby altering the shape of the olecranon fossa to create a symmetrical (equilateral) triangular cavity. In these straight-legged cursors, the olecranon process of the ulna penetrates deeply into the olecranon fossa of the humerus, in some creating a perforate fossa accompanying the fully extended ulna.

A transformation in the shape of the distal humerus from the state seen in D. superbus to a state closely approaching the wolf occurs in the Borocyon lineage. The distal humerus of the B. niobrarensis holotype (ACM 3452, Loomis, 1936: fig. 4), a probable male, shows a modest reduction of the medial epicondyle and a somewhat asymmetric olecranon fossa. Humeri of B. robustum from the Bridgeport Quarries (UNSM 26210, 26260), from the Hemingford Quarries (UNSM 26420, Bx-12), and from the Suwanee River (KU 113751) show an even more reduced medial epicondyle and symmetrical olecranon fossa as in wolves. Jenkins (1973) observed that flexor muscles arising from an extended medial epicondyle exert a torque at the elbow joint not adequately compensated by the extensors. One of the options employed by cursorial carnivores to resist this torque involves diminished flexor musculature and reduction of the medial epicondyle. The narrow distal humerus of B. robustum is considerably modified from the plesiomorphic state evident in D. superbus and contemporary amphicyonines, such as Amphicyon galushai , and indicates a decreased torque and less eversion at the elbow and a more straight-legged stance with parasagittal alignment of the forelimb.

RADIUS- ULNA

The association of radius-ulna with humerus and scapula in the holotype of Borocyon niobrarensis (ACM 3452) establishes the proportions of the forelimb in this species relative to large living carnivorans (fig. 19A). A similar amount of limb elongation is estimated for B. robustum and B. neomexicanus , approaching that seen in the wolf and cheetah. There is some anatomical similarity to the forelimb of Panthera leo , particularly in the shoulder, yet significant differences exist in the antebrachium and forepaw of Borocyon . ACM 3452 is the most robust individual of B. niobrarensis , considered a male, yet the radius and ulna of the holotype ( Loomis, 1936: fig. 4) are more slender and slightly more elongate than those elements of the lion. An associated radius and ulna in another B. niobrarensis individual (UW 10004), here interpreted as a female, are even more slender and gracile than in the male holotype.

The eight radii of B. robustum from the Hemingford Quarries are all longer than radii of B. niobrarensis (appendix 1), demonstrating that the lengthened distal forelimb has accompanied an increase in body size within the lineage. Although more robust than the radii of cheetah and wolf, the similarity in form of the B. robustum radius to the radii of these carnivores is striking (fig. 22). A complete ulna of B. robustum has not been recovered, but in B. niobrarensis the distal ulna (fig. 23) with its prominent rounded styloid process for articulation with the carpal cuneiform contrasts with the reduced and flattened process of the wolf and cheetah, suggesting a more mobile ulnarcuneiform articulation at this joint. However, the distal ulnar process articulating with the radius is reduced, presaging the diminished state seen in the cheetah and wolf.

The radius of the wolf and cheetah differ from radii of large felids and ursine ursids in that the shaft is straighter, anteroposteriorly thin and bladelike, with reduction of both the styloid process and the width of the distal end of the bone. The distal ulna becomes more closely applied to the radius as mobility of the distal radioulnar joint is diminished. Furthermore, when manually articulated, the radius and scapholunar are more congruent in the wolf and cheetah than in large felids and ursids, creating a closely registered joint. Borocyon robustum and B. niobrarensis do not achieve this exacting fit of the wolf and cheetah but make a close approach. The radius of B. robustum in particular approximates the form of the cheetah radius (fig. 22). The shaft is elongate, anteroposteriorly thin and bladelike, with reduction in the width of the distal end as in the cheetah, although not to the extent seen in the wolf. The size of the radial head is reduced relative to the breadth of the diaphysis, as in wolf and cheetah, but it is clear that the ovate shape of the radial head and the extent of its articular surface with the ulna show that less than 90 ° of supination was possible.

An associated radius and proximal ulna (UNSM 25554) and two additional radii (UNSM 25553, 44721) of B. robustum , in which the articular circumference of the radial head is well preserved, permit an estimate of the probable amount of rotation of the radial head in the trochlear notch of the ulna. The proximal articular surface of the radius is ovoid, similar in form to that of the lion, but not as concave. In B. robustum this is due to the flatter, less rounded capitulum of the distal humerus, which is also seen in the living bears where the proximal articular surface of the radius is nearly flat; in the cheetah there is a much more concave, circular radial head. The anteromedial lip of the radial head in Borocyon , the wolf, and in large living felids and ursids is slightly elevated as a small process (the capitular eminence of Davis, 1964: 97) that allows better registration between the radius and capitulum of the distal humerus during flexion at the elbow. Davis considered the capitular eminence a bony stop that limited the rotary movements of the radius; it is quite reduced in the cheetah.

The radius is bound to the ulna at the elbow by the annular ligament that anchors the head in the radial notch. The radial notch of both B. robustum and B. niobrarensis is shallow, similar to that of the wolf, more so than the slightly deeper notch in large living felids. The amount of potential rotation of the radial head in the notch during pronation and supination is indicated by a semicircular transverse facet situated on the articular circumference of the radial head. In Canis lupus this facet is both sharply demarcated and quite limited in its transverse extent, which greatly impedes pronation-supination in the wolf. In B. robustum the border of the facet is not as sharply indicated nor is the facet as transversely limited as in the canid, however manipulation of the associated radius and ulna of the beardog shows that the extent to which the radius can rotate in the radial notch exceeds that seen in the wolf but appears somewhat limited relative to Panthera leo . When this is considered togeth- er with the rather flat capitulum of the distal humerus, the less concave radial head, and the shallow radial notch, these features suggest a slightly more restricted capability for pronation-supination of the radius in the beardog.

Canis lupus and Borocyon robustum display similar placement of the scar for the interosseous ligament binding radius and ulna. The scar for the ligament and interosseous membrane extends from immediately below the bicipital tuberosity of the radius for twothirds of the length of the radial diaphysis along its ulnar margin. The ligament is confined to the proximal diaphysis directly beneath the bicipital tuberosity. Dimensions and location of the narrow interosseous scar of the beardog are most similar to those of the wolf and cheetah: scar width relative to length of the radius calculates to 2.8 to 3.5% for wolf and cheetah and 3.0% for B. robustum (N57); it is much broader in the large living felids. The ligament in these carnivores restrains the rotation of the radius about the ulna during pronation–supination, and the location of the narrow interosseous scar on the ulnar margin of radii of B. robustum and the bladelike form of the elongate radius shows that the ligament must have functioned much as in the wolf and cheetah. In a recent analysis of the cheetah antebrachium, Ohale and Groenewald (2003) describe the interosseous ligament as effectively limiting pronation and supination of the radius in this running carnivore.

CARPALS

Close examination of the carnivoran carpus demonstrates that the principal movement in the wrist occurs at the articulation of the proximal carpals with radius-ulna (proximal carpal joint) with only subordinate adjustment at the mid-carpal joint. As Yaldon (1970) has shown for the larger running carnivorans, the carpus is chiefly a flexion hinge, with very limited ulnar and radial deviation restricted to the proximal carpal joint. When the forefoot strikes the ground, the carpus goes into full extension and is locked by virtue of carpal stop facets and binding ligaments and remains thus through limb retraction. Then, as protraction begins, carpal flexion is initiated and attains its climax during the protraction phase of the gait. The extent that the manus can be flexed varies in the different carnivoran families and is minimal at the mid-carpal joint. In felids ( Felis , Acinonyx ), canids ( Canis , Vulpes ), and a hyaenid ( Crocuta ) studied by Yalden (1970), flexion at the mid-carpal joint was always much less (,40 °) than at the proximal carpal joint (90 ° –120 °).

In the living carnivorans studied by Yalden (1970), extension of the wrist at the proximal carpal joint is arrested by binding ligaments that necessarily act to prevent extreme hyperextension at the moment when the forefoot contacts the ground, bears weight, and propels the animal forward. Hyperextension of the wrist in Carnivora occurs only in the proximal carpal joint (bones of the mid-carpal joint are locked) and does not exceed,40 degrees in canids and felids, with,55 degrees reported in Ursus ( Yalden, 1970) . Yalden thought that the ligaments binding the proximal carpal joint were likely stretched in extension, which seems a probable consequence of forefoot impact with the substrate. Stretching of these ligaments must play a major role in preventing hyperextension of the wrist at the proximal carpal joint in Borocyon , acting as a shock absorber just as in living large Carnivora .

The carpals are preserved in the associated B. niobrarensis forefoot (fig. 11) and in a composite foot of B. robustum (figs. 24, 25). The most evident innovation appearing in the carpus is an increase in the size and proximodistal height of the scapholunar from Daphoenodon superbus through B. niobrarensis to B. robustum . In B. robustum it is particularly massive and robust, with its increased height accompanying elongation of the radius and reflecting the mass of the forequarters of this large carnivore. In living Carnivora the convex articular surface of the scapholunar conforms to the concave distal radius, permitting flexion and extension at the proximal carpal joint of the wrist.

Extension/flexion of the Borocyon manus at the proximal carpal joint is governed by these transversely aligned articular surfaces of scapholunar and radius that restrict the joint to fore–aft motion around a transverse axis of rotation. Based on the dimensions and extent of the proximal articular surface of the scapholunar, this surface in B. robustum appears somewhat flatter and less convex than those of large felids of similar body size such as Panthera leo and P. tigris , and displays a convexity more like that of the wolf and cheetah. Moreover, when the width of this articular surface in B. robustum is compared to that of, for example, Panthera leo , the scapholunar in the beardog is mediolaterally shorter, more abbreviate, corresponding to the more narrow, distal radius of the wolf, which lacks the distal breadth evident in the lion.

Flexion involves a relatively unimpeded transverse folding movement at the proximal carpal joint of the wrist until, as Yalden (1970) describes, the distal radius comes into contact with a volar process on the scapholunar that forces some amount of ulnar deviation of the paw as flexion continues. In all living carnivorans, on the radial side of the scapholunar at the posteromedial corner, there is a knoblike volar process (fig. 25, vp), the scapholunar tubercle of Yalden (1970). In living canids, hyaenids, and felids, a bony ridge develops at the junction of this process with the body of the scapholunar. This ridge, acting as a bony stop (figs. 24, 25, #1), redirects the forward movement of the scapholunar on the radius during flexion. As the manus flexes during protraction of the forelimb, the ridge comes into contact with the posteromedial border of the distal radius, diverting the manus. The positive arcuate form of the bony ridge fits a complementary negative concavity on the radius, forcing the scapholunar (hence the paw) to move in ulnar deviation. This action has been noted previously in living carnivorans by both Taylor (1974) and Yalden (1970).

In the cursorial wolf, cheetah, and hyaenids ( Crocuta crocuta , Hyaena brunnea ), the bony ridge is particularly well developed. This ridge on the scapholunar is absent in living ursids and most other living arctoid carnivorans (a small ridge occurs in Procyon lotor ), where, if ulnar deviation occurs to any extent, it is due to contact of the distal radius with the base of the volar process itself. Yalden (1970) viewed some form of ulnar deviation as common to all Carnivora , helping to prevent the protracting manus from contacting the contralateral forelimb as the limb swings forward.

In the species of Borocyon where the carpus is known, the scapholunar with its bony stop is developed similarly to the wolf and cheetah (figs. 24, 25, #1). In Borocyon niobrarensis the scapholunar ridge is not quite as pronounced as in some individuals of B. robustum where the ridge is indeed prominent and the volar process itself is more massive.

Grooves on the distal surface of the scapholunar for the magnum and unciform control movement at the mid-carpal joint and are aligned essentially fore–aft in living Carnivora . The magnum serves as the principal guide for flexion of the distal on the proximal carpal row (fig. 25, d). It can move only a limited distance fore–aft in its scapholunar groove, perhaps,20 ° at most, determined by the alignment of its proximal ridge. The groove for the magnum in the scapholunar is deep and similarly configured in felids (including the cheetah) and ursids, but is shallower and nearly in the same plane as the adjacent groove for the unciform in canids and Borocyon . The magnum of large felids and ursids is posteriorly broad (fig. 25), but in Borocyon robustum , despite the overall geometric similarity of magnum, unciform, and trapezoid to Ursus arctos (fig. 24, anterior view), the magnum has become narrow and mediolaterally compressed as in the wolf, suggesting an incipient adaptation for more restricted fore–aft motion.

