Novltates PUBLISHED by the AMERICAN MUSEUM of NATURAL HISTORY CENTRAL PARK WEST at 79TH STREET, NEW YORK, N.Y

AMERICAN MUSEUM Novltates PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, N.Y. 10024 Number 3146, 37 pp., 9 figures, 2 tables August 15, 1995

Phylogeny of the Caninae (Carnivora: Canidae): the Living Taxa

TEDFORD,I BERYL E. TAYLOR,2 XIAOMING WANG3

ABSTRACT Fifty-seven characters of the skull, mandible, Canini clade (with Fennecus) on chromosome dentition, and postcranium distributed among 122 morphology or in a basal unresolved multicho- character states obtained from specimens repre- tomy with other canines on allozyme evidence. senting 15 living genera of the canid subfamily More contentious is the position ofthe Asian rac- Caninae (67% ofwhich are monotypic) were sub- coon-dog Nyctereutes, placed as a sister taxon of jected to cladistic analysis assisted by a maximum- the South American crab-eating "fox" Cerdocyon parsimony computer program (HENNIG86). The in our analysis but allied with Vulpes on karyo- program found a single tree, 90 steps in length logical evidence or a part ofthe basal canine mul- with a consistency index of65 and retention index tichotomy with regard to the allozyme results. The of78. The reconstruction delineates two sister taxa: South American bush-dog Speothos, a hypercar- the foxlike tribe Vulpini, and the wolflike and South nivore, is placed on osteological grounds in a clade American taxa, tribe Canini. This division is also with the rest of the South American genera in supported by karyological and biomolecular stud- agreement with chromosome evidence although ies although the composition ofeach group varies allozymes relate it to the Canis group. Despite with the evidence used. The osteological evidence these individual cases there is reasonable concor- leads to a more fully resolved relationship than dance in the conclusions drawn from the three presently available from other systems. Problem lines ofphyletic inference. Previous neontological taxa include the foxes Urocyon and Otocyon, con- and paleontological studies of canines have not sidered sister taxa and members of the Vulpini clarified relationships within this group. clade osteologically, but either as members of the

' Curator, Department of Vertebrate Paleontology, American Museum of Natural History. 2 Curator Emeritus, Vertebrate Paleontology, American Museum of Natural History. 3 Research Associate, Vertebrate Paleontology, American Museum of Natural History.

INTRODUCTION Some years ago the first and second authors "zorros" of South America together, al- ofthe present contribution undertook a phy- though they lack a uniting synapomorphy ex- logenetic analysis ofthe living wild canids as cept for those characterizing the South Amer- a step toward an analysis that would include ican canines as a whole. Likewise Canis lacks the fossil record of New World species. Al- a synapomorphy but contains a suite of spe- though that work is in advanced stages of cies that are closely similar morphologically completion, the present authors thought that and lack hypercarnivorous specializations the analysis ofthe living taxa was ofsufficient seen in its sister group Cuon + Lycaon. Such interest that its results should not be delayed conventions may be perceived as weaknesses further. Interest in the phylogeny of living in the analysis as they affect 61 percent ofthe canids has intensified in recent years, es- species on which the analysis is based. Nev- pecially as molecular techniques of analysis ertheless they are not a priori groupings as have become available (Wayne et al., 1989). we believe there is sufficient morphological In previous studies, beginning with a pre- coherence among the species in such groups liminary account by Tedford in 1976, two of to justify treating them as units in the anal- us have explored the basis for the monophyly ysis. Some karyological and biochemical ev- of the family Canidae (Wang and Tedford, idence, discussed below, also confirms their 1994) and the larger phylogenetic framework relationship. (Wang and Tedford, in press) which substan- In the cladistic analysis of the Caninae tiates the recognition ofa three-taxon system which follows, we have employed dental, cra- as characteristic of canid phylogeny (fig. 1). nial, and postcranial characters. Some ofthese The present work focuses on analysis of the characters were resolved into several char- Caninae using the other clades (Hespero- acters states and some of the latter seemed cyoninae and Borophaginae) as outgroups. to form morphoclinal series. We used a total Species ofthe paraphyletic fossil canine Lep- of 57 characters representing 122 character tocyon, stem group for the Caninae, serve as states in our analysis. The emphasis on dental a third outgroup in this analysis. characters reflects the relative uniformity of Another convention is adopted in pre- the skull and skeleton within the Caninae and senting our analysis concerning the level of the practical necessity of eventually bringing the operational taxonomic units employed. the evidence from fossil and living taxa into Data-gathering was at the specific level (see a common framework. Table 1 lists the char- appendix 1); our analysis is at the generic acters used in this analysis, numbered in the level. This is justified because nearly 70 per- order in which they appear at branching points cent of living genera are monotypic and the on the cladogram (fig. 2), which is constructed remainder (Vulpes, Urocyon, "Pseudalopex," from the character matrix in table 2. The 57 and Canis) are not markedly heteromorphic. numbered characters used in the analysis are Vulpes zerda and V. lagopus, are commonly treated by osteological regions to facilitate separated as monotypic subgenera but they comparison. are contained within the range ofinterspecific Institutional abbreviations used are: variation shown by this diverse genus. They AMNH(M), American Museum of Natural also possess a defining synapomorphy (short History, Department ofMammalogy; ANSP, nasals, character 19, table 1) that groups them. Academy of Natural Science, Philadelphia; Furthermore, mitochondrial DNA analyses F:AM, Frick Collection, American Museum (Geffen et al., 1992) confirm the monophyly of Natural History; UCMP, University of of Vulpes species including the fennec and California, Museum of Paleontology. arctic fox. "Pseudalopex" species (e.g., "P." gymnocercus, "P." griseus, "P." sechurae) are a group not consistently differentiated from ACKNOWLEDGMENTS either Pseudalopex culpaeus or, more com- We thank the Department of Mammalogy monly, from Dusicyon australis in previous ofthe American Museum ofNatural History analyses. We group these small paraphyletic for access to the specimens used as the basis 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE 3

for this study. Two reviewers, Dr. Robert CYNOIDEA ARCTOIDEA Hoffmann and Dr. Annalisa Berta materially Caninae assisted the preparation ofthis work through Ma 0 Borophaginae their insightful comments and criticisms. Mr. Raymond Gooris skillfully prepared most of 10 the figures. Hesperocyoninae CHARACTER ANALYSIS 20- The synapomorphies discovered in comparative study of the Caninae are examined by osteological region in an effort to examine 30- their possible functional significance and the B morphoclinal relationships between charac- co 40 0.. ter states. We comment on cases ofapparent 0 A parallelism implied by the most parsimonious cladogram consonant with all these data. In the following paragraphs boldface num- 50 T bers in parentheses refer to the characters/ states (assuming the derived state "1" for the character numbers alone) as listed in table 1 60 and to their postulated occurrence on the Fig. 1. Cladogram indicating the phyletic rela- cladogram (fig. 2). This provides a useful cross tionships ofthe major divisions ofthe Canidae and reference for this discussion and related ma- their temporal ranges. Relationships with other Ca- terial. niformia as developed in Wang and Tedford (1994: SKULL: The gross morphology ofthe canid fig. 9). Letters identify branch points in fig. 2. skull is stereotyped and primitive on the whole when compared to other Carnivora. Most canids have mesocephalic skulls where In the majority ofCaninae the nasal bones the antorbital part of the skull constitutes (19) reach posteriorly to or beyond the most about 40 percent of the total skull length. posterior extent of the maxillary-frontal su- Although domestic breeds of Canis lupus ture. Short nasals, ending in front ofthe max- familiaris show the extremes of antorbital illary-frontal suture, are a synapomorphy skull length possible while still maintaining found in all species of Vulpes (fig. 3A). This a functional system, wild canid species rarely character occurs only in Atelocynus (fig. 3F) show such adaptations under natural selec- and Speothos among the Canini where, along tion. Lengthening ofthe antorbital portion of with their very small frontal sinuses, it con- the skull results in a narrow muzzle, and an stitutes a synapomorphy phyletically linking increase in the diastemata between the pre- these genera. molars. This is an autapomorphy known only Huxley's (1880) study of the frontal sinus in three living species, a vulpine (Vulpesfer- (32) in living Canidae and his selection of it rilata), and two canines (Chrysocyon brachy- as the key character in subdivision of that urus and Canis simensis). Shortening of the family has been widely criticized (most com- antorbital part of the skull is part of a func- pletely by Matthew, 1924) mainly on the tional complex including widening ofthe pal- grounds that its presence and volume is ba- ate (41) (especially the anterior region which sically linked to increasing body size. Huxley may range up to 20% of skull length), pre- anticipated this criticism and tried to dem- molar crowding, and loss ofposterior molars onstrate that such a correlation did not ex- (51). This complex has characterized the most actly hold. We have reinvestigated this char- hypercarnivorous ofliving canines (Speothos acter in some depth in living and especially venaticus, Cuon alpinus, and Lycaon pictus) fossil Canidae and confirm that enlargement in addition to some Neotropical fossil species ofthe frontal sinus (sensu Miller et al., 1964) (Berta, 1988). occurs as a derived state in all subfamilies 4 AMERICAN MUSEUM NOVITATES NO. 3146

Caninae

Vulpini Canini

o0)0~~~~~~~q

~ ~ ~ ~ ~ ~ 0

0 N z -N0

C)03-00 I 04 0U)

0~ C14 04cl)

o10 U)~~~~~~~~~7U N~~ ~ ~ ~ ~ ~ U') co ~ ~ ~ ~ co It~~o NN~~~~~~~U

r-~~~~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~

N4 -~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I

C-4~ ~ ~

0~~~

Li)~~ ~ ~ ~~~

Fig. 2. Cladogram of the Canidae showing phylogenetic relationships of genera within the subfamily Caninae. Characters considered synapomorphies and autapomorphies are indicated at branching points and terminations by number (left) and state (right) in reference to the list in table 1. Characters indicated by closed bars are synapomorphies or autapomorphies. Those indicated by open bars are found at two or three points on the cladogram and are considered to represent parallelisms or reversals (state 0). Lettered branching points represent important cladogenetic events discussed in the text. 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE within the Canidae and is not related to body Primitively the form of the supraoccipital size. Figure 3 illustrates the development of bones in canids are rectangular or fan-shaped the sinus in living Caninae. The primitive when the skull is viewed from the rear and state (lack of the sinus) is manifested exter- the inion usually fails to overhang the occip- nally by a shallow depression or groove creas- ital condyles to any degree in lateral view (fig. ing the dorsal surface of the postorbital pro- 4A-D). In Canis and sister taxa the supraoc- cess. The primitive condition is common to cipital shield is triangular in shape, the inion all the living foxes (tribe Vulpini, fig. 3A, B). is usually pointed and consistently overhangs Enlargement of the frontal sinus that occurs the occipital condyles [(55), fig. 4E, F]. The within members of the tribe Canini includes posterior production of the inion and estab- lateral expansion into the base of the post- lishment of the triangular form of the su- orbital process, inflation of this process and praoccipital shield is correlated with the rel- posterior and ventral expansion of the sinus ative size of the interparietal part of the sag- along the anterior surface of the braincase ittal crest. This feature is thus not simply [32(1), fig. 3A, B]. Inflation ofthe base ofthe correlated with large size since the smaller postorbital process usually removes the de- jackal species show this condition and they pression on its dorsal surface except in Chry- clearly overlap in size the Pseudalopex spe- socyon. In the more derived state [32(2), fig. cies which have the primitive form of the 3G, H], shown by Canis and related genera, supraoccipital shield (compare fig. 4C and E). the sinus penetrates the postorbital process These differences were recognized by Pocock to its tip and the sinus is expanded posteriorly (1913) in his comparison of Dusicyon aus- along the anterodorsal surface of the brain- tralis and Canis latrans. This character shows case ultimately to the frontoparietal suture. no homoplasy in the Caninae and hence ap- This inflation of the frontal region gives a pears to be ofsignificance in phyletic analysis. markedly convex outline to the dorsal surface In most canids the palatine bones extend ofthe cranium above the orbits and it widens to orjust anterior to the end ofthe tooth row. the postorbital constriction. In some mem- Two exceptions are the South American ca- bers of the Canini, including Nyctereutes, nines and the closely related Nyctereutes Atelocynus, and Speothos, the frontal sinus is which have long palatines that extend the small and little expanded (fig. 3F). This is palate to and behind the tooth row (26). In indicated by the lack of a depression on the Cerdocyon, however, the cladogram suggests dorsal surface of the postorbital process de- that the palatine is secondarily shortened. The spite the superficially foxlike appearance of palatine ofChrysocyon is also short, probably the process. Hence this feature is not scored owing to the relatively elongated upper tooth- as a reversal in table 2. Cerdocyon, sistergroup row, as seen in other dolichocephalic taxa of Nyctereutes, retains a very large and lat- such as Canis simensis and Vulpes ferrilata. erally inflated sinus (fig. 3D), which, as in The highly autapomorphous Otocyon, alone other South American Canini, does not pen- among the vulpines, has a posteriorly ex- etrate the relatively small postorbital process. tended palate. Otocyon is largely an insectiv- The sinus in Cerdocyon is not markedly ex- orous canid and hence shares that feature with panded posteriorly nor is it markedly inflated other mammals in which termites are an im- dorsally, but the cavity is extended lateral to portant food (Ewer, 1973). the ethmoturbinates. In lateral view, canids primitively have a In the hesperocyonines and borophagines, moderately arched or nearly flat zygoma. In a postparietal foramen (11) is primitively dorsal view the jugal flares markedly from present on the dorsolateral aspect ofthe tem- the side ofthe skull; the orbital border ofthis poral region at the parietal/supraoccipital su- bone is everted and shelflike. The squamosal ture. In canines except Leptocyon this suture process of the zygoma is relatively deep (corresponding to the tentorium on the inner throughout its length. In the Caninae the zy- side ofthe brain cavity) is pushed posteriorly goma is rather strongly arched dorsoventrally due to enlargement of the cerebrum and the (4), the jugal lacks the degree of lateral flare postparietal foramen becomes extremely re- from the side ofthe skull, and its orbital bor- duced or lost. der is not everted (35) as in the Vulpini. In 6 AMERICAN MUSEUM NOVITATES NO. 3146

It; 1) q

./1

Urocyon cinereoorgenteus Lycalopex vetulus

Vu/pes vulpes IOmm il,

r"' If

Cerdocyon thous Ate/ocynus microtis

Pseudolopex cu/poeus Fig. 3. Comparative dorsal views of the skulls of living canid species sectioned to show the extent and form of the frontal sinus (character 32, table 1); frontal turbinates stippled, frontal sinus shaded. A, Vulpes vulpes, AMNH(M) 60C. B, Urocyon cinereoargenteus, AMNH(M) 536 (partly restored after AMNH(M) 121276) which shows the primitive lack of a frontal sinus common to the Vulpini; C, Lycalopex vetulus, AMNH(M) 391, a form similar in size to the vulpines in A and B, showing a large frontal sinus that does not penetrate the postorbital processes (character state 32(1), table 1); D, Cerdocyon thous, AMNH(M) 36503, and E, Pseudalopex culpaeus, AMNH(M) 66737, also show character state 32(1) typical of most South American Canini; F, Atelocynus microtis, AMNH(M) 100095, shows prim- 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE

V 10~~~~~~lmm

G ~~~~~~~H

Conis lofrons

Conis lupus itively small frontal sinuses (character state 32(1), table 1) confined to the lateral walls of braincase. G, Canis latrans, F:AM uncataloged, and H, Canis lupus, F:AM uncataloged, showing the large frontal sinus especially its posterior extension to the frontal-parietal suture in C. lupus (character state 32(2), table 1). the South American canine clade, as first not- caudal and small rostral elements); a partial ed by Berta (1988: 26), the long scar for the septum formed from the entotympanic; and origination ofthe medial masseter muscle on an internal carotid artery that lies in a groove the ventrolateral surface of the zygoma be- on the medial side of the caudal entotym- comes uniquely enlarged (36), particularly in panic (see Hunt, 1974; Wang and Tedford, the anterior segment. In these taxa the inser- 1994 for further discussion and figures). The tion of the medial masseter is correspond- auditory region undergoes little change dur- ingly enlarged, forming a triangular scar being the history of the Canidae except for the low the masseteric fossa and anterolateral to development ofthe external auditory meatus the angular process of the mandible. which is produced laterally as a short tube Characters in the auditory region that de- from the porus acousticus, especially in the fine the Canidae include: large entotympanic living genera of vulpines and canines. Only that ossifies from at least two centers (large some South American canines and the allied 8 AMERICAN MUSEUM NOVITATES NO. 3146

A \4

Pseudolopex 31(1) Fig. 4. Comparative views of the left upper incisors and canine (anterior view above lateral view) and the left side of the braincase (lateral view to the left of posterior view, mastoid process stippled, paroccipital process accentuated to show several features used in phyletic analysis (characters states numbered to correspond to table 1). A, Leptocyon vafer, incisors, F:AM 27412, cranium, UCMP 77703, show the primitive condition in incisor and canine form and relative size plus supraoccipital, mastoid and paroccipital form and relative size; B, Vulpes vulpes, AMNH(M) 29055, I3 lacks median cuspule [character state 17(1)], paroccipital process broad, closely appressed to bulla, (character 18); C, Pseu- dalopex culpaeus, AMNH(M) 66737, showing large mastoid process (character 33) among character states common to Canini in this member ofthe South American clade; D, Cerdocyon thous, AMNH(M) 30628, 12-3 separated by diastema (character state 23), upper canine short and slender (character 27), primitive narrow and salient paroccipital process and derived large mastoid (character 33) in this small canine; E, Canis aureus, AMNH(M) 88712, 13 enlarged with posterior cingulum (character state 53). 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE 9

Conis

Lycoon lm ~27 3 1(2) Supraoccipital shield triangular in shape, inion overhangs condyles (character 55); F, Lycaon pictus, AMNH(M) 164162, shares characters with Canis. All drawings to a common size for ease ofcomparison, specimens selected to conform to a median size to minimize allometric relationships imposed by individuals of markedly differing relative size.

Nyctereutes show a significant departure in process in canids is closely associated with this general trend by possessing a short me- the posterodorsal surface of the large bulla atus surrounding a relatively small diameter with the process usually having a consider- porus. able portion ofits length free from the bulla. Primitively the base of the paroccipital This free tip is directed posteriorly in prim- 10 AMERICAN MUSEUM NOVITATES NO. 3146

Urocyon cinereoorgenteus Otocyon mego/otis

Nyctereufes procyonoides

Atelocynus microtis Speothos venoticus Fig. 5. Comparative internal views of the left ascending ramus of the mandible of representative canids to show the subangular lobe (character 24, table 1), form ofthe angular process, and distribution of rami of the median pterygoid muscle. Muscular insertions represented by roughened areas on the angular process, symbolized in stippled pattern. Inf., fossa for inferior ramus, median pterygoid; Mass., process for pars reflexa external masseter; Sup., fossa for superior ramus, median pterygoid (character states 38(l)-(2), table 1), and the form of the coronoid process (character 37, table 1). A, Vulpes vulpes, AMNH(M) 185649, primitive state of characters 24, 37 and 38; B, Otocyon megalotis, AMNH(M) 161153, showing subangular lobe but primitive angular process ("Type A" ofGaspard, 1964); C, Urocyon cinereoargenteus, AMNH(M) 4270, showing subangular lobe and "Type A" angular process with expanded base; D, Pseudalopex culpaeus, AMNH(M) 67088, derived angular process ("Type B" ofGaspard 1964) and coronoid process; E, Cerdocyon thous, AMNH(M) 14627, derived angular process ("Type D" ofGaspard, 1964) and low subangular lobe; F, Nyctereutes procyonoides, AMNH(M) 57113, derived angular process ("Type D" of Gaspard, 1964) and large subangular lobe, and low coronoid; G, Canis aureus, AMNH(M) 81041, derived angular process ("Type C" of Gaspard, 1964) with primitive short coronoid relative to height; H, Speothos venaticus, AMNH(M) 98560, and I, Atelocynus microtis, AMNH(M) 76579 show the derived angular process ("Type D" of Gaspard, 1964) and long coronoid. 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE I1I itive hesperocyonines and borophagines, and icate attenuated process in the vulpines in turned ventrally in canines (10), a feature also being rounded in form and more robust even independently acquired in some hespero- though the species of all of these genera are cyonines and borophagines. Later canines closely similar in size. Accompanying the de- usually show increasing union between the velopment of the subangular lobe is the an- bulla and paroccipital process, especially in teriorly converging superior and inferior forms in which the bulla is relatively inflated, margins ofthe horizontal ramus ofthe man- with the base of process hollowed out for dible. reception of the bulla, and the free tip con- The Canidae primitively show an angular sequently shorter in length (fig. 4). A further process of the mandible which is expanded modification is seen in all the Vulpini in which medially to form a broad shelf for the inser- the base of the process is laterally expanded tion of the branches of the median pterygoid to cover most of the posterior surface of the muscle. This shelf separates the insertion of bulla. The free tip ofthe process is very short the superior branch from the smaller fossa and turned laterally, not extending markedly for the inferior branch which lies on the ven- below the body of the process contacting the tral surface ofthe shelf adjacent to the inser- bulla [(18), fig. 4B]. A broad paroccipital pro- tion of the pars reflexa of the external mas- cess is also found in two species of jackals, seter. In the primitive Caninae a shelf for Canis adustus and C. simensis, but the free insertion of the superior branch of the me- tip is longer and directed ventrally as in other dian pterygoid is lacking and the whole pro- Canini. According to this analysis the state cess is slender, attenuated and often termi- of this character in the Vulpini appears to be nates in a dorsal hook. a unique feature separating the foxes and their A conspicuous and unique expansion of allies from all other Caninae. A posteriorly the angular process is characteristic of the expanded paroccipital process [31(1)] tends South American Canini and the allied Nyc- to characterize the Canini; such an expansion tereutes, excluding the stem taxon "Pseudal- is carried to an extreme condition to produce opex" ("P." gymnocercus, etc.). In this group a longitudinal plate [31(2)] in the hypercar- the primitive slender angular process with its nivorous Speothos and Lycaon. dorsal hook and small pterygoid fossae ("Type The Canini show enlargement of the mas- A" of Gaspard, 1964: fig. 24; fig. 5A-C of toid process (33) that is evident even in small this paper) is modified by dorsoventral ex- species similar in size to those in the Vulpini pansion, reduction of the dorsal hook, and (fig. 4C, D). Important axial muscles that ex- conspicuous enlargement of the fossa for the tend and control the lateral movement ofthe inferior ramus of the medial pterygoid mus- head insert on the mastoid process. The en- cle [38(1), "Type B" of Gaspard, 1964: fig. largement of this insertion may imply cor- 24; fig. SD of this paper]. Associated with respondingly more powerful movement ofthe these features is a coronoid process that is head. anteroposteriorly long at its base relative to MANDIBLE: Most canids have relatively its height (37) rather than the shorter based deep and thick horizontal rami of the lower and high coronoid found widely distributed jaw, but the Caninae are typified by possess- within the Canidae (compare figs. 5D and G). ing shallow and thin jaws (5) that support the In the Cerdocyon, Nyctereutes, Atelocynus, extended and relatively delicate premolar se- Speothos clade, both fossae for the medial ries. In a few taxa of the Caninae there are of pterygoid muscle are expanded on the inter- unique enlargements the digastric inser- nal tion on the horizontal ramus into a distinct side of the large angular process [38(2), process called the "subangular lobe" (24) by "Type D" of Gaspard 1964: fig. 24; fig. 5E, Huxley (1880: 251). This is markedly devel- F, H, I ofthis paper]. A large angular process oped in vulpines with complex molars, spe- in which the fossa for the superior branch of cifically Urocyon and Otocyon (fig. SB, C). A the medial pterygoid is expanded [(54), "Type subangular lobe is also present among Canini C" ofGaspard 1964: fig. 24] characterizes the in Cerdocyon (fig. SE), and is especially genus Canis and its sister taxa (fig. 5G). This marked in Nyctereutes (fig. SF). The angular type of process seems merely a modification process in these canines differs from the del- of the primitive "Type A", and therefore is 12 AMERICAN MUSEUM NOVITATES NO. 3146 coded as a separate character (54) instead of it is in Otocyon. In Canis and sister taxa the one state of character 38. 13 is enlarged [(53), fig. 4E, F], the postero- The above references to Gaspard (1964) medial cingulum is enlarged, and a sharp crest refer to her recognition of four functional appears on the medial side of the tooth that groups of mandibular structure in the living passes from the crown apex to merge with canids, based on the pattern of insertion of the medial cingulum. In addition, a shelflike the two rami ofthe medial pterygoid and the posterior cingulum is developed on 11-2 pars reflexa of the external masseter. We il- which includes the bases of the enlarged lat- lustrate these four types from actual speci- eral and medial cuspules to form a yokelike mens (fig. 5) and contend that the patterns crest encircling the posterior part of the in- have phylogenetic significance in that canines cisor crown. These modifications ofthe upper with "Type A" angular processes are rela- incisive battery are mirrored to a lesser extent tively plesiomorphic, while types B through by the lowers and together they represent a D have more limited distribution within the functional complex paralleled by some ofthe Caninae as discussed above. Borophaginae. CANINES AND INCIsORS: Although in the PREMoLARS: The Caninae are character- Canidae sexual dimorphism is manifested in ized by the presence of narrow and elongate relative canine size there are some clades premolars (6) with weak or absent posterior characterized by relatively small canines in cusplets on P3 and p2-3 (7) set in a shallow both sexes. Canines of small diameter and ramus (5) and separated by short diastemata short crowns (27), are present in Cerdocyon [(8), fig. 6]. The primitive condition, retained and Nyctereutes (fig. 4D) on the one hand and by the Borophaginae (fig. 6), includes broad Cuon and Lycaon (fig. 4F) on the other. Rel- and short premolars with strong cusplets atively small canines are also possessed by forming a nearly closed tooth row and a rel- the Vulpini that have mandibles bearing the atively deep jaw. In some Caninae the pre- subangular lobe (Urocyon and Otocyon). This molar row is further modified by changes in condition is thus associated with dental and the form ofthe premolar teeth. They become mandibular features indicating both greater short and relatively high-crowned [(20), fig. omnivory and carnivory. 7B, C] and the p2 is isolated by longer dia- Primitively the incisors of canids show a stemata (21) in some ofthe Vulpini (Urocyon gradual increase in size from I1 to I3 but the and Otocyon). Anterior cusplets derived from 13 is usually enlarged disproportionally in the cingulum on the premolars are primi- most forms. The Caninae show even greater tively retained in some Caninae (Leptocyon) enlargement of the 13 relative to 11-2. The but the vulpines and all canines lack these I1-2 are primitively flattened conical teeth stylar cuspules. In the Canis-group in general, with medial and lateral cusplets, without pos- a second posterior cusplet on p4 [(34), fig. terobasal cingula. The 13 is primitively a high 7G-I] is present. In the canine genera Lycaon cone with a medial cusplet (fig. 4A) and a and Cuon, the p4 has an anterior cusplet [(22), posteromedial cingular cuspule. The upper fig. 7H, I] and in Lycaon, the p2-3 have strong incisors in the Caninae are simplified, the anterior cusplets (57). These cusplets have medial cusplet on I3 is absent [17(1), fig. 4B] differentiated from the anterior crest of the and in the Urocyon-Otocyon clade the cus- premolars rather than from the cingulum. plets on I 1-2 are very weak or absent [17(2)]. They are not homologous with the stylar cus- These teeth maintain an apically interlocking plets possessed by other canids. occlusion with the lower incisors. In Otocyon CARNAssIAs: In canids the carnassials have an unusual short diastema occurs between 12 changed very little from the primitive form and I3 (23), that arises because the 11-2 have common to most Caniformia. The upper car- a more anterior position than the 13. An ap- nassial in the Caninae is more elongate and proach to this condition is also present as narrower across the paracone (9) than in the individual variation in Nyctereutes and Cer- Borophaginae where the primitive robust docyon (fig. 4D) but the diastemata are short- form is retained and accentuated by increas- er so that the crown of 12 is still not com- ing the width to length ratio. The protocone pletely seen in front of 13 in lateral view as is relatively reduced in the Caninae; the an- 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE

Hesperocyon gregarius D I '

B :- 10 mm

2(1)

lOmm

K >

- ~~~~~~~~~~~~8 N/ Leptocyon gregorii

L 2(1)

6/ lOmm Fig. 6. Comparative views of the cheek teeth of the three major groups within the Canidae. A-E, Hesperocyoninae: Hesperocyon gregarius (F:AM 76177, Orellan, early Oligocene); A, lateral and B, occlusal views ofthe left upper cheek teeth; C, lateral, D, medial and E, occlusal views of the left lower cheek teeth. F-J, Borophaginae: Cormocyon leptodus (F:AM 49045, Arikareean, medial Oligocene); F, lateral and G, occlusal views of the left upper cheek teeth; H, lateral, I, medial and J, occlusal views of the left lower cheek teeth. K-O, Caninae: Leptocyon gregorii (F:AM 49063, Arikareean medial Oligocene); K, lateral and L, occlusal views ofthe left upper cheek teeth; M, lateral, N, medial and 0, occlusal views ofthe left lower cheek teeth. Drawn to the same size as dentitions in fig. 7. Numbers refer to characters listed in table 1 and indicate features that distinguish these very similar forms. 14 AMERICAN MUSEUM NOVITATES NO. 3146

terior cingulum is correspondingly reduced, entoconid is the enlargement ofthe posterior often leaving a strong notch between the an- cingulum (or hypoconulid shelf) which often terior base of the paracone and protocone. encloses a small posterior basin behind the Only in Otocyon among the Caninae do we talonid cusps and may bear a hypoconulid see molarization of the P4, involving expan- (13). A hypoconulid shelf appears as an in- sion of the buccal cingulum and the proto- dividual variation in Leptocyon species, but cone, and reduction in size of the metacone it otherwise occurs consistently only the Vul- blade. More common in the Caninae is a con- pini and Canini. Among the more hypocar- dition in which the upper carnassials are rel- nivorous vulpines the talonid is conspicu- atively small as compared to the size of the ously wider than the trigonid (fig. 7B, C) and molars that succeed them. Huxley (1880: 248) a protostylid may occur on the trigonid. A recognized this relationship but measured it protostylid (29) also occurs on the carnassials with reference to the basicranial axis and ap- of Urocyon (fig. 7B) and some South Amer- plied the term "microdont" to the condition ican canines (fig. 7D), sometimes accompa- in which the P4 was proportionally small nied by a mesoconid on the cristid obliqua. (25%) relative to this axis. We have found it Two independent adaptations for hyper- more practical, especially in working with carnivory are seen in Speothos and in the fossils, to express the relationship ofthe length sister genera Lycaon and Cuon. These groups of P4 to the sum of the lengths of Ml and increase the sectorial function of the carnas- M2. The term "small" here refers to upper sials by reduction or loss of the entoconid carnassials less than 80 percent the combined [(42), (43), fig. 7H, I] and often the metaconid length of the upper molars. Such dentitions, as well. Speothos venaticus has lost the M2 useful at the specific level diagnosis, are high- and m3 (51), but the late Pleistocene S. pa- ly homoplastic in the higher level phylogeny civorous retains a small double-rooted M2 ofthe Canidae where they are associated with (Berta, 1988). Fossil taxa otherwise closely complex molar structure and other features resembling Cuon suggest the path of talonid that may be related to increased omnivory. reduction as having evolved from a taxon The form of the lower carnassial (ml) is with a bicuspid talonid. especially instructive in phyletic analysis. Since the carnassials form a vital function- Certain changes in its structure can be re- al unit in carnivores and canids achieve a solved as steps in morphoclinal trends to- complexity of carnassial structure unique in ward talonid enlargement and complexity many ways, their importance is often em- (hypocarnivory) on the one hand, and talonid phasized in phyletic analyses (e.g., Matthew, reduction and simplification (hypercarnivo- 1924, 1930). However, as discussed above, ry) on the other. Beginning with the bicuspid the functional requirements of hyper- or hy- talonid [2(1)], synapomorphous for the sis- pocarnivory, constrained by the primitive ter-taxa Borophaginae and Caninae, the en- form of the canid carnassial, place limits on toconid is a discrete cusp or crest and a nearly the directions of morphological transforma- complete buccal cingulum is present (fig. 6). tion. Such orienting influences force similar In the Caninae the entoconid enlarges, the solutions to these specialized feeding modes base of which may become large enough to resulting in considerable homoplasy in dental coalesce with that of the hypoconid to block form. the talonid basin [2(2)]; the buccal cingulum UPPER MoLARs: The primitive canid upper is reduced. Most Vulpini have achieved the molar is transversely elongate with a well- latter condition and populations of Vulpes developed labial cingulum bearing a promi- and Urocyon species may include a sizable nent parastyle connected to the paracone by proportion of individuals in which the en- a crista; a lingual cingulum with a hypocone toconid and hypoconid are joined by cristids usually posterolingual to the protocone; and from these cusps to form a transverse crest subequal para- and metacones and small para- [2(3), fig. 7B]. This transverse crest is the final and metaconules borne on well developed step in the morphocline leading to the car- cristae that link the protocone with the prin- nassial talonid structure typical ofliving Can- cipal labial cusps. Changes in this funda- ini. Associated with the increasing size ofthe mental form include: increase in size of the 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE

6 13 14(l) 12 AVA vulDoes 7 _ 2(3) B I -_-r_2~~~~5 Urocyon 2 12

C Otocyon A uI IN~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

29 2(3) D Pseudolopex E Cerdocyon

F Nyctereutes 13 2(3) G, Canis 42 43 34 22\ 1 3

2:~~~~~~~~~~~~~~~~~~d234 Lycoon 51 34 42 43 51

Cuon Fig. 7. Comparative views ofpart ofthe lower dentition (p4 m 1-3) ofrepresentative Caninae showing in sequence from left to right in each row, the lateral view of the left side, medial view of the right side and occlusal view of the left side of the cheek teeth. Numbers refer to characters and character states listed in table 1. A, Vulpes vulpes, AMNH(M) 29055; B, Urocyon cinereoargenteus, AMNH(M) 100224; C, Otocyon megalotis, AMNH(M) 161153; D, Pseudalopex culpaeus, AMNH(M) 66737; E, Cerdocyon thous, AMNH(M) 30628; F, Nyctereutes procyonoides, AMNH(M) 57113; G, Canis aureus, AMNH(M) 88712; H, Lycaon pictus, AMNH(M) 164162; I, Cuon alpinus, AMNH(M) 54842. All drawn to the same scale as Vulpes vulpes, bar equals 10 mm. hypocone [(1), fig. 6]; reduction in size ofthe decrease in transverse diameter so that the conules, especially the paraconule; great re- teeth are very narrow for their length [(30), duction in size of the parastyle [(3), it is usu- as seen in the more hypocarnivorous South ally very weak to absent in living Canini]; American taxa and Nyctereutes]; increase in 16 AMERICAN MUSEUM NOVITATES NO. 3146 height of the labial cusps; disproportionate docyon and Nyctereutes, are also typified by enlargement ofthe paracone until it becomes having relatively short ears. On the other conspicuously larger than the metacone [(45), hand, Chrysocyon has a conspicuously elon- in Canis and its close relatives] often accom- gated limb (48). Cerdocyon, Nyctereutes, and panied by reduction in the buccal cingulum Speothos also have a straight or nearly straight (46). In hypercarnivorous taxa the hypocone caecum, a character shared with Chrysocyon. becomes very reduced (44), even lost; and The distribution ofthis latter character is not finally the M2 is also lost (50). fully determined from the data at hand, but LOWER MoLARs: Primitively the canid m2 its polarity can be established by reference to (fig. 6) possesses a complete buccal cingulum; other Carnivora (Ewer, 1973). In a compar- a small, but discrete, paraconid; a metaconid ative study of the calcanea of canids, Stains equal to or slightly smaller than the proto- (1975) found that the most distinctive taxa conid; and a talonid with a hypoconid and were Nyctereutes, Atelocynus, Speothos, and an entoconid crest. The Caninae have re- Chrysocyon. Of these four genera only Nyc- duced buccal cingula in which only the an- tereutes and Atelocynus showed a clear re- terobuccal segment is retained (fig. 6M, 0). semblance to one another. Speothos was Enlargement of the anterobuccal cingulum unique and Chrysocyon was most similar to (12) is a universal trend among all Caninae Canis. From the fossil record we were able (fig. 7). In certain Vulpini and among the to determine that the metatarsal (MT) I is South American Canini and Nyctereutes the reduced to a proximal rudiment (15) in Can- buccal cingulum is extended posteriorly to inae above the level or organization of Lep- the hypoconid (fig. 7). In addition, Urocyon tocyon (although Vulpes and domestic dogs has a protostylid [(28), fig. 7B] developed from occasionally have a MT I as an atavistic vari- the buccal cingulum at the base of the pro- ant). Lycaon is unique among canids in re- toconid. Enlargement of the metaconid (14) duction of the metacarpal I to a rudiment occurs early in the history of the Caninae at (56). The entepicondylar foramen of the hu- about the same stage as expansion of the merus is lacking (16) in all canines above anterobuccal cingulum. The metaconid may Leptocyon (although this primitive state may be equal to or larger than the protoconid and also occur as a variant in other Caninae). its swollen base gives a decidedly convex outline to the lingual border ofthe tooth (fig. 7). Otocyon is also unique in having both a com- PHYLOGENY RECONSTRUCTION pletely molariform m3 and a small oval m4 Table 1 lists the morphological characters (25) (fig. 7C). discussed above; character numbers and POSTCRANLAL SKELETON AND SoFr PARTs: character state numbers in parentheses cor- Most canids, including all vulpines and some respond to those cited. Character polarities canines, have epipodials that are shorter in are determined by using Hesperocyon and the front limbs than in the hind. This is in- Cormocyon as successive outgroups. Not only dicated by the ratio of the radius to the tibia are these two genera known by nearly com- which is less than 90 percent in those forms. plete cranial, dental, and postcranial mate- In derived canids (Chrysocyon, Canis and al- rials, but they are probably the closest rep- lies), the radius equals or exceeds 90 percent resentatives of the primitive morphotype of ofthe length ofthe tibia (47). Values between their respective subfamilies, Hesperocyoni- 80 and 90 percent are seen in several species nae and Borophaginae (Tedford, 1976; Wang, of Vulpes (e.g., V. vulpes, V. macrotis, V. la- 1994; Wang and Tedford, 1994, in press). gopus) and Otocyon among the vulpines and A single shortest tree is found by nearly all other Canini. The South American HENNIG86 (Version 1.5; Farris, 1988) on Canini are relatively short-legged, and among the 18 taxon character matrix (table 2). All them, Speothos and Atelocynus show excep- characters are nonadditive and unweighted. tionally short limbs (49) with respect to length The tree is calculated using the option of"im- of the spinal column from cervicals to lum- plicit enumeration" (i.e., a search algorithm bars (Hildebrand, 1952: fig. 6; Langguth, designed to find trees of minimal length). 1969). These same genera, along with Cer- Aided by CLADOS (Version 1.2; Nixon, 0o 0 O Y