Adjacent to the magnum, the unciform of Borocyon also participates in flexion of the wrist at the mid-carpal joint. This flexion is significantly inhibited in the wolf by stop facets developed between scapholunar, carpal cuneiform, and unciform (fig. 25). First, the posterolateral corner of the scapholunar makes a particularly broad contact with the carpal cuneiform. But of most significance in the wolf, the unciform has added a posterior process that locks against the posterior margin of the scapholunar immediately medial to the facet between scapholunar and carpal cuneiform ( C. lupus , fig. 25, # 3). A similar ridgelike unciform process that inhibits flexion occurs in large felids, but in them the process contacts more of the posterolateral corner of the scapholunar ( P. leo , fig. 25, # 3) than it does in the canid. The ridgelike unciform process of felids attains its maximum development as a bony stop in the cheetah where the top of the ridge forms a concave facet into which the posterolateral corner of the scapholunar fits and locks.

In both ursine ursids and the species of Borocyon the unciform lacks a posterior process and so the amount of flexion involving the unciform is not as restricted as it is in the wolf, lion, and cheetah, where stop facets greatly inhibit flexion of the midcarpal joint ( Fig. 25 ). Despite the absence of these stop facets in B. robustum , the limited extent and more horizontal nature of the proximal articular surfaces of the unciform and especially the magnum suggest very restricted flexion at the mid-carpal joint.

Medial to the groove on the scapholunar for the magnum of Borocyon is a broad facet for the trapezoid. In B. niobrarensis the trapezoid is a blocky, wedge-shaped bone (length, 20.3 mm; greatest transverse width, 19.7 mm). Its medial planar facet for the trapezium is rectangular (length, 11.7 mm; height, 6.6 mm). Posterior to the trapezoid on the distal surface of the scapholunar is a slightly concave, subcircular facet ( B. robustum : length, 6 mm; width, 7 mm; B. niobrarensis : length, 5 mm; width, 5 mm) for the reduced trapezium. The Borocyon trapezium is proportionately much smaller than the large element in ursids and even smaller than the much differently configured trapezium of large felids. The trapezium itself is known only in B. niobrarensis (ACM 3452, greatest length, 13 mm), where it is a triangular element situated directly behind the trapezoid. Its placement and size are much like that of the wolf. It articulates with the slender, reduced first metacarpal (, 5 cm in length). The size of the trapezium, first metacarpal, and its phalanges suggest that the extensor, flexor, abductor, and adductor muscles of digit I were as modified as those in the canid, where the first digit was similarly reduced.

Yalden (1970) observed that when the midcarpal joint in Carnivora is fully extended, all stop facets are brought into contact and the joint is locked in full extension. Furthermore, flexion of metacarpus on carpus heavily depends on the heads of paraxonic metacarpals 3–4 (with participation of metacarpal 5) fitting into the corresponding concave facets of magnum and unciform. These facets are aligned to allow only very limited flexion and extension. There appears to be almost no movement between the trapezoid and second metacarpal in many large felids ( Panthera leo , P. tigris , Acinonyx ), due to the nearly flat intervening joint surfaces and to a bony stop formed between the head of the second metacarpal and a round pit on the posteromedial side of the magnum. The second metacarpal of these felids is held largely immobile, locked between magnum and trapezium. In Borocyon robustum and in the wolf the second metacarpal is somewhat more mobile since the trapezium is essentially nonfunctional, and the facet on the magnum for the metacarpal is present but is not developed as a locking device.

Large felids also employ a bony stop involving the magnum to arrest extension at the mid-carpal joint. In the lion, tiger, puma, and jaguar the descending process on the anterior face of the scapholunar fits into a facet on the upper surface of the magnum (fig. 24, # 2) that locks the two bones and prevents hyperextension of the distal carpals on the scapholunar. In the cheetah the magnum facet has further evolved to form a deep pit for the tip of the scapholunar so that the locking mechanism is better developed than in any other felid. Yalden (1970) noted that when the forefoot strikes the ground in the cheetah, hyperextension is limited to the radioulnar-proximal carpal joint; any extension at the midcarpal joint is prevented by this locking mechanism. In Borocyon the contact between magnum and scapholunar is like that of many arctoid Carnivora and canids in which the apposed joint surfaces are firmly held by ligaments but are not as tightly locked as in felids.

Alignment of the distal carpal row with the scapholunar and carpal cuneiform in Borocyon is developed as in the caniform Carnivora and lacks the ‘‘stepped’’ placement of the unciform, magnum, and trapezoid of felids (including Acinonyx ) resulting from the offset heads of their overlapped proximal metacarpals (fig. 24, P. leo ). This stepped configuration is characteristic of all large living felids and reflects the intricate interlocked character of the distal carpals and proximal metacarpals, a unique attribute of the felid manus not present in arctoid Carnivora .

METACARPALS

The metacarpals of Borocyon , while similarly paraxonic and elongate, differ in lacking the overlapping, ‘‘interlocked’’ proximal ends of MC2 on MC3, and MC3 on MC4, evident in the lion and cheetah, and also characteristic of the tiger, jaguar, and puma. The proximal metacarpals of Borocyon articulate without overlap and in this respect are typical of arctoid Carnivora . In large felids the interlocked character of the metacarpals can involve the distal carpals, best exemplified by the insertion of the proximal second metacarpal between magnum and trapezium. Here a palmar process of the metacarpal fits into a pit in the medial face of the magnum, locking the metacarpal between that bone and the trapezium. This does not occur in arctoid Carnivora or in hyaenids.

The cheetah and canine canids show close appression of the proximal shafts of the metacarpals, whereas Borocyon and the lion do not. Borocyon displays a distal separation or splay of the metacarpals much like Panthera leo (fig. 11) but emphasizes the weight-bearing paraxonic metacarpals 3–4 with somewhat greater reduction of metacarpals 2 and 5. Large living felids other than the cheetah also show this distal separation of the metacarpals. The forepaws (figs. 11, 26) of Borocyon niobrarensis and B. robustum must have resembled the broad paws of large living felids, possibly used to deliver a blow to prey, manipulate objects including food to a limited extent, and even excavate a burrow, with these behaviors having been witnessed for P. leo by Schaller (1973).

PHALANGES

Phalanges are associated with individuals of both B. niobrarensis and B. robustum . Proximal, intermediate, and ungual phalanges are present in the forefoot of the B. niobrarensis holotype (fig. 11). A proximal and an intermediate phalanx occur with the partial skeleton of F:AM 107601 and four proximal phalanges with UW 10004 (both B. niobrarensis ). The B. robustum holotype (CM 1918) described by Peterson (1910) included five proximal and three intermediate phalanges but no unguals. Overall, the proportions of the phalanges are most comparable to those of ursids such as Ursus arctos , but the detailed anatomy of the metacarpal-phalangeal joints are more similar to those of large felids.

The holotype forefoot of B. niobrarensis (fig. 11) includes five proximal phalanges correctly assigned to the corresponding metacarpals by Loomis (1936), based on comparison with the articulated forefeet of Daphoenodon superbus . Two proximal phalanges are symmetrical (about their long axes) and elongate (lengths,, 31–32 mm) and therefore belong to the paraxonic metacarpals 3–4. The proximal phalanx of the third digit is slightly more robust than that of the fourth. Another symmetrical proximal phalanx (length,,30.4 mm) fits metacarpal 2, and two smaller, asymmetric phalanges belong to metacarpals 1 and 5. Metacarpal 1 (41.8 mm) and its proximal phalanx (27.1 mm) are much reduced, although these bones are slender and gracile. Prominent bony keels on the distal metacarpals insert between and guide a pair of sesamoids for each digit. These keels compare with those of living large felids and are much more developed than the metapodial keels of ursine ursids.

Anatomically significant features of the proximal phalanges of digits 2–5 (fig. 26) include (1) a notch or groove in the base of the palmar surface for the metacarpal keel and the combined tendons of the superficial and deep flexors of the digits, similar to those of large felids; (2) a pair of prominent bony tubercles on either side of this notch for attachment of the sesamoid and interosseous tendons—the interosseus muscles are flexors of the metacarpal-phalangeal joints but also contract to support these joints to prevent hyperextension when the carnivore places weight on the paw; (3) a concave base of the proximal phalanx congruent with the distal metacarpal so that the joint is firmly registered and further secured by sesamoid and collateral ligaments. Flexion is significantly limited by the height of the metapodial keels and adjoining sesamoid bones.

The intermediate (middle) phalanges are much shorter than the proximal ones and all have a strong proximal dorsal process for extensors of the digits (fig. 26). When the intermediate phalanx is plantar-flexed,45 °, the dorsal process fits a corresponding groove in the distal proximal phalanx. This restricts the joint to only flexion and extension, assisted by the biconcave base of the intermediate phalanx that, together with the dorsal process, prevents axial rotation of the digit. Some of these intermediate phalanges display slightly asymmetric shafts but the degree of asymmetry does not approach that seen in living felids where the trait is related to hyper-retractility of the claws ( Gonyea and Ashworth, 1977; Wang, 1993). The distal articular surface of the intermediate phalanges extends onto the dorsal surface of the bone, demonstrating that the ungual phalanges could be slightly retracted. However, the ungual phalanges are not specialized for hyper-retractility and instead are slender, long, and nearly straight (not curved), as in living ursids.

Unguals are generally rare but six wellpreserved unguals of the associated forefeet of B. niobrarensis (ACM 3452) range in length from 23.6 to 25.2 mm, averaging 24.2 mm (fig. 11); they appear quite elongate relative to the short intermediate phalanges. The foreclaws of large ursids are longer than the hindclaws, and this was possibly true of Borocyon (no unguals of the hindfoot can be certainly identified). These unguals show slight development of a bony hood and the presence of a dorsal process for attachment of the tendon of the common digital extensor.

The angulation of the phalangeal joints indicates the presence of thick fibroelastic metacarpal and digital pads cushioning impact with the ground during footfall (fig. 26). Undoubtedly, thick connective tissue pads supported and cushioned both the fore- and hindfeet, extending from the distal metapodials forward beneath the digits ( Alexander et al., 1986).

Rather short intermediate phalanges emerge as a distinctive trait of Borocyon robustum and B. niobrarensis , confirmed in the former by associated phalanges in the holotype hindfoot (CM 1918) and in the latter species by associated forefeet in its holotype (ACM 3452). These short phalanges were unexpected since they are atypical of living large Carnivora : in the wolf the intermediate phalanges of the paraxonic digits 3–4 range from,67% to 71% of the length of the proximal phalanges, and the intermediate phalanges of the side toes (digits 2 and 5) fall at,58%–63%. In B. niobrarensis (ACM 3452), where associated digits are available, the intermediate phalanges of digits 3–4 (N 5 4) all fall from,59% to 67% and the side toes (digits 2 and 5, N 5 3) at,55%–58%. These values for B. robustum are,61% for a paraxonic digit and,58% for a side toe.

In contrast, ursids ( Ursus arctos , U. americanus ) fall at,74%–77% for the principal weight-bearing digits. The lion and puma show similar values of,71%–77% for paraxonic digits 3–4, and the cheetah has values somewhat lower at,65%–66%. Despite these proportional differences, the shape of the phalanges of Borocyon are most like the phalanges of large bears such as Ursus arctos , presumably because of the shared arctoid ancestry of ursids and amphicyonids, whereas the more derived anatomy of the metacarpal-phalangeal joints parallels that of large felids like the lion. This joint anatomy is an adaptation to a more digitigrade stance.

HINDLIMB

Proportions of the Borocyon hindlimb do not correspond to those of living canids or the cheetah, falling closer to the lion (fig. 19B) and differing surprisingly little from plesiomorphic amphicyonids such as Daphoenodon superbus and Daphoenus . Caution is necessary here in that no associated hindlimb of Borocyon robustum is known (there is a possible association of femur and tibia in a small B. niobrarensis female, UW 10004), and so the estimate of hindlimb proportions is taken from unassociated femora and tibiae. This estimate, however, cannot be far from the mark.

INNOMINATE

The innominate is known only in B. niobrarensis (F:AM 107601) where its form differs to some extent from that of large living felids and ursids. The acetabulum is deep, as in the large cats, but the ilium and ischium are proportionately slightly shorter. A broader ilium than seen in these felids (but much like wolves) suggests that gluteal muscles for retracting the femur were well developed for propulsive thrust. Extension and medial rotation of the femur at the hip joint performed by the gluteal group is balanced by the caudal hip muscles (gemelli, obturators, quadratus femoris) that laterally rotate the femur, resulting in the parasagittal path of the hindlimb. Extent and position of the articular surface of the head of the femur within the acetabulum are as in the large living felids, wolves, and cheetah (roof of the acetabulum a horizontal surface of similar depth and area with the same anterior and posterior ventral extension).