cn I0 - C0) CU 0 ~C o O 0 , "'

'0 o 1'0 U C 0 *CC)C)U 0) C-~ 0 rI 0 a) , o -; 04 Y ~~ a)7tU CUC 4) ,~ ED~ D 0 ~~~ 4) co cis CU'S _~ # CUa) a)Cu S * 1S*,a ~~~ r -- 3 7a 0 CU CU 0 a) _C I-- __U 0 0 -. _CU _a ~ I--

00 02a)ci '00 CU 1- .> r ._. CU CU_ 0 0 0 a) 1C 0 - 0 0 I- CU a) a 0 CU 0 0 ~ 0 ..:j1 CU o a) 04 ~ 8 Q co I - ,- 0 I- 0 _ C4 0 H Cd 'C U 0 0 0 o CC.) ',o Q2 r02 04 4- v CU _ '0 OaO CU .C " ° 2 *,lU __ 00 AC0 33CU4 _, 2 0o -4 O cU 4-'wDA 2 o oco _____0 0 06.9 CU _ o __ _dCU_ xi 0 ,; ._ _ 0 * 0 Oa) CU 0,-I a).

0) CU O o CU~~~~~~~~~~~

u CIOhi Cd~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~CC4-2S , S !. E 8 a S 0 CU C.)0CUd I- o en o CU 0en Cd Cl''o' , C) 0

CU .0 SC_ E - -4 I t 0 oa o R e E e E 0a)1-0 014 --) CU t 0 CU4 U 0 C c

17 C.) _CU .CU co '0" C)

co 0 o

Cl) '0 '0 co CX .DU C) '0 CU '0 CU NaCU CU CU - wE o. C) '0 C)- CU

c ZA 0~~~ .CX ts _ _ _ _ c4o CUe -'0 '. _C_U 1-4 - O.- . C

2 en C)

CU CU ~~~ ~~~' .- E a - CU0 E OC E 0 '0 - ~~~ ~-

CU6 F- 0ow CU 0 0 ~0 CUCUC.. m CU '-o . .0 Ct)C CU00 S^ - o ., co w9 o

Q 'OC U, o oen C3Uq o 0 o 0 000 0 00 U, C) Q C o C)~

CU COCUC CU 0. C) CUXf Q vo; CU 0.° * E , 0~~~~~~~0 C)I. 0. C. C) U) (UCU C) -C CU_ N_ C4 e .) o,

CU 0 CU' 0. .r .*. .0.e CU CU _N N tCUUC eF0 iO-0 N N NNN^^ e

18 '0 I.- 0 CU 4) 0 O ~C 0 0 0- 0 -4 CU0 CU .r o IOU~~~~~~C '0CU 4) ) o cc c CU s '0 0 0, 4)'0~~~~~ 6 I4) 4- co 'Q *e*r'M^ o coU .- I- 4) Q O00Q O O 0O cO E 0 CU o Y '0eyf, , , Ee 4) "~~~~~~C)0, ~ ~ ~ ~A.CUd.~ r- CU N ' 'e t CU 4 .4 _U4 _ _ _ _ 4) .dC .60 4)C 04 _ __00 4 __ 0, _ _ _ _ _ Cd I.. d _ _ 4.) ,O 0

4.4. 0 ~0 0o CU CU o cd 14) '0 CO s 4~0 *C 0 4)~ O0 IC 40, 0j r o 0 ,0cis ;. 3 -4) 0 0 .E '0C4 0.' 0s 0 w .E CUCUU CU C CUc, CU 0 C od oa CU 000 0 4- .4) 4L) _d CU5 0 4) 0. Cd - - 0 0 - 0 *-" * E.1 CU CU _U _U CU CU CU 0, 0000000000.000000 ,O 0~ 00 0 0C0 0 0

- 0,

C.-

0 0 0 0 0 0. .r .r a ".4 +U cn~ 0 4)& o CIu 'o 5-CU1 *s °~ CU o" o. 40 o

CU4 0. t t I I I , co ^ cU 0 O

5 -Y

0* ~ C .C oy r^ u £ o oR 6"- Ne

19 0- 0

cI-A

CD) 0 0 0 00(D000 0 CD LO 0 000000 0O00C0 00OO00 0 0 0 ~40 0 0000000OO 0O0 0 00~ C,0 0 0 0 00C00000O0O O0 0 O0 a) 0 000 0 00 C-O0 0 0 -4 0O0 OD0 0O0 0 000e-:0~-40 0 00C00 000 0- - (D0(0 0 0 0 00C-0OO- 00O00 )0 000O00 0 0 0 00 - U' 00 000 0 0 0 0 0 0O0 0 0

0 0 0 0 000 4- (0 C0 0 0000 444400000000 0 0 IC) 0 O( 0 0 040 C0

') OD 0 CO0 0000 ' 0 0 t- 0 0 0 0 C0000004000O 000 0 0 0 00OO0-4-4-4-4 - -4000C>:00 0 a 0 00 0 000 00 O0 0 0 0 0 0 -4 -40-00000O0000 co0 0 0 0 -4-0000000O 0 0 0 0 0 0.4 a) 0 0 0 - 00004-4--4- 0- - Ia ) - 0 ~ 000D0 -40 C C 2 -4 - -4 0 0 -4

cO 0 0 14 14 ICD

) 0 0 0 0 0>C

C\j000 0 ~ ~ ~ ~ L.

20 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE

1992), the character distribution ofthis single cingulum of m2, also increased the breadth tree was further explored and plotted as shown of this tooth across the trigonid. in figure 2. The tree statistics (not counting Apparently, the Vulpini developed a dis- the autapomorphies) are: tree length = 90, tinctive paroccipital process early in their his- consistency index = 65, and retention index tory which appears at Node E. The breadth = 78. of the process, cupping the posterior surface Nodes A and B represent the fundamental ofthe bulla, and the very weak development cladistic relationships of the Canidae. The ofthe horn of the process, stand in notewor- characters shown at Node C are those com- thy contrast to those ofall other canids. Vulpes mon to all members ofthe monophyletic sub- is plesiomorphous in most ofits dental char- family Caninae. Primitive states of these acters. However, living species of this genus characters are present in the extinct sister tax- show a high frequency of development of a on, the Borophaginae. Leptocyon includes a transverse crest on ml through the coales- group of species completely plesiomorphous cence ofcristids from the hypoconid and en- with respect to the rest ofthe Caninae. Later toconid, thus dividing the talonid basin into occurring species ofthe genus show some de- larger anterior and smaller posterior sub-ba- rived character states within dental mor- sins. This development seems to represent phoclines that suggest closer phyletic rela- the ultimate level of elaboration of the ml tionship to the Caninae represented by Node talonid, a condition also achieved by most D, such as: lengthening and increasing height Canini (Node G) as well as the Borophaginae. of crown of the premolars, enlargement of A group ofomnivorous foxes, here termed the m 1 entoconid and presence of a weak the Urocyon group, represents a clade (Node hypoconulid shelf, and a short m2 relative to F) in which the dentition is modified for in- ml with enlargement of the metaconid and creasing hypocarnivory. These taxa show anterobuccal cingulum. The suggestion of a characteristically short upper carnassials, specific patristic relationship of Leptocyon quadrate upper molars, and relatively wide species with later vulpines is thus a permis- lower molar talonids. Node F represents the sible conclusion in developing the genealogy vicariance ofa stock characterized by having ofthe phyletic tree. In cladistic analysis, how- simple incisors lacking cusplets, short, tall ever, our knowledge ofLeptocyon is not suf- crowned premolars isolated by diastemata ficient to recognize any ofthe species as sister (especially around p2), and an ml talonid taxa of the more derived species of canines. wider than its trigonid. The New World rep- Thus the genus Leptocyon remains a para- resentative, the living Urocyon, has a highly phyletic group awaiting further analysis. distinctive mandible bearing a subangular Node D represents a fundamental split lobe and high condyle, a transverse crest on within the Caninae into foxlike (tribe Vul- the ml talonid, and relatively small canines. pini) and wolflike (tribe Canini) sister taxa. This same combination of characters occurs The derived characters possessed by both of in the highly apomorphous Otocyon that rep- these groups and presumably uniquely typi- resents the Old World vicar. Otocyon is fying their common ancestor, include only a uniquely distinguished by a diastema be- few, seemingly minor, characters ofthe lower tween 12 and I3, the presence of M3, a sub- molars and the loss of the postparietal fora- molariform P4, and a posteriorly extended men, the hallux and the entepicondylar fo- palate, as well as the mandibular characters ramen ofthe humerus. The dental characters mentioned above. involve increasing molarization of the ml In addition to the presence of a frontal si- and m2 through enlargement of the ml pos- nus, the common ancestor ofall Canini (Node terior cingulum (hypoconulid shelf) and the G), is hypothesized to have modified the variable presence of a hypoconulid closing primitive zygoma. The jugal has lost its lat- the talonid basin posteriorly. Enlargement of eral flare; its lateral surface has come into line the m2 metaconid to a size equal to or ex- with the sides of the muzzle when viewed ceeding the protoconid, thus markedly in- dorsally and the zygomatic process of the creasing the convexity of the lingual border squamosal is reduced in depth anteriorly. The ofthe tooth. Enlargement ofthe anterobuccal angular process of the mandible is also en- 22 AMERICAN MUSEUM NOVITATES NO. 3146 larged in different ways for insertion of the gest-legged canid, Chrysocyon, and the short- increasingly important medial pterygoid est-legged canids Atelocynus and Speothos. muscle. Consistent dental modifications An unresolved trichotomy occurs at Node K: present at this point are the transverse crest the Lycalopex, Chrysocyon, and the Cerdo- between hypoconid and entoconid on ml and cyon-Speothos clade. Except its slight ten- the elimination or great reduction ofthe Ml- dency toward hypocarnivory, Lycalopex ve- 2 parastyles. tulus is perhaps closest to the primitive mor- The sister taxa that emerge at Node G rep- photype of the taxa in Clade K. Chrysocyon, resent on the one hand a mainly Neotropical on the other hand, is highly autapomorphous group ofsmall to medium-sized forms (Node with its very elongate skull, with a short pal- H), and on the other the large wolflike species atine relative to an elongated tooth-row, and (Node 0). Synapomorphies uniting members extremely elongated limbs. Beside such au- of the former group include unique features tapomorphies, Chrysocyon possesses an in- of the ascending ramus, i.e., a wide and low triguing combination ofcharacters that make coronoid process, enlarged muscle attach- its phylogenetic position difficult to assess. ments for the medial masseter on the zygoma The frontal shield still has a depression on and on the anterolateral face of the angular top of the postorbital process even though process, a palate that extends to or behind the frontal sinus may be extended into the the tooth row, post-orbital processes not in- base ofthe process (on one side ofAMNH(M) vaded by the frontal sinus. Also uniting 36962) and its caecum is straight like the oth- members of clade H are a suite of less con- er Canini grouped at Node L. In addition, its sistent dental features (hence not coded in elongated legs also hint at membership with this analysis) particularly in the m2 in which the Canis-Cuon-Lycaon clade. In fact, such the buccal cingulum is extended to the hy- a hypothesis would require only two more poconid, a mesoconid is present, and the par- steps than in the shortest tree of figure 2. acristid forms a strong blade although the The apomorphous sister group signaled by paraconid is lacking or nearly so. Node L demonstrates an interesting trend to- Within the South American canids, species ward frontal sinus reduction, further enlarge- of "Pseudalopex" ("P." gymnocercus, "P." ment of the angular process, the presence of griseus, and "P." sechurae) are left at the base short pinnae (and correspondingly short and ofthe clade due to their plesiomorphous na- small-caliber external auditory meati), and ture. Clade I is slightly more derived owing short legs. In those taxa investigated anatom- to an enlarged angular process bearing a rel- ically, the caecum is straight or nearly so in atively large fossa for the inferior branch of contrast to the convoluted state primitive in the medial pterygoid. Beginning in Pseudal- the Canidae. These taxa can be resolved into opex culpaeus, the more primitive species of two sister groups showing contrasting dental Clade J, the metaconid ofm2 becomes lower adaptations. Hypocarnivorous taxa (Cerdo- than the protoconid, and the P4 protocone is cyon and Nyctereutes at Node M) possess short reduced. This trend continues in Dusicyon diastemata between I2 and I3, a short P4 australis and more derived fossil taxa in which compared to the length of the molars, rela- the P4 protocone is nearly lost and the p3-4 tively small canines, and a mandible with the have developed a posterior tilt (Berta, 1988). subangular lobe also seen in Urocyon and The remaining South American canines Otocyon. Cerdocyon is plesiomorphic in re- plus Nyctereutes (Node K) are weakly held taining a large frontal sinus, but it is auta- together by their narrow (relative to labial pomorphic in having lost the cuspules on Il- length) upper molars. This clade contains a 2. The recognition that Nyctereutes, a form diverse group of taxa ranging from moder- restricted to the Palearctic, is the sister genus ately hypocarnivorous Nyctereutes, Cerdo- of Cerdocyon was an interesting conclusion cyon, and Lycalopex, to more or less meso- from this analysis. It shows reduction in the carnivorous Chrysocyon, and Atelocynus, to size ofthe frontal sinus although fossil species completely hypercarnivorous Speothos. The ofNyctereutes have a slightly larger sinus than range of their postcranial morphology is the living form. The other sister group united equally broad, as this group contains the lon- at Node N represents a hypercarnivorous ad- 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE 23 aptation culminating in the unique loss ofM2 sorial carnivore of open grassland and sa- and m3 in Speothos. Its plesiomorphous sis- vanna. The dental modifications in this genus ter taxon Atelocynus likewise nearly lacks a parallel those of some of the more derived frontal sinus, and both have short nasals that Borophaginae and also converge on the type rarely extend to the level of the most poste- of dentition possessed by primitive Hyaeni- rior position of the maxillary-frontal suture. dae. All the large wolflike Canini are united by possession of very large frontal sinuses that are extended laterally into the postorbital PREVIOUS PHYLOGENETIC processes and posteriorly to the parietal su- HYPOTHESES ture (Node 0). They too have enlarged an- NEONTOLOGY gular processes ofthe mandible, but they emphasize differential development of the su- Despite the voluminous literature on the perior branch of the medial pterygoid rather living canids (the Caninae ofthis review) there than the inferior branch as in the South have been few neontological studies of the American forms. In addition the front limbs phyletic relationships of these taxa. Histori- are elongated relative to the hind limbs, par- cal proposals to group species into genera ticularly the epipodials (expressed as a radi- constitute attempts at limited phyletic anal- us/tibia ratio ofgreater than 90%). Canis and ysis and these will be specially noted in the its hypercarnivorous sister taxa Cuon and Ly- paragraphs that follow. Huxley's (1880) di- caon also possess a number of dental syna- vision of the canines into foxlike (alopecoid) pomorphies including enlargement ofI3 (rel- and jackal-like (thooid) groups, primarily ative to 11-2), enlargement of its cingulum, based on the presence or absence ofa frontal and development of a medial crest; enlarge- sinus, was an attempt to group the species of ment ofthe paracone ofM1-2 relative to the Canis as that genus was used in the last cen- metacone; reduction of the buccal cingulum tury. However, Huxley's conclusions were not on the upper molars; presence of a second used as a basis for classification by subse- posterior cusplet on p4; and lack of the par- quent workers, including the first mono- aconid plus presence ofa weak paracristid on graphic review of the Canidae by Mivart m2. In addition the sagittal crest becomes (1890). Mivart relegated the foxes and dogs more important, the interparietal portion es- to the genus Canis but accorded separate ge- pecially so, such that the supraoccipital shield neric rank to Cuon, Lycaon, Otocyon and assumes a triangular shape and the pointed Speothos. The simple key used to character- inion markedly overhangs the condyles. The ize these genera was based on digital count latter characters are usually more pro- and dental formula. nounced in old males, but they constitute an Recognition ofthe foxes as a genus Vulpes, obvious difference in form even when com- distinct from Canis, gained almost universal paring some ofthe smallerjackal species with favor in the 20th century although debate plesiomorphous forms of similar size, such continued as to the generic status ofthe more as Dusicyon australis or Pseudalopex cul- distinctive forms: the fennec (Fennecus zer- paeus. da), arctic fox (Alopex lagopus), and gray fox Cuon and Lycaon are closely related to Ca- (Urocyon cinereoargenteus). Likewise the nis, but diverge (Node P) in aspects of ad- South American canids, generally separated aptation to a more predatory way oflife. Their from Canis in the 19th century, were later molars emphasize shearing functions by re- united under Canis (except Speothos) by Mi- ducing the lingual cusps, the hypocone on vart (1890), and again separated generically M 1-2, and the metaconid and entoconid on in the early 20th century. Thomas (1914), m 1-2. Their canines are also small relative Kraglievich (1930), Cabrera (1931) and Os- to the size oftheir cheek teeth. Lycaon shows good (1934) were among the principal revi- the most derived state ofhypercarnivory with sors. An early test of the classical grouping its large multicuspid premolars, unusually ofthe South American forms with Canis was wide palate, and unique reduction ofthe pol- a comparison of C. antarcticus and C. latrans lex to a vestige as befits a pack-hunting, cur- by Pocock in 1913. Pocock concluded that 24 AMERICAN MUSEUM NOVITATES NO. 3146 the affinities of C. antarcticus (the type spe- subgenera; (3) Chrysocyon; and (4)Lycalopex. cies ofDusicyon H. Smith, 1839) lay with the Later (1975) he relegated the "zorros de cam- South American forms whereas the coyote po" to the genus Canis with Dusicyon and was related to the wolf and jackal. Thomas Pseudalopex as subgenera concluding that (1914), and later Cabrera (1931), developed these taxa represented "generalized dogs" in morphological keys for identification of the the same way as other species attributed to South American species. They recognized Canis. The "zorros de monte," on the other several genera (Lycalopex, Dusicyon, Cer- hand, represent "differentiated dogs" that docyon and Pseudalopex), the first two of Langguth intimates form a monophyletic which were considered monotypic and Cer- group that originated in the Brazilian high- docyon became so by transfer of C. microtis lands. These forms were united morpholog- to a new genus Atelocynus Cabrera, 1940. The ically using some of the characters we use to peculiar Chrysocyon was also accorded ge- support the same grouping. Chrysocyon and neric rank and Speothos was still placed in a Lycalopex were too specialized in their own separate subfamily Simocyoninae by such re- ways for Langguth to clearly discern their cent authors as Simpson (1945). Some (e.g., phyletic relationships. Osgood, 1934) believed that Urocyon had a In a dissertation study oflarge Quaternary special affinity with the other South Ameri- South American canids, published in 1988, can forms rather than with Vulpes. Among Berta was the first to employ cladistic meth- the attempts to group these genera were the ods as a main tool for phylogenetic recon- studies of living and fossil taxa by Kragliev- struction. Her cladistic framework ofthe liv- ich (1930) who recognized three genera: Ly- ing South American canids was influenced by calopex (for L. vetulus only), Canis with four an earlier manuscript ofthis project, and has subgenera (Cerdocyon, Chrysocyon, Pseudal- reached somewhat the same conclusions. opex, and Dusicyon), and Speothos (as a sim- Many ofthe same characters have been used ocyonine). Lycalopex vetulus was regarded as in support of her phylogeny as well as ours. a basal taxon. Osgood (1934) likewise chose Our conclusion, however, differs from those to recognize the South American forms as a of Berta's in two aspects: (1) placement of group, but he concurred with Thomas and Chrysocyon and Lycalopex in the same clade Cabrera in recognizing their generic distinc- as Cerdocyon group; (2) separation ofseveral tion from Canis. He advocated the use ofthe species of "Pseudalopex" (e.g., "P." gym- subgenus as a means of showing the affinities nocercus) from the genotypic species P. cul- of these forms among themselves. Thus he paeus and their placement at the base of the recognized two genera (apart from Urocyon South American canid clade. and Speothos): Chrysocyon and Dusicyon, the Clutton-Brock et al. (1976) undertook a latter with three subgenera, Cerdocyon, Ly- phenetic study of 35 of the 39 living species calopex and Dusicyon. The first two subgen- by numerical methods. These workers scored era were monotypic, following Osgood's 90 characters in their analysis, including transfer of microtis from Cerdocyon to Dus- morphological features of skull (13 charac- icyon. Simpson (1945) adopted Osgood's ters), teeth (13), pelage ofhead and body (17), classification. pelage ofextremities (plus other external fea- There have been few changes in the no- tures, 17), postcranial skeleton and internal menclature until the group was reviewed by anatomy (14), and behavior (16). Only about Langguth (1969, 1970, 1975). Langguth com- 10 percent of the same characters were used prehensively compared the biology of the in our analysis. The phenetic analysis was South American canids as a whole and with presented as two-dimensioned plots of the special reference to Chrysocyon. He conclud- principal coordinates algorithm for all char- ed in 1969 and 1970 that there were four acters and for the skull and teeth only. A list generic groups: (1) "zorros de campo," placed of "near-neighbor" similarity figures for the in the genus Dusicyon with Dusicyon and same combinations of data was used in as- Pseudalopex as subgenera; (2) "zorros de sessing the details of species relationships. monte," using the genus Cerdocyon and plac- The principal results illustrated the isolated ing Cerdocyon, Atelocynus and Speothos as positions ofthe monotypic living genera Ly- 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE 25 caon, Cuon, Speothos, Otocyon, Nyctereutes, cies of Canis, Dusicyon, and Vulpes form an Chrysocyon, and more weakly Alopex (from unresolved multichotomy under Canis. Ca- Vulpes). Urocyon and Fennecus had high nis in turn has an unspecified relationship to similarity to other Vulpes species (this com- the six monotypic genera that follow it in his parison showed V. vulpes to be atypical), and classification. In contrast to these phenetic Cerdocyon and Atelocynus were closely sim- analyses, cladistic analysis is able to offer hy- ilar to species of Dusicyon (sensu Simpson, potheses about the phyletic relationships of 1945). In addition to the monotypic genera, these groups as well as illuminate the possible three larger genera were recognized: Canis, phylogenetic relationships of the more apo- Vulpes, and Dusicyon, although these were morphous forms that appear so distinctive found to be closely related. In a sense their on phenetic analysis. conclusions represent a return to the scheme used by Mivart (1890), and at one point Clut- PALEONTOLOGY ton-Brock et al. (1976: 140) remarked that There have been few paleontological con- "no objective analysis of the results of this tributions to the phylogeny