Pronounced eversion of the ischium and ischial tuberosity in living carnivorans corresponds to a parasagittal orientation of the hindlimb whereby the action of the hamstrings is brought into a fore–aft alignment. The cheetah and wolf show an extreme degree of ischial eversion among living Carnivora . Borocyon displays a degree of ischial eversion corresponding to the large felids and does not approach canine canids or the cheetah. When the innominates of B. niobrarensis and a Sumatran tiger of the same body size are compared, the ischium is of similar form and length and the tuberosity of the same size and degree of eversion, but the ilium is somewhat shorter and broader in the beardog.

FEMUR

Borocyon robustum is represented by two complete femora (UNSM 26435, KU 113645). UNSM 26435 is from a very large animal (length,, 38 cm), probably male, and was found at Shimek’s Quarry, one of the upper Runningwater localities. This large femur is most similar to those of living lion and tiger but with some important distinctions. There is a close correspondence in orientation of the proximal femoral head, neck, and trochanters, the margins and shape of the femoral hemisphere, the location of the fovea capitis femoris (fig. 27) for the femoral ligament, and the form of the diaphysis. The form and amount of separation of the distal condyles are as in these large living felids; the condyles are not placed as close together as in the wolf. Femora of Ursus arctos and Thalarctos maritimus differ from those of B. robustum in that the femur of these ursids is more strongly abducted; the femoral neck and head are proximally elevated and the diaphysis rotated outward—this is the abducted ursid femur of Jenkins and Camazine (1977).

The anatomy of the femur provides useful information on hindlimb stance and excursion, particularly the orientation of the femoral head relative to the diaphysis; the configuration of the margin of the articular surface of the head relative to the diaphysis and acetabular rim; and the location of the fovea capitis femoris for the femoral ligament ( Jenkins and Camazine, 1977). Jenkins and Camazine (1977) showed that the excursion of the femur of a domestic cat during walking describes a more adducted path than in a small canid ( Vulpes fulva ) and raccoon ( Procyon lotor ). When stationary, the femur of the fox was abducted,10 °, and the cat aligned in a parasagittal plane, whereas that of the ambulatory raccoon was abducted 25 ° –35 °. Jenkins and Camazine (1977) observed a relationship between the alignment of the anterior margin of the articular surface of the femoral head and the amount of femoral abduction in these carnivores. In the domestic cat and foxes studied by them, the more parasagittal stance correlated with the alignment of the anterior margin of the articular surface of the femoral head. In these small carnivores, the anterior margin formed a more acute angle with the diaphysis than in ambulatory carnivores with more abducted femora where the angle is greater. The anterior margin of the articular surface in Borocyon robustum (UNSM 26435) forms an angle with the diaphysis slightly greater than in the domestic cat but more acute than in the ambulatory raccoon and ursids. In addition, the femoral head of B. robustum is aligned nearly at a right angle with the diaphysis, suggesting only a limited amount of femoral abduction. The orientation of the articular margin of the femoral head corresponds most closely to carnivores with more parasagitally aligned femora. Procyon lotor and living ursine bears have strongly abduct- ed femora where the amount of abduction is greater than in any other family under discussion here, including amphicyonids.

In large living canids such as Canis lupus and Canis latrans , orientation of the femur is determined by the upward-angled femoral head ( Jenkins and Camazine, 1977: fig. 8F), which is absent in felids and amphicyonids, in which the head is aligned nearly at a right angle to the diaphysis. The femoral anatomy of the wolf and coyote indicate not only greater femoral abduction than in the domestic cat and foxes but also in comparison to large felids ( P. leo , P. tigris ). The anterior, dorsal, and posterior margins of the femoral head are quite similar in B. robustum (UNSM 26435) and in the lion and tiger, yet in B. robustum the anterior margin is aligned at a somewhat more acute angle to the diaphysis than seen in the large cats. Also, B. robustum is characterized by a transversely broad proximal femur, in that the distance between greater trochanter and head is more extended, thicker, and straighter than in living felids. This condition would seem to favor a more parasagittal stance and is approached by some lions.

Whereas the amount of femoral abduction might be best measured in living or recently deceased animals, here, using only associated innominates and femora from osteological collections, a proxy measure was taken: the angle between a parasagittal plane aligned parallel to the dorsal acetabular rim and the long axis of the femoral diaphysis, when the femur is oriented at 45 ° below the horizontal. In this position the margin of the femoral head is made congruent with the dorsal acetabular rim. This is not a position that would be normally adopted in the living animal, but it allows a comparable measurement in large living felids, cheetah, wolf, and B. niobrarensis (no pelvis of B. robustum has been found). The abducted angle in the cheetah is,29 °, the tiger 34 ° –40 °, the wolf,47 ° –52 °, and in B. niobrarensis somewhere in the range of,28 ° – 37 ° (although the femur [KU 113645] and innominate [F:AM 107601] of B. niobrarensis are not from a single individual, the measurement was considered acceptable because the femoral head was congruent within the acetabulum).

The placement of the fovea capitis femoris on the femoral head lies in a position nearly identical to that of the lion and tiger (fig. 27) so that, together with alignment of the proximal femur relative to the diaphysis as well as the orientation of the articular margin of the femoral head, a slightly abducted yet essentially parasagittal orientation of the hindlimb seems probable.

Certain additional observations derive from this comparison: the congruence of the margins of acetabulum and femoral head in Borocyon niobrarensis is most similar to that of cheetah and Sumatran tiger among skeletons at my disposal. As the femur starts to retract at the beginning of the propulsive phase, the anterior margin of the head is largely congruent with the dorsal rim of the acetabulum, but as retraction proceeds, a strip of the articular surface of the head remains outside the dorsal rim as a more parasagittal femoral alignment is adopted. When the femur of Borocyon is fully retracted (end of propulsive phase), the posterior acetabulum overlaps an area on the femoral head that lies beyond the articular surface, which approximates the limit of femoral retraction as described in the domestic cat and fox by Jenkins and Camazine (1977).

In all of these carnivorans, only in extreme abduction of the fully retracted femur could the articular margin of the femoral head ever be congruent with the posterior dorsal rim of the acetabulum; this is a position unlikely to occur in normal locomotion. Consequently, a portion of the femoral head’s articular surface is commonly exposed during the terminal phase of femoral retraction in many living Carnivora , especially in those animals (e.g., procyonids, mustelids) where the dorsal rim of the acetabulum has not been extended laterad over the femoral head.

TIBIA

B. robustum is represented by three tibiae (fig. 28), one (UNSM 25558) from UNSM locality Bx-7; a second (UNSM 44719) from Hovorka Quarry (Bx-21), both probably males; and a third (UNSM 26260) from the Bridgeport Quarries. The Runningwater tibiae are of similar size, whereas the Bridgeport tibia is more gracile and may belong to a female. A smaller tibia of B. niobrarensis is known from Horse Creek Quarry (UW 10004). In all of these individuals, the prominent tibial tuberosity and cnemial crest indicate a pronounced ability to extend the knee on the thigh; they form an anteriorly extended ridge on whose lateral side is the deep concavity for the anterior tibial muscle. In daphoenines the tendon of this muscle would extend across the dorsal surface of the tarsus to attach to a ridge on the medial base of metatarsal 2 and a rugose area on the base of metatarsal 1 and adjoining entocuneiform. It dorsiflexes the tarsus on the tibia. The tendency of the anterior tibial muscle to not only flex the tarsus, but to rotate the paw laterally, is balanced by the peroneal muscles that also flex the tarsus but rotate the paw medially. The peroneal muscles, the cranial tibial, and long digital extensor are all innervated by the peroneal nerve and perform an integrated action in dorsiflexion and rotational balancing of the hindpaw. Inversion and eversion of the hindfoot of B. robustum is only possible because of the ball-and-socket articulations of astragalus with navicular, and calcaneum with cuboid, which allow axial rotation within the tarsus. The tibial-astragalar and tarsalmetatarsal joints do not permit axial rotation. Therefore, the ability of the Borocyon hindfoot to adjust to an uneven substrate was regulated by the coordinated action of the anterior tibial and peroneal muscle complex at the mid-tarsal joint.

Digitigrady in living carnivorans has been discussed relative to the pattern of muscle scars on the posterior surface of the tibia ( Ginsburg, 1961; Wang, 1993). Proximally, four deep muscles are involved: flexor hallucis longus (FHL) laterally, popliteus muscle medially, and between them the tibialis posterior (TP) and flexor digitorum longus (FDL). In the wolf and domestic dog ( Evans, 1993: 375), the TP and FDL are reduced and their registration on the tibia is restricted to a very narrow elliptical area, a thin lenticular scar trending downward along the upper third of the diaphysis toward the medial side of the bone (fig. 29). In contrast, living ursids such as Ursus arctos and Ailuropoda melanoleuca display a well-developed area of attachment for these muscles that forms a broad, prominent scar running down nearly the entire length of the posterior tibia ( Davis, 1964: 116). Moreover, the popliteus is restricted to the proximal tibial diaphysis in canines but extends down nearly two-thirds of the shaft in living ursids.

A prominent broad scar for TP-FDL is also characteristic of most digitigrade felids. This scar is centered on the posterior surface of the upper diaphysis and, although reduced in area relative to that of living ursids, it is broader than in the wolf. However, the cheetah is an exception among the large cats and retains a very narrow scar pattern most like the wolf.

Although these contrasting scar patterns have been linked to either a digitigrade or plantigrade stance, it is likely that the reduced area for tibialis posterior and the long digital flexor in the wolf and cheetah is related to the reduction in size and functional importance of the tibialis posterior. The short flexor muscles of the crus end in long tendons, and short-fibered muscles ending in long tendons indicate considerable elastic extension in the tendons ( Alexander, 1988: 21), suggesting that reduction in the posterior tibial muscle and the long tendons of the other flexors are adaptations to efficient flexion-extension of the tarsus and digits, contributing to a fore–aft path of the hindlimb. The posterior tibial muscle inverts (outwardly rotates) the foot, an action less necessary in digitigrade carnivores adopting a more fore–aft motion of the hindfoot. Also, the tendon of FDL joins that of FHL to form a conjoined tendon, becoming the deep digital flexor (flexor digitorum profundus) of the digits, and because FHL becomes the principal phalangeal flexor, FDL is less necessary to that role. The lenticular scar for tibialis posterior decreases in width and in area from that in Daphoenodon superbus (CM 1589) through B. niobrarensis (UW 10004) to B. robustum (UNSM 26361): in D. superbus like that of living wolves, and in B. robustum and B. niobrarensis nearly identical to the pattern seen in Acinonyx jubatus (ZM 16913). When tibial length is plotted against the width of the TP-FDL scar for a variety of living large carnivorans, Borocyon groups with the wolf and cheetah (fig. 30).

The tibiae of cheetah and wolf share a trait found in no other living carnivoran examined in this study. The proximal tendon of the long digital extensor, originating on the lateral condyle of the femur, travels downward to occupy a marked groove (sulcus muscularis, fig. 28, ZM 16913, Acinonyx ) immediately anterior to the lateral condyle of the tibia. The tendon continues distad to a muscular belly, then transforms to a distal tendon to extend the digits and participate in flexing the tarsus. The sulcus muscularis is absent in living arctoid Carnivora , in amphicyonids (with the exception of some temnocyonines), and in hyaenids, and it does not occur in the other living felids, where a more shallow embayment is present farther anteri- or on the tibial crest. In the only tibia of B. robustum (UNSM 26361) in which the proximal region is intact, the sulcus is absent.

TARSALS: ASTRAGALUS- CALCANEUM

Among the tarsals, the astragalus and calcaneum of Borocyon robustum are particularly diagnostic (figs. 31, 32) and differ from those of large contemporary amphicyonines. The astragalus is proximodistally elongated, the trochlea is narrower, and the distal condyle (head) migrates beneath the trochlea rather than remaining in its more plesiomorphic medial location as seen in Amphicyon and Daphoenodon superbus (fig. 33). Additionally, the sustentacular facet for the calcaneum on the posterior face of the astragalus is restrict- ed to a more proximal location and does not extend distad to the base of the condyle as in Amphicyon and other large amphicyonines.