of the living can- study would produce these three genera as ids based on first hand studies other than presently composed but nor would it produce those of Matthew (1924, 1930). His views any other clear-cut grouping of species"! have been very widely accepted and are now Van Gelder's (1978) review of the canid well entrenched in textbooks. Matthew was classifications of Langguth (1975) and Clut- one of the principal proponents of the con- ton-Brock et al. (1976) attempted to reconcile cept of the unicuspid talonid "dogs" or Sim- these schemes in light of his contention that ocyoninae as a monophyletic group. Simpson the genus can be defined as a collection of (1945: 109-110, 223-224) retained this group species capable of interbreeding and produc- with reservations, and listed the genera at- ing hybrids. Applying this rule of thumb to tributed to it. Only 7 (Enhydrocyon, Philo- the few data regarding canid hybrids (usefully trox, Euoplocyon, Protocyon, Speothos, Cuon, summarized in Gray, 1971), Van Gelder and Lycaon) out of a total of 15 genera as- (1978) concluded that all living species cur- signed by Simpson to this group belong to rently assigned to Canis, Dusicyon and Vulpes the Canidae as we have defined it. The re- should be included within the genus Canis, mainder, including the fossil Simocyon itself, thus advocating the taxonomic suggestion belong to arctoid families. A growing dissat- made by Clutton-Brock et al. (1976) and isfaction with the phyletic basis for this view quoted above. The monotypic Nyctereutes, has been indicated by others (e.g. Clutton- Chrysocyon, Speothos, Cuon, Lycaon and Brock et al., 1976) and the superficial basis Otocyon were retained as distinct genera al- for grouping the canids in this manner has though no interbreeding tests ofthese species been rather strongly refuted by the data pre- were cited to substantiate this view. Van sented herein. Matthew also believed that Gelder (1978) used the subgenus to group the species ofthe genus Tomarctus lay at the root species ofhis Canis into presumably phyletic of the phylogeny of Vulpes and Canis. He units and the resulting classification differs indicated that "Nothocyon" (now Cormo- very little from that of Clutton-Brock et al. cyon, and partly including species referred to (1976) or, for that matter, from that ofMivart Leptocyon) was in turn ancestral to Tomarc- (1890). The ability to hybridize can be viewed tus. Although we could agree that some spe- as a symplesiomorphy, a primitive feature, cies placed in "Nothocyon" by Matthew have for the ability to interbreed between two spe- a phylogenetic relationship with Vulpes, we cies must reflect the retention ofmuch ofthe deny any close phylogenetic relationship of genome oftheir common ancestral stock. This Tomarctus with Vulpes or Canis. Tomarctus view explains the lack of resolution of phy- is a typical borophagine with an overlay of letic affinities to be gained from interbreeding apomorphic features typical oflater members studies, a point elaborated by Rosen (1979: of that group (complex incisors and premo- 275-278) in discussing the relevance of hy- lars, massive molars, strong premaxillary- bridization data in cladistic analysis. Thus, frontal union, etc.). An unprecedented in Van Gelder's (1978) classification the spe- amount of evolutionary reversal would have 26 AMERICAN MUSEUM NOVITATES NO. 3146 to accompany the origin ofany living canines fission mechanism that he held to be a rare from that source. The evidence seems clear centric misdivision leading to an increase in that the Caninae have a phyletic relationship diploid number. Thus, low diploid number with the vulpines and, along with Leptocyon, would be primitive in his view. Todd de- constitute a monophyletic group. The fossil duced that the primitive diploid number for record of the Caninae, to be discussed else- the Carnivora is 2n = 38 from inspection of where, does not alter this view and fossil gen- the distribution of diploid numbers among era inserted into the cladogram based on Re- karyotyped Carnivora. cent forms amplify some groups but does not Comparing the karyotypes among the can- change the branching pattern indicated in fig- id genera led Chiarelli (1975) to the following ure 2. However, canid fossils do serve three general conclusions: important functions. First is the correct in- (1) Canis (including C. aureus, C. fami- terpretation of homoplasies. Apparent syn- liaris, C. latrans, C. lupus, C. mesomelas, C. apomorphies inferred from living taxa alone rufus), Lycaon, Chrysocyon, Atelocynus, Ly- can turn out to be independently derived when calopex (L. vetulus), "Pseudalopex" griseus fossil taxa are included in the analysis. The (Gallardo and Formas, 1975), "P." gymno- second function of fossil canids is their tran- cercus (Brum-Zorilla and Langguth, 1980) and sitional positions in phylogeny, which breaks Speothos all have only acrocentric autosomes, up seemingly long lists of synapomorphies a high diploid number (74-78), share mor- into smaller components or intermediate phologically identical X chromosomes, and states. Although this latter function has the some show similar autosome morphologies. appearance ofweakening the support for each (2) Vulpes species (V. bengalensis, V. pal- node (diluting the synapomorphies), it ac- lida, V. ruppelli, V. vulpes) have a range of tually strengthens the overall phylogeny by diploid numbers (34-60) but the same fun- displaying evidence ofthe actual sequence of damental number ofautosomes (72). The ex- character acquisition, thus further corrobo- tremes, V. vulpes (34- 40) and V. bengalensis rating hypotheses of character transforma- (60; the "basic-fox" of Clutton-Brock et al., tions. The third function is that the direct 1976), may hypothetically be differentiated historical record provided by fossils yields by centric fusion mutations in Chiarelli's data on the temporal succession oftaxa, their view. zoogeographic relationships, and predictions (3) Urocyon and Fennecus (Vulpes zerda) regarding the history of the group. differ slightly in diploid number (66 and 64 respectively) but have an identical fundamental number (70) and have two morpho- KARYOLOGY logically similar chromosomes marked by an In recent years, increasingly comprehen- achromatic region near the centromere. sive karyological data have become available (4) Alopex( Vulpes lagopus), Otocyon, Cer- for the Canidae. Chiarelli (1975) summarized docyon, and Nyctereutes have karyotypes not these data and used the aspects of chromo- completely comparable with other canids in some morphology to advance phylogenetic terms of chromosome morphology. Alopex conclusions. His conclusions were dependent shows chromosome polymorphism with dip- on the acceptance of a particular hypothesis loid numbers 48, 49, 50, and 52 having been of chromosome evolution. Chiarelli (1975) observed, but the fundamental number is 88 posited a Robertsonian or centric-fusion due to the large number of meta- or sub- mechanism and hence regards higher chro- metacentric chromosomes. Nevertheless, mosome number as representative of the close karyological relationship exists with primitive state. In his view, Canis has the Vulpes vulpes as shown by comparative chro- more generalized karyotype (2n = 78), Vulpes mosome studies (Miikinen and Gustavsson, (V. vulpes 2n = 34-38; V. bengalensis 2n = 1892; Yoshida et al., 1983) and reciprocal 60) the more specialized one. In another con- crosses of these species in which the hybrids tribution using the same data base, Todd show reduced fertility (Gray, 1971). Otocyon (1970), in an innovative view of canid chro- shows a high diploid number (72) and large mosome evolution, postulated a karyotypic fundamental number (80). Although the 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE 27 karyotype seems to be dominated by acro- siderable similarity in chromosome mor- centric chromosomes, the number of meta- phology despite the variation expressed by centrics or submetacentrics appeared to differ missing chromosomes and unique additions in different animals studied. Todd (1970) and rearrangements. The gray wolf served as suggested, without discussion, a relationship a basis of comparison in reconstruction of to the karyotype of Vulpes bengalensis. Cer- the phylogenetic relationships of the taxa in- docyon, karyotyped by Wurster-Hill in 1973, vestigated. A three-taxon statement resulted was found to have a unique chromosome in which the gray wolf had the South Amer- complement with diploid number of 74, but ican canids (Speothos, Cerdocyon and Chry- a fundamental number of 110 due to a nearly socyon) as a sister taxon and they in turn were equal number of submetacentric and meta- the sister group to a clade including Urocyon, centric chromosomes (36) and acrocentrics Fennecus and Otocyon. Four lower diploid (38). This is the highest fundamental number canids (2n < 64), in which metacentric chro- known within the Carnivora. Partial genetic mosomes predominate, were compared by compatibility with Canis is shown by pro- G-banding techniques. Wayne et al. (1987b) duction of sterile hybrids in crosses with the found that Alopex and Vulpes macrotis have domestic dog (Gray, 1971). Nyctereutes races identical G-banding patterns and chromo- show considerable karyotypic variability. some morphology. The red fox, V. vulpes, has Nyctereutes procyonoides viverrinus (the Jap- only two chromosomes in common with Al- anese race) has a low diploid number (42) opex and V. macrotis but shows extensive and fundamental number of 68 and mor- chromosome arm similarity with these foxes phologically unique chromosomes (X is the justifying a postulated close phyletic rela- largest acrocentric and Y is a small, satellited tionship between these species. The raccoon- acrocentric). Todd (1970) deduced that the dog, Nyctereutes procyonoides (2n = 42) was chromosome complement of Nyctereutes found to share some chromosomes with these corresponds to that required to yield the New foxes and also with Canis lupus. In fact Wayne World canine array through the fissioning et al. (1987b) reported that approximately mechanism, and hence that genus is a sur- three-fourths of the gray wolf chromosomes vivor of an early stage in the differentiation (2n = 78), or abut 85 percent of the length of the Canini. However, recent studies by of the wolf karyotype, have similarity to a Maikinen and Fredga (1980), Ward et al. chromosome arm or entire chromosome of (1987) and Wurster-Hill et al. (1986, 1988) the raccoon-dog. About three-fourths of the indicate that the mainland nominate race, N. chromosome arms of Nyctereutes are also p. procyonoides, has a more typical Canini morphologically similar to those of the red karyotype (2n = 56, NF 68) with an X chro- and arctic foxes but there is less correspon- mosome, a medium-sized metacentric and a dence in terms ofthe total length ofthe karyo- small Y metacentric with satellites. The au- type than the case of the wolf. tosomes consist of 10 metacentrics to sub- Despite the uncertainty about the precise metacentrics and 44 acrocentrics. relationship of Nyctereutes, Wayne et al. In a more recent summary, Wayne et al. (1987b) placed the raccoon-dog as a sister (1987a, 1987b) used