The ectal facet of the astragalus in B. robustum is moderately concave, more so than in Amphicyon and other large amphicyonines, in order to receive the ectal prominence of the calcaneum. This ensures a close registration of astragalus and calcaneum that somewhat restricts motion between the two bones, much as in living Panthera tigris and P. leo where this same type of registration occurs. In canine canids such as Canis lupus and C. latrans the ectal prominence is strongly protuberant, forming a ridge deeply inserted into the posterior surface of the astragalus, creating a tight, immovable registration between the two bones. The interlocked astragalus and calcaneum in these canines represents an extreme expression of this trait in Carnivora . In the cheetah this locked relationship of astragalus and calcaneum is less developed than in the wolf and coyote but is more developed than in any other living large felid.

When compared to the Amphicyon calcaneum, the calcaneum of B. robustum is distally narrower with a smaller sustentaculum and less developed coracoid process. A distal extension of the calcaneum accompanies elongation of the astragalus. The length of the Borocyon calcaneum relative to its distal width differentiates it from distally broad, more robust calcanea of contemporary Amphicyon and other large amphicyonines (fig. 34), and the posterior surface of the sustentaculum is usually more deeply grooved by the flexor tendon of the digits.

The astragalus of Borocyon robustum is differently proportioned relative to Amphicyon galushai and two large contemporary amphicyonines from the Bridgeport Quarries and from Florida (fig. 33). These large amphicyonines, found together with Borocyon , all display a short astragalus with a wide trochlea and broad head relative to the elongate astragalus with narrow trochlea and head of Borocyon robustum . The astragali from the Rose Creek Member of the John Day Formation and from the Suwanee River locality are indistinguishable from those of B. robustum (fig. 33). The holotype of B. niobrarensis lacks an astragalus; however, two referred B. niobrarensis astragali and associated calcanea (UW 10004, F:AM 107601) are similar to B. robustum but are smaller and less elongate.

The dimensions of the holotype astragalus (fig. 3, CM 1918) of Borocyon robustum plot with astragali of the upper Runningwater and Bridgeport samples (fig. 33) and hence support the inclusion of the B. robustum holotype with that hypodigm. The holotype astragalus is associated with a mandibular fragment, an isolated m2 (not reported by Peterson), and the navicular, ecto-, meso-, and entocuneiform (fig. 3).

There is a clear parallel between the astragalus-calcaneum of Borocyon robustum and those of large living felids. Distally extended, narrow calcanea of P. leo and P. tigris show proximal placement of the sus- tentaculum, a tall, narrow tuber calcis, and a tendency for flexor tendons to groove the posterior surface of the sustentaculum. These derived features, present in Borocyon robustum , are less developed in B. niobrarensis . Together they confirm an incipient specialization of the tarsus for digitigrady and more restricted fore–aft motion, a trend that independently reaches its culmination in canine canids. The tarsus, heavily bound by its investiture of ligaments, can only flex and extend at the tibia-astragalar and distal tarsal-metatarsal joints but retains limited axial rotation at the intratarsal joint between astragalus-navicular and calcaneum-cuboid. The similar form of these tarsals in B. robustum to those of P. leo and P. tigris suggests a hindfoot capable of the same range of movements.

TARSALS: CUBOID AND CUNEIFORMS

The cuboid is readily distinguished from the cuboids of living ursids and contemporary large amphicyonines. In Borocyon robustum and B. niobrarensis the cuboid is relatively tall. Its proximal surface that articulates with the distal calcaneum is nearly horizontal, and the anterior face of the bone is vertical, forming a right angle with the upper surface. Living ursids and contemporary large amphicyonines have a low cuboid in which the proximal surface is not horizontal but is inclined laterad, and the anterior face of the bone descends at a sloping angle. Because the amphicyonine and ursine cuboid is proximodistally short, the medial cuboid facets for the navicular are continuous (or nearly so) with the subjacent facet for the ectocuneiform. In Borocyon , the ectocuneiform facet is an isolated ellipse well distad (7–9 mm) of the navicular facets that are situated along the upper rim of the cuboid.

The tall Borocyon robustum cuboid is most closely paralleled by the elongate cuboids of the cheetah (ZM 16913, height, 21.6 mm) and wolf (ZM 15596, 19.1 mm), although the beardog cuboids are more robust. A Sumatran tiger cuboid (ZM 14602) is not as tall as in Borocyon nor is the anterior face quite as vertical. The B. robustum cuboid (UNSM 25562, locality Bx-7) is much larger and taller (32.8 mm) relative to that of B. niobrarensis (F:AM 107601, 26.8 mm) and a Sumatran tiger (ZM 14602, 23.6 mm). Cuboids of Amphicyon galushai (UNSM 26390, 26394) from UNSM locality Bx-12, an upper Runningwater site, measure only 28.3 and 27.6 mm.

The ectocuneiform is preserved in the B. robustum holotype in articulation with the meso- and entocuneiform and navicular (fig. 3), but its plantar half is missing. Nonetheless, an isolated ectocuneiform (UNSM 44720) of B. robustum from UNSM locality Bx-12 is complete and represents a larger animal than the holotype. Its upper surface is horizontal, more so than in the holotype. An elliptical cuboid facet (14.0 X 8.8 mm) situated nearly at the anterior margin of the upper rim of the bone is the only facet on the lateral face. There is a robust plantar condyle as in large felids (including the cheetah), which is absent in ursine ursids and much smaller in the wolf. On the medial face are an upper facet for the mesocuneiform and a lower facet for the elevated base of metatarsal 2. The mesocuneiform is a short, blocky element tapering backward to a point. It contacts medially the tall, bladelike entocuneiform, which is situated posteromedial to the mesocuneiform on the inner side of the tarsus. The concave distal end of this large entocuneiform (height, 23.5 mm; anteroposterior width, 18.1 mm) receives the base of a slender, short metatarsal 1. All cuneiforms and the cuboid have slightly concave ventral articulations with the corresponding metatarsals.

Alignments of facets joining the tarsal cuneiforms and cuboid are largely in the vertical plane, suggesting little movement between them when bound by investing ligaments. Intratarsal movement was chiefly limited to the articulations between astragalus and navicular on the one hand and cuboid and calcaneum on the other hand, which allowed the foot to adjust to an irregular substrate. These articulations indicate very little intratarsal flexion-extension.

The wolf tarsus (ZM 15596) shows some parallels with that of Borocyon robustum : (1) elongate calcaneum; (2) astragalar head situated more under trochlea; (3) astragalar trochlea narrow; (4) elongate cuboid with vertical anterior face and flat upper articular surface (cuboid facets, however, are typically canine and differ from those of Borocyon — there is an upper navicular facet, a middle circular navicular facet, and a lower elongate ectocuneiform facet); and (5) volar process of the ectocuneiform not as developed as in Borocyon but its relationship to a reduced meso- and entocuneiform similar. In the wolf, a thin, quadrate entocuneiform for attachment of a vestigial first metatarsal is much more reduced than in Borocyon (in the wolf the tiny entocuneiform is applied to a small convex facet on the posteromedial corner of the navicular). Only a small nubbin of bone represents the wolf’s vestigial first metatarsal. The tarsus of Canis lupus shows a greater reduction of tarsal elements associat- ed with the internal digits of the foot than does Borocyon .

The tarsus of living large felids shows many similarities with the Borocyon tarsus. This is evident in the form of calcaneum and astragalus, cuboid, navicular, and the geometry and placement of the tarsal cuneiforms. On the other hand, ursid tarsals differ markedly from the tarsals of Borocyon , more so than those of any other large living carnivorans. In Ursus arctos (ZM 17888) the meso- and entocuneiforms are not reduced and remain large blocky elements, and the three tarsal cuneiforms are all of the same height (thickness) and are aligned sideby-side; bears lack the offset mesocuneiform and reduced entocuneiform of living felids and canids. This contrasts with Borocyon in which the ectocuneiform is deep, mesocuneiform dorsoventrally compressed and elevat- ed, and entocuneiform elongate and bladelike. The inner cuneiforms show less reduction in Borocyon relative to the cheetah and other large felids because the first digit of the hindfoot is not as reduced as in these cats and in canine canids, retaining an elongate, small metatarsal 1 with at least one phalanx remaining and probably two. In living felids the first metatarsal is a tiny vestigial spur of bone without associated phalanges. In living canids, the first metatarsal is a small, nearly spherical nubbin of bone attached by ligament to the distal terminus of the entocuneiform and the posterior surface of the second metatarsal. Hence, the Borocyon tarsus differs from the tarsus of all living felids and canids in the retention of a less reduced first digit of the hindfoot, yet one entirely distinct from this robust digit in living bears.

METATARSALS

The metatarsals of Borocyon demonstrate a paraxonic, elongate hindfoot. Cheetah and wolves show close appression of the proximal shafts of the metatarsals, and the metatarsus of Borocyon parallels this, but to a lesser extent. The paraxonic metatarsals 3–4 are robust and straight, as in living canids, not curved as in many felids; they are flanked by shorter metatarsals 2 and 5 whose shafts display a flattened internal face, indicating that they were closely applied to the central paraxonic metatarsals. Some distal separation of the metatarsals occurs but this is much less, relative to the metacarpals. The distal metatarsal keels are pronounced as in felids. Large living felids other than the cheetah show a slight distal separation of the metatarsals, yet this appears even less evident in Borocyon .

Borocyon ’s metatarsals lack overlapping proximal ends as seen in the metacarpals of large living felids (MC2 on MC3, MC3 on MC4). Otherwise, the fit of the metatarsal articulations are anatomically similar. The proximal head of MT5 inserts into MT4, and the head of MT4 into MT3; the head of MT2 is raised above that of MT3 so that MT2 can articulate with the ectocuneiform. The relative lengths of MT3–MT4 to the shorter, flanking MT2 and MT5 are similar in the lion, tiger, and Borocyon . The fourth metatarsals of amphicyonids display a large, elongate ligament scar on the posterior diaphysis a little above mid-shaft. This scar appears to represent the insertion of a plantar tarsal ligament that reinforces and stabilizes the tarsal-metatarsal joint—this scar is particularly thick in Borocyon robustum , and more distally situated than in ursids, suggesting that the ligament may be important in maintaining a digitigrade stance. The paraxonic digitigrade hindfoot of Borocyon is doubtless adapted to fore–aft motion, contributing to the propulsive thrust provided by the hindlimb.

The principal weight-bearing metapodials of the paraxonic fore- and hindfeet of Borocyon are metacarpals 3–4 and metatarsals 3–4. These metapodials show an increase in absolute length from B. neomexicanus through B. niobrarensis to B. robustum (table 5), accompanying the overall size increase seen in these species. Borocyon metapodials also are dimorphic, which is interpreted as sexual and is documented at several localities (table 6).

Length measurements of the paraxonic metacarpals 3–4 and metatarsals 3–4 from living large felids, canids, and ursids, when compared with those for Borocyon , show that the metatarsals in many large cats (including the cheetah Acinonyx ) are,20%–40% longer than the metacarpals (fig. 35; appendices 2, 3). In Panthera leo and P. tigris this disparity between fore- and hindfoot is less (,11%– 23%). But in wolves and coyotes, and in Borocyon robustum and B. niobrarensis , the paraxonic metapodials of the hindfoot are only,8%–17% longer than those of the forefoot (fig. 35; appendix 4).

AXIAL SKELETON

There is no representation of the vertebral column, sternum, or ribs of Borocyon robustum except for three caudal vertebrae (fig. 3) belonging to the holotype described by Peterson (1910). He realized that these caudals indicated that a long tail was present as seen in Daphoenodon superbus . A partial vertebral column does accompany the Horse Creek skeleton (UW 10004) referred here to Borocyon niobrarensis . Its nine vertebrae (axis, two cervicals, three thoracics, two lumbars, and a caudal vertebra) are nearly identical in all respects to the column of D. superbus described by Peterson (1910). A capability to flex and extend the back is evidenced by inclination of the neural spines of the thoracic and lumbar vertebrae directed toward a centrally situated anticlinal vertebra.

SUMMARY

The largest individuals of Borocyon robustum , all with markedly elongate limbs, occur in several of the Hemingford Quarries considered the youngest in the Runningwater Formation on biochronologic criteria. Sites in the lower Runningwater Formation yield on average smaller individuals referred to B. niobrarensis . When viewing limb elements of the largest individuals of Borocyon robustum , particularly the radius and paraxonic metacarpals, the size and elongation of the forelimb are striking.

To estimate the relative degree of limb elongation of the species of Borocyon , linear measurements of limb elements were compared with those of the plesiomorphic Da- phoenodon superbus (fig. 36), based on Peterson’s (1910) genoholotype (CM 1589). The slightly smaller size of B. niobrarensis relative to B. robustum is reflected in these percentages. Forelimb segment lengths for B. niobrarensis show increases of,27%–46% over D. superbus . The increase in segment lengths for B. robustum is clearly more substantial, from as low as,35% for the humerus to as much as 76% for the radius. The increase in length of the radius (hence the ulna), femur, and tibia of B. robustum relative to D. superbus is especially obvious. Fore- and hindlimb segment lengths of B. robustum exceed those of B. niobrarensis for all elements.

Despite the scarcity of associated limb material and small sample sizes, the available data suggest that the fore- and hindlimbs of Borocyon robustum are proportionately longer than those of B. niobrarensis . Although limb lengths of large male B. niobrarensis probably overlap those of small females of B. robustum , the proportions of these limbs differ, such that B. robustum limbs appear consistently more elongate.

Figure 37 illustrates a hypothetical restoration of a male B. robustum together with the holotype skeleton of Daphoenodon superbus . These two carnivores represent the two end-member states of this New World daphoenine amphicyonid lineage. Daphoenodon (D.) superbus demonstrates the plesiomorphic condition, probably retained in most or all species of D. ( Daphoenodon ), and Borocyon robustum represents the culmination of limb development in the subgeneric Borocyon lineage. Although the D. superbus holotype is a female, the males of its probable ancestor, Daphoenodon notionastes from the Arikareean of Florida, would have been similar in size and are known to exhibit the same absence of limb elongation. Hence, the two skeletons adequately illustrate the transformation in skeletal dimensions and limb proportion that occurred during evolution to larger body size in these daphoenines. Such a transition is unknown in Old World amphicyonid carnivores.

The anatomy of the limbs demonstrates the following features of Borocyon bearing on stance and gait: nine amphicyonids, indicated by transverse narrowing of the distal humerus involving suppression of the medial epicondylar area for the origin of flexor muscles and deepening and realignment of the olecranon fossa for extension of the ulna—the elbow in Borocyon robustum is only slightly everted and is parasagittally oriented as in living wolves; form of the radial head and distal ulnar structure, indicates a more restricted capability for pronation-supination than found in the large living felids, but not as markedly developed as in the wolf; thus, a cuffing or swatting stroke of the paw could have been employed in bringing down prey;

proximal phalanges somewhat elongate and largest on the paraxonic metapodials 3–4; intermediate phalanges short (with developed dorsal extensor processes) and lacking the marked degree of asymmetry indicative of claw retractility in felids; ungual phalanges of the forefoot long, laterally compressed, and tapering, hence similar to those of living ursids, suggesting that digging was possible (lions dig warthogs from burrows despite their specialized phalangeal anatomy); articular alignment of phalanges indicates presence of cushioning fore- and hindfoot pads; hindlimb providing a strong forward propulsive thrust as in large living felids, in which the form of Borocyon innominate and femur best approximate the form of those bones in Panthera leo and P. tigris among large living carnivores; posterior placement of the fovea capitis femoris for the femoral ligament like that of large living felids, and lacking the more central location seen in living wolf and coyote or in small domestic cats with parasagittal hindlimbs—hence, slight abduction of the Borocyon femur and hindlimb as in lion and tiger; ischial tuberosity of the pelvis only moderately everted as seen in large living felids, without the extreme degree of ischial eversion of wolves and cheetahs; muscle scar pattern of the posterior tibia for flexor muscles of the hindfoot as in the wolf and cheetah, suggesting a proximal concentration of muscle with long distal tendons; tibia much more massive in the larger Borocyon species due to greater body size, and a tibialastragalar joint indicating a hindfoot emphasizing flexion-extension;

hindlimb proportions similar to those of Panthera leo , P. tigris , and New World amphicyonines and lacking the distal elongation of tibia-fibula seen in canine canids and the cheetah— Borocyon and these large cats also correspond in tarsal anatomy, especially in the form of astragalus and calcaneum, reflecting tarsal elongation and specialization for parasagittal motion of the ankle and hindfoot; tarsal elongation in the terminal species, B. robustum , is especially evident from the shape and marked elongation of the cuboid and adjacent tarsal elements; a limited ability for inversioneversion of the hindfoot is retained at the astragalonavicular and calcaneocuboid joints; absence of a complete vertebral column prevents detailed analysis of the participation of the back as it contributes to stride length; however, correspondence of the few known vertebrae of Borocyon to those of Daphoenodon superbus leaves little doubt as to the presence of a similar column in which an anticlinal vertebra serves as a fulcrum for extension and flexion of the spine during the running/bounding gaits, contributing to an increase in stride length—this inference receives support from the partial vertebral column of B. niobrarensis (UW 10004) from Horse Creek Quarry in Wyoming, which matches the column of D. superbus ;

(10) absolute limb lengths increasing steadily from Oligocene Daphoenus vetus to early Miocene Daphoenodon superbus , continuing through Borocyon neomexicanus , B. niobrarensis , and ending with B. robustum , paralleling the increase in body size in the daphoenine lineage; however, a proportional increase in length of the distal forelimb also occurs in daphoenines, beginning with Daphoenus vetus (humeroradial ratio,84%), Daphoenodon superbus (86.7%), and concluding with the species of Borocyon (,95%–97%); Borocyon robustum displays the most evolved lower forelimb, relative to the propulsive felidlike hindlimb, of any New World daphoenine or amphicyonine beardog;

(11) dimorphism evident where sample size is adequate: examples include the mandible (with dentition), radius, astragalus, calcaneum, and metapodials; large males and smaller females of both Borocyon robustum and B. niobrarensis are inferred from the presence of dimorphic (robust and gracile) metapodials of both fore- and hindfeet coming from four different localities (table 6).

GAIT AND STANCE

The musculoskeletal structure of a quadrupedal carnivoran predator reflects the locomotor strategy used to obtain its prey. Anatomical innovations recorded in the skeleton as preserved in the fossil record can be described and measured by the investigator, whereas physiological and behavioral attributes that accompany skeletal adaptations usually cannot be identified in fossils. As a result, we lack important information such as estimates of metabolic efficiency and endurance bearing on the economy of pursuit. These require knowledge of oxygen uptake, muscle cross-sectional area and fiber orientation, energy economy of various gaits, as well as evidence of social behaviors such as group hunting. Despite such limitations the musculoskeletal configuration and proportions of the limbs and feet of fossil carnivorans provide insight into the locomotor adaptations aiding in prey capture. To overtake prey, a predator benefits from enhanced speed, acceleration, maneuverability, and endurance, and these attributes combine in differing proportions in the various species of living Carnivora , where they are reflected to a greater or lesser degree in the skeleton. Thus, in fossils the appendicular skeleton remains a useful albeit incomplete guide. Lengthening of the limbs and adoption of a more upright stance, accompanied by skeletal modifications promoting a more restricted fore–aft motion of the limb and foot, are obvious adaptations influencing the gait, stride, and energy budget of Cenozoic carnivorans.

The term ‘‘cursorial’’ has been commonly used to describe carnivorans seemingly adapted for sustained, often rapid, efficient locomotion. These ‘‘running’’ mammals display anatomical specializations of the skeleton that are thought to contribute significantly to this cursorial ability. Some recent investigators have regarded the use of the term as ill-defined and nonspecific as to the morphological and physiological attributes that are essential to the definition of a cursorial mammal ( Stein and Casinos, 1997; Carrano, 1999).

It has been generally agreed that some discussions of cursoriality in mammals confuse speed and endurance, whereas these two aspects merit separate consideration. There is no a priori reason why a very fast mammal should also exhibit great physiological endurance and, similarly, the converse is true. Some ambush predators may be very fast yet cannot sustain this speed for any distance, whereas other mammals may travel great distances but at rather average rates of speed. Taylor (1989) identified three types of ‘‘cursors’’: (1) running carnivorans that possess both speed and endurance (e.g., gray wolf); (2) sprinting carnivorans that employ an initial rapid acceleration, capturing prey by their speed over a shorter distance; and (3) ‘‘trotting’’ carnivorans that engage in prolonged pursuit but at only moderate speeds.

Some suggested definitions of the term that have arisen as a result of recent biomechanical studies cannot be employed for fossil mammals because they emphasize physiological characteristics that are indeterminate for extinct species. As a result, paleontologists have tended to emphasize morphological features that distinguish rapid and metabolically efficient living mammals, most of large body size, usually. 10 kg. Here I use the term cursorial to refer to carnivorans with morphological specialization of their postcranial skeleton favoring parasagittal alignment of the limbs and demonstrated ability for fore–aft movement of the limb segments and feet, indicating increased stride length and locomotor efficiency. There is the tacit assumption that physiological endurance and an economy of gait in some form are the likely accompaniments to this morphology.

The skeletal morphology of species of Borocyon is reasonably compared with the anatomical adaptations of living large carnivorans of similar body size, while conceding that physiological attributes, whether mundane or remarkable, must remain unknown. However, from this comparison comes the realization that Borocyon had evolved a skeletal morphology unique to large carnivorans of its time, prefiguring many parallel adaptations in large predatory Carnivora of the late Cenozoic and Recent. Moreover, in the terminal species, B. robustum , we find a mosaic of skeletal features not seen in combination in any large living carnivoran. The existence in the fossil record of carnivores with a blend of adaptive traits of the locomotor system not found in any living carnivoran was recognized by Van Valkenburgh (1987).

Lengthening the limbs, particularly the lower limb and foot, are essential to the endurance strategy by increasing stride length. Recent studies of mechanical design in mammals suggest advantages (mass-specific muscle forces are reduced by increasing their mechanical advantage) that often accompany lengthening of the limbs and attaining a more upright stance ( Biewener, 1989 a, 1989 b, 1990). Energy is further conserved by concentrating muscle mass in the proximal limb segments, thereby reducing the limb’s moment of inertia and the energetic cost of locomotion ( Myers and Steudel, 1985). Thin distal tendons transmit the force of a much thicker, proximally placed muscle while effectively reducing the mass of the lower limb.

Table 4 (fig. 19A, B) presents fore- and hindlimb proportions for amphicyonids relative to those of living ursids, felids, and canids based on the length ratios of radius/ humerus and tibia/femur. Ursids, regardless of size, consistently have a tibia,83% (many,75%) the length of the femur. Felids possess a tibial length.83% the length of the femur, hence always longer than ursids, and certain felids (the puma, clouded leopard, and cheetah) exceed 90%, with the cheetah.100%. Canids (Caninae) also have markedly elongated the tibia, having ratios.96% and often reaching 107% in several species. Against this background, the ‘‘shortlimbed’’ amphicyonids have ratios generally falling between,80% and 90%, with Borocyon estimated to fall within this range at,85%–90%. Thus, hindlimb proportions of Borocyon are much like those of Panthera leo and P. tigris and do not show the extreme ‘‘cursorial’’ adaptations exhibited by the wolf and cheetah. The B. robustum hindlimb is capable of powerful forward thrust while nonetheless incorporating a number of skeletal features favoring fore–aft motion in all hindlimb segments.

In the forelimb, proportional relationships among these carnivoran families differ from those evident in the hindlimb (table 4, fig. 19A). Ursid ratios range from 79% to 86% in most bears, but the black bear data suggest a tendency to lengthen the lower forelimb (,82%–92%) in this species. Most amphicyonid ratios fall below 90%. Howev- er, among amphicyonids, the species of Borocyon are remarkable in the lengthening of the distal forelimb. Values of,95%–100% exceed the ratios of almost all cats except for the cheetah, which has a ratio similar to Borocyon . Other than the cheetah, the only other large felid exhibiting distal forelimb elongation is the lion with ratios consistently.90%. Living canids exhibit pronounced elongation of the radius/ulna with ratios from 97% to 108%.

From these observations it is evident that the forelimb can lengthen independently from the hindlimb, and within a family (such as felids), limb proportions of species may vary while reflecting a common anatomical pattern: (1) ursids usually do not elongate the lower limb bones in either fore- or hindlimb but maintain short lower limb segments; (2) living canids are in direct opposition to ursids in that they elongate both the lower fore- and hindlimb segments; (3) felids exhibit a variety of possible options in limb elongation—the lion and tiger show similar hindlimb ratios, whereas the cheetah clearly parallels canids in elongation of both fore- and hindlimb, and the lion (relative to the tiger) evolved a lengthened forelimb apparently as an adaptation for pursuit of prey on the open African plains.