the Giemsa-banding taxon to the foxes and joined the low and patterns of chromosomes to more closely high diploid trees at the base in the "most compare their morphology in a selected group parsimonious" arrangement. Nyctereutes was ofcanids. These studies delineated two major compared with representative feline chro- groupings based on diploid number. Those mosomes and found to contain a number of species having a high diploid number (2n > morphologically similar features resulting in 64) and largely acrocentric chromosomes their assessment that that genus possessed the correspond to the first group discussed by most primitive canid karyotype. Wayne et al. Chiarelli (1975) plus Cerdocyon, Otocyon and agree with Todd (1970) that the low diploid Chiarelli's third group (Urocyon and Fenne- number of the raccoon-dog implies that fis- cus). Similarities in chromosome morphol- sion mechanisms must be involved to yield ogy revealed by G-banding techniques al- higher diploid numbers. In addition, fusion, lowed Wayne et al. (1987a) to recognize con- translocation and terminal addition mecha- 28 AMERICAN MUSEUM NOVITATES NO. 3146 nisms must also be involved to yield the phy- group placed as a sister taxon to Canis and logenetic pattern discerned from the karyo- Lycaon. Speothos is separated from the South type data. American taxa and allied with Canis and Ly- Comparing the cladistic relationships ofthe caon, an association that would require five investigated taxa deduced from the karyo- additional steps in our morphological anal- logical data with those deduced from the ysis. This result also demands significantly morphological evidence advanced in this pa- greater karyotypic resemblance of Speothos per (fig. 8), we see three major areas ofagree- and Canis than shown by the chromosome ment: (1) the Canis and Vulpes groups ofspe- evidence. cies represent the primary sister taxa; (2) that The phylogenetic position of Nyctereutes the South American taxa examined consti- also remains unresolved in the allozyme tute a monophyletic group allied to Canis and comparisons although it now becomes a have the same branching pattern; and (3) that monotypic element in the pentachotomy the Vulpes species, including Alopex are a rather than a part of the vulpine group as in monophyletic group. Areas of disagreement the karyotype phenogram. Likewise the allo- are also threefold: (1) in the Wayne et al. zyme data do not clarify the relationships of analysis Nyctereutes is a member of the vul- Otocyon or Urocyon to each other or other pine clade; (2) Otocyon and Urocyon are in a canids, but Fennecus is grouped with the oth- clade associated with the South American er Vulpes species as in the morphological re- canids and Canis; and (3) Fennecus is asso- sults. ciated with the gray and bat-eared foxes rath- In 1989 Wayne et al. reported the results er than Vulpes. Despite these details it is en- ofDNA hybridization among the same canid couraging to see that both sets of data group taxa and found it yielded limited information most canids into vulpine and canine clades beyond confirming the grouping of the wolf- and that the South American taxa are allied like canids (wolf and jackals) and the foxlike to the canines. canids (red fox, arctic fox and fennec). Other taxa could not be resolved because the stan- dard error ofthe measurements (melting tem- BIocHEMIcAL SYSTEMATICS peratures) is too large to permit clear dis- Wayne and O'Brien (1987) used electro- crimination of further groupings of taxa. phoretic methods to examine the protein A DNA sequence study by Geffen et al. products of 51 genetic loci from blood and (1992) used mitochondrial DNA restriction tissue preparations. Eleven of 15 canid genera fragment and restriction site data plus the (Cuon, Dusicyon, Atelocynus, and Pseudalo- sequence variation among 402 base pairs of pex omitted) were examined. The electro- cytochrome b to resolve the relationships of phoretic data were analyzed quantitatively the foxlike canids. The results confirmed the using several methods for calculating genetic monophyly ofAlopex, Vulpes and Fennecus distance and illustrated by two methods of and some details ofrelationship among these tree construction (UPGMA and distance and other species of Vulpes, but it left the Wagner). These phenograms reflect overall relationships of Otocyon and Urocyon to one and specific similarity which may or may not another or to other foxes unresolved. represent phylogeny. The data have the same significance in phylogeny reconstruction as the karyological data discussed above. They CONCLUSIONS summarized their results in a strict consensus In the preceeding pages, features of the os- tree which depicted phenetic relationships less teology ofthe Recent canine carnivores have resolved than the chromosome data (fig. 9). been used to generate a phylogenetic synthe- The tree is rooted using Ursus arctos as the sis based on 57 characters distributed among outgroup, but the canid relationships form a 122 character states. Polarity was established pentachotomy with the relationships of Uro- from examination ofthree extinct outgroups cyon, Otocyon and Nyctereutes completely that represent successive sister taxa of the unresolved. The South American canids (less living Caninae. Fifteen canine taxa were cho- Speothos) are, however, grouped and that sen as the basis for the analysis, ten ofwhich 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE 29

Morphology

U1)

) -0 U) Q1) 0 -0 z~ 0 1~~~ C)~) -0 ci) ci) I- I 0 C) 60a- .) l)

Chromosomes

Fig. 8. Cladograms depicting canine relationships based on Wayne et al. (1987b) chromosomal data compared with the morphological data provided by the present study. Array oftaxa limited to comparable forms. were monotypic genera and four polytypic Phylogenetic reconstruction by a maxi- taxa were considered monophyletic because mum parsimony program (HENNIG86, Ver- of close morphologic resemblance of their sion 1.5, Farris, 1988) was analyzed using species, bolstered in the case of Canis and CLADOS, Version 1.2, Nixon, 1992). A sin- Vulpes by chromosomal and molecular evi- gle tree was found with length = 90 (excluding dence ofpropinquity. Character analysis was autapomorphies), consistency index = 65 and however based on the array ofspecimens list- retention index = 78 (fig. 2). The reconstruc- ed in Appendix 1. tion delineates two sister taxa (fig. 2, Node 30 AMERICAN MUSEUM NOVITATES NO. 3146

U)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~".

S) Qc EQ)8 c3 U) - (~ ~ ~ Q)rz ,,

Allozymes

Fig. 9. Cladograms depicting canine relationships based on Wayne and O'Brien's (1987) allozyme data compared with the morphological data provided by the present study. Array of taxa limited to comparable forms.

D): the foxlike Caninae, tribe Vulpini, and any other system (figs. 8, 9). The position of the wolflike and South American taxa plus the foxes Urocyon and Otocyon are variously the raccoon-dog, tribe Canini. This funda- portrayed in the phenetic analyses of chro- mental subdivision is supported by karyo- mosome morphology (a sister taxon of the logical and biomolecular studies although the fennec and ofthe Canini clade) or ofallozyme precise composition ofeach group varies with alleles (as part of the basal pentachotomy of the evidence used. In all cases the morpho- canine relationships). Even more contentious logical evidence leads to more fully resolved is the position ofthe Asian raccoon-dog, Nyc- relationships than presently available from tereutes, which is firmly included within the 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE

South American clade as a sister taxon ofthe it in a trichotomy (fig. 2, Node K) with mem- crab-eating "fox," Cerdocyon, in our analy- bers of the South American clade in agree- sis, but is allied with Vulpes based on kar- ment with both chromosome and allozyme yology or joins Otocyon and Urocyon in the evidence. basal canine pentachotomy in the allozyme There is thus reasonable concordance of results. The position of the South American results with the three methods ofphyletic in- bush-dog, Speothos, also has different rela- ference as to the basic framework of canine tionships in the chromosome analysis versus phylogeny. Chromosomal and molecular the allozyme data. Chromosome morphology methods have not been applied to the whole and osteology agree in placing Speothos with- array ofRecent species so the analysis is more in the South American clade, but allozymes comprehensive in the case ofthe osteological ally it with the wolf group as a sister taxon investigation. It is hoped that further biomo- of Canis lupus plus the hunting-dog, Lycaon lecular data will eventually allow as compre- pictus. In an earlier version ofthe osteological hensive a sample and that sequencing studies synthesis (Berta, 1988) the maned-wolf, may provide additional insights into the phy- Chrysocyon, was held to be a sister taxon of logenetic relationships of the Recent Cani- the wolf group, but the present results place dae.

REFERENCES Berta, A. Gaspard, M. 1988. Quaternary evolution and biogeogra- 1964. La region de l'angle mandibulaire chez phy of the large South American Can- les Canidae. Mammalia 28: 249-329. idae (Mammalia: Carnivora). Univ. Geffen, E., A. Mercure, D. J. Girman, D. W. Mac- California Publ. Geol. Sci. 132: 1-149. donald, and R. K. Wayne Brum-Zorilla, N., and A. Langguth 1992. Phylogenetic relationships ofthe fox-like 1980. Karyotype of South American pampas canids: mitochondrial DNA restriction fox Pseudalopex gymnocercus (Carniv- fragment, site and cytochrome b se- ora, Canidae). Experimentia 36: 1043- quence analysis. J. Zool. Soc. London 1044. 228: 27-39. Cabrera, A. Gray, A. P. 1931. On some South American canine gen- 1971. Mammalian hybrids-a check-list with era. J. Mammal. 12: 54-67. bibliography. Commonw. Bur. Anim. Chiarelli, A. B. Breed. Genet. Tech. Commun. no. 10 1975. The chromosomes of the Canidae. In (revised), Edinburgh. M. W. Fox (ed.), The wild canids, their Hildebrand, M. systematics, behavioral ecology and 1952. An analysis of body proportions in the evolution, pp. 40-53. New York: Van Canidae. Am. J. Anat. 90: 217-256. Nostrand. 1954. Comparative morphology of the body Clutton-Brock, J., G. B. Corbert, and M. Hills skeleton in recent Canidae. Univ. Cal- 1976. A review of the family Canidae, with a ifornia Publ. Zool. 52: 399-470. classification by numerical methods. Hunt, R. M., Jr. Bull. Br. Mus. (Nat. Hist.) Zool. 29: 119- 1974. The auditory bulla in Carnivora: an an- 199. atomical basis for reappraisal of carni- Evans, H. E., and G. C. Christensen vore evolution. J. Morphol. 143: 21-76. 1979. Miller's anatomy of the dog. Philadel- Huxley, T. H. phia: Saunders. 1880. On the cranial and dental characters of Ewer, R. F. the Canidae. Proc. Zool. Soc. London 1973. The carnivores. Ithaca, NY: Cornell 16: 238-288. Univ. Press. Kraglievich, L. J. Farris, J. S. 1930. Craneometria y clasification de los can- 1988. Hennig 86, Version 1.5. New York: Port idos sudamericanos especialmente los Jefferson Station. Argentinos actuales y f6siles. Physis 10: Gallardo, M., and J. R. Formas 35-73. 1975. The karyotype ofDusicyon griseus (Car- Langguth, A. nivora, Canidae). Experimentia 31: 639. 1969. Die Siudamerikanishen Canidae unter 32 AMERICAN MUSEUM NOVITATES NO. 3146