Certain species of living carnivoran ‘‘cursors’’ that employ trotting gaits sustained for long intervals lengthen both hindlimb and forelimb, as exemplified by canine Canidae . Others such as the lion and cheetah that engage in rapid, often short, but what in some situations can evolve into a prolonged pursuit, elongate the lower forelimbs, but not to the degree seen in Caninae. And the lion, while exhibiting some lengthening of its forelimb (fig. 19A), still maintains a rather short hindlimb for powerful thrust as it explodes from ambush.

Borocyon is exceptional among amphicyonids and even large living felids in elongation of the lower forelimb (fig. 19A; table 4). Whereas the Borocyon hindlimb is similar in its proportions to that of Panthera leo , P. tigris , as well as some amphicyonids, forelimb proportions are comparable to the cheetah and wolf.

In addition to increasing stride length, economy of motion is also achieved by confining the limb to a fore–aft path. This is evident in modification of skeletal structure, particularly the shape of limb and foot bones and congruent joint surfaces. Living canine canids ( Canis lupus , C. latrans ) exemplify many skeletal traits seen in limbs modified for fore–aft movement: these include closely registered joint surfaces in the lower limb segments; by bringing the elbow inward to lie under the shoulder; by increasing extension of the lower forelimb on the humerus to produce a more erect stance; by reduction of the carpal extensors and flexors arising from the distal humerus; by close apposition of the radius and ulna so that pronation/supination is more limited; by tight appression of the metapodials; by development of pronounced keels on the distal metapodials to lock the proximal phalanges; by interlocking of astragalus and calcaneum; and by the deeply grooved astragalar trochlea inserted in the distal tibia-fibula, practically limiting tarsal motion to a single degree of rotational freedom about the tibia. These morphological specializations contribute to the seemingly effortless habitual trotting gait of a canine pack on the hunt.

The species of Borocyon lack many of these canine skeletal adaptations for fore–aft motion of the limbs but they have evolved enough in parallel to suggest a facultative ability for some form of limited pursuit. Of these, the most significant are: (1) inward reorientation of the elbow demonstrating a more parasagittal alignment of the forelimb, and an elbow joint restricted to parasagittal flexion/extension; (2) modified form of the distal humerus indicating an erect forelimb stance with proximal ulna deeply inserted in olecranon fossa of the humerus, and a diminished medial epicondyle of the humerus indicating reduced torque at the elbow joint; (3) lengthening of the radius-ulna, closely united by a narrow interosseous ligament and membrane, suggesting less capability for pronation/supination of the manus, and a slender bladelike radius with reduced proximal and distal ends, as seen in radii of the wolf and cheetah; (4) paraxonic digitigrade fore- and hindfeet with straight-shafted elongate metapodials; (5) pronounced distal keels on the metapodials; (6) various integrated adaptations of femur, tibia, and hindfoot that promote propusive thrust but that also reflect an incipient parasagittal alignment and limited fore–aft motion of the hindlimb; these adaptations include evidence of femoral adduiction, proximal shift and reduction of posterior tibial musculature, an altered form of proximal tarsals and metatarsals in keeping with an erect digitigrade stance, and lesser emphasis on eversion-inversion of the hindfoot.

These skeletal features indicate that Borocyon robustum was probably capable of an economy of gait involving a lengthened digitigrade stride and parasagittal fore–aft limb and foot motion differing from that of all contemporary amphicyonine beardogs. A powerful propulsive thrust provided by the hindlimbs, together with the robust muscled shoulders and lengthened forelimbs, conveys an image of a large, formidable carnivoran capable of strong acceleration and an interval of prolonged pursuit, but without the extreme skeletal and perhaps physiological adaptations of large canine canids.

Whether Borocyon species were able to employ an extended trotting gait for long intervals over open terrain requires a currently unavailable knowledge of physiology and behavior. However, the fact that these species occupied semiarid, open plains in the North American midcontinent during the early Miocene seems to implicate that environment in selecting for such traits, at least locally. It is suggestive in this regard that Borocyon from the Suwanee River of the Florida Gulf Coast does exhibit somewhat shorter limbs than its Great Plains counterpart.

Once these large carnivores overtake their prey, adaptations for the kill come into play, involving anatomical specialization of the teeth, skull, jaws, and forelimbs. The killing strategy may require that the forelimbs adopt a compromised structure so that they function both in locomotion and in prey capture. This is evident in a large felid such as the lion in which the forelimbs are essential both in pursuit and in holding and dispatching the prey. However, in canine cursors the potential use of the forelimb in prey capture is sacrificed to the fore–aft gait, a compromise favoring endurance, and is accompanied in wolves and African hunting dogs by reliance on pack hunting. Sustained gaits of the canine type must be powered by aerobic locomotion, and since the metabolic rate of active striated muscle achieves maximum efficiency at a speed of contraction somewhat less than the speed of maximum power output ( Alexander, 2003), a gait such as the sustained trot seems well adapted to endurance pursuit in these canine cursors. The limb morphology of Borocyon robustum occupies an ‘‘intermediate’’ position between limbs of a large predatory felid such as Panthera leo and those of fully specialized cursors, the wolf and cheetah. Such a large intermediate morph does not exist among the extant carnivoran fauna.

From this perspective, Borocyon robustum emerges as an exceptional predator, not only among amphicyonids but when compared to large carnivorans throughout the Cenozoic. When considering the totality of its skeletal adaptations, this beardog was likely capable of some facultative form of sustained pursuit. Because the species of Borocyon lack the retractile claws of large living felids ( Panthera leo , P. tigris ), grasping and holding the prey during the kill would not have been an option. The massive skull with canidlike dentition suggests that the jaws and teeth were employed much as in large wolves. Its skeleton, with parasagittally aligned limbs anatomically specialized for more restricted fore–aft locomotion, suggests that a wideranging trotting gait emphasizing endurance may have been part of the adaptive strategy of this carnivore.

Physiological unknowns limit the degree of confidence that can be placed on conclusions derived solely from skeletal morphology ( McNab, 1990). For example, an estimate of a carnivore’s metabolic rate can be based upon the difference in consumption of oxygen between its resting rate and the maximum rate that can be achieved; this is termed the metabolic scope ( Schmidt-Nielsen, 1984). This ratio has been found to approximate a value of 10 for most mammals ( Hemmingsen, 1960; Taylor et al., 1981; Schmidt-Nielson, 1984) and is independent of body size. From these investigations, it emerged that both domestic dogs and wild canids such as the wolf possessed a metabolic scope three times greater than the general mammalian value ( Langman et al., 1981), a 30-fold increase in oxygen consumption above the resting rate. This result has a significant bearing on the endurance of these carnivorans and influences their style of locomotion and preferred gaits. Such information remains largely beyond reach for extinct amphicyonids, yet if attainable would influence our evaluation of their locomotor pattern.

The species of Borocyon were not the first lineage of Carnivora to evolve limbs special- ized for fore–aft motion and a lengthened stride. Temnocyonine amphicyonids developed elongate limbs and feet suggestive of a pursuit mode similar to that of living canids, yet the evidence of their postcranial anatomy is relatively sparse. Only five individuals are represented by skeletons that nonetheless convey a relatively complete picture of the anatomy of both fore- and hindlimbs. A few additional isolated bones from other localities can be directly compared with the more complete skeletons to provide a somewhat larger sample size. At least one partial postcranial skeleton assigned to each of the three currently recognized genera ( Temnocyon , Mammacyon , and an undecribed genus) establishes that elongation of the limbs and feet was typical of end-member species of all of these lineages in the late Oligocene–early Miocene. Taken together, the species of Borocyon and the few temnocyonines represent the first Neogene experiments evolving a long-limbed, mid-sized to large carnivoran predator within North America.

DENTITION AND FEEDING

The dental formula of 3-1-4- 3 in upper and lower jaws is consistent for nearly all species of North American amphicyonids except the temnocyonines, Borocyon robustum , and Clarendonian Ischyrocyon , which have lost M3. The migrant amphicyonines ( Amphicyon , Ysengrinia , and Cynelos ) that entered North America from Eurasia in the early Miocene all retain the full complement of molars (M1– M3/m1–m3) but reduce the size of their anterior premolars (P1–P3, p1–p3). North American daphoenine and temnocyonine amphicyonids retain well-developed premolar batteries in contrast to amphicyonines that live alongside them. Except for the presence of M3, the teeth of amphicyonids are similar in number and form to those of canids, so much so that they were for many years classified as Canidae (e.g., Simpson, 1945; Romer, 1966). Despite the similarity of the dentition, the morphology of the mandibular symphysis of daphoenine amphicyonids differs in some respects from Scapino’s (1965) description of living Canis . However, as in canids, the symphysis of amphicyonids remains unfused, with the exception of some very old adults.

JAW REGISTRATION

In Daphoenodon superbus the proportions of the holotype skull and mandibles (CM 1589) are only slightly distorted so that dental relationships during occlusion can be observed. Here, as in the wolf, the upper and lower toothrows are in contact in central occlusion yet the carnassials are not tightly registered. As the jaws are brought together in the wolf ( Canis lupus ), neuromuscular mechanisms first approximate the carnassial blades, followed by interlocking of the canines and I3 ( Scapino, 1965; Mellet, 1981). Because canine contact is retarded relative to that of the carnassial pair, the wolf does not utilize the autocclusal bite mechanism described by Mellett (1984) in which the canines establish contact prior to the carnassials. In Canis lupus the canines do not come into firm contact until after the carnassials have begun to shear. However, in young D. superbus , the upper and lower canines make contact earlier in the bite, although not yet in a tightly locked relationship. As the carnassials wear over time, manual occlusion of the teeth in these beardogs suggests that interlocking of the canines eventually coincides with initial carnassial contact, suggesting that autocclusion employing the canines becomes more prevalent with age. This situation also apparently obtains in Borocyon robustum and B. neomexicanus where dental and mandibular dimensions correspond to those of the smaller D. superbus . Unfortunately, all skulls of B. robustum are distorted by crushing and lack canines so that the mandibles cannot be reliably occluded with the upper teeth. Nonetheless, certain aspects of its jaw mechanics can be described with some confidence.

In carnivorans with sectorial carnassials and a flexible mandibular symphysis, registration of the working-side mandible with the corresponding upper toothrow involves a lateral shift of the mandible as closure takes place. As the jaws close, the edges of the carnassials are brought into shearing registration by a slight lateral adjustment of the working mandible. Scapino (1965) described a coupling action in a canid between the zygomandibularis and medial pterygoid muscles of the working mandible that maintains the carnassials in close registration. The precise registration of the carnassial blades, however, is facilitated by the shape of the craniomandibular joint and a flexible mandibular symphysis, which allows a final outward movement of the rear of the mandible to bring the m1–P4 into tight occlusion. At the end of the closing stroke, interlocking of the canines and I3 brings the mandibles into centric occlusion. As an individual ages, the molars and carnassials develop subhorizontal to flat wear surfaces in both canids (e.g., Canis lupus ) and in amphicyonids. In old age, then, a precise occlusion was no longer possible or necessary because the shearing blades were worn away, leaving only broad crushing surfaces. At this point, some aged beardogs fused the mandibular symphysis and employed a jaw mechanism devoted entirely to crushing occlusion.

MANDIBULAR SYMPHYSIS

Precise registration of the teeth is aided by a mobile mandibular symphysis in many species of Carnivora . The anatomy of the mandibular symphysis was analyzed in depth by Scapino (1965, 1981). Carnivoran symphyses were classified according to the way soft tissues in the symphyseal space related to the form of the mandibular symphyseal plates in histologic sections. Although in living Carnivora the form of the symphysis can differ with increasing body size and even within a species, the smallest species of Daphoenodon from the Great Plains and the largest Borocyon have similarly configured mandibular symphyses. These symphyses do not precisely correspond to any of Scapino’s (1981) symphyseal types (Classes I–IV). However, the conjoined hemimandibles approximate his Class II symphysis where each symphyseal plate exhibits a smooth rectangular area anterosuperiorly and the remain- der of the plate is characterized by numerous prominent, randomly oriented bony rugosities. The flat, non-rugose symphyseal plates of the wolf, considered a highly flexible Type I symphysis by Scapino, differ from these rugose, interdigitating plates evident in the symphyses of Daphoenodon and Borocyon species (fig. 38A, C).

The mandibular symphyseal plate of a female B. robustum (fig. 38A, UNSM 27002), in which the symphysis is well preserved, shows the anterosuperiorly placed, rectangular zone, with a smooth surface presumably for the attachment of a single, elongate fibrocartilage (length, 17.8 mm; width, 8.0 mm). In living carnivorans this fibrocartilage facilitates compression and expansion of the symphyseal joint. Below and posterior to the fibrocartilage zone, the remaining surface of the plate displays prominent interdigitating bony rugosities that show random alignment. These rugosities interlock with those of the opposing plate to prevent translation of the hemimandibles. The lower border of the mandible displays foramina leading into valleys between the rugosities, suggesting the possible presence of a venous network within the symphysis as described by Scapino (1981).

Additional intact mandibles of B. robustum (UNSM 25548, 25684, 26417) were carefully prepared and these exhibit a morphology of the symphyseal plates as in UNSM 27002. The mandibles of Borocyon neomexicanus (F:AM 49239) and B. niobrarensis (ACM 3452) that preserve an intact symphysis exhibit a similar morphology. Because the symphyseal plates in individuals of Daphoenodon superbus display the same pattern, this type of mandibular symphysis probably characterized the Borocyon lineage from its inception.

Of the carnivoran symphyses discussed and illustrated by Scapino (1981: fig. 3), the gross form of the symphysis of B. robustum seems at first to most closely approach Ursus americanus (fig. 38B). However, in B. robustum the fibrocartilage zone is larger and better developed than in the bear, and the robust, interdigitating rugosities lack the pronounced anteroposterior alignment evident in Scapino’s (1981: fig. 3F) example of Ursus (some mandibles of Borocyon robustum do show an anteroposterior alignment of a few rugosities in the posterior symphyseal plate). The ursid symphyseal plate is elliptical in form, similar to that of Borocyon robustum , except that in ursids there is no evidence of a single, large anterosuperiorly placed fibrocartilage pad. In ursids the pad is narrow and irregular in form ( Scapino, 1981) and lacks the smooth subrectangular surface situated along the upper part of the symphyseal plate that indicates its presence in the species of Borocyon and in Daphoenodon superbus .

Compression and expansion of the fibrocartilage pad contributed to flexibility of the symphyseal plates in Borocyon . In carnivorans with flexible symphyses, opening and closing of the posterior symphysis occurs about a vertical axis through the compressible fibrocartilage ( Scapino, 1981). Slight rotation of the mandibles about their long axes, guided by the canines as the jaws close, also places the pad under compression as the carnassials are brought into contact during the bite. The fibrocartilage in B. robustum , just as in Canis ( Scapino, 1965) , probably accommodated movements of the mandible about its long axis during carnassial shear and crushing by the molars (m2–m3, M1– M2). The only articulating pair of hemimandibles of B. robustum belonging to a single individual (UNSM 25548, 25684) shows that rotation about a vertical axis through the symphysis results in opening of the posterior symphysis, and also permits slight rotation of these mandibles about their long axes. Furthermore, the presence of a single, large fibrocartilage pad clearly indicated in several individuals (UNSM 27002, 26417, 25548) indicates an ability for compression about a vertical axis through the pad and hence some symphyseal mobility.

The considerable height of the guiding canines and carnassials in B. robustum places a constraint on side-to-side shift of the jaws during the closing phase of the power stroke. Scapino (1981) described this action involving flexible symphyses of Types I and II, which is applicable in modified form to B. robustum . As the working mandible moves laterad, the posterior symphyseal space widens, with this action occurring around a vertical axis passing through the fibrocartilage pad. As the mandible comes to a stop,

+

plates prevent translation of the hemimandibles more effectively than the smoother symphyseal plates of the wolf. The ursid lacks an extensive fibrocartilage and has a more rigid mandibular symphysis.

constrained by the tissues of the craniomandibular joint, the carnassials are aligned but not yet in contact and the canines have met to guide the closure of the jaws. As the working mandible closes, it rotates inward slightly as m1 slides upward against P4. This compresses the fibrocartilage pad, which will return to its initial shape as the jaws open.

Scapino (1981) observed that a strongly interdigitated symphysis with limited or absent fibrocartilage pad was a stiff symphysis. He recognized this type of symphysis in large living felids and ursids. In ursids the interdigitating rugosities and various binding ligaments make dorsoventral or anteroposterior translational shear of the symphyseal plates impossible. However, manipulation of the hemimandibles of Ursus americanus shows that a very limited amount of rotation about the long axis of the mandible can occur. On the other hand, manipulation of the hemimandibles of B. robustum (UNSM 25548, 25684) demonstrates much more symphyseal mobility, involving rotation around the long axis of the mandible and also about a vertical axis through the fibrocartilage. Under Scapino’s classification, a mandibular symphysis as seen in B. robustum would differ from those of ursids and should be similar to Class II with limited flexibility. The condition of the mandibular symphysis and dentition in B. robustum accords with Scapino’s (1981: 370) summary analysis whereby a mobile symphysis and large carnassials might be expected in carnivorans processing hard materials such as bone. In durophagous carnivorans of large size, the complex interdigitating, rugose symphyseal plates prevent mandibular dislocation while interposed soft tissues still permit some flexibility.

In summary, the sample of mandibles of Borocyon robustum demonstrates the probable presence of a single, large fibrocartilage pad; of limited movement about the long axis of the mandible and around a vertical axis through the symphysis when hemimandibles are conjoined; and of foramina entering the rugose bone at the ventral border of the symphysis. These traits taken together suggest (1) a compressible symphyseal pad; (2) posterior opening and closing of the symphysis around a vertical axis through the pad; (3) the ability to adjust carnassial registration by limited mandibular axial rotation; and (4) perhaps incorporation of venous blood within the symphysis involved in volume adjustments (expansion–contraction) of symphyseal connective tissues. Some such movement at the symphysis is essential to precise registration of the large Borocyon carnassials, as evidenced by wear patterns on these teeth.

Scapino (1981) considered a flexible symphysis with a fibrocartilage pad, intermandibular cruciate ligaments, and a venous plexus as primitive for carnivorans. The mandibular symphysis in Borocyon robustum possibly still reflects this primitive condition, but with adjustment for increased body size, here considered to be the pronounced rugose interdigitation preventing translation of the symphyseal plates while still permitting some flexibility in the joint.

DENTAL OCCLUSION AND TOOTHWEAR

It is possible to accurately describe dental occlusion and toothwear in several North American amphicyonids where adequate samples exist. A sample of Daphoenodon superbus from the den site at Agate Fossil Beds National Monument ( Hunt et al., 1983), and from coeval waterhole deposits ( Hunt, 1990), provides a series of ontogenetic stages from juveniles to aged adults, totaling at least 14 individuals. Mandibles of this species range in ontogenetic age from a young individual with only milk teeth in place to aged males with the molars nearly worn flat. Wear on the teeth, both premolars and molars, is consistent with patterns observed in other species of the genus—these species probably employed the teeth during feeding in the same manner.

As an individual of D. superbus ages, its teeth show the greatest degree of wear in two areas of the dentition: the canines and the carnassial/molar group. Incisors are gradually blunted by wear over time, yielding flatsurfaced pegs. Premolars (p1–p4, P1–P3) are not heavily worn, even in the oldest individuals in which wear is limited to blunting the principal cusp on each tooth (on p4, both the principal and posterior accessory cusps are eventually worn flat). The degree of wear increases as one proceeds backward in the toothrow, culminating in heavily worn carnassials and molars. Thus, it is evident that food taken into the mouth was transferred to the rear where the focus of mastication occurred.

The dentition of both juveniles and adults of the species of Borocyon shows near identity of dental pattern and wear relative to the teeth of the D. superbus sample; a similar dental/mandibular function and ontogeny of tooth usage and wear is inferred for these species, presumably retained through descent from populations of late Arikareean Daphoenodon . Toothwear facets on the carnassials and molars in D. superbus provide evidence of shear as well as the blunting of cusps. The shearing blade of m1, formed by the labial surfaces of paraconid and protoconid, is nearly vertical in young adults in which wear has just been initiated (this wear surface forms a ‘‘V’’ on the labial face of the trigonid, the apex downward, ending at the cingulum). A reciprocal vertical wear surface on P4 appears on the inner surface of the metastylar blade and paracone. As m1 slides past P4 during shear, m1 is shifted slightly linguad as it moves along the inner face of P4 and the lingual surfaces of the M1 paracone and metacone.

Without food between the teeth, full closure of the jaws results in wear facets produced by direct contact of upper and lower canines, incisors, and p4–m3 with P4– M3. There are several contact points along the toothrow: (1) the interdigitating alignment of the canines and third incisors produces deeply grooved vertical facets on these teeth, but only flat wear surfaces appear on the blunted tips of the inner incisors; (2) insertion of m1 into the embrasure between P4 and M1 takes place without contact of the m1 protoconid with the maxilla—the upward trajectory of m1 is arrested by contact of its talonid basin with the M1 protocone, and by contact of m2–m3 with the posterior half of M2 and M3; and (3) the principal and accessory cusps of the large p4 produce a vertical facet on the anterior face of P4, and the principal p4 cusp also creates a facet on the inner heel of P3. The more anterior premolars usually do not directly occlude; food is trapped between their principal cusps and worked back to the carnassials and molars by the tongue—the horizontal, blunt wear surfaces limited to the principal cusps of the anterior premolars show that pressure contact with their tips is the principal cause of tooth wear.

Examination of an ontogenetic series of wolf ( Canis lupus ) dentitions shows that there is an established sequence of wear facets that develop on the carnassials, molars, and premolars as the animal ages. This sequential wear pattern is closely approximated by mandibles of Daphoenodon . Initially, in the wolf, only the tips of cusps are slightly blunted by wear, chiefly on the carnassials. Next, significant wear appears on the carnassial pair as a narrow strip of dentine where the enamel has been breached along the cutting edges of P4 and the m1 trigonid; this strip extends from paraconid to protoconid on m1, and from paracone to metastylar blade on P4. Wear on P4 is often pronounced in the carnassial notch itself. This wear pattern is consistent with the Every effect, the interlocking of upper and lower carnassials by resistant food material ( Mellett, 1981). Other than the wear on P 4 in the upper teeth, only the para- and metacone of M1 show flat wear facets from slight blunting of these cusps. This stage is followed by a confluence of horizontal facets from increased blunting of the cusps on m1 proto- and paraconids and the metaconid. P4 shows further wear on paracone and metastylar blade creating a planar surface that is slightly medially inclined. The paracone and metacone of M1 are next worn to more extensive flat surfaces but remain tall, and the principal cusps of the anterior premolars (p2–p4, P1– P3) now show initial wear. Finally, in late wear, the m1 protoconid and paraconid are heavily worn to a nearly continuous planar surface—hypoconid and entoconid are worn to a flat platform as is the m2 protoconid, producing a uniform crushing surface extending from the m1 talonid to include the entire m2. The P4 paracone is also heavily worn to form a lingually inclined surface that extends onto the metastylar blade. Wear on the para- and metacone of M1 has produced a continuous horizontal surface. Wear on the tips of the anterior premolars yields blunt cusps, with wear most extreme on the more posterior teeth (P3, p4).

If the mandibles of wolves are manually occluded at each of these various wear stages, it becomes evident that tooth-to-tooth contact involving carnassial shear during food processing is not the principal agency producing wear facets on the teeth. The only possible cause of this wear pattern during ontogeny is the processing of hard food objects that serve to blunt the teeth. It is well known that wolves will process and often consume bone of their prey along with meat ( Mech, 1970; Mech et al., 1971; Haynes, 1982, 1983; Gazzola et al., 2005), and it is bone and other resistant connective tissues that are responsible for the blunting of the cusps over time. Young and Goldman (1944: 245) reported the contents of the stomach of a male Alaskan wolf that bears on this point: ‘‘The stomach was about half full of the hair, skin and leg bones of a Dall sheep. These bones had been bitten into small sections about one inch in length by the strong carnassial, or chopping teeth, of the wolf.’’ Van Valkenburgh (1988) found that when processing hard materials, tooth breakage is a frequent occurrence among large living carnivores, with bone-processors most prone to tooth fracture. Of the eight B. robustum mandibles, two show damaged teeth subsequently abraded through later wear.

The same pattern of blunting of the cusps on carnassials, molars, and premolars of the wolf is evident in Borocyon robustum , the only species of Borocyon where the sample of mandibles is adequate to judge ontogenetic wear. However, in these Borocyon mandibles the m2 is lengthened relative to m2 of Canis lupus , creating a longer, more developed crushing platform that extends from the m1 talonid to m3. This platform in B. robustum is 35%–36% (N 5 6) of the functional toothrow length (p2–m3). It is,26% in the wolf (N 5 7). It is this platform that serves as the locus for bone processing in the African hunting canid Lycaon pictus ( Van Valkenburgh, 1996) . Thus, the wear pattern of the dentition, the presence of a crushing platform at the rear of the toothrow, the structure of the mandibular symphysis and craniomandibular joint, as well as the robust skull with massive temporomandibular musculature suggest that durophagy was probably important in food processing by these large, mobile carnivorans.

Scapino (1981) considered a flexible symphysis to be a practical adaptation in large durophagous carnivorans crunching hard bone, allowing the teeth to avoid sudden damaging impacts when breaking hard material, a behavior made possible by a compensating rotation of the mandible about its long axis. Borocyon robustum represents an end-member of a lineage that came to rely on carnassial/molar crushing for the processing of hard materials.

The method of Therrien (2005) was used to create a mandibular force profile for Borocyon robustum (table 7). The hemimandible (UNSM 25684) that best occluded with the paratype cranium (see frontispiece) was measured because of the lack of distortion of its mandibular corpus and dentition. This is a large individual (UNSM 25684), probably male, but not the largest of the available hemimandibles of the species. Bite force calculated at successive points along the length of the mandible reflects adaptation to load-bearing ( Biknevicius and Ruff, 1992 a, 1992b). Bending strength is represented by the section modulus (Z) calculated from measurements of mandibular depth and width at these points.

Therrien (2005) argued that the mandibular force profile derived from these values for jaw depth and width contributed insight into feeding behavior. He applied beam theory to the hemimandible, which was modeled in cross-section as a solid ellipse of bone. Biknevicius and Ruff (1992a) developed a similar model that accounted for the absence of bone within the mandibular cylinder (medullary space), a method requiring planar radiographs of the mandible to measure cortical bone thickness.

The mandibular force profile for B. robustum (fig. 39) was compared to those for canid, felid, and hyaenid carnivores presented by Therrien (2005: figs. 3–7). Mandibular depth and width were measured at the same interdental loci utilized by Therrien (2005: fig. 2). Bending strength in the dorsoventral (Zx) and labiolingual (Zy) planes was calculated from these measurements as well as relative mandibular force (Zx/Zy). The values for Zx and Zy, relative to the distance (L) from the articular condyle to each interdigital locus, allow a comparison among

TABLE 7

Mandibular Force Profile for Borocyon robustum (UNSM 25684) Calculated According to Therrien’s (2005) Method of Beam Analysis

carnivore species of the magnitude of bending strength along the length of the mandible.

Zx/Zy values are proportional to the ratio of the depth and width diameters. Ratios.1.00 demonstrate resistance to dorsoventral loading (a deep, narrow mandible); a ratio,1.00 indicates a resistance to labiolingual torsion or bending. Therrien (2005) suggested that the ability to tolerate high dorsoventral loading enabled the animal to exert considerable bite force on its prey. Resistance to labiolingual forces permitted the carnivore to withstand strong torsional and transverse forces that result from the processing of hard materials or from the violent actions of struggling prey.

The mandibular force profile of B. robustum closely corresponds to those of living canids such as the grey ( Canis lupus ) and dire wolves ( Canis dirus ), but it parallels living hyenas in the strength of the symphyseal region (fig. 39). Therrien (2005) considered that the high Zx/Zy values of hyaenids (,1.20) at the canine reflect a well-buttressed symphysis, on occasion involved in bone processing in Crocuta . The Zx/Zy value for B. robustum at the canine is also 1.20 but it is unlikely that the symphyseal region was routinely involved in bone processing. Although the highly interdigitated, rugose bone of the Borocyon symphysis attests to its strength relative to that of the wolf, the force profile suggests that durophagous processing was accomplished by the post-carnassial battery that forms the crushing platform at the back of the mandible. As in canids studied by Therrien, the slope for the dorsoventral force values (Zx/L) was steeper than that for the relatively uniform labiolingual values (Zy/ L, table 7), indicating that dorsoventral bending strength is pronounced beneath the posterior cheek teeth. The Zx/Zy values similarly indicate increased dorsoventral buttressing of the mandible beneath the crushing molars.

A deep mandibular corpus beneath the molars was interpreted by Therrien (2005) as an adaptation for bone processing, and it supports the inference that B. robustum was a durophagous feeder (note the marked increase in Zx/Zy behind the carnassial from 2.56 to 3.33, which is similar to Canis lupus at 2.66 to 3.24). When considering labiolingual forces (Zy/L) applied to the mandible during torsion and the application of transverse stresses, the mandible of B. robustum is apparently equally strong at the canine and in the post-carnassial region, reflecting that the emphasis in mandibular reinforcement is principally focused on a highly developed resistance to dorsoventral bending in the postcarnassial mandibular corpus.

Application of Therrien’s method to B. robustum does not provide absolute values for bite force but yields a useful relative comparison with living carnivores that suggests a feeding style involving quick, shallow but powerful bites, a strongly buttressed mandibular symphysis, and the processing of hard materials by the posterior cheek teeth, hence a composite of canid and hyaenid attributes, and in overview the possibility of a pack hunting lifestyle. No

(in part from Therrien, 2005). Values represent relative mandibular force (Zx/Zy).

large living carnivore is an exact comparator to Borocyon robustum in its mandibular morphology and force profiles—its feeding style appears to have combined adaptations distributed today among large canids and hyaenids. Perhaps most striking, however, is that the forces generated by its mandibular biomechanics were indeed formidable.

In the totality of its mandibular and dental characteristics, Borocyon robustum remains somewhat outside the morphologic spectrum represented by living large carnivorans, a predator whose anatomical mosaic of craniodental and postcranial traits existed only briefly in the North American early Miocene interval.

CONCLUSIONS

The daphoenine amphicyonid Borocyon robustum was the dominant long-limbed predatory carnivoran of the North American early Miocene. The species occurs from the Pacific Northwest through the Great Plains to the Florida Gulf Coast, a geographic range made known only recently. Although previously unrecognized as a keystone species of its predator guild, due to scarcity of fossil remains, currently available craniodental and postcranial material provide a nearly complete account of the skeleton. The principal skeletal modifications evolved within the Borocyon lineage were identified by compar- ison with the plesiomorphic species Daphoenodon superbus and D. notionastes , which are both ‘‘short-legged’’ daphoenine amphicyonids.

Borocyon robustum represents the conclusion of a remarkable experiment within the daphoenine Amphicyonidae , evolving a longlimbed pursuit predator, capable of securing its prey through enhanced muscular strength and an economic gait that probably emphasized endurance over speed. The lengthened forelimb, equaling proportions of efficient living runners such as wolves, attests to this. The forelimb is remarkable for an early Miocene amphicyonid, and its musculoskeletal characteristics are distinct from all New World amphicyonines. Elongation of the distal forelimb is most similar to that of living wolves and cheetahs and is more pronounced than seen in living lions and tigers. The hindlimbs, adapted for forward thrust with a parasagittal alignment, powered the engine of locomotion. A proportional increase in stride length, together with limb elements emphasizing efficient fore–aft motion, suggests a capability for stamina and even speed in pursuit of prey.

Craniodental morphology of B. robustum includes a canid-like dentition with welldeveloped canines, premolars, carnassials, and crushing post-carnassial molars. The wear pattern of these teeth gives evidence of processing hard food materials while retaining a capability for carnassial shearing during the early and prime periods of the carnivore’s lifespan. In old age the crushing function predominated, based on the tendency for molars and carnassials to be worn to blunted, flat subhorizontal surfaces. The form of the mandibular symphysis suggests a degree of mobility during mastication, with this flexibility adjusting the jaws for processing durophagous materials such as bone.

Borocyon robustum , as the terminal species of the lineage, emerges as a mosaic of craniodental and postcranial traits not evident in any other extinct or living large carnivoran. This predator evolved skeletal features that precede similar parallel adaptations seen today in several of the large canid and felid lines. Yet the total array of skeletal features occurring in B. robustum represents its own unique adaptive solution to the environments of the North American early Miocene. Among the large predatory Carnivora of the Oligocene and early Miocene, Borocyon robustum and the temnocyonine amphicyonids are the closest anatomical parallels to the longer-limbed pursuit predators of the Pleistocene and Recent.

The species of Borocyon frequented semiarid terrain of New Mexico and the Great Plains in the early Miocene and it seems probable that the lengthened limbs and anatomical adaptations for an economy of gait evolved in response to the drying climate and more open terrain of the North American midcontinent in the late Arikareean– early Hemingfordian interval. Its size and efficiency as a dominant member of its predator guild surely contributed to its subsequent appearance in the Pacific Northwest and the Gulf Coast.

Figures

Fig. 7. Comparison of dentitions of Borocyon robustum (A, B) and B. neomexicanus (C–E), end member species of the Borocyon lineage. A, UNSM 25547, P2, P4–M2, alveoli for C, P1, P3; B, UNSM 25684, c, p1–p4, m1–m3; C, F:AM 49241, juvenile, p2, p4–m3; D, F:AM 49239, P1–P4, M1–M3; E, F:AM 49239, c, p1–p4, m1, m3, alveoli of m2.

Fig. 8. Occlusal views of p4–m3 of the end-member species of the Borocyon lineage. A, B. robustum, UNSM 25552; B, B. neomexicanus, F:AM 49241. The squared posterior border of p4 is characteristic of species of the subgenera Borocyon and Daphoenodon.

Fig. 12. Holotype cranium of Borocyon neomexicanus (F:AM 49239), Standing Rock Quarry, Zia Sand Fm., Sandoval Co., New Mexico. A, Dorsal view; B, ventral view. This is the only known cranium and is an old adult with worn teeth: P1–P4, M1–M3. Although reduced in size, an M3 persists in this species.

Fig. 13. Basicranium of Borocyon neomexicanus (F:AM 49239), Standing Rock Quarry, Zia Sand Fm., Sandoval Co., New Mexico. A, Right auditory region; B, left auditory region. Despite crushing, the auditory bulla and surrounding basicranium are anatomically similar to the bulla and basicranium of Daphoenodon superbus. The bulla of the Standing Rock amphicyonid is a thin-walled, flask-shaped, single chamber, as is the rudimentary auditory bulla of D. superbus.

Fig. 14. Elongate radii of (A) female and (B) male Borocyon neomexicanus, Standing Rock Quarry, Zia Sand Fm., Sandoval Co., New Mexico. Note large distal radial exostosis (arrow) of the male, a trait also present in Daphoenus vetus and Daphoenodon superbus but not developed in B. robustum despite a sample of eight radii from the Hemingford and Bridgeport Quarries. The single radius of Borocyon cf. B. robustum from Florida retains a small exostosis.

Fig. 15. Elongate metacarpals 4 and 5 of (A) Borocyon neomexicanus (F:AM 68242) from Standing Rock Quarry and (B) B. robustum (F:AM 68254) from Blick Quarry, Sandoval Co., New Mexico, demonstrating that the Borocyon lineage persisted in the southwestern United States from latest Arikareean into the early Hemingfordian. The only Hemingfordian record of Borocyon in the Southwest are these two metacarpals from Blick Quarry.

Fig. 16. Upper dentition and palate of Daphoenodon superbus (A) and Borocyon neomexicanus (B). Although the teeth of B. neomexicanus are worn, both species show similar unreduced premolar form, reduction of M3, and shearing P4. The New Mexican beardog has developed the ‘‘folded’’ M2, a synapomorphy of Borocyon species, not yet evident in D. superbus.

Fig. 17. Comparison of (A) M1 and (B) M2 dimensions of species of the daphoenine subgenera Borocyon and Daphoenodon and the amphicyonine Amphicyon galushai from the early Miocene of North America.

Fig. 18. Comparison of (A) m1 dimensions and (B) m1 length relative to m2 length for species of the daphoenine subgenera Borocyon and Daphoenodon from the early Miocene of North America.

Fig. 25. Comparison of the carpus (plantar view) of Borocyon robustum and those of Panthera leo, Canis lupus, and Ursus arctos. See text for discussion. Gray tone indicates extent of articular surfaces for scapholunar on magnum and unciform; in ursid and B. robustum the scapholunar moves largely unimpeded over these surfaces during intracarpal flexion but in lion and wolf there is a bony stop (3) between unciform and scapholunar that arrests this movement (this stop is even more developed in the cheetah). Abbreviations as in figure 24: vp, volar process.

Fig. 37. Hypothetical restoration of the skeleton of Borocyon robustum (A) compared to the holotype skeleton of Daphoenodon superbus (B, from Peterson, 1910), representing the two end-member species of the Borocyon lineage during the early Miocene in North America. Fossils show that Borocyon robustum, the terminal species of the subgenus, was widely distributed from the Pacific Northwest to the Gulf Coast of Florida by the end of the early Hemingfordian.