besonderer Berucksichtigung des Maih- Tedford, R. H. nenwolfes Chrysocyon brachyurus Illi- 1976. Relationship of pinnipeds to other car- ger. Z. Wiss. Zool. 179: 1-88. nivores (Mammalia). Syst. Zool. 25: 1970. Una nueva clasificacion de los canidos 363-374. sudamericanos. Actas IV Congr. Lati- Thomas, 0. noamericano de Zool. 1: 129-143. 1914. The generic and subgeneric names of 1975. Ecology and evolution in the South South American Canidae. Ann. Mag. American canids. In M. W. Fox (ed.), Nat. Hist., ser. 8, 13: 350-352. The wild canids, pp. 192-206. New Todd, N. B. York: Van Nostrand. 1970. Karyotypic fissioning and canid phylog- Makinen, A., and K. Fredga eny. J. Theor. Biol. 26: 445-480. 1980. Banding analysis of the somatic chro- Van Gelder, R. G. mosomes of racoon dogs; Nyctereutes 1978. A review of canid classification. Am. procyonoides, from Finland. Fourth Eu- Mus. Novitates 2646: 10 pp. ropean Colloquium on Cytogenetics of Wang, X. Domestic Animals, pp. 420-430. 1994. Phylogenetic systematics of the Hes- Makinen, A., and I. Gustavsson perocyoninae (Carnivora: Canidae). 1982. A comparative chromosome-banding Bull. Am. Mus. Nat. Hist. 221: 207 pp. study in the silver fox, the blue fox, and Wang, X., and R. H. Tedford their hybrids. Hereditas 97: 289-297. 1994. Basicranial anatomy and phylogeny of Matthew, W. D. primitive canids and closely related 1924. Third contribution to the Snake Creek miacids (Carnivora: Mammalia). Am. fauna. Bull. Am. Mus. Nat. Hist. 50: Mus. Novitates 3092: 34 pp. 59-210. In press. The initial diversification ofcanids in 1930. The phylogeny ofdogs. J. Mammal. 11: the White River Group, Eocene and 117-138. Oligocene of North America. In D. R. Miller, M. E., G. C. Christensen, and H. E. Evans Prothero and R. J. Emry (eds.), The ter- 1964. Anatomy of the dog. Philadelphia: W. restrial Eocene-Oligocene transition in B. Sanders, i-xii, 1-941 pp. North America. Cambridge Univ. Press. Mivart, St. G. Ward, 0. G., D. H. Wurster-Hill, F. J. Ratty, and 1890. Dogs, jackals, wolves and foxes. A Y. Song monograph of the Canidae. London, i- 1987. Comparative cytogenetics of Chinese xxxvi, 1-216 pp. and Japanese racoon dogs, Nyctereutes Nixon, K. C. procyonoides. Cytogenet. Cell Genet. 45: 1992. ClaDos, Version 1.2. Ithaca, NY: Cor- 177-186. nell Univ. Wayne, R. K., and S. J. O'Brien Osgood, W. H. 1987. Allozyme divergence within the Cani- 1934. The genera and subgenera of South dae. Syst. Zool. 36: 339-355. American Canidae. J. Mammal. 15:45- Wayne, R. K., W. G. Nash, and S. J. O'Brien 50. 1987a. Chromosomal evolution of the Cani- Pocock, R. I. dae, I. Species with high diploid num- 1913. The affinities ofthe Antarctic Wolf(Ca- bers. Cytogenet. Cell Genet. 44: 123- nis antarcticus). Proc. Zool. Soc. Lon- 133. don, 1913: 382-393. 1987b. Chromosomal evolution ofthe Canidae, Rosen, D. E. II. Divergence from the primitive car- 1979. Fishes from the uplands and intermon- nivore karyotype. Cytogenet. Cell Gen- tane basins of Guatemala: revisionary et. 44: 134-141. studies and comparative geography. Wayne, R. K., R. E. Beneviste, D. N. Janczewski, Bull. Am. Mus. Nat. Hist. 162: 267- and S. J. O'Brien 376. 1989. Molecular and biochemical evolution of Simpson, G. G. the Carnivora. In J. L. Gittleman (ed.), 1945. The principles of classification and a Carnivore behavior, ecology and evo- classification of mammals. Bull. Am. lution, pp. 465-494. Ithaca, NY: Cor- Mus. Nat. Hist. 8: 1-350. nell Univ. Press. Stains, H. J. Wozencraft, W. C. 1975. Calcanea of members of the Canidae. 1993. Order Carnivora. In D. E. Wilson and Bull. South. California Acad. Sci. 74: D. M. Reeder (eds.), Mammal species 143-155. ofthe world: a taxonomic and geograph- 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE 33

ic reference, 2nd ed. pp. 279-348. 1988. Fragile sites, telomeric DNA sequences, Washington, D.C.: Smithson. Inst. Press. B. Chromosomes, and DNA content in Wurster-Hill, D. H. racoon dogs, Nyctereutes procyonoides, 1973. Chromosomes ofeight species from five with comparative notes on foxes, coy- families of Carnivora. J. Mammal. 54: ote, wolf, and racoon. Cytogenet. Cell 753-760. Genet. 49: 278-281. Wurster-Hill, D. H., 0. G. Ward, H. Kada, and Yoshida, M. A., M. Takagi, and M. Sasaki S. Whittlemore 1983. Karyotypic kinship between the blue fox 1986. Banded chromosome studies and B (Alopex lagopus Linn.) and the silver fox chromosomes in wild-caught racoon (Vulpes fulva Desm.) Cytogenet. Cell dogs, Nyctereutes procyonoides viverri- Genet. 35: 190-194. nus. Cytogenet. Cell Genet. 43: 85-93. Wurster-Hill, D. H., 0. G. Ward, B. H. Davis, J. P. Park, R. K. Moyzis, and J. Meyne 34 AMERICAN MUSEUM NOVITATES NO. 3146

APPENDIX 1 List of Specimens of Living Canids Examined The sample was chosen such that interspecific variations were evaluated in genera with multiple species. Specimens listed below constitute the total sample from which morphological observations, illustrations, or measurements were made. The classification follows that by Wozencraft (1993) except as noted below. The arctic fox Alopex lagopus is here considered a species of Vulpes. We restrict the usage of Pseudalopex to its genotypic species P. culpaeus. The small South American "zorros" placed under the genus Pseudalopex by Wozencraft (1993) are a paraphyletic group in our analysis. Three of these species are thus without a proper generic name, here indicated by quotation marks: "Pseudalopex" gymnocercus, "P." sechurae, and "P." griseus. Lycalopex is available for L. vetulus as a monotypic genus. We were not able to examine the following species: Urocyon littoralis, Vulpes cana, V. ferrilata, and V. rueppelli. An asterisk following the catalog number indicates specimens whose frontal bone is cut open to expose the frontal sinus area or whose frontal sinus (or the lack of it) can be observed from natural breakages. Number Sex Locality Vulpes bengalensis: AMNH(M) 54517 M Kheri Forest, Uttar Pradesh, India AMNH(M) 54526 F Kheri Forest, Uttar Pradesh, India Vulpes chama: AMNH(M) 81760 Orange Free State, South Africa Vulpes corsac: AMNH(M) 85021 Gobi Desert, Mongolia Vulpes (Alopex) lagopus: AMNH(M) 15637 Northwest Territory, Canada AMNH(M) 40979 M Melville Bay, Greenland Vulpes macrotis: AMNH(M) 589 M Arizona, United States Vulpes pallida: AMNH(M) 82198 M Jobelein, Sudan Vulpes velox: AMNH(M) 15871 M Alberta, Canada AMNH(M) 100190 M New York Zoo, United States Vulpes vulpes: AMNH(M) 60C* No data AMNH(M) 295* No data AMNH(M) 29055 M Northwest Territory, Canada AMNH(M) 88713 M Turkomen Desert, Iran AMNH(M) 98223 F Alberta, Canada AMNH(M) 141148 M New Jersey, United States AMNH(M) 166938 F New Jersey, United States AMNH(M) 185649 M Georgia, United States AMNH(M) 243093* M Florida, United States AMNH(M) 253187 M Florida, United States Vulpes (Fennecus) zerda: AMNH(M) 70126 M Egypt AMNH(M) 80019 M New York Zoo, United States Urocyon cinereoargenteus: AMNH(M) 536* M California, United States AMNH(M) 4270 Chihuahua, Mexico AMNH(M) 5366 F Florida, United States AMNH(M) 26016 M Jalisco, Mexico 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE 35

APPENDIX 1-(Continued) Number Sex Locality AMNH(M) 29488* San Rafael del Norte, Nicaragua AMNH(M) 100224 Florida, United States AMNH(M) 121276* F California, United States AMNH(M) 243095 F Florida, United States Otocyon megalotis: AMNH(M) 111* no locality AMNH(M) 114270 M Masi Sand River, Kenya AMNH(M) 161153 Kondoa, Tanzania "Pseudalopex" gymnocercus: AMNH(M) 205778 M Soriano, Uruguay AMNH(M) 205783 M Treintay Tres, Uruguay AMNH(M) 205786 F Treintay Tres, Uruguay AMNH(M) 205787 F Treintay Tres, Uruguay AMNH(M) 205792 M Treintay Tres, Uruguay "Pseudalopex" sechurae: AMNH(M) 46525 M Portoviejo, Ecuador AMNH(M) 46528* M Portoviejo, Ecuador '"Pseudalopex" griseus: AMNH(M) 14081 F Patagonia, Argentina AMNH(M) 33292* M Temuco, Maguehue, Chile AMNH(M) 41510* F Lavalle, Santiago del Estero, Argentina AMNH(M) 90050 F New York Zoo (from Chile) Pseudalopex culpaeus: AMNH(M) 66737* Upper Rio Pila, Ecuador AMNH(M) 67088 F Sincha, Ecuador AMNH(M) 147547 M New York Zoo, United States AMNH(M) 262663 M Oruro, Bolivia Dusicyon australis: ANSP 588 (cast) Falkland Islands Lycalopex vetulus: AMNH(M) 391* Chapada, Brazil AMNH(M) 100100 F New York Zoo, United States Chrysocyon brachyurus:

AMNH(M) 36962* M Urucum, Mato Grosso do Sul, Brazil AMNH(M) 120999 Alegrete, Rio Grande de Sul, Brazil AMNH(M) 133940 F Mato Grosso, Brazil AMNH(M) 133941 M Mato Grosso, Brazil Cerdocyon thous: AMNH(M) 14623 Bonda, Colombia AMNH(M) 14627 Cacaqualito, Colombia AMNH(M) 14636 Onaca, Colombia AMNH(M) 30628 F La Bomba, Venezuela AMNH(M) 36503* M Trinidad, Paraguay AMNH(M) 63942* Caqueta, Colombia AMNH(M) 133934 M Mato Grosso, Brazil AMNH(M) 133935 M Mato Grosso, Brazil AMNH(M) 209122 M Beni, Bolivia AMNH(M) 209124 F Beni, Bolivia AMNH(M) 234221 M Misiones, Paraguay 36 AMERICAN MUSEUM NOVITATES NO. 3146

APPENDIX 1-(Continued) Number Sex Locality Nyctereutes procyonoides: AMNH(M) 22792* F New York Zoo, United States AMNH(M) 43139 F Fujian, China AMNH(M) 45333 F Zhejiang, China AMNH(M) 57113 M Sichuan, China AMNH(M) 59323 M Fujian, China AMNH(M) 249766 F Honshu Island, Japan AMNH(M) 249767 F Honshu Island, Japan Atelocynus microtis: AMNH(M) 76031 Urubamba, Peru AMNH(M) 76579 M Lagarto, Loreto, Para, Peru AMNH(M) 95284 M Rio Tapajoz, Pari, Brazil AMNH(M) 95285 F Rio Tapajoz, Para, Brazil AMNH(M) 98639 F Iquitos, Loreto, Peru AMNH(M) 100095* F New York Zoo, United States Speothos venaticus: AMNH(M) 98558 M Iquitos, Loreto, Peru AMNH(M) 98560* F Iquitos, Loreto, Peru AMNH(M) 167846 M New York Zoo, United States AMNH(M) 184668 New York Zoo, United States Canis adustus: AMNH(M) 83767 M Tabora, Tanzania Canis aureus: AMNH(M) 54033 Ethiopia AMNH(M) 54515 F Hardwar, India AMNH(M) 81041 M Billata River, Ethiopia AMNH(M) 88712 M Dar Kaleh, Iran Canis latrans: AMNH(M) 21269* Durango, Mexico AMNH(M) 24860 M Sinaloa, Mexico Canis lupus: AMNH(M) 69452 no locality Canis mesomelas: AMNH(M) 34732* M Uasin Gishu, Kenya AMNH(M) 180109* F Lotome, Karamoja, Uganda AMNH(M) 233009 M Windhoek, Otjikundua, Namibia Canis rufus: AMNH(M) 4609 M Illinois, United States Canis simensis: AMNH(M) 81001 F Kordofan Plateau, Sudan AMNH(M) 81025 F Kordofan Plateau, Sudan AMNH(M) 81034* F Kordofan Plateau, Sudan AMNH(M) 214798* F Africa AMNH(M) 214799 Africa Cuon alpinus: AMNH(M) 54842 M Dharwar, India AMNH(M) 99636* New York Zoo, United States AMNH(M) 101882 M Het Zang Plateau, Java 1995 TEDFORD ET AL.: PHYLOGENY OF LIVING CANINAE 37

APPENDIX 1-(Continued) Number Sex Locality Lycaon pictus: AMNH(M) 114248 M Masi Sand River, Kenya AMNH(M) 114254* F Masi Sand River, Kenya AMNH(M) 164162 M Mourage, Chad

Recent issues of the Novitates may be purchased from the Museum. Lists of back issues of the Novitates, Bulletin, and Anthropological Papers published during the last five years are available free of charge. Address orders to: American Museum of Natural History Library, Department D, Central Park West at 79th St., New York, N.Y. 10024. TEL: (212) 769-5545. FAX: (212) 769- 5009. E-MAIL: [email protected]

L-

i) This paper meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper).