AMERICAN MUSEUM

Nov itates PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, N.Y. 10024 Number 2725, pp. 1-64, figs. 1-9, tables 1, 2 April 15, 1982

Phylogeny of Early Tertiary Carnivora, With a Description of a New Species of Protictis From the Middle Eocene of Northwestern Wyoming

JOHN J. FLYNN1 AND HENRY GALIANO2

ABSTRACT Since Cope erected the Miacidae in 1880, all for the suprageneric taxa Carnivora, Feliformia, early Tertiary carnivorans have generally been Caniformia, Didymictidae (Didymictida) (new), grouped in this family because of their overall Aeluroida (new), Feloidea, and Viverravidae. primitive morphology. A historical review of the Protictis, including the subgenera P. (Protictis), classifications and definitions of the Miacidae is P. (Bryanictis), and P. (Protictoides) (new), is provided. Discovery of a new subgenus of Pro- diagnosed and discussed. Simpsonictis is consid- tictis from the Uintan (mid Eocene) of north- ered a valid genus and is allied with Viverravus western Wyoming led to a reevaluation of the in the Viverravidae (Aeluroida, Feliformia). Pa- phylogeny of the Carnivora. The Miacidae (Mia- laeogale is removed from the Mustelidae (Cani- coidea), as previously constructed, is a paraphy- formia) and is considered a close relative of the letic assemblage of all the early Tertiary Carniv- Viverravidae. Ictidopappus is placed incertae ora. Phylogenetic analysis of the distribution of sedis within the Aeluroida. "Plesiomiacis" and cranial, dental, and postcranial morphology indi- Quercygale, which were previously considered cates that the bipartite division recognized in liv- members of the "Viverravinae," are considered ing Carnivora may be traced to the earliest Ter- members of the Caniformia. The sabretoothed tiary representatives of this order. Most taxa "ipaleofelid" Nimravidae are excluded from the previously included in the "Viverravinae" ("Mi- Feloidea (Aeluroida, Feliformia) and are included acidae") are members of the Feliformia and those instead within the Caniformia. A summary of the placed in the "Miacinae" ("Miacidae") are Can- proposed phylogeny and a classification of the iformia. Diagnoses and discussions are provided Carnivora are presented.

INTRODUCTION Carnivorans have been represented in since the nineteenth century. Much valuable most collections of early Tertiary faunas descriptive and interpretive work has been

1 Ph.D. candidate, Department of Geological Sciences, Columbia University; Department of Vertebrate Paleontol- ogy, American Museum of Natural History. 2 Department of Vertebrate Paleontology, American Museum of Natural History.

Copyright ©) American Museum of Natural History 1982 ISSN 0003-0082 / Price $4.15 2 AMERICAN MUSEUM NOVITATES NO. 2725 published on the carnivorans from these fau- predominance of primitive morphologies in nas in the past, although little has been writ- the Miacidae, but it is also a result of the ten recently on the phyletic relationships of failure of most previous workers to distin- these animals. While many investigators guish advanced characters that might ally have pursued intensive studies of the inter- various miacid taxa to later Carnivora. This relationships of the living Carnivora, the crit- paper attempts to decrease the gap in our ical Paleocene and Eocene Carnivora have understanding of carnivoran phylogeny. The remained largely ignored. Miacidae are no longer treated as a miscel- There is a pressing need for a critical re- laneous collection of primitive, early Ter- view of the paraphyletically constructed tiary Carnivora. Instead we apply a phylo- family Miacidae. Since Cope established the genetic methodology to an analysis of Miacidae, this family has been considered a "miacid" genera (particularly the "Viver- basal stock of primitive Carnivora. Regret- ravinae") in an attempt to determine the re- tably, most workers have justified this prac- lationship between these forms and the later, tice by stressing that most of the early Ter- advanced carnivorans. The Miacidae were tiary forms lack the diagnostic features of the not a closely allied group that formed, in its later, more advanced carnivorans. In gener- entirety, the primordial stock for all later al, the grouping of all primitive carnivorans Carnivora. The differentiation into the two has greatly hindered the determination of the distinct lineages (Caniformia and Feliformia) specific interrelationships of the major car- recognized in living carnivorans, occurred nivoran groups. In spite of this, a few at- early in carnivoran phylogeny and is defini- tempts have been made to include miacid tively reflected in the phylogenetic alliances genera in the analysis and arrangement of of the early Tertiary Miacidae. carnivoran lineages (Wortman and Matthew, The early Teritiary Carnivora have gen- 1899; Wortman, 1901; Gregory and Hellman, erally been classified in a single horizontal 1939), but such efforts have not been gen- group, the Miacidae, although many workers erally accepted. have recognized that various "miacids" are Tedford (1976) presented the first attempt more closely related, phylogenetically, to at distinguishing the major carnivoran clades. other carnivoran taxa than to each other (see Through an analysis of derived character especially Matthew, 1909, p. 354). This study states in the Carnivora he proposed a test- analyzes the phylogenetic relationships of able hypothesis of relationships for many the early Tertiary carnivorans to other car- groups of living and fossil carnivorans. This nivorans in greater detail than did prior stud- phylogeny of the Carnivora represents an im- ies, concentrating particularly on the distri- portant beginning in the advancement of our bution of shared derived dental features in understanding of the evolutionary history of these taxa. The results of this phylogenetic this group. Tedford's brief discussion of the analysis are then used to construct a cladistic relationships of Simpson's Miacoidea em- classification of the Carnivora. The inclusion phasizes the lack of knowledge of the distri- of the early Tertiary Carnivora in a detailed bution of derived characters within the Mi- phylogenetic analysis and vertical classifi- acidae and the essential need for an analysis cation of the entire Carnivora is very differ- of the phylogenetic relationships of the mi- ent from the treatment the "mi- acid genera with other Carnivora: "The fact acids" have received in earlier studies. We that very few miacoids have been identified believe that this is an important step in dis- as phyletically related to members of the cerning the precise relationship of the archa- modern superfamilies only increases the iso- ic to the living carnivorans. In any study an lation of the archaic and modern carnivore important distinction must be made between families .. .. This represents one of the larg- hypotheses of phylogeny and methods of est gaps in our knowledge of the phylogeny classification. We propose here a testable of the Carnivora" (Tedford, 1976, p. 364). hypothesis of phylogenetic relationships for Such a gap exists partly because of the the Carnivora, and then use this phylogeny 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 3 to construct a vertical classification that re- were particularly generous in providing ac- flects and is consistent with the proposed cess to specimens, equipment, funds, and phylogeny. moral support for this project. Mr. 0. Si- The impetus for the present study came monis patiently prepared the material of P. from the discovery of new carnivoran ma- (Protictoides) from a very stubborn and un- terial from the Uintan Tepee Trail Formation forgiving matrix. John J. Flynn was support- of northwestern Wyoming (see McKenna, ed by a Columbia University Graduate Fac- 1980; MacFadden, 1980). An attempt to de- ulties Fellowship during the course of this termine the precise taxonomic affinities of study. these dental specimens immediately pro- duced some interesting biostratigraphic and ABBREVIATIONS systematic results. All previous carnivorans AMNH, Department of Vertebrate Paleontol- from the North American Uintan had been ogy, American Museum of Natural History, referred to the Miacinae, as the Viverravinae New York. (except possibly Plesiomiacis, see the later USNM, United States National Museum, Smith- discussion of this genus) appeared to reach sonian Institution, Washington, D.C. extinction in the Bridgerian. However, even AC, Pratt Museum, Amherst College, Amherst, the limited material available from the Tepee Massachusetts. Trail Formation indicated that this form was truly a Uintan "viverravine," whose closest TERMINOLOGY relative was the Torrejonian genus Protictis. The dental terminology used in this paper Furthermore, it became obvious that since is shown in figure 1. Additional terminology the "Miacidae" was so clearly separable into not on this figure follows the usage of Van two morphologically distinct subfamilies Valen (1966), Maclntyre (1966) and Szalay (recognizable even on the basis of only lim- (1969). We have chosen not to use molar ter- ited dental material), the family was most minology for P4 morphology because we do likely a paraphyletic assemblage of generally not believe that there is a simple homology primitive carnivoran species. An analysis of between the features of the lower molars and the character distributions and the construc- premolars in the Carnivora. Because of the tion of a hypothesis of relationships for the uncertainty of developmental origins and "Miacidae" (and the "Viverravinae," in confusion about the primitive "molariform" particular) followed in an attempt to deter- or "premolariform" nature of the eutherian mine whether these were truly monophyletic P4, we prefer to use a descriptive premolar taxa. The possible relationship between var- ious "miacids" and more advanced Carniv- terminology. ora was also investigated at that time. This paper represents the conclusions drawn dur- MEASUREMENTS ing various stages of analysis of the series of An EPOI "Shopscope" and an IKL "Mi- problems that arose from the initial attempt crode" Precision Digital Positioner accurate to identify and to determine the phylogenetic to 0.001 mm. were used to make measure- relationships of the Tepee Trail carnivoran. ments and are recorded to 0.01 mm. Mea- surements follow the orientations described ACKNOWLEDGMENTS by Maclntyre (1966, p. 126). We thank E. Manning, N. Neff, A. Gold, J. Eaton, R. Cifelli, C. Tarka, L. Meeker, HISTORICAL REVIEW OF MIACID and Drs. G. Maclntyre, M. Novacek, R. CLASSIFICATION Tedford, and M. C. McKenna for many stim- Flower (1869, 1883) proposed the division ulating discussions, helpful suggestions, en- of the living Carnivora into the Arctoidea, couragement, and valuable criticism of this Cynoidea, and Aeluroidea. This arrangement manuscript. Drs. Tedford and McKenna has been followed, in one modified form or 4 AMERICAN MUSEUM NOVITATES NO. 2725

PAS

METW

A PAC - MEC

PACL - MECL

- TB

PCG PCNG PSCG AAC

D

B

FPAC

PAD MED

LAC C E FIG. 1. Dental terminology. Diagrammatic representation of carnivoran right P44 and M1', indicating the descriptive terminology used in this paper. A) Occlusal view of P4, B) lingual view of P4, C) occlusal view of P4, D) occlusal view of M1, E) occlusal view of M,. AAC-anterior accessory cusp; BTCG- basal trigonid cingulum; CN-carnassial notch; CO-cristid obliqua; END-entoconid; FPAC-first posterior accessory cusp; HYD-hypoconid; HYLD-hypoconulid; LAC-lingual accessory cusp; LCG-lingual cingulum; MC-main cusp; MECL-metaconule; MED-metaconid; METB-metastyle blade; METN-metastyle notch; METW-metastyle wing; PAC-paracone; PACL-paraconule; PACN-posterior accessory cusp notch; PAD-paraconid; PALD-paralophid; PAS-parastyle; PASW-parastyle wing; PCG-precingulum; PCNG-posterior cingulum; PRC-protocone; PSCG- postcingulum; PRD-protoconid; PRLD-protolophid; SPAC-second posterior accessory cusp; TB- talonid basin. Anterior toward left of page. Buccal toward top of page in A, C, D and E.

another, by most subsequent workers. Simp- sidered the morphology and systematics of son (1945) provided the most recent compre- the living, and some fossil, Carnivora in a hensive review of the various classifications phylogenetic framework. In that study Ted- proposed for living carnivorans.3 Tedford ford reviewed previous morphological and (1976), in his discussion of pinnipeds, con- classificatory studies of Carnivora and pro- posed a bipartite division into Caniformia 3 We follow Van Valen's (1969) suggestion that, in and Feliformia. order to avoid terminological ambiguity, the common Authors of previous studies of fossil Car- term "carnivoran" be used to distinguish members of nivora (Matthew, 1901a, 1909, 1915; Simp- the order Carnivora from other carnivorous mammals. son, 1937; Hough, 1948, 1953; Maclntyre, 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 5

1966) have invariably considered all Paleo- family Miacidae. The Viverravidae were cene and Eocene carnivoran genera as mem- considered members of a lineage distinct bers of the Miacidae. The Miacidae were from the Canidae, and assumed to be ances- erected by Cope in 1880a (p. 78) basically as tral to the living Viverridae, as proposed a grouping of all the primitive Camivora, and much earlier by Wortman (1886) and Flower included the genera Miacis and Didymictis and Lydekker (1891). (see table 1 for a detailed summary of the Wortman's (1901) study of the Eocene classifications discussed in this paper). Cope Mammalia from the Yale Peabody Museum believed in a horizontal scheme of classifi- Marsh Collection expanded on ideas pre- cation in which taxa of a generally primitive sented jointly with Matthew in 1899. Wort- nature were grouped with even more primi- man proposed and defined the carnivoran tive "stem-groups." Even if members of a suborder Carnassidentia, equivalent to the given taxon shared a few derived characters suborder Carnivora Vera of Flower and Ly- (synapomorphies) with, and were clearly re- dekker (1891). Wortman emphasized the im- lated phyletically to, more advanced forms, portance of dental characters in defining car- they were classified with more primitive taxa nivoran groups, and realized that the unfused because of greater overall primitiveness. The state of carpal elements was primitive, that Miacidae were therefore included by Cope it may occur in many different lineages, and (1880a, 1883) within the paraphyletic Creo- that its occurrence was therefore not indic- donta, which also included the Arctocyoni- ative of close phyletic relationship. Together dae, Leptictidae, Oxyaenidae, Amblyctoni- with the Creodonta and Pinnipedia, the Car- dae, and Mesonychidae. nassidentia formed a tripartite division of the Trouessart's (1885) catalogue of carni- Carnivora. Wortman (1901) further consid- vores listed the Miacidae within Cope's ered the relationships of the Eocene Canidae (1880a) Creodonta, and added a new subfam- and Viverravidae; more rigorously defined ily Miacinae. This first attempt at dividing the Viverravidae; named O5dectes as an the Miacidae below the familial level did not Eocene viverravid; distinguished the Viver- include a taxonomic diagnosis or discussion, ravidae from the contemporary Canidae, the nor was there another taxon proposed of successor Viverridae, and the Mustelidae; equivalent rank to the Miacinae. and criticized Matthew's (1901a) arrange- Flower and Lydekker (1891) divided the ment of the Creodonta. order Carnivora into the three suborders Matthew's (1901a, p. 59) classification of Carnivora Vera, Pinnipedia, and Creodonta. the Creodonta followed Schlosser's (1888) Wortman and Matthew (1899) considered subdivision of the group into Creodonta the phyletic relationships of living carnivo- Primitiva (including Oxyclaenidae), Creo- rans and determined that the ancestry of the donta Adaptiva (including Palaeonictidae, Canidae and Viverridae could be traced di- Viverravidae, and Arctocyonidae), and rectly back to the Eocene "Creodonta." Creodonta Inadaptiva (including Oxyaeni- They synonymized the genotype of Miacis dae, Hyaenodontidae, and Mesonychidae). Cope with Vulpavus Marsh, included most He included the Viverravidae in the Creo- of the species of Miacis under Uintacyon donta Adaptiva, and tentatively excluded Leidy, and concluded that Uintacyon was Vulpavus and Uintacyon from the Viverrav- ancestral to and should be included within idae and Creodonta by retaining them in the the Canidae. Wortman and Matthew also Canidae, but he also suggested that Vulpa- proposed a new family Viverravidae to in- vus and Uintacyon might later be placed in clude the species previously assigned to the Creodonta ifthey proved to have the creo- Didymictis Cope, which they synonymized dont tarsus. with Viverravus Marsh. By including Uin- Matthew (1909), in his extensive mono- tacyon (Miacis) in the Canidae and placing graph on the fossils of the Bridger Basin Viverravus (Didymictis) in a new family, Eocene, continued to utilize Schlosser's these workers effectively abandoned the (1888) and his own (1901a) concept of Creo- 6 AMERICAN MUSEUM NOVITATES NO. 2725

TABLE 1 Synopsis of Selected Portions of the Classifications Discussed in the Text

COPE, 1880 Palaeonictidae Creodonta All the living (fissiped) families (Vulpavus, Miacidae Uintacyon, Miacis within Canidae) Didymictis Creodonta Miacis Pinnipedia Arctocyonidae Leptictidae Oxyaenidae MATTHEW, 1909 Amblyctonidae Creodonta Mesonychidae Eucreodi TROUESSART, 1885 Arctocyonidae Miacidae Creodonta Miacinae Miacidae Cynoidei Miacinae Miacis, sensu stricto Miacis M. (Lycarion) Didymictis M. (Harpalodon) Viverravus M. (Prodaphaenus) Harpalodon Uintacyon, sensu stricto Ziphacodon U. (Miocyon) Triacodon Vassacyon Cercoleptoidei WORTMAN AND MATTHEW, 1899 Oodectes Creodonta Vulpavus Viverravidae Palaearctonyx Viverravus (including Didymictis) Viverravinae Carnivora Viverravus Canidae Didymictis Vulpavus (including Miacis, in part) Pseudocreodi Uintacyon (including Miacis, in part) Hyaenodontidae Etcetera Triisodontidae Creodonta, incertae sedis MATTHEW, 1901 Oxyclaenidae Creodonta Primitiva Oxyclaenidae Adaptiva GREGORY AND HELLMAN, 1939 Palaeonictidae Carnivora Viverravidae (Creodonta Cope, in part) Arctocyonidae Procreodi Inadaptiva Arctocyonidae (including Oxyclaenidae, Oxyaenidae Triisodontidae) Hyaenodontidae Acreodi Mesonychidae Mesonychidae Pseudocreodi WORTMAN, 1901 Oxyaenidae Carnivora Hyaenodontidae Carnassidentia Amphicreodi Viverravidae Palaeonictidae 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 7

TABLE 1-(Continued)

(Fissipedia; Carnassidentia Wortman) Feliformia Eucreodi Nimravidae Feloidea (Aeluroidea Flower) Felidae Viverridae (including Miacidae) Megantereontidae Herpestidae Machairodontidae Hyaenidae Caniformia Felidae Hyaenoidea (suborder) Musteloidea Viverridae Mustelidae Hyaenidae Canoidea (in part; Arctoidea Flower, in part) Canoidea Canidae Proaeluroidae (superfamily) Procyonidae Proaeluroidae Ursidae Quercygalidae (Pinnipedia) Ursoidae Otarioidea Procyonidae Phocoidea Canidae Ursidae SIMPSON, 1945 Agriotheriidae Carnivora Amphicyonidae Fissipeda Mustelidae Miacoidea Leptarctidae Miacidae Ailuridae Viverravinae Ailuropodidae Didymictis Caniformia, incertae sedis Ictidopappus Miacidae Viverravus Plesiomiacis Miacinae MACINTYRE, 1966 Uintacyon Carnivora Miacis Fissipeda Tapocyon Miacidae Vulpavus Viverravinae Vassacyon Protictis Oodectes P. (Bryanictis) Palaearctonyx P. (Simpsonictis) Pleurocyon P. (Protictis) Canoidea Ictidopappus Feloidea Etcetera KRETZOI, 1945 Creophaga (order) Pseudocreodi CRAY, 1973 Hyaenodontidae Miacidae Oxyaenidae Viverravinae Deltatheridioidea Quercygale Acreodi Didymictis Mesonychidae Ictidopappus Ptolemayidae Viverravus Triisodontidae Protictis 8 AMERICAN MUSEUM NOVITATES NO. 2725 donta Adaptiva and Inadaptiva in his subdi- subfamily designation for the Viverravinae vision of the Creodonta into the Eucreodi was new, whereas the use of Miacinae was (=Adaptiva; including Arctocyonidae and a resurrection of Trouessart's (1885) taxon Miacidae), Pseudocreodi and Acreodi (to- for a different and more restricted group of gether = Inadaptiva; including Hyaenodon- miacids. This arrangement brought together tidae, Oxyaenidae, Mesonychidae, and Ox- the primitive relatives of the later carnivo- yclaenidae). Although he retained the rans, but retained, somewhat, the concept of Eucreodi within the Creodonta, Matthew an early split of the living forms into two subsequently (1909, pp. 330, 339-340) stated distinct lineages. Matthew considered inval- that the Miacidae, only, were definitely an- id the synonymies of Didymictis with Vi- cestral to later Fissipedia. Furthermore, verravus and Miacis with Vulpavus (as pro- Matthew (1909, p. 354) suggested that the posed by Wortman and Matthew, 1899, and Viverravinae and Miacinae may have been followed by Wortman and Matthew 1901a). ancestral to the Aeluroidea and Arctoidea, In his 1915 revision of the fauna of the respectively, although they remained grouped Wasatch and Wind River Eocene, Matthew in the same family as an expression of their keyed the families of Creodonta (p. 5), the recent common ancestry and shared primi- genera of Miacidae (p. 18) and the species of tive morphological structure. Didymictis (p. 20), and considered the phy- Although Matthew (1901a, 1909) followed letic affinities and stratigraphic distributions Cope's classification of the Miacidae in the of the Miacidae. Creodonta, he dealt more explicitly with the The monograph by Matthew on the Paleo- philosophical considerations of horizontal, cene of the San Juan Basin, published post- gradal classifications, as is seen in this pas- humously in 1937, included further discus- sage: "It will be observed that the definition sion and definition of the Creodonta, of the suborder Creodonta is in accord with Miacidae, and Didymictis. The classification the general views of its author upon its clas- and ideas presented there differed little from sification and phylogeny, according to which those elaborated in 1909 and 1915, although the suborder represented a structural group, Matthew did propose a new subgenus Pro- defined by the possession of certain common tictis for Didymictis haydenianus. characters indicative of relationship. The al- The arrangements presented in the 1909, ternate theory of classification, that an order, 1915, and 1937 papers were more consistent family or other division represents a phylo- with Matthew's gradistic philosophy than genetic line, including animals often of was the one proposed in 1901a. The two clas- widely different structure but which in the sifications of Matthew and Wortman are opinion of the classifier are genetically con- strikingly different, and emphasize the meth- nected, will of course involve a radical odological and philosophical split between change in our conception of the Creodonta, the two. Matthew's classification was essen- regarding the inadaptive forms as pseudo- tially horizontal and stressed the retention of Carnivora, and transferring the adaptive a majority of primitive characters by the Vi- group to the Fissipedia" (Matthew, 1909, p. verravidae, and possibly the recent common 320). ancestry between this family and the remain- Matthew (1909) also resurrected the Mi- der of the Creodonta, even though he re- acidae as a diagnosable family of primitive, tained the Eocene genera Uintacyon and early Tertiary carnivorans classified within Vulpavus within the Canidae. Wortman's the Eucreodi of the Creodonta. Uintacyon scheme was more vertical in emphasis; he and Vulpavus were removed from the Can- included the Viverravidae with more ad- idae and were considered members of the vanced carnivorans on the basis of shared Miacidae. Matthew (1909) proposed and di- derived dental morphology, in an attempt to agnosed the two subfamilies Viverravinae distinguish the course of phyletic diversifi- and Miacinae, for the Viverravidae and the cation. The inclusion of the primitive Uin- Eocene "Canidae," respectively. The tacyon and Vulpavus in the more advanced 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 9

Canidae was more consistent with Wort- ory and Hellman paper suggested the radical man's methodology than with Matthew's. proposals of a subfamily Miacinae of the Simpson described a new viverravine ge- Viverridae, which included the genera pre- nus, Ictidopappus, and two new species of viously contained in Matthew's Miacinae, Didymictis, D. tenuis and D. microlestes, whereas the Viverravinae were abandoned from the Fort Union Paleocene of Montana and Viverravus and Didymictis were placed in 1935. Simpson later (1937) considered the within the subfamily Viverrinae (with the liv- relationships of these species, as well as oth- ing civets). Viverravus, Didymictis, and Ic- er Viverravinae, and provided diagnoses for tidopappus were believed to be very close the Viverravinae, Didymictis, and Ictido- relatives of the living Viverridae, and the pappus. His arrangement of the Viverravi- Miacinae were considered to be more distant nae followed that of Matthew (1909), where- allies, but both groups were thought to be as his discussion of Ictidopappus suggested closer to the ancestry of the Feloidea Simp- that this genus might be ancestral to Viver- son (=Aeluroidea Flower) than to the Mus- ravus or that Viverravus might be an off- teloidea Gregory and Hellman and Canoidea shoot of Didymictis. Simpson. The concept of the Miacidae had thus re- The most extensive, recent consideration mained as espoused by Matthew (1909, 1915, of the Miacidae was in Simpson's (1945) clas- 1937) and followed by Simpson (1935, 1937), sification of the mammals. This arrangement until 1939 when Gregory and Hellman pub- followed Matthew in recognizing the Miaci- lished a study on the evolution and classifi- nae and Viverravinae as subfamilies of the cation of the Viverridae. That paper pre- Miacidae. The family Miacidae was the only sented a radical departure from previous member of a superfamily Miacoidea (equiv- classifications and notions of relationships. alent to Matthew's Eucreodi). Together with Gregory and Hellman adapted Matthew's the superfamilies Canoidea (=Arctoidea plus classifications of the Carnivora by dividing Cynoidea of Flower, 1869) and Feloidea the order into three suborders and five in- (=Aeluroidea of Flower, 1869) the Miacoi- fraorders-the Procreodi, Acreodi, Pseudo- dea formed a coordinate, tripartite division creodi, and a new Amphicreodi (within the of the Fissipeda. Such an arrangement rep- Creodonta); and an expanded concept of Eu- resented an important advance in the inter- creodi (including the suborder Fissipedia = pretation of the phylogeny of the Fissipeda. Carnassidentia, and the suborder Pinnipe- The classification of the Miacidae within this dia). The Eucreodi were classified as an in- group explicitly recognized the closer rela- fraorder of the Fissipedia, although the text tionship between the Miacidae and the more (Gregory and Hellman, 1939, p. 382) states advanced Carnivora than between the Mia- that the Eucreodi could be expanded to in- cidae and Creodonta. An assumption of di- clude both Fissipedia and Pinnipedia, since rect phyletic continuity between the Miaci- "all the families of modern Carnivora were dae and other Fissipeda was proposed, and derived from one common ancestral stock removal of the Miacidae from a position ...." The suborders were referred to par- within the Creodonta allowed for the possi- enthetically in the Gregory and Hellman bility that the Miacidae might not share a classification, and it is unclear whether this close phylogenetic history with the Creodon- was for historical reference or for actual ta. taxonomic utilization. Gregory and Hellman Simpson's (1945) discussion of the Car- rejected the conservative horizontal system nivora (including Creodonta, Fissipeda, and of classification used by Matthew, and they Pinnipedia) provided a selected history of the believed that their arrangement better re- classificatory arrangements utilized until flected a phylogenetic approach to classifi- 1945, and a consideration of the philosophy cation. The concept of the "Miacidae" was he used in constructing the classification pre- retained; however, it was placed in its en- sented therein. Simpson dealt only briefly tirety into the family Viverridae. The Greg- with most of the previous classifications, but lo AMERICAN MUSEUM NOVITATES NO. 2725 considered those of Matthew, and Gregory nomic change that did not alter the previous and Hellman in more detail. Most of Simp- concepts of miacid phylogeny. son's work on the Miacoidea, and much of The only taxonomic study of the Miacidae his classification of the Creodonta was ac- that Simpson (1945) discussed in any detail knowledged as being based heavily on Mat- was that of Gregory and Hellman (1939), pri- thew's numerous and extensive treatments marily because it proposed the most signifi- of the Carnivora. Gregory and Hellman's cant change in the contemporary systematic (1939) arrangement of the Carnivora was usage. The suggestion that the Miacidae be considered the only radical departure from included within the Viverridae was consid- standard classifications, and was therefore ered radical, and was rejected by Simpson. thoroughly analyzed by Simpson. Simpson believed that only primitive char- The threefold division of the Carnivora, as acters allied the miacids with the Viverridae, used by Simpson and most previous work- and the group as a whole could be "struc- ers, was replaced by a fivefold division by turally ancestral" (Simpson, 1945, p. 221) to Gregory and Hellman (1939), as described the Feloidea and Canoidea.4 above. Simpson accepted this as both logical Maclntyre (1962) briefly discussed the and theoretically acceptable as a statement morphology and systematics of the Miaci- of carnivoran phylogeny. He believed, how- dae, and named a new genus, Simpsonictis, ever, that the tripartite division was equally based on material of Didymictis tenuis Simp- consistent and acceptable, more convenient, son. His discussion pointed out that the Mi- and more familiar. "Logic in classification is acidae were generally considered the primi- not solely concerned with making coordinate tive stock for the modern Carnivora. The units of like antiquity or arising from the earliest miacids were members of the Viv- same dichotomy-indeed this ideal is im- erravinae (although they were derived in the practical. There are other essential logical loss of M3), but the Miacinae probably rep- considerations, for instance, the desirability resented a closer approximation to the mor- of approximate balance as to evolutionary phology of the ancestral miacid stock. The status and distinctness .... This is a case Viverravinae consisted of Viverravus, Di- where a compromise between horizontal and dymictis and Ictidopappus. The description vertical classification is usual and seems to and discussion of the genus revealed that me advisable if not necessary" (Simpson, Simpsonictis could be considered a viverra- 1945, p. 218). vine with various insectivorous dental spe- Simpson, like Matthew, stressed the im- cializations, but with similarities to smaller portance of compromise in favor of horizon- species of both Viverravus and Didymictis. tal classification. Arrangements which em- Later Maclntyre (1966) published a thor- phasized the vertical aspects of phylogeny ough and extensive review of the systematics and classification were lauded, but ultimately of Ictidopappus and Protictis as the first part rejected. of an intended series on the evolution of the Simpson's (1945) discussion of the Mia- Miacidae. MacIntyre raised Matthew's sub- coidea points out the dichotomy between genus Protictis to generic rank, and divided those that would classify the Miacidae hori- zontally with the Creodonta (including Cope, Schlosser, Matthew, and Simpson prior to 4Although the text reads "I prefer to express the re- 1945), and those that included the Miacidae lationships by placing the Miacidae in a separate super- within the Fissipeda (Winge, Wort- family from which both Feloidea and Viverroidea vertically arose" (Simpson, 1945, p. 222), there is no such super- man, Scott, Osborn, and Simpson in 1945). familial taxon as Viverroidea in Simpson's classification Simpson here placed the Miacidae in the Fis- (or any other author, as far as we are aware). We presume sipeda because he believed it was the ar- that since Simpson was discussing the relationship of rangement "increasingly favored by my the Miacidae to the Viverridae (Feloidea) and Canidae [Simpson's] colleagues" (p. 221), although (Canoidea) in the preceding sentences, he mistakenly he believed that this was only a formal taxo- used the term "Viverroidea" instead of Canoidea. 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 11 this genus into three subgenera-Protictis member of the Viverravinae and placed the (Protictis), P. (Simpsonictis), and P. species in the genus Viverravus. In 1945 (Bryanictis). The designation of Simpsonic- Kretzoi erected a new genus, Quercygale, tis as a subgeneric member of Protictis, and for "Viverravus" angustidens. Apparently the erection of a new subgenus P. (Bryan- unaware of the paper by Kretzoi, de Beau- ictis) to include "Didymictis" microlestes mont (1965) designated "V." angustidens as and a new species P. (Bryanictis) vanvaleni, the type species of a new genus, Humbertia, greatly broadened the concept of the taxon and named another species, H. helvetica. Protictis and restricted the scope of Didymic- The next year de Beaumont (1966) realized tis. that Humbertia was synonymous with Kret- Maclntyre (1966) considered the relation- zoi's Quercygale. Nevertheless, de Beau- ship of the Miacidae to the Creodonta and mont's 1965 paper included an excellent Fissipeda, and favored the vertical classifi- comparative chart of the many differences cation of the family with the Fissipeda, as distinguishing "Humbertia" from "Viver- advocated by Simpson (1945). Even though ravus"; proposed the hypothesis that "Hum- he stated that he believed a vertical classifi- bertia" was derived from North American cation best portrayed the phyletic relation- Didymictis, rather than Viverravus; and rec- ships of the Miacidae, he did not attempt to ognized that "Humbertia" was very similar analyze or organize the members of the para- morphologically to Tapocyon. phyletic Miacidae using a vertical or phylo- Van Valen (1967) included a brief refer- genetic methodology. ence to the relationship of Quercygale in his Maclntyre (1966, pp. 195-205) in his gen- broader paper on the classification of the in- eral discussion listed characters that could sectivorans. Van Valen felt that Quercygale be hypothesized as primitive in the common should be considered a subgenus of, or mem- ancestor of the Miacinae and Viverravinae. ber of the genus, Tapocyon. The lack of an Many of these are probably primitive for eu- M3 in a miacine (Quercygale) led Van Valen therians, in general, although with further to conclude that the Viverravinae could not analysis some of the dental and cranial fea- be unambiguously distinguished from the tures might be determined to be derived for Miacinae (which he presumably differentiat- the Miacidae (and possibly other Carnivora, ed only on the presence or absence of M3). as well). Following that section, Maclntyre Van Valen also considered Kretzoi's (1945) considered the actual possibilities for miacid Quercygalidae a synonym of the Miacinae. ancestry in the known Cretaceous taxa (p. In his discussion of the origins of placental 203). The problem of the interpretation of P4 carnivores, Van Valen (1969) compared the morphology was thoroughly and accurately morphology of various "miacids" with the discussed, but no definitive conclusion was Cretaceous genera Protungulatum, Cimo- drawn as to whether a simple or a "molari- lestes, and Procerberus. He also erected a form" P4 was primitive for the Carnivora. A new tribe Ictidopappini within the Miacidae simple P4 had been assumed primitive for to include Ictidopappus, and resurrected therians by most workers, and Maclntyre Simpsonictis as a valid genus distinguishable provisionally accepted this convention for from Protictis. the ancestral carnivoran. Under that scheme An analysis of the fauna from the Upper Cimolestes appeared to exhibit the closest Beds of approximation to the expected ancestral Eocene Headon England by Cray morphology for all later Carnivora. (1973) included a discussion of Quercygale A number of other shorter papers have and the addition of "Viverra" hastingsae been written which considered the relation- Davies as a species of Quercygale. Cray dis- ships of various genera of supposed mem- regarded Van Valen's suggestion that Quer- bers of the Viverravinae. cygale be considered a subgenus of the mi- Teilhard de Chardin (1915) recognized acine Tapocyon by retaining Quercygale as "Viverra" angustidens from France as a a distinct genus within the Viverravinae. 12 AMERICAN MUSEUM NOVITATES NO. 2725

A PROTICTIS (PROTICTIS) G P. (PROTICTOIDES)

B P. (BRYANICTIS) H MIACI S

C DIDYMICTIS I UINTACYON

C. C.)

D VIVERRAVUS J VULPAVUS

E SIMPSONICTIS K OODECTES

F ICTIDOPAPPUS L PROCYNODICTIS

17..

FIG. 2. Comparison of lower dentitions of early Tertiary Carnivora. The first molar of each series is drawn to approximately the same size, and the dentitions are aligned vertically along the anterior edge of the first molar. The species and specimen numbers used as representatives for the genera are: A) P. (P.) haydenianus, AMNH 3371, 16540; B) P. (B.) microlestes, AMNH 11521, USNM 9302; C) D. protenus, AMNH 2831; D) V. sicarius, AMNH 11521; E) S. tenuis, AMNH 35349 and 35350, USNM 9097; F) I. mustelinus, USNM 9296; G) P. (P.) aprophatos, AMNH 105371; H) M. parvivorus, AMNH 11496; I) U. vorax, AMNH 11507, 12620; J) V. profectus, AMNH 12626; K) 0. herpestoides, AMNH 11495; L) P. vulpiceps, AMNH 2514. This figure is partially adapted from Matthew, 1909, p. 347. Buccal view of right dentitions; anterior toward right of page. Specimens A, B, E (in part) and L have been reversed during drawing. 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 13

G MIACIS A PROTICTIS (PROTICTIS)

H UINTACYON B P (PROTICTOIDES) I\'

C P. (BRYANICTIS) I OODECTES

\NOAJ ~-' O

D DIDYMICTIS J VULPAVUS

A-0 E VIVERRAVUS K PROCYNODICTIS''K:"

F ICTIDOPAPPUS

FIG. 3. Comparison of upper dentitions of early Tertiary Carnivora. The first molar of each series is drawn to approximately the same size, and the dentitions are aligned vertically along the anterior edge of the first molar. The species and specimen numbers used as representatives for the genera are: A) P. (P.) haydenianus, AMNH 3368, 16540; B) P. (P.) aprophatos, AMNH 97701; C) P. (B.) mi- crolestes, AMNH 35361, 35362; D) D. protenus, AMNH 2831; E) V. sicarius, AMNH 11521; F) I. mustelinus, USNM 9295; G) M. parvivorus, AMNH 11500; H) U. vorax, AMNH 11507, 12620; I) 0. herpestoides, AMNH 11495; J) V. profectus, AMNH 12626; K) P. vulpiceps, AMNH 2514. This figure is partially adapted from Matthew, 1909, p. 344. Crown view of right dentitions; anterior toward right, buccal toward top of page. Specimens A, C, F, and K have been reversed during drawing. 14 AMERICAN MUSEUM NOVITATES NO. 2725

PREVIOUS DEFINITIONS OF THE pressed" (Cope, 1883, pp. 304-305; 1875a, p. MIACIDAE 11; and 1877, p. 123). Wortman and Matthew's (1899) proposal The previous classifications of the Miaci- of a new family Viverravidae stressed the dae are thoroughly discussed in the Histori- similarity of the members of this family to cal Review section of this paper. However, the Viverridae in the basicrania, dental for- consideration of the characters used by var- mula and pattern, atlas and vertebral artery, ious workers to define the Miacidae and its and limb structure. They considered distinc- subdivisions is interesting and useful to the tive the deep vertical notch in the metastyle present categorization of the various "mi- blade of P4 and the presence of a large P4 acid" taxa. Those authors did not distinguish parastyle, in both families. The Viverravidae between derived and primitive features in were distinguished from the creodonts by the their taxonomic definitions. The original def- presence of P4/M1 sectorial, from "early initions and characterizations of relevant Canidae" (=Miacinae of Matthew, 1909) by taxa are presented in this section, whereas the parastyle on P4 and presence of only two features uniquely derived and diagnostic for molars, and from the Viverridae by the re- various carnivoran taxa are discussed in tention of the primitive free scaphoid, lunar more detail later in this paper. Figures 2 and and centrale. 3 provide comparative drawings of the den- Wortman (1901) included the Viverravidae titions of representative genera of early Ter- and Fissipedia (including the "miacines") in tiary Carnivora. Many of the features dis- the carnivoran suborder Carnassidentia. His cussed here and in the systematics section definition of the suborder was carnassial below are also illustrated. teeth present and always consisting of P4/M1; In Cope's (1883) discussion of the Creo- scaphoid, lunar, and centrale generally donta he defined the Miacidae as creodonts united; ungual phalanges compressed and with tubercular upper molars, sectorial last pointed. The Viverravidae were character- upper premolar, and a tubercular sectorial ized by the lack of a hypocone and a molar first lower molar. He subdivided the Miaci- formula of either 3/3 or 2/2 (since Oodectes, dae into the genera Miacis and Didymictis with three molars, was considered a viver- and characterized them as having, respec- ravid), in addition to the characters cited tively, two inferior tubercular molars and above from Wortman and Matthew (1899). four premolars, and one inferior tubercular The Viverravidae were distinguished from molar and four premolars. The discussion of the contemporary "Canidae" by the perfo- Miacis included this description: lower den- ration of the transverse process of the atlas tition consisting of I? Cl PM4 M3; single- for the passage of a vertebral artery and a rooted first premolar; M1 with a broad heel, well-developed P4 parastyle, both of which one edge submedian and a little elevated were considered "entirely distinctive" (p. above the other; and M2.3 tubercular with 143) characters. Furthermore, the lack of conic tubercles anteriorly and a short heel. curvature in the femur, and the total absence The description of Didymictis included: "In- of a hypocone on the upper molar differen- ferior molars six, consisting of four premo- tiated the viverravids from the canids. lars and two true molars. True molars, a Matthew (1901a) considered the Palaeo- posterior tubercular, and an anterior nictidae, Arctocyonidae, and Viverravidae tubercular-sectorial, i.e., with three elevated as members of the Creodonta Adaptiva de- cusps and a posterior heel. Premolars with fined by P4/M1 carnassial, when present; a lobe behind the principal cusp. The canine claws of modern type and carried free of the teeth are directed forwards, and are very ground; and scapholunar-centrale early be- close together, so that it is doubtful whether coming united. In the Viverravidae the post- there were any incisors. An ungual phalange carnassial teeth were assumed to be "be- of the typical species is strongly com- coming tubercular" (Matthew, 1901a, p. 59). 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 15

The extensive monograph by Matthew the only family placed in the Eucreodi, were (1909) provided a good deal of information distinguished by the lack of fissuring in the on miacid taxonomy. Matthew included the ungual phalanges, and were characterized by Miacidae and Arctocyonidae in the Eucreodi the carnassials being P4 and M1 (sic-we as- of the Creodonta. The Miacidae possessed sume he is referring to M1, and that this is the following creodont characters: scaphoid, just a printer's error in the publication). lunar and centrale generally distinct; third Within the Miacidae the genera were split trochanter on the femur; entepicondylar fo- into two major groups corresponding to the ramen on the humerus; small brain; unossi- Viverravinae and Miacinae, although the two fied tympanic bulla; and toes 5-5. The diag- groups were not explicitly named in the key. nostic characters of the family Miacidae One group (Viverravinae) included Didy- included P4/M1 carnassials as in modern Car- mictis and Viverravus and was distinguished nivora; postcarnassial teeth generally tuber- by the prominent antero-external cusp of P4, cular; metacone smaller than protocone; un- the elongate oval shape of M1, and the pres- gual phalanges compressed and unfissured at ence of two upper and lower molars. The the tips; and paraxonic digit symmetry. other group (Miacinae) contained Uinta- Within the Miacidae, the members of the cyon, Miacis, Vassacyon, Oodectes, Vulpa- subfamily Viverravinae were distinguished vus, and Palaearctonyx and was distin- by the molars reduced to 2/2; large antero- guished by the absence or weak development external cusp (parastyle) on P4 and posterior of the antero-external cusp of P4, the short- accessory cusp on P4; elongate skull; pres- oval or round shape of M2 and M3, and the ence of a calcaneal fibular facet; and other presence of three upper and lower molars. features of the limb morphology. In contrast, Simpson's (1945) new superfamily Miacoi- the Miacinae had a full molar complement of dea did not include any formal diagnosis or 3/3; the antero-external cusp of P4 either ru- description of the taxon, except a statement dimentary or absent; skull length moderate that it was proposed to include the Miacidae, or short; absence of a calcaneal fibular facet; the primitive precursors of the later Carniv- and other differences in limb structure. The ora. Miacinae were further divided into Cerco- All the workers cited above have pre- leptoidei and Cynoidei; the first had the car- sented morphological characters that may nassial only weakly differentiated, and the ally the Miacidae with the more advanced molars were all of similar form, whereas the fissiped Carnivora. Furthermore, they all di- second group had well-differentiated molars chotomously divided the Miacidae into two and tubercular posterior molars. Matthew major groups. Each of these groups was be- regarded the split of the Miacidae into the lieved to be ancestral to one of the two major Miacinae and Viverravinae as the beginning branches of living carnivorans by some of the differentiation into the arctoid and ae- workers. However, all the above studies re- luroid Carnivora, respectively. The Cynoidei tained all the early Tertiary Carnivora within were further believed to be the precursors of a single, horizontally arranged family. The the Cynoidea, whereas the Cercoleptoidei members of this family Miacidae were either were considered ancestral to the Arctoidea classified with, and considered ancestral to, proper. the Fissipeda, or placed within the Creodon- Matthew's (1915) paper included keys to ta because of their overall morphological the families of Creodonta, genera of Miaci- similarity to and supposed recency of com- dae, and species of Didymictis. The key for mon ancestry with other Creodonta. We be- the Creodonta included the divisions Pro- lieve that the family Miacidae is a paraphy- creodi, Eucreodi, Pseudocreodi, and Acreo- letic grouping of generally morphologically di, distinguished by differences in the ungual primitive but not closely related taxa. The phalanges, and further characterized by fea- taxa previously included within the Miacidae tures of the carnassials. The Miacidae were have clear phyletic relationships with the lat- 16 AMERICAN MUSEUM NOVITATES NO. 2725 er, more advanced Carnivora. Therefore, we tinct cusps, precingulum and postcingulum propose abandonment of the use of the fam- not continuous around lingual base of pro- ily Miacidae to include all primitive, early tocone; small P4 anterior accessory cusp; P4 Tertiary Carnivora. Furthermore, the use of lingual accessory cusp absent; entotympanic the Miacidae, or the term miacid, as a waste- bulla unossified; scaphoid, lunar, centrale basket taxon in future taxonomic, phyloge- separate; calcaneal fibular facet present; netic, and evolutionary studies is discour- scapula without teres major process or sec- aged. The retention of a non-monophyletic ondary scapular spine. taxon should be avoided, and we propose a DIscuSSION: Although Flower (1869) lim- hypothesis of relationship for all the taxa ited the concept of the group Carnivora to previously included within the Miacidae. the living Aeluroidea, Arctoidea, and Cy- Concise, explicit, and testable definitions of noidea, the earliest workers to consider the taxa and hypotheses of relationships are es- relationships of fossil taxa included the Creo- sential to the understanding of the evolution- donta as a suborder of the Carnivora (for in- ary history of the Carnivora. stance, Cope, 1880a; Trouessart, 1885). Van Valen (1966) and Tedford (1976) have re- SYSTEMATICS cently excluded the Creodonta from the Car- nivora; we agree with this exclusion and re- ORDER CARNIVORA BOWDICH, 1821 strict the concept of Carnivora to those taxa INCLUDED TAXA: Feliformia and Canifor- previously included within the Miacidae, mia. Fissipeda, and Pinnipedia. The construction DISTRIBUTION: ?Early Paleocene to Re- and definition of a monophyletic taxon, Car- cent. nivora, that includes both the living and fos- DIAGNOSIS: Eutherian mammals differing sil "miacids," "pinnipeds," and "fissipeds," from all other mammals in the following de- but excludes the Creodonta and all other rived characters-P4/M1 modified as the prin- mammals, is very difficult. It is essential, cipal carnassial teeth, lacking a migratory lo- however, to construct monophyletic groups cus for the carnassial as found in the that are explicitly and rigorously diagnosed Creodonta, but with retention of some sub- and are readily subject to critical review by sidiary shear on other teeth; P4 protocone other workers. Tedford's (1976) consider- located anterolingually, far forward of the ation of the relationships of the carnivorans paracone; extremely elongate P4 metastyle was an important first attempt at such an ar- wing with a well-developed metastyle blade rangement. However, his diagnoses of the and carnassial notch; well-developed, elon- order Carnivora, and suborders Feliformia gate P4 talonid with at least two posterior and Caniformia excluded the members of the accessory cusps; M2-3 reduced in size; Miacidae from a phyletic position within processus hyoideus formed by a ventrome- these groups, as all of the known miacids dial prolongation of the squamosal. appear to lack an ossified entotympanic. In Primitive features for the Carnivora: reten- the above diagnosis, we characterize the or- tion of Ml-3: P4 with a small, but distinct der Carnivora in such a way that the taxon parastyle cusp; P3 lacking a protocone; all remains a monophyletic assemblage that in- molar talonids short, basined, with a large cludes the Miacidae, but excludes the Creo- hypoconid as the highest talonid cusp, and donta. For a graphic representation of the an obliquely oriented cristid obliqua; molar phylogeny hypothesized for all Carnivora, trigonids high, with the paraconid < meta- see figure 4. conid < protoconid size; molar paraconids The phylogeny and taxonomic diagnoses and metaconids moderately closely ap- presented in this paper were determined by pressed, resulting in a closed-V between the a cladistic analysis of the entire Carnivora, paralophid and protolophid in the trigonid; and for lower level taxa within the Carniv- upper molars with both a paraconule and ora. The primitive to derived polarity of metaconule; molar hypocones absent as dis- the characters used in this analysis was ini- 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 17

C-,~~~~~~~~~~~~~~~~~~~~~~~~~,

oC q, C- C., 0 a h0 *11

PROTICTIS

DIDYMICTIDA

FELIFORMIA

CARNIVORA * 1 FIG. 4. Proposed phylogenetic relationships of the Carnivora. The numbers at the nodes refer to taxa and character distributions summarized in the Phylogeny section of this paper (see Systematics section for detailed discussions). Genera and species are in lightface type and higher level taxa are in boldface type. tially determined through comparison to the contradicted hypothesis of phylogeny. In character distributions in a number of pos- cases in which polarity assignments resulted sible outgroup taxa (including Creodonta, in character distributions that were ex- Metatheria, Cimolestes, Didelphodus, Pro- tremely incongruent, we considered the orig- cerberus, Gypsonictops, Protungulatum, inal hypotheses of polarity refuted rather Avunculus, Purgatorius, Leptonysson, Didy- than accepting numerous independent acqui- moconus, and others). sitions or losses for these features. Following the generation of an initial hy- Occasionally, two character states found pothesis of phylogeny based on these char- within the Carnivora will also be found in acter distributions, we analyzed the congru- outgroup taxa; in such cases, the conflict is ence of polarity determinations and resolved by restricting the outgroup (gener- distribution of characters, with the phyloge- ally to Cimolestes and the Creodonta under ny. Some of the initial polarity assignments the testable assumption that these taxa may resulted in character distributions that were be more closely related to the Carnivora than incongruent with the phyletic arrangement of are the other outgroup taxa) or by internal taxa supported by the distribution of the ma- parsimony (minimizing the number of char- jority of the other characters. The most par- acter conflicts between taxa within the Car- simonious arrangement of character distri- nivora). These conflicts emphasize the need butions among taxa was accepted as the least for more intensive analysis of possible char- 18 AMERICAN MUSEUM NOVITATES NO. 2725 acter reversal or nonhomology (conver- nictops (AMNH 105158), Didelphodus gence) in these features. Our final hypoth- (AMNH 4228), Cimolestes (AMNH 101992, eses of character polarity, character 105156, 105152), and the Creodonta (e.g., distribution and phylogeny are presented be- Proviverra, AMNH 13142) the protocone is low, and any contradictions of the a priori directly medial, or slightly anterolingual to polarity assignment based on outgroup com- the paracone. The extreme anterior position parisons are noted in the text. of the protocone of P4 is derived, and is char- Watrous and Wheeler (1981) provide a acteristic of primitive Carnivora. good formalization of a similar method of The extreme elongation of the metastyle determining character polarity and analyzing wing on P4 and concurrent development of character distributions in a cladistic analysis. an elongate shearing blade and distinct car- We particularly agree with their (1981, p. 9) nassial notch is derived for the Carnivora. belief that cladistic analysis is generally a This distinctive metastyle morphology is step-by-step process in which many charac- found in the molars of the Creodonta in con- ter polarity and distribution determinations nection with the carnassial function of those are dependent on preceding polarity deci- teeth. The metastyle of P4 in the Creodonta sions and taxonomic clustering (clustering is also modified, but to a lesser degree, by based on other character distributions-their slight elongation into a shearing blade and "Functional In-Group (FIG)/Functional Out- development of a shallow carnassial notch. Group (FOG)" concept). This character may ally the Creodonta to the The development of P4 and M1 into the Carnivora, as no other mammals elaborate principal shearing pair, or carnassials, is the metastylar region of P4 in the same man- highly modified from the condition found in ner, or to the same extent. Cimolestes and creodonts and various "insectivores." In Didelphodus elongate the metastyle to a other sectorial mammals, shear may be de- slight to moderate degree, but do not develop veloped on all molars and some premolars, a distinct blade and notch. The Carnivora or isolated on a single carnassial pair. In var- develop a metastyle blade that is more elon- ious true insectivores and Cretaceous euthe- gate and a notch that is deeper than that rians, shear is distributed throughout the found in any of the Creodonta or the Creta- cheek tooth row, without the isolation of a ceous eutherians. single carnassial pair of teeth (in particular, An elongate, well-developed talonid on P4 P4/M1). The Hyaenodontidae and Oxyaeni- is here considered a derived character unit- dae in the Creodonta develop principal car- ing all Carnivora, and therefore is primitive nassial pairs of MI and MI, respectively, for the taxa included within the order. The although subsidiary shear is retained on the talonid is relatively broad, rimmed poste- more anterior teeth. Furthermore, in the riorly and lingually by a strong cingulum, and Creodonta, these teeth represent only the there are two posterior accessory cusps pres- terminal pair of carnassials, in a sequence in ent along the labial ridge running from the which the primary shear locus moves pos- posterior margin of the main cusp. The pres- teriorly during ontogeny (Mellett, 1977). In ence of these two posterior accessory cusps the Carnivora, the sole carnassial pair is results in the formation of two distinct notch- uniquely developed as P4/M1 and shear is es, between the main cusp and the first pos- maintained at this locus only, throughout terior accessory cusp, and between the first life. and second posterior accessory cusps (see The protocone on P4 of carnivorans is lo- fig. 1 for premolar terminology). A small cated far anterolingually on the tooth crown, cusp or cuspule may have been developed far forward of the apex of the paracone, and from the posterior cingulum in the earliest frequently medial or anterolingual to the Carnivora, but the patchy distribution of this parastyle. In Didymoconus and Kennalestes character throughout the Carnivora makes it the protocone of P4 is located posterolingual unclear whether this is a derived feature for to the paracone apex, whereas in Gypso- the entire Carnivora, or whether it is evolved 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 19 independently as an advanced character sonictops M1_3 are approximately equal in within some carnivoran taxa. size, although M1 is slightly smaller than M2 A simple P4 has been hypothesized as in both genera, and the trigonids are high on primitive for carnivorans (see discussion in all the molars. Procerberus and Cimolestes Maclntyre, 1966, p. 204), particularly in re- have M1 approximately equal in size to M3, gard to the trigonid of P4. Outgroup compar- both of which are slightly smaller than M2 in ison did produce an a priori hypothesis that Procerberus and slightly larger than M2 in a simple P4 is primitive. An elaborate P4 tal- Cimolestes. In Protungulatum and Purga- onid is not found in any of the outgroups torius the M1 and M2 are about equal in size, Cimolestes, Didelphodus, Leptonysson, or and in both genera these teeth are smaller the Creodonta; the talonids are very short, than M3. Members of the Paenungulata also narrow, and generally have only a single, primitively have M1 < M2 < M3 size. The posteriormost cingulum cusp. The arrange- extreme reduction of M2-3 size (relative to ment of the cusps on the talonid of P4 in car- MI) in Carnivora is obviously derived over nivorans lacks the structure normally devel- the primitive condition exhibited by the taxa oped in the molar talonids, as they are mentioned above. aligned in a straight, or slightly arcuate, All the Carnivora may be united by the anteroposterior line and are without a simple development of a processus hyoideus as a analogy to the molar cusps. In the earliest process of the squamosal, which extends Carnivora P4 was elaborated, particularly in ventromedially to contact the promontor- the talonid, but it was not a molariform ium. This squamosal process forms the an- tooth. terior margin of the stylomastoid foramen, However, the results of this study indicate and the facial nerve exits the auditory region that an elaborate P4 talonid may be the prim- posterior to the process. A squamosal pro- itive state for Carnivora. The elaborate tal- cessus hyoideus is found in the primitive car- onid structure described above is found nivorans Viverravus, Vulpavus, and Didy- widely in the early Tertiary carnivorans pre- mictis and in more derived Carnivora, viously assigned to the "Miacidae" in both whereas in most other mammals the proces- the "Viverravinae" (members now assigned sus hyoideus is formed, instead, from the to the Feliformia, see below) and the "Mi- mastoid portion of the petrosal. Work in pro- acinae" (members assigned to the Canifor- gress by R. L. Cifelli indicates that devel- mia, see below), although various members opment of this squamosal process may be of both suborders may later simplify and re- uniquely derived for the Carnivora. The duce P4. Protictis (Protictis) haydenianus, elaboration of a processus hyoideus in mam- P. (Bryanictis) microlestes, Miacis hargeri, mals may range from a process that com- Procynodictis vulpiceps, Uintacyon masse- pletely bridges over the facial canal to con- tericus, and a new subgenus of Protictis tact the promontorium, to a simple swelling from the Eocene of northwest Wyoming all of the petrosal external to the facial canal. have an elaborate P4 talonid with two pos- The tympanohyal, the most cranial element terior accessory cusps. Many other "miac- of the hyoid apparatus, is either fused to, or ids" also have a well-developed, elongate P4 articulates with the processus hyoideus.5 talonid, although the second posterior acces- sory cusp is not present in these forms. 5 The developmental relationship between the tym- All Carnivora have greatly reduced the panohyal and the processus hyoideus is complex, and size of M2-3 relative to the size of P4 and the distribution of modes of development of the two is highly variable within the Mammalia (see van der Ml. The trigonid height on M2-3 versus M, Klaauw, 1931). Because of this workers have variably is also reduced relative to the primitive eu- called a bony swelling or process in this area a processus therian condition, as is the ratio of M2-3 tri- hyoideus, a tympanohyal, a tympanohyal process, etc. gonid height/tooth height. Primitively Mz-3 We prefer the use ofprocessus hyoideus as a descriptive are approximately equal in size, or Ml < morphologic term, without inherent developmental im- M2 < M3 size. In Kennalestes and Gyp- plications. 20 AMERICAN MUSEUM NOVITATES NO. 2725

Ursids differ from other Carnivora in the de- elaboration of the posterolingual cingulum, velopment of a processus hyoideus posterior or development of an antero- or medio-lin- to the exit of the facial nerve. The process gual protocone would be derived within the in ursids thus forms the posterior margin of Carnivora. the stylomastoid foramen. In all the outgroups investigated the molar Comparison of the morphology of the taxa trigonids are high, the paraconid size < united by the derived features discussed metaconid size < protoconid size, the para- above, with that of outgroup taxa permits a conids and metaconids are fairly closely ap- hypothesis of primitive character states that pressed, and the angle between the paralo- are retained in the Carnivora. The full euthe- phid and protolophid forms a closed-V in the rian molar complement Ml-3 is retained in trigonid. The trigonid on M2, however, is early carnivorans; although the members of much lower than the trigonid on M1, and this the feliform branch lose M3, many cam- may represent a character derived for the forms retain M3 throughout their phyletic Carnivora, with respect to the primitive con- history. dition in the outgroups in which all the tri- Some members of both the feliform and gonids are subequal in height. The molar tal- caniform lineages retain a small, distinct, an- onids are always shorter in length than the terolabial basal parastyle cusp on P4. A small trigonids, the hypoconids are robust and are parastyle is present on P4 in all the outgroups the largest talonid cusps, the talonids are examined, and this morphology may be hy- well basined lingually, and the cristid obliqua pothesized as primitive for Carnivora. Later are aligned oblique to the anteroposterior carnivorans modify the parastyle in at least axis of the tooth. This morphology is also two different and distinctive manners. The widespread in both primitive Feliformia and large protocone found on P4 of all the out- Caniformia, and may be considered the prim- groups would also be primitive for all carniv- itive molar morphotype for the Carnivora. oran taxa. The upper molars of most carnivorans re- P3 would lack a distinct protocone cusp, tain the metaconule and paraconule found but would have a well-developed posterolin- widely in eutherian molars. Molar hypo- gual cingulum at the base of a slope and shelf cones are not developed as distinct cusps on that extends lingually from the posterior base the posterolingual margins of the protocone. of the paracone. Gypsonictops and Didel- However, a well-developed postcingulum phodus have a protocone cusp on P3, in Cim- and precingulum were probably present in olestes the protocone of P3 is variably pres- the earliest Carnivora, but the two cingulae ent in different species, but the protocone is were not continuous around the base of the absent in the Creodonta. In Proviverra and protocone. This hypothesized upper molar Sinopa P3 is very similar to the morphology morphology would be similar to that found proposed for the early Carnivora and found in early eutherians that have not yet strongly in most "miacines"; a protocone is absent, modified the molars from the primitive tri- and a large cingulum is developed on the pos- tubercular mammalian pattern. terolingual margin of the tooth. In known The carnivoran P4 probably retained the primitive Carnivora a protocone on P3 is primitive condition of a small anterior acces- found only in Protictis, while all other gen- sory cusp ("paraconid") and lack of a lingual era lack this cusp. Therefore, the polarity for accessory cusp ("metaconid") found in most the presence or absence of the protocone on early eutherian mammals. Cimolestes, Did- P3 for the Carnivora is uncertain. However, ymoconus, Purgatorius, Avunculus, and the limited distribution of this character Creodonta lack a lingual accessory cusp on within the Carnivora and the absence of the P4; Didelphodus has this cusp variably de- protocone in the Creodonta and some Cim- veloped; whereas Protungulatum has a large olestes would support the hypothesis that the lingual cusp on a molariform trigonid. When absence of the protocone is a primitive re- it is present, all the groups have a small, low tention for carnivorans. Any subsequent anterior accessory cusp, or only an anterior, 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 21 basal cingulum as in many creodonts. Within would be primitive for the Carnivora. There- the Carnivora, most species have a very fore, based on present evidence of the dis- small, basal anterior accessory cusp, while tribution of ossification in the ear regions only Protictis (Bryanictis) vanvaleni has a within the Carnivora, and in the available distinct lingual accessory cusp. We agree Cretaceous outgroups, it appears that an with Maclntyre's (1966, p. 204, and above) unossified bulla would be characteristic of conclusion that a simple P4 trigonid is a prim- the earliest carnivorans. itive eutherian condition retained in car- In the feet, the Carnivora primitively re- nivorans, but the carnivoran talonid is tain an articular facet on the calcaneum for primitively elaborated in a unique way. the fibula, and separate ossification of the Comparison with outgroups reveals a con- scaphoid, lunar, and centrale bones. Both of flicting character polarity, dependent on the these conditions are found in primitive eu- particular taxon that is chosen as an out- therians and metatherians, as well as early group. We recognize this conflict, but think carnivorans. Marsupials have a well-devel- that the presence of a simple P4 in Cimo- oped articulation between the calcaneum and lestes, Didymoconus, Purgatorius, Avuncu- a lateral process of the fibula, as do Provi- lus and the possible close relatives of the verra (and other Creodonta), Protungulatum Carnivora, the Creodonta, and the derived (Szalay and Decker, 1974), Asioryctes (Kie- state of the very molariform premolars in lan-Jaworowska, 1977), and the primitive fe- Protungulatum, provide strong evidence liform Didymictis. Both Novacek (1980) and that a simple P4 trigonid may be primitively Szalay and Decker (1974) consider the pres- retained in the Carnivora. ence of a large calcaneal fibular facet to be An unossified entotympanic bulla is pre- primitive for the Theria: "The articulation of sumed a primitive retention in the Carnivora. the fibula with the tarsus in Protungulatum The little available evidence from Creta- is with both the astragalus and calcaneum, ceous eutherians indicates that an unossified and this is undoubtedly primitive for the bulla was widespread in early mammals. Eutheria and other Mammalia" (Szalay and Furthermore, the absence of an ossified bulla Decker, 1974, p. 237).. The calcaneal fibular in known skulls of early Caniformia and Fe- facet is well developed in primitive feliforms, liformia ("miacids") such as Vulpavus and and in some living Feloidea, as well. Didy- Viverravus indicates that the common an- mictis retains the primitive morphology for cestral stock of these two lineages would this feature; there is a well-developed, wide also have lacked an ossified entotympanic articular surface on the calcaneum, and a bulla. However, the present sample of well- large facet on the laterodistal end of the fib- preserved basicrania in early carnivorans is ula for articulation with the calcaneum. This very small and the absence of intact ossified is very similar to the condition in marsupials bullae may be an artifact of preservational and creodonts, although the distal end of the and sampling biases. It is possible that larger fibula is simpler and more reduced in the Fel- samples of extremely well-preserved mate- iformia. Paguma, Herpestes, and Viverra rial of early carnivorans will reveal an ossi- also retain this calcaneal fibular articulation, fied bulla, as has happened in the past with although it is not as strongly developed in other groups (e.g., leptictids, see Novacek, these forms as in Didymictis, and the surface 1977). Large collections of the basicrania of area of the facet on the calcaneum is much the Nimravidae indicate only partial ossifi- smaller. The degree of ossification of the cation of the bullae,6 which would also rein- scaphoid, lunar and centrale has served as force the possibility that an unossified, or an important character in the early classif- only partially ossified, entotympanic bulla ications of the Carnivora (including the Creo- donta, see Historical Section). It is obvious, 6 Hunt (1974a), however, believes that this is only an however, that the separate ossification of artifact of preservation, and that the Nimravidae would these bones is a primitive condition found in have had a totally ossified bulla in life (see below). metatherians, all early eutherians, Creodon- 22 AMERICAN MUSEUM NOVITATES NO. 2725 ta, and early feliform and caniform Carniv- elongate, entire tooth with elongate oval out- ora. Fusion of the scaphoid and lunar into a line; M2 hypoconulid larger than, or equal to, single scapho-lunar bone probably occurred hypoconid size (particularly height). independently in both lineages of Carnivora, Members of the Feliformia primitively re- but may possibly be used to unite taxa within tain these carnivoran characters: lack of mo- these lines. lar hypocones as distinct cusps, precingulum The primitive carnivoran scapula would and postcingulum not continuous around lin- have a large primary scapular spine separat- gual base of protocone; P4 protocone large; ing the infraspinous and supraspinous fos- calcaneal fibular facet well developed; P4 an- sae, possess a straight axillary border, and terior accessory cusp small; P4 talonid elon- lack a teres major process and any indication gate with two posterior accessory cusps; up- of a secondary scapular spine (see Canifor- per molar conules present; molar trigonids mia section for further discussion of scapular high, paraconid < metaconid < protoconid morphology). This primitive scapular mor- size, paraconid and metaconid closely ap- phology is found in many members of mam- pressed forming a closed-V trigonid; bullae malian outgroup taxa such as the Creodonta, unossified; scaphoid, lunar and centrale sep- Metatheria, Rodentia, Perissodactyla, Ar- arate; scapula lacking teres major process tiodactyla, and "condylarths." and secondary scapular spine. The previous discussion of characters in DISCUSSION: Kretzoi (1945) abandoned the the earliest Carnivora is critical to an under- order Carnivora and placed all the Carnivora standing of the distribution of morphologies into two other orders, Feliformia and Cani- and the phyletic history of the included car- formia. Kretzoi's Feliformia included only nivoran taxa. Once the general morphotype the Nimravidae, Felidae, Megantereontidae, for the Carnivora has been hypothesized, in- and Machairodontidae; whereas the Miacoi- cluding both primitive characters retained dea, Canoidea, and remaining Feloidea (of from earlier evolutionary stages and char- Simpson, 1945) were placed in the Canifor- acters uniquely derived for members of the mia. Tedford (1976) considered the Felifor- Carnivora, all of these characters may be as- mia and Caniformia subordinal divisions of sumed to be primitive for taxa within the or- the order Carnivora, and he revised these der Carnivora. The distribution of derived taxa to include the Pinnipedia within the characters is analyzed below in order to de- Caniformia, and all of Simpson's Feloidea termine the distribution of monophyletic within the Feliformia. We agree strongly taxa included within the Carnivora. The re- with a natural division of the Carnivora into mainder of this systematic section is based two clades and we follow Tedford's (1976) on such an analysis, as is the construction of arrangement of the Carnivora. However, we the hypothesis of relationship presented later amend his characterizations so as to include in this paper. all the members of the paraphyletic "Miaci- dae" with their closest relatives in the ap- propriate carnivoran suborders. SUBORDER FELIFORMIA KRETZOI, 1945 All feliforms enlarge the parastyle of P4 INCLUDED TAXA: Infraorders Didymictida over the condition found in the various out- and Aeluroida. group taxa. In some groups (e.g., Hyaeni- DISTRIBUTION: ?Early Paleocene to Re- dae) the parastyle is extremely enlarged and cent. adds a significant component to the shearing DIAGNOSIS: Carnivorans differing from all or crushing surface area of the tooth. How- other Carnivora in the development of the ever, in all cases the parastyle is an elabo- following derived features-P' parastyle en- ration of the basal cingular cusp found in larged; very deep, narrow, slitlike carnassial most eutherians, even when the cusp is in- notch in the metastyle blade of P4; M3 lost corporated in an enlarged shearing blade, as in all taxa; M2 reduced in size, and more sim- in the sabretoothed felids. Development of ple in morphology; M2 talonid extremely a large cusp in a similar topographic position 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 23 may occur in some Caniformia, but this is of the talonid of M2, with talonid length equal not homologous with the enlarged parastyle to, or greater than, the trigonid length. This cusp of feliforms (see following discussion results in a narrow, elongate, rectangular or on the Nimravidae). oval outline of the entire tooth crown. The Wortman and Matthew (1899) used the narrow, elongate M2 talonid and the elon- presence of a deep vertical notch in the meta- gate, oval outline of M2 are also found in style blade of P4 in the Viverravidae and Vi- some living viverrids (Feloidea) that retain verridae to distinguish these taxa from con- a well-developed M2. In all the outgroups M2 temporary Canidae (including "miacines"). talonid length is shorter than the trigonid This appears to be a good character uniting length. The primitive caniforms previously the Feliformia, all of which have the deep, classified in the Miacinae also have the prim- vertical, slitlike notch and not the primitive, itive state of an M2 talonid that is shorter in broader, shallower notch characteristic of length than the trigonid, and in some cases other carnivorans, creodonts, and some the talonid length is extremely reduced. Cimolestes (e.g., C. magnus). Even in forms that parallel the early feli- The upper and lower third molars are lost forms in the loss of M3 (e.g., Quercygale, in all Feliformia. All of the outgroups ex- "Plesiomiacis," etc.) the talonid of M2 is amined, as well as early Carnivora (e.g., ear- very short and never equal to or greater than ly Tertiary Caniformia), have the full molar the trigonid in length. complement of M'-3 found in primitive There is also an enlargement of the hypo- Eutheria. The Feliformia reduced the num- conulid on M2 in conjunction with the length- ber of molars to MI-2 early in their phy- ening of the talonid. The hypoconulid of M2 letic history, as all the known feliforms, be- is primitively a large, high and distinct cusp ginning in the early Paleocene, have lost M3. in all Feliformia, and is developed separately Later, more advanced Caniformia may also from, and posterior to, the entoconid and reduce their molar numbers in parallel, al- hypoconid. The hypoconulid is equal in size though this can be hypothesized to have oc- to, or larger than, the hypoconid, although curred independently in a number of phyletic the hypoconid retains its primitive large size lineages (e.g., advanced members of the and robust, conical morphology. The hypo- groups "Miacinae," Nimravidae, Canidae, conulid on M2 of the outgroup taxa is gen- Mustelidae, Procyonidae, etc.). erally smaller than the hypoconid and is not M2 is reduced in size and complexity in the elaborated as a large, distinct cusp; rather it feliforms, when compared to the morphology is part of a continuous ridge that encircles of the second upper molar in the outgroups the talonid basin and incorporates the cristid and primitive caniforms. M2 is much smaller obliqua, large hypoconid, hypoconulid, and than M1, the tooth is narrow anteroposte- entoconid. The hypoconulid on M2 of prim- riorly, and the metastylar region is very re- itive caniforms is very reduced and forms an duced. Again, advanced Caniformia may indistinct cusp or swelling on the low ridge analogously reduce M2 as they reduce the that surrounds the talonid basin. Even in the size and number of molars at the posterior few derived caniforms that independently end of the cheek tooth row. Advanced Feli- expand and lengthen the talonid of M2, such formia may lose M2 entirely, and similarly as the Procyonidae, the morphology is dif- reduce M1. ferent; the hypoconulid is not extremely en- Matthew's (1915) key to the genera of Mi- larged concurrently with the talonid length- acidae introduced another character that ening and the "entoconid" is a series of supported a bipartite split of the Miacidae. cuspules and cusps along the lingual tooth The elongate, oval shape of M2 distinguished margin. No such multiplication of cusps is Didymictis and Viverravus (corresponding seen in the Feliformia. to his 1909 Viverravinae) from the remainder The Feliformia includes most of the taxa of the Miacidae. This character may be de- assigned by Simpson (1945) to the Viverrav- fined in more detail as an extreme elongation inae (Miacidae) and to the Feloidea, and by 24 AMERICAN MUSEUM NOVITATES NO. 2725 other workers to the equivalents of these metaconid size; P4 with a well-developed groups. However, a number of taxa are ob- notch between the anterior accessory cusp viously excluded from the Feliformia, as de- and the main cusp; P4 talonid broad; canines fined in the diagnosis above. These include multigrooved. the Nimravidae (Nimravinae and Hoplopho- The members of this family are character- neinae of de Beaumont), Plesiomiacis, Quer- ized by a number of additional, primitive cygale, and Oodectes, which are alternative- characters: P4 retains a long talonid with at ly assigned to the Caniformia, and are least one well-developed, and possibly a sec- considered later in this study. In contrast, ond moderately developed posterior acces- the equivocal genus Palaeogale appears to sory cusp; M1 metaconid not lower than the be a feliform, as suggested by Hunt (1974b), low paraconid (as is found in some Viver- rather than a mustelid caniform as proposed ravus, whereas the metaconid is lost in some by Simpson (1946). Therefore, the Felifor- feloids); molar trigonids with paraconid and mia, as defined here, would include Protic- metaconid relatively closely appressed, re- tis, Didymictis, Ictidopappus, Viverravus, sulting in closed trigonids; molar cristids Simpsonictis, Palaeogale, and the Feloidea oblique and talonids basined, not trenchant (exclusive of the Nimravidae). as in some Viverravidae; M2 sulcus obliquus Our definition of the Feliformia is partly deep and basined, rimmed ventrally by a bas- based on characters presented in the litera- al cingulum; M2 hypoconulid large, lingual, ture by other workers. However, many of closely appressed to entoconid, as opposed these characters were initially used to define to the relatively larger hypoconid of cani- the "Viverravinae," and their utility in also forms and the posterocentral hypoconulid, characterizing the Feloidea remained unrec- well separated from the entoconid, found in ognized. By distinguishing between the de- the Viverravidae and Ictidopappus; conules rived and the primitive characters used by on the upper molars always present and well earlier workers, a hypothesis of relationship developed, not lost as in all other Feliformia; may be proposed that incorporates the Fe- cingulae well developed, except in P. (Pro- loidea and "viverravine Miacidae" on the tictoides), in contrast to weak cingular de- one hand, and the Canoidea and "miacine velopment in other feliforms. Miacidae" on the other, into rigorously de- DISCUSSION: This family is proposed for fined, monophyletic taxa. The Feliformia the members of one branch of the Feliformia comprises both early and late Tertiary taxa that remain conservative in most of their of one of the two major carnivoran clades. dental and skeletal morphology, but that are united by a number of derived features. We follow Maclntyre (1966) in recognizing Mat- DIDYMICTIDA, NEW INFRAORDER thew's subgenus Didymictis (Protictis) as a DIDYMICTIDAE, NEW FAMILY distinct genus, with the morphologically dis- TYPE GENUS: Didymictis Cope, 1875a. tinct subgenera P. (Protictis) and P. (Bryan- DISTRIBUTION: Middle Paleocene to late ictis). We exclude Maclntyre's subgenus P. Eocene. (Simpsonictis) from Protictis and the Did- INCLUDED GENERA: Didymictis Cope, ymictidae, and include it as a resurrected ge- 1875a; Protictis (Matthew, 1937), including nus in the Viverravidae (see below). We do the subgenera P. (Protictis) (Matthew, not follow Rigby's (1980) suggestion that Bry- 1937); P. (Bryanictis) Maclntyre, 1966; and anictis be generically differentiated from Pro- P. (Protictoides), new subgenus. tictis, as our analysis has clearly shown that DIAGNOSIS: Feliform carnivorans differing the two taxa share a number of derived fea- from all other Feliformia in the following de- tures and are best considered members of the rived characters-M2 paraconid extremely, same genus (see below). We also propose a and metaconid moderately, reduced, para- new subgenus P. (Protictoides) from the mid- conid much smaller than, or subequal to, dle Eocene of Wyoming. 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 25

Both Didymictis and the various subgen- reduce, or lose entirely, M2. In Didymictis era of Protictis share a number of specialized the trigonid is lower than in other Didymic- dental characters that unite them in a mono- tidae and the trigonid cusps are extremely phyletic group, for which we propose the reduced. name Didymictidae. The distinctive taxon of All Didymictidae develop a distinct, nar- primitive carnivoran perceived by early row, deep carnassiform notch between the workers in their broadly constructed, para- anterior accessory cusp and the main cusp phyletic grouping of species within Didy- on P4. The notch in didymictids is deeply ex- mictis is retained in the usage of a family cavated and V-shaped, and the anterior ac- Didymictidae, whereas the diversity and cessory cusp is closely appressed to the main monophyletic nature of the included material cusp. The anterior cusp may remain small, is still recognized in the taxonomic distinc- primitively, but the notch is clearly elabo- tion of the various included genera and sub- rated in all included taxa. Other feliforms genera. We also propose a new taxon, the lack this notch, as the anterior accessory infraorder Didymictida, that includes only cusp is well separated and offset lingually the members of the family Didymictidae. from the main cusp by a wide, shallow, This higher level taxon represents a group of U-shaped valley. equivalent rank to the Aeluroida (see below), The talonid of P4 in the didymictids is and we favor its use to differentiate one ma- broadened over the primitive state for feli- jor evolutionary branch of the Feliformia. forms, although the two posterior accessory The diversity represented within the Didy- cusps and the posterior talonid cingulum mictida is much less than in the Aeluroida, characteristic of all Carnivora is retained. but it is clear that the two taxa represent a The increase in transverse talonid width oc- fundamental, dichotomous split in the phy- curs largely in the lingual portion of the tal- logenetic history of the Feliformia. onid; the talonid basin is enlarged, deepened, The Didymictidae are generally primitive and rimmed lingually by a higher lateral wall. members of the Feliformia, except for the No other feliforms or caniforms develop few derived characters discussed below. The such a broad, deep, basined talonid on P4. family represents an early offshoot from the The canines are multigrooved in the only feliform stock that evolved certain autapo- Didymictidae for which they are known, P. morphies in the various included taxa, but (Protictis) and Didymictis. Grooving of the that largely remained primitive in both the canines may be derived for the Didymicti- skeleton and the dentition. Protictis (Protic- dae, as it is not present in Viverravus. How- toides) represents the last occurrence of the ever, the distribution of this character may family in the earliest Uintan, and the group be biased by the lack of preserved canines has no close living descendants in the mod- in many early feliforms, and it is possible ern carnivoran fauna. that grooved canines are primitive for the The paraconid of M2 is extremely reduced Feliformia. The grooving exhibited in the in the Didymictidae, whereas the metaconid Didymictidae is a multiple series of closely is more moderately reduced. This results in spaced, deep grooves extending along the a very low M2 trigonid, with poorly distinct entire height of the tooth crown, and most paraconid and metaconid cusps. The trigonid deeply excavated near the base of the crown. is much lower than that found primitively in Similar grooving may be found in some Fe- feliforms and represents a reduction in both loidea, but is not present in any of the out- talonid height and cusp development over groups. Therefore, the polarity of this char- the primitive state for Carnivora. Other feli- acter is uncertain. forms retain the primitive condition of a Didymictis is derived in a number of fea- moderately high trigonid and well-developed tures, including: reduction of P4 posterior cusps, with the paraconid lower than than accessory cusps to one; close appression of the metaconid, except in those Feloidea that the paraconid and metaconid on the lower 26 AMERICAN MUSEUM NOVITATES NO. 2725 molar trigonids; great elaboration of cheek the features characterizing the latter genus. teeth basal cingulae; and reduction of M2 tri- These features in Protictis include: lack of gonid. extremely well-developed cingulae on the The talonid of P4 remains elongate in Did- cheek teeth; M2 paraconid and metaconid ymictis but the number of posterior acces- not so extremely closely appressed; M2 tri- sory cusps and notches is reduced to one. gonid higher; retention, primitively, of two Didymictis modified its trigonids quite dif- posterior accessory cusps on the talonid of ferently from all other feliforms, through ex- P4. tremely close appression of the paraconids Protictis (Protictis) differs from all other and metaconids, ultimately resulting in near Protictis in: extremely narrow M2 talonid; fusion of the two cusps in extremely derived transversely narrow, elongate parastyle lobe species (e.g., D. protenus). on P4; strong "backward tilting" of para- Didymictis elaborated on the cingulae cone, and straight, vertical posterior margin primitively present in this group, and of paracone on P4; first posterior accessory evolved extremely pronounced basal cingu- cusp located vertically high on posterior lae rimming the cheek teeth. The cingulae flank of main cusp on P4. became very broad and developed numerous DIscUSSION: Matthew's (1937) recognition rugosities and cuspules that are very distinc- that "Didymictis" (Protictis) represented a tive features of this genus. clearly distinct, morphologically definable The Didymictis lineage evolved toward taxon within Didymictis was an essential larger size and greater occlusal surface area step in the separation of monophyletic taxa on the cheek teeth. This was achieved by an from a diverse paraphyletic assemblage. increase in the area, length, and crenulation In 1966 MacIntyre formally designated a of the cingular enamel, as well as a reduction new genus, Protictis, for the material includ- in the trigonid height and cusp development ed within Matthew's subgenus. Most of the of M2. This resulted in a nearly continuous, characters used to define and describe the flat, platelike surface extending from the taxa included here in Protictis are well de- base of the M1 trigonid to the posterior mar- scribed and illustrated in MacIntyre's paper, gin of the M2 talonid, and is found in Didy- and that study should be consulted for a mictis only. more detailed analysis of the morphology of these groups. PROTICTIS (MATTHEW, 1937) Maclntyre's (1962) recognition of the TYPE SPECIES: Protictis (Protictis) hay- unique morphology represented by Simp- denianus (Cope, 1882). son's (1935) "Didymictis" tenuis also added INCLUDED SUBGENERA: Protictis (Protic- to our knowledge of the diversity present tis) Matthew, 1937; P. (Bryanictis) Mac- early in carnivoran evolution. However, we Intyre, 1966; and P. (Protictoides) new sub- believe that MacIntyre's initial decision to genus. provide generic status for Simpsonictis ten- DISTRIBUTION: Middle Paleocene to mid- uis is valid, and we therefore do not follow dle Eocene. his latest (1966) inclusion of Simpsonictis as DIAGNOSIS: Members of the Didymictidae a subgenus of Protictis. Rigby (1980) and differing from other taxa included in the fam- Van Valen (1969) have resurrected Simpson- ily by the following derived features-ante- ictis, and we agree with this action. We do rior accessory cusp on P4 enlarged and very so, however, because the members of the high; lingual margin of P4 straight anteropos- taxon lack the diagnostic, derived features of teriorly, not curved; P3 develops a weak, but Protictis and the Didymictidae, and instead distinct, protocone cusp anterior to the apex share derived features with Viverravus. of the paracone; P4 main cusp situated on the Simpsonictis is an unusual, derived, early buccal tooth margin; main cusp on P4 leaf- Tertiary member of the Feliformia that is ge- shaped in outline. nerically separable from all other carnivo- Protictis also differs from Didymictis in rans and is closely related to Viverravus, 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 27 rather than Protictis (as suggested by Rigby occlusal outline, uninterrupted by constric- and Maclntyre, in Rigby, 1980). tions or expansions of the lingual tooth As mentioned previously, we do not agree crown. This suite of characters results in a with Rigby (1980) that P. (Bryanictis) rep- P4 in Protictis that has an extremely smooth resents a genus distinct from Protictis (Pro- occlusal outline, a wide area of lingual basin- tictis). There are a number of features that ing and broad sloping off the shoulders of the exclusively ally the two taxa, as will be elab- cusps, and a steepening of the buccal tooth orated below, and nothing to support a ge- crown. neric separation of the two. Rigby does not All the species of Protictis that preserve provide a new diagnosis that would serve to P3 show the development of at least a weak differentiate P. (Bryanictis) at the generic protocone cusp located anterior to the para- level. Protictis (Protictis) and P. (Bryanic- cone. Protictis is the only feliform that de- tis) are very closely related, and without bet- velops a protocone on P3. Other Carnivora ter evidence for an alternative hypothesis of have a well-developed cingulum on the pos- relationship, we see no reason for separating terolingual margin of P3, at the base of a P. (Bryanictis) as a new genus. deep, lingual paracone slope or shelf. Pro- Protictis is a distinctive, monophyletic tictis develops a distinct cusp at the ante- taxon within the Didymictidae. All the taxa riormost edge of this lingual cingulum, there- included within Protictis greatly enlarge the by broadening P3 and adding another lobe to size of the anterior accessory cusp on P4. the tooth. The protocone is anterior to the This cusp becomes high and is clearly de- apex of the paracone in all cases. The polar- rived over the small, low anterior accessory ity of this character is debatable, as some of cusp characteristic of the Feliformia, Cani- the Cretaceous outgroups have a protocone formia, and all the Cretaceous outgroups. In on P3, although it is not anterior to the apex connection with the enlargement of the cusp, of the paracone (see discussion on Carnivora the notch between the anterior accessory above). If the presence of a protocone on P3 cusp and the main cusp becomes much deep- is primitive, then all other feliforms and the er than the notch in Didymictis. caniforms must have lost it independently, The main cusp on P4 is situated far buc- while Protictis remained primitive in its re- cally, along the buccal edge of the tooth. The tention of this cusp. Based on the arguments entire cusp row is thus aligned in a smooth given previously, however, we regard the arc, convex buccally, and is located close to presence of an anteriorly positioned P3 pro- the buccal margin of the tooth. The buccal tocone as uniquely derived for Protictis. The margin of the tooth is also smooth and con- development of a distinct, lingual protocone tinuous, and is buccally convex. This con- cusp may be correlated with the elaboration trasts with the primitive morphology exhib- of the anterior accessory cusp on P4, as these ited by Didymictis. In Didymictis the main two cusps would provide complementary oc- cusp is on the lingual edge of the tooth clusal surfaces for increased shear. crown. The cusp row does not form a con- The species included within Protictis all tinuous, smooth line or arc, and the outline possess a main cusp on P4 that is anteropos- of the buccal margin of the tooth is inter- teriorly constricted on its anterior and pos- rupted by a number of bulges and constric- terior flanks. This constriction occurs about tions along its anteroposterior length. In the midway up the height of the cusp, at the Aeluroida, there are constrictions of the notches that mark the junctions between the crown at the bases of all the cusps on P4, main cusp and the anterior and first posterior and lateral bulgings where the cusps are the accessory cusps. As a result of this constric- most robust. The cusps are not arranged in tion, the main cusp has a "leaf-shaped" out- a linear or arcuate manner, and they tend to line in lateral view. This is derived with re- lie at the center or lingual portion of the spect to the primitive, inverted-V shaped crown. The lingual margin of P4 in Protictis main cusp found on P4 in other Carnivora is derived in its anteroposteriorly straight and eutherian outgroup taxa. 28 AMERICAN MUSEUM NOVITATES NO. 2725

Maclntyre's (1966) diagnosis of Protictis tictis in the following derived features-P4 (Protictis) indicated that the talonid of M2 in anterior accessory cusp tremendously en- this subgenus was relatively much narrower larged and modifed for shearing; expansion than in any other early Tertiary feliform. and very deep basining of the lingual portion Protictis (Bryanictis) and taxa which serve of the talonid of P4; large, well-developed P3 as outgroups to the Protictis clade (e.g., protocone; all P4 cusps very strongly later- Didymictis and the Caniformia) all have a ally compressed; alignment of the cusps on much wider M2 talonid. Therefore, this fea- P4 in a straight, anteroposterior line on the ture distinguishes P. (Protictis) from all oth- buccal tooth margin, buccal margin of the er Protictis. tooth straight anteroposteriorly, or only The parastyle lobe of P4 in P. (Protictis) slightly convex buccally. is very narrow transversely and is elongated DISCUSSION: Maclntyre's (1966) diagnosis in an anterobuccal/posterolingual direction. and description of this subgenus, and its in- This results in an elongate, oval outline for cluded species, provides the most compre- the parastyle lobe and cusp that differs from hensive treatment of this taxon. Our diag- the shape found in all other carnivorans. nosis is presented only to elaborate on the Didymictis, P. (Bryanictis), P. (Protic- derived morphology present in this subge- toides), and Viverravus all have a more nus, and as evidence to document the mono- rounded and conical shape for this lobe; in phyletic nature of P. (Bryanictis). Both P. the Caniformia, however, the parastyle cusp (B.) microlestes and P. (B.) vanvaleni share and lobe are extremely small or absent. derived characters that ally them more Protictis (Protictis) has an extremely closely to each other than to any other taxon, straight, vertically oriented posterior margin which strongly supports the validity of re- on the paracone of P4. The anterior slope of taining these two species within their own the paracone is long and extends far poste- distinct subgenus. The arguments for not riorly to the paracone apex. This results in considering these species as members of a a paracone profile that points posteriorly and distinct genus have been given above. appears to be "tilting" in a posterior direc- Most of the characters used to ally P. (B.) tion. The paracone apex is relatively farther microlestes and P. (B.) vanvaleni, are anterior, and the anterior paracone slope is shared in a more primitive condition by all shorter, in all other Protictis, Didymictis, Protictis. Protictis (Bryanictis) elaborated and early Tertiary Carnivora. greatly on these features and evolved a mor- The first posterior accessory cusp on P4 of phology that is more derived than that ex- P. (Protictis) is positioned very high on the hibited by other Protictis. The protocone on posterior flank of the main cusp. This cusp P3 of P. (Bryanctis) is greatly enlarged and is higher in P. (Protictis), relative to the base forms a distinct, bulbous, high cusp on a lin- of the tooth crown and the talonid rim, than gual lobe of P3. The protocone is larger, bet- in any other early Tertiary carnivoran. ter developed as a distinct cusp, and invari- ably present in this subgenus, as opposed to PROTICTIS (BR YANICTIS) MACINTYRE, 1966 the more variable intraspecific presence and smaller size of this cusp on P3 in P. (Protic- TYPE SPECIES: Protictis (Bryanictis) mi- tis) haydenianus. The protocone is well an- crolestes (Simpson, 1935). terior to the apex of the paracone, as in P. INCLUDED SPECIES: P. (Bryanictis) mi- (Protictis). P3 is unknown in P. (B.) van- crolestes (Simpson, 1935) and P. (B.) van- valeni and P. (Protictoides). valeni Maclntyre, 1966. The anterior accessory cusp on P4 in both DISTRIBUTION: Middle and ?late Paleo- species of P. (Bryanictis) is elaborated into cene. a large, vertical shearing blade at the anterior DIAGNOSIS: Members of the genus Protic- end of the tooth. As mentioned in the dis- tis that share all the derived characters of cussion of Protictis, the enlargement of the that genus, but that differ from all other Pro- anterior accessory cusp appears to be cor- 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 29 related with an increase in the size of the of the lingual portion of the talonid of P4 is protocone of P3. This certainly seems to be a derived feature for both species of P. the case in P. (Bryanictis), as both the pro- (Bryanictis). This is a greater elaboration of tocone of P3 and the anterior accessory cusp the condition found in other Protictis, in on P4 are greatly enlarged and would have which the talonid is moderately wide and ba- acted as an efficient shearing pair during oc- sined. The talonid basin of P4 in P. (Bryan- clusion. The very deep, narrow, carnassi- ictis) is extremely wide and deep and is en- form notch on P4, between the anterior ac- tirely encircled by a high basin rim. The rim cessory cusp and the main cusp, would also is composed of the posterior accessory cusp have been involved in the increased shear wall buccally, and a high cingulum on the between P3 and the anterior part of P4. The posterior and lingual margins of the tooth. anterior accessory cusp of P4 in P. (Bryan- The deep basining is best developed on P4 of ictis) is unique among feliforms, in that it is P. (B.) vanvaleni. much taller than the first posterior accessory cusp. Even in other Protictis, in which the anterior accessory cusp is also enlarged, the PROTICTIS (PROTICTOIDES), NEW SUBGENUS anterior cusp height is never greater than the TYPE SPECIES: Protictis (Protictoides) height of the apex of the first posterior cusp. aprophatos, new species. In all Protictis species, however, as in all INCLUDED SPECIES: The type species other carnivorans, the main cusp on P4 is al- only. ways the tallest and dominates the tooth DISTRIBUTION: Earliest Uintan (middle crown. Eocene). All the cusps on P4 in P. (Bryanictis) are ETYMOLOGY: Subgeneric name from a very strongly laterally compressed. This is combination of the Greek -oides, meaning particularly extreme in P. (B.) vanvaleni, in like or resembling, and Protictis; in refer- which the cusps are very thin transversely ence to the morphological similarlity and hy- and the flanks of the cusps are nearly verti- pothesized close phyletic relationship to cal. This results in an extremely sharp, nar- Protictis (Protictis). row, continuous ridge on the buccal tooth DIAGNOSIS: Member of the genus Protic- margin, that runs the entire length of the tis, differing from all other Protictis in the tooth and connects the crests of the cusps. following derived features-P4 anterior mar- The alignment of P4 cusps in P. (Bryan- gin extremely straight, not embayed; P4 cin- ictis) is derived over the primitive condition gulae very reduced or absent; P4 labial mar- for Protictis; the cusps are aligned straight gin straight; P4 occlusal outline triangular, anteroposteriorly along the extreme buccal with all borders straight; P4 cingulae reduced margin of the tooth. This contrasts with the or absent; P4 with very strong, high posterior more arcuate arrangement of the cusps on transverse talonid ridge (cingulum). the crown in other Protictis. The cusp row CHARACTERIZATION: The known ele- in P. (Bryanictis) forms a very straight, high ments, P4 and P4 of P. (Protictoides) apro- wall on the buccal tooth margin, bordering phatos approximate the size of late Paleo- a wide surface area on the lingual side of P4. cene Didymictis. Protictis (Protictoides) In outline, the buccal margin of the tooth greatly exceeds the size of Simpsonictis, P. crown is also smooth and straight anteropos- (Bryanictis), P. (Protictis) and Viverravus teriorly, or only slightly convex buccally. (see table 2). The presence of an extremely The buccal tooth margin in Protictis is also large, trenchant P4 parastyle and protocone smooth, although it is much more strongly (see fig. 5) clearly distinguishes P. (Protic- buccally convex, whereas the crown outline toides) from P. (Bryanictis), P. (Protictis), in Didymictis is very irregular in shape with Didymictis and Viverravus. The basined numerous constrictions and expansions of area labial to the protocone, between the the margin. paracone and the parastyle, is much broad- A transverse expansion and deep basining ened. The labial base of the paracone is con- 30 AMERICAN MUSEUM NOVITATES NO. 2725

TABLE 2 Measurements (in Millimeters) of Early Tertiary Feliformia Data on Protictis (Protictis), Protictis (Bryanictis), Ictidopappus, and Simpsonictis from Maclntyre, 1966 P4 P4 P4 (p4) (P4) (P4) length, metastyle length, premolar premolar premolar labial (length) lingual width length width P. (Protictis) haydenianus (Torrejonian localities, New Mexico) 7.77 + 0.12 2.70 + 0.10 9.22 ± 0.09 5.85 ± 0.19 6.95 ± 0.09 3.16 + 0.03 P. (Protictoides) aprophatos (Wind River Basin, Wyo- ming) 12.86 4.48 14.29 8.39 8.86 3.91 P. (Bryanictis) microlestes (Gidley Quarry, Montana) 4.86 ± 0.07 1.73 ± 0.06 5.98 ± 0.13 3.72 ± 0.05 4.31 ± 0.06 2.03 ± 0.04 P. (Bryanictis) vanvaleni (Torrejonian, New Mexico) 4.6 2.35 Ictidopappus mustelinus (Gidley Quarry, Montana) 5.0 2.7 5.1 3.7 3.0 2.0 Simpsonictis tenuis (Gidley Quarry, Montana) 2.48 + 0.07 1.12 ± 0.02 Viverravus gracilis (genotype) (Bridger Basin, Wyoming) 6.13 2.43 6.95 3.70 5.61 1.93 Viverravus sicarius (type) (Bridger Basin, Wyoming) 9.22 3.41 9.99 6.12 7.69 3.34 vex, rather than concave, giving P4 a however, to the unmodified condition also straighter occlusal profile than that found in found in P. (Protictis), P. (Bryanctis) and the above mentioned genera. The anterior various "miacines" in which the presence of and lingual margins of the tooth are also ex- a second posterior accessory cusp creates a tremely straight, resulting in a strongly tri- notch between this cusp and the more ante- angular occlusal outline. The basal cingulum riorly located first posterior accessory cusp. is very weakly developed, being most pro- The anterior accessory cusp, although pro- nounced at the lingual base of the metastyle portionally smaller than in P. (Bryanictis), and at the anterolabial base of the paracone. is stronger and positioned closer to the main The metastyle blade is elongated anteropos- cusp than in Simpsonictis, Viverravus, and teriorly into a well-developed, straight shear- Didymictis. It most closely resembles P. ing crest. All the cusps are laterally com- (Protictis) in this feature. The cusps are pressed into trenchant piercing points and transversely compressed and have a leaf- ridges, rather than bulbous and rounded as shaped outline in lateral view, similar to the in Didymictis and Viverravus. condition in P. (Protictis), Didymictis, and P4 retains the primitive condition of a well- the extreme elaboration found in P. (Bryan- developed second posterior accessory cusp ictis). The basal cingulum is extremely weak distinct from the posterior cingular ridge. on P4, with significant development only on This differs from the talonid structure of P4 the posterior border of the tooth where it in Simpsonictis, Didymictis, Ictidopappus, forms a pronounced, buccolingual ridge be- and Viverravus which have reduced the sec- hind the second posterior accessory cusp. ond posterior accessory cusp. It is similar, The comparative description provides ad- 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 31 ditional details and elaboration on diagnostic COMPARATIVE DESCRIPTION characters of P. (Protictoides) aprophatos. Comparison to other primitive feliforms: Protictis (Protictoides) aprophatos, (1) P4: The protocone of P4 in P. (Protic- new species toides) aprophatos is broad, stands high, and Figures 2, 3, 5, 6 is so prominent as to create a sharp drop immediately posterolabial to the cusp (where HOLOTYPE: AMNH 97701; isolated right the anteromedian edge of the paracone and P4. the protocone meet). The summit projects TYPE LOCALITY: Bone Bed A of Love upward almost as high as the metastyle, (1939), type section of the Tepee Trail For- when viewed from the medial side and when mation, East Fort Basin, 17 miles northeast the base of the crown is placed horizontally of Dubois, northwest portion of the Wind (see fig. 5). This strongly contrasts with P4 River Basin, Fremont County, Wyoming. protocone morphology in the "miacine" AGE: Earliest Uintan (middle Eocene-see genera Miacis, Uintacyon, Vassacyon, Vul- McKenna, 1980; Berggren et al., 1978). pavus, Procynodictis, and Tapocyon in DIAGNOSIS: Same as for the subgenus. which the protocone is small, bulbous, low ETYMOLOGY: Species name from the crowned, and noticeably lower than the Greek aprophatos, meaning unforetold or metastyle. The P. (Protictoides) protocone unexpected; in reference to the unusual oc- morphology is similarly exhibited in P. currence of a species of Protictis in the Uin- (Bryanictis), Viverravus, P. (Protictis), and tan (middle Eocene). Didymictis, but in these genera the relative REFERRED MATERIAL: AMNH 105371, height of the protocone to the height of the right ramal fragment bearing P4 and the al- parastyle is less when compared to P. (Pro- veolus for the posterior root of P3. Also from tictoides). The protocone of P4 in P. (Protic- Bone Bed A, and possibly belonging to the toides) is more trenchant than in Didymictis. same individual as the holotype. In Didymictis and P. (Bryanictis) the pro- Association of P4 and P4: The assignment tocone is more rounded and bulbous than in of the two Tepee Trail specimens to the same P. (Protictis) (based on AMNH 17059) and species is unquestionable. Both specimens P. (Protictoides) aprophatos. Protictis were discovered in the same locality and per- (Protictis), P. (Bryanictis), Viverravus, and fectly complement each other in size. They Didymictis all have a strong, distinct cingu- occlude well with one another, indicating a lum rimming the base of the crown, as op- possible individual association as well, al- posed to the very weak cingulum develop- though this cannot be determined with cer- ment in P. (Protictoides) aprophatos. tainty. The occlusal surface labial to the proto- The poor dental representation of P. (Pro- cone, between the paracone and parastyle tictoides) aprophatos is compensated by the (area of P4 talonid occlusion) is proportion- fact that these two teeth are quite diagnostic ally much more expanded on P4 in P. (Pro- carnivoran elements. P44s in carnivorans ex- tictoides) than in P. (Bryanictis), P. (Protic- hibit great diversity at all taxonomic levels. tis), Didymictis and Viverravus. In occlusal The striking morphologic differences seen in view, the anterolingual portion of the para- P. (P.) aprophatos are considered diagnos- cone in P. (Protictoides) aprophatos ap- tic at the subgeneric level. As detailed in lat- pears to be more convex than in the above er sections, the characters analyzed do not genera, thereby filling out the indentation intergrade (form part of a variability se- between the parastyle and the protocone. quence) with those of other carnivoran gen- The parastyle of P. (Protictoides) apro- era. The derived nature of many of the char- phatos (as in P. (Bryanictis), P. (Protictis) acters in P. (Protictoides) tends to further and Viverravus) lacks the cingulum encir- emphasize the distinction of the Tepee Trail cling the base of the cusp that is found in form from other carnivorans. Didymictis. However, a faint cingulum orig- 32 AMERICAN MUSEUM NOVITATES NO. 2725

A

B

E LeU

C

FIG. 5. Protictis (Protictoides) aprophatos, AMNH 97701, type specimen, right P4. Stereophoto- graphs of occlusal (A) and lingual (B) views. Photograph of labial (C) view. Anterior direction toward top of page in A, and toward right of page in B and C. Scale is 1.0 cm. 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 33

C.' t.. 1I a, i i

. 4

A .4

E

U1 A

4w. Ol,,,. I ,'P A R'' 'N 4

~. _ } B C

FIG. 6. Protictis (Protictoides) aprophatos, AMNH 105371, referred specimen, right P4 and partial mandible. Stereophotographs of occlusal (A) view. Photographs of labial (B) and lingual (C) views. Anterior direction toward top of page in A, right of page in B, and left of page in C. Scale is 1.0 cm. inates at the posterolabial base of the para- style of P4 in P. (Protictis) is proportioned style, runs along the labial margin of the similarly to that in P. (Protictoides). The tooth, and terminates at the metastyle. This large, deeply notched P4 parastyle is unique condition of the cingulum is similar to that to P. (Protictoides) and is not found in any seen in P. (Bryanictis), P. (Protictis) and other North American Eocene carnivoran Viverravus. Proportionate to the protocone, genus. the parastyle is much stronger and taller in The lingual and labial cingulae of P4 in P. P. (Protictoides) than in P. (Bryanictis), (Protictoides) aprophatos are faint. This Viverravus, and Didymictis (the parastyle in condition more closely resembles the cingu- Didymictis is short and bulbous). The para- lar development found in P. (Bryanictis), P. 34 AMERICAN MUSEUM NOVITATES NO. 2725

(Protictis), and Viverravus than that ob- rather leaf-shaped primary cusp. As in Vi- served in Didymictis, which has distinctly verravus, Didymictis, P. (Protictis), and stronger cingulae. The enamel at the base of Procynodictis the first posterior accessory the protocone in AMNH 97701 is slightly cusp of P4 in the Tepee Trail specimen stands damaged, making it difficult to determine the significantly taller than the anterior acces- extent of the lingual cingulum in P. (Protic- sory cusp. The opposite condition is found toides). The cingulum begins at the meta- in Tapocyon and P. (Bryanictis), in which style, but it does not appear to have encir- the anterior accessory cusp of P4 is slightly cled the protocone as clearly as in Didymictis. taller than the first posterior accessory cusp. In occlusal view, the base of the paracone The first posterior accessory cusp in P. (Pro- on P4 in P. (Protictoides) projects more la- tictis) and P. (Protictoides) does not domi- bially than in P. (Bryanictis), P. (Protictis), nate the talonid, nor is it positioned as low Didymictis, and Viverravus. On both sides of as in Didymictis or Viverravus. Instead, it is the paracone/metastyle notch in AMNH in positional alignment with the second pos- 97701 the edges of the shearing ridge are terior accessory cusp, and is positioned damaged, making comparisons of the height higher on the tooth crown. In P. (Protictis) and wear of that area of P4 to other taxa dif- the P4 first posterior accessory cusp is pro- ficult. portionally smaller, relative to the talonid (2) Ramus and P4: A mental foramen is size, than that of P. (Protictoides). The pos- present beneath the alveolus of the posterior terior region of the talonid is dominated by root of P3 in the referred ramal fragment of two features in P. (Protictoides): a small, P. (Protictoides) aprophatos. Comparison but distinctly developed second posterior ac- shows that a mental foramen may be located cessory cusp, and an internally broad, high, in the area below P3 in most primitive car- ridge-forming posterior cingulum, which to- nivorans. The ramal fragment is somewhat gether occupy an area equal to that covered narrow transversely and slender as com- by the first posterior accessory cusp. The pared to the deep and thick jaws of Tapo- second posterior accessory cusp and the pos- cyon, Uintacyon, Vassacyon, and Vulpavus. terior cingulum form the margins of a well- The ramal proportions are similar to those developed, concave lingual basin. The strong observed in Viverravus. This suggests that development of both the first and second the Tepee Trail form may have been rather posterior accessory cusps results in the for- long-snouted with an uncrowded premolar mation of a second, distinctive, deep and dentition. elaborate talonid notch, a feature that is sim- P4 is practically undamaged, with slight to ilarly developed in P. (Protictis). moderate oblique wear on all the cusps (sim- A posterior labial cingulum is barely ob- ilar to the degree of wear observed in the servable on the P. (Protictoides) P4, which upper carnassial). The anterior cusp is pro- is similar to the condition of the cingulum in portionally larger, and positioned closer to Viverravus and P. (Bryanictis). the primary cusp, than in Didymictis. Vi- The Ictidopappus P4 is very autapomor- verravus, Didymictis, and Simpsonictis have phous and differs from P. (Protictoides) relatively small anterior accessory cusps on aprophatos in the following features: its P4 when compared to those of P. (Protictis), much smaller size; protocone located medial P. (Protictoides), and P. (Bryanictis) (the to the paracone rather than the parastyle; cusp is relatively largest in the last two gen- absence of a basin between the protocone, era). The lateral profile of the primary cusp parastyle and paracone; paracone lower; a on P4 in P. (Protictoides) becomes constrict- "wavy" buccal tooth margin; total lack of ed anteriorly and posteriorly at the notches cingulae; less distinct, more trenchant cusps; between this cusp and the anterior and first better developed shear facets; less distinct posterior accessory cusps. Both species of paracone-metastyle blade notch; and more P. (Bryanictis), and to a lesser degree P. deeply embayed anterior margin. (Protictoides) and P. (Protictis), share this The Ictidopappus P4 is also very derived 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 35 and differs from P. (Protictoides) in: its parastyle, short metastyle blade, concave la- much smaller size; lack of talonid develop- bial margin and moderately well-developed ment; straight anterior and posterior edges cingulae on the Prodaphaenus P4 quickly of the main cusp, in profile; lack of a notch differentiate it from P. (Protictoides). The between the anterior accessory cusp and the teeth in Prodaphaenus appear to be gener- main cusp; weak posterior, transverse, cin- ally adapted for omnivory, as opposed to the gular talonid ridge; larger main cusp; sub- dentition in P. (Protictoides) which seems triangular, rather elongate-rectangular, oc- best suited for hypercarnivory. clusal outline. The general arrangement and proportions P4 in Simpsonictis differs from P. (Protic- of the cusps on P4 in P. aprophatos are most toides) by: its much smaller size; anterior similar to the morphology found in Procy- accessory cusp lower and more anteriorly lo- nodictis (including Plesiomiacis, see below). cated; first posterior accessory cusp much Protictis (Protictoides) differs from Procy- smaller and lower on the base of the main nodictis in its much larger size; the large size cusp; main cusp relatively higher; tooth less of the anterior accessory cusp on P4; the lack elongate and rectangular; talonid less broad; of cingulae on P4; the development of a lingual talonid basin less well developed; strong, transverse posterior talonid ridge; lack of the second posterior accessory cusp and the more shallow depth of the jaw be- and notch. neath P4. In addition, the structure of P4 is Comparison to Uintan caniforms: extremely different between P. aprophatos Miacis7 lacks a prominent parastyle, has and Procynodictis in that the latter lacks a a proportionally smaller P4 protocone, has parastyle and has a very reduced protocone. relatively weaker and less distinct posterior Tapocyon differs from P. (Protictoides) in accessory cusps on P4, has a deeper ramus its more reduced and less trenchant proto- beneath P4, and averages smaller in size than cone, smaller parastyle, and shallower car- P. (Protictoides) aprophatos. nassial notch on P4 and its lack of a second In Uintacyon and Vassacyon the depth of posterior accessory cusp on P4. Tapocyon is the mandible below P4 is proportionally similar, however, in the development of a much deeper, the talonid of P4 is never as large anterior accessory cusp on P4. large, and the protocone and parastyle of P4 are never so large and trenchant as in P. (Protictoides). A significant feature distin- AELUROIDA, NEW INFRAORDER guishing P. (Protictoides) from Uintacyon INCLUDED TAXA: Ictidopappus, the Vi- and Vassacyon is the lack of the second pos- verravidae, Palaeogale, and the Feloidea. terior accessory cusp and talonid notch in DISTRIBUTION: Early or middle Paleocene the latter, but the retention of the primitive to Recent. state in P. (P.) aprophatos. DIAGNOSIS: Members of the Feliformia The broadened protocone, absence of a which differ from all other feliforms in the following derived features-conules absent on all upper molars; M1 paraconid and meta- 7 Simpson (1945) placed the genus Mimocyon Peter- conid distantly separated, resulting in an son, 1919 in Miacis. We disagree with this synonymy. open-V on the trigonid of M1; loss of the sec- The holotype of Mimocyon longipes is neither Miacis, ond posterior accessory cusp on P4, talonid nor even a carnivoran. This late Uintan form is ex- simpler; reduction of molar cingulae. tremely similar to the Bridgerian hyaenodontid species The Aeluroida primitively retain the fol- Proviverra (?) major Wortman, 1902 (see Matthew, lowing feliform characters that are modified 1909, p. 473, pl. XLV, fig. 4), but is much larger and in has a slightly larger posterior cusp on P4. Thus, the P. the other Feliformia, the Didymictida: P4 (?) major lineage is continued into the late Uintan of anterior accessory cusp very small and well North America, when it is represented by "Mimocyon" separated from the main cusp by a broad, longipes. We do, however, include the Chinese species, U-shaped valley; P4 talonid relatively narrow Miacis lushiensis Chow, 1975, in the genus Miacis. and shallowly basined or with lingually slop- 36 AMERICAN MUSEUM NOVITATES NO. 2725 ing shelf; anterior and posterior ridges from paracone, metacone, and protocone of M1-2, the apex of the main cusp are straight and and the loss of the molar conules appears to slope directly downward to the accessory be associated with the great emphasis on hy- cusp, and are not constricted in any way; P3 percarnivory and shearing in the molar tri- protocone absent; lingual and buccal margins gons. The buccal ends of the preprotocrista of P4 irregular in outline; M2 trigonid cusps and postprotocrista are weakly developed, well-developed, trigonid relatively high. indistinct, and low and do not add signifi- DISCUSSION: The Aeluroida is a monophy- cantly to the topographic relief of the crown letic clade that includes living carnivorans in the trigon basin. and former members of the primitive, early The second posterior accessory cusp and Tertiary "viverravine" carnivorans. The Ae- notch on P4 are not present in any Aeluroida, luroida seem to be evolving a more hyper- and their loss is assumed to be derived from carnivorous morphology, such as that exhib- the primitive condition of the presence of ited by many feloids. The hypothesis that all this cusp in the Feliformia. As discussed ear- the taxa included within this group are more lier, the presence of the second posterior ac- closely related to each other than to any oth- cessory cusp is assumed to be primitive for er taxa indicates that it may be possible to the Carnivora and Feliformia, and reduction more precisely determine the phyletic rela- or loss of this cusp would be derived for any tionships of the early Tertiary Carnivora. given taxon. Didymictis parallels the Aelu- Based on the distribution of derived char- roida in the loss of this cusp. The simplifi- acters within the Feliformia, it is possible to cation of the talonid of P4, and the loss of distinguish two clearly different and well-de- posterior accessory cusps occurs in many fined lineages, consisting of the totally ex- groups of Carnivora. This must have been tinct Didymictidae (Didymictida) as one derived independently in various lineages, branch, and the much more diverse fossil based on its distribution in groups that are and living Aeluroida as the other. The mor- clearly not closely related, such as the Ae- phology of the Aeluroida is extremely var- luroida, some Didymictidae and many Can- ied, but the derived characters presented in iformia. However, because the Aeluroida the diagnosis seem to ally all the diverse can be assumed to be monophyletic on the members of the infraorder. The taxa placed basis of the most parsimonious distribution in the Aeluroida in this paper are: Ictidopap- of derived characters found within the Car- pus, the Viverravidae (including only Vi- nivora, there is no reason to make the ad hoc verravus and Simpsonictis), Palaeogale and assumption that the second posterior acces- the Feloidea (of Simpson, 1945 and most sory cusp on P4 is lost independently in the subsequent workers, but excluding the Nim- individual Aeluroida taxa. For this reason we ravidae). However, the precise phyletic re- consider the loss of the second posterior ac- lationships of the controversial genera Ic- cessory cusp on P4 to be a derived character tidopappus and Palaeogale within the uniting the Aeluroida. Aeluroida are uncertain, as is discussed thor- A very distinctive character uniting the oughly later in the systematic sections of this Aeluroida is the anteromedial position of the paper. paraconid of M1, and the relatively distant Conules are absent on both of the upper separation of this cusp from the metaconid. molars of the Aeluroida. All the taxa includ- The wide separation of the paraconid and ed within this group have simplified the metaconid results in an open-V on the tri- structure of the trigon basin on MI and M2, gonid, formed by the paralophid and proto- as both the paraconule and metaconule are lophid. Primitively in feliforms, and car- lost. This is an extremely derived condition nivorans in general, the paraconid and in comparision to the large molar conules metaconid of M1 are closer together and form found in the Didymictidae, all primitive Car- a much more closed-V on the trigonid. All nivora and other mammalian outgroups. Aeluroida have a much more open trigonid, Large shear facets are well developed on the and some groups within the Aeluroida (e.g., 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 37

Palaeogale and the Felidae) may become were assumed to be parallelisms evolved for even more derived in the opening of the tri- hypercarnivory. gonid-V by reduction of the metaconid and Ictidopappus seems to share a number of anteroposterior alignment of the paraconid derived features with the Feliformia and the and protoconid. Some advanced Caniformia Aeluroida, although the precise homologies of parallel the Aeluroida through a similar the characters are sometimes suspect. The ge- opening of the trigonid of M1. nus is believed to have reduced the molar All Aeluroida reduce the cingulae on the dentition to 2/2, as stated by Simpson (1935, upper and lower cheek teeth. The lingual and 1937) and Maclntyre (1966). However, both posterior talonid cingulae on P4 are weak, as the upper and lower jaws are broken imme- are the basal cingulae of the upper and lower diately posterior to M2, and the presence or molars. This contrasts with the primitive re- absence of M3 cannot be determined with tention of well-developed cingulae in the certainty. If the dental formula truly was Didymictidae (except Protictis aprophatos). Ml-2, this would correspond to the prim- The Aeluroida retain a suite of characters itive dental formula found in all feliforms. that are primitive for Carnivora and/or Feli- The trigonid of M2 is broken and represented formia, and are also found in the other only by an alveolus, but the talonid appears major branch of the Feliformia, the Didy- to be moderately elongate relative to the pre- mictidae. However, there are a number of sumed size of the trigonid. The inferred ratio primitive features that are maintained in the of M2 talonid/trigonid length appears to be Aeluroida, but are modified by the Didy- lower than that found in most feliforms, but mictidae. These characters (outlined in the higher than that found in the Caniformia. diagnosis, above) serve to distinguish the Also, the elongate oval outline of the talonid Aeluroida from the Didymictidae, but only of M2 is very similar to the feliform condi- by enabling one to state that a taxon retain- tion. There is a relatively large hypoconulid ing these features is not a member of the on M2 in Ictidopappus, that is equal to, or Didymictidae. greater than, the height of the hypoconid, as in other Feliformia. M2 is reduced and is much smaller than Ml, as in all feliforms that ICTIDOPAPPUS SIMPSON, 1935 retain M2. In the Caniformia, M2 is smaller DISCUSSION: Simpson (1935) first de- than M1, but it is not as transversely narrow scribed this very unusual, poorly known and anteroposteriorly shortened, as in the (represented by only one upper and one low- Feliformia. er jaw) carnivoran. The animal's dentition P4 in Ictidopappus is highly modified for was obviously well adapted for shear, as was shearing and is extremely worn and fractured discussed in detail by Maclntyre (1966). on the only known maxilla. The parastyle is These two are the only workers that have very large, high and an integral part of a long considered the morphology and possible re- shearing ridge on P4. It is separated from the lationships of Ictidopappus. Simpson (1937) paracone only by a shallow concavity in the believed that Ictidopappus would be an ideal ridge and a narrow valley on the buccal por- ancestor for Viverravus, except for the un- tion of the tooth crown. The parastyle is usual proportions and simple structure of P4 higher and more robust than in most feli- in the former genus. Simpson also hypothe- forms, but is similar in size (relative to the sized that P4 in Ictidopappus could be struc- paracone) to the parastyle in some Protictis, turally ancestral to that of Viverravus, al- Felidae, and Hyaenidae. However, the incor- though he saw no good evidence in support poration of a large "parastyle" into a shear- of this idea. MacIntyre (1966) criticized the ing ridge also occurs in some Nimravidae notion that Ictidopappus might be ancestral (here considered members of the Canifor- to or closely related to Viverravus as he saw mia). The anterobuccal "cusp" in these nim- no special resemblance between the two gen- ravids is not considered homologous with the era. Any similarities between the two forms large parastyle of the Feliformia (see discus- 38 AMERICAN MUSEUM NOVITATES NO. 2725 sion below). It is possible that a similar par- primitive for Carnivora, and therefore does allel development of a "parastyle" occurred not ally the various taxa that lack it. How- in Ictidopappus, but the general morphology ever, the Caniformia develop a continuous and topography of the cusp is similar to that lingual cingulum, which excludes Ictidopap- in other feliforms. In the absence of evidence pus from a position within that group. to falsify this homology, we reject the ad hoc P3 in Ictidopappus has a small lingual lobe assumption that this character is an adapta- for a protocone, although wear has obliter- tion for hypercarnivory evolved indepen- ated any manifestation of this cusp. How- dently in Ictidopappus. Instead, we consider ever, the position of this cusp is very differ- the enlarged parastyle a derived character ent from the position of the protocone for Ictidopappus and the Feliformia. developed in Protictis. In Ictidopappus the There is a lingual lobe on P4 for the pro- lingual lobe is directly lingual, or posterolin- tocone, although wear has removed all evi- gual to the apex of the paracone, whereas dence of the cusp. The size of the lobe in- the protocone in Protictis is located antero- dicates that the protocone may have been a lingually on the crown of P3, far anterior to well-developed cusp, a primitive carnivoran the apex of the paracone. The development character retained by the Feliformia, but the of an internal expansion, and possibly a pro- complete removal of this cusp by wear may tocone cusp, on P3 appears to be different alternatively signify that the cusp was small- and independently derived in both Ictido- er than in most feliforms. The protocone lobe pappus and Protictis. of Ictidopappus is unusual for a carnivoran, Within the Feliformia, Ictidopappus ap- in that it is relatively far posterior on the pears more closely allied to the Aeluroida crown of P4. Within Carnivora the primitive than to the Didymictidae. The genus has an protocone position is far anterior to the para- open-V on the trigonid of M1 formed by a cone, forming a distinct anterolingual projec- wide separation of the metaconid and the an- tion on the tooth crown. The functional and terocentrally located paraconid. This condi- phyletic significance of this character is un- tion is derived for all Aeluroida and is not clear, but it may be an autapomorphy for Ic- found in any of the Didymictidae. Ictidopap- tidopappus. pus also lacks conules on the upper molars, The structure of the carnassial notch of P4 as in the Aeluroida, but the large degree of is difficult to discern. There is a shallow, wear on the molar crowns makes this asser- broad notch present in Ictidopappus, but tion less certain, and subject to modification there is a break in the tooth at, or slightly through the discovery of new material. anterior to, the position of the deep, slitlike There are a number of characters in Ictid- portion of the notch found in the Feliformia. opappus that do not appear to be feliform, If this break is in the position of the vertical, and others that are more similar to those of slitlike notch in feliforms, Ictidopappus may the Caniformia. As mentioned above, var- have had a similar notch. However, the ious P4 characters of Ictidopappus are diffi- available evidence indicates that there was cult to discern and alternative interpretations not a deep, vertical notch in this genus, and to those given above are considered here. the notch consisted only of a low concavity The protocone of P4 is located medial to the in the paracone-metastyle ridge. Both prim- paracone and more posterolingually than in itive feliforms and caniforms have at least all other carnivorans, but is similar to the some vertically oriented notching on P4, position in the Creodonta. The total lack of therefore, the notch on P4 in Ictidopappus any trace of the protocone cusp on the well- may be very different from both of these con- worn crown indicates that the protocone ditions. may have been reduced in size, as is found Ictidopappus resembles other Feliformia in the Caniformia. The structure of the car- in the lack of a lingual cingulum continuous nassial notch in P4 is equivocal, and is equal- around the entire base of the protocone on ly dissimilar to the notch in both feliforms the upper molars. This is a character that is and caniforms. Homology of the "parastyle" 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 39 of P4 is uncertain; although we consider it a lingual accessory cusp. This is similar to the true, enlarged, feliform, cingular parastyle condition in P. (Bryanictis) vanvaleni, cusp, it is possible that Ictidopappus par- which has a definite lingual accessory cusp, alled some nimravids in the partial elabora- and differs from all other Carnivora, which tion of the anterior ridge of the paracone into lack any indication of this cusp. Overall, the a parastyle-like crest on a shearing blade (see trigonid of P4 in Ictidopappus appears much fig. 9). more molariform than the trigonid of any The relative size of the talonid and trigonid other carnivoran (except P. (Bryanictis), of M2 is uncertain, but the moderate elon- which is only molariform in the development gation of the talonid may be intermediate be- of a lingual accessory cusp): the anterior ac- tween the conditions exhibited by the Feli- cessory cusp is high and far lingual (para- formia and Caniformia. Ictidopappus may, conid-like), rather than anterior, low and therefore, represent a caniform with a rela- cingular; the lingual accessory cusp is tively long talonid. rudimentary, but in the position of a lingual Other workers have assumed that the mo- metaconid, rather than entirely absent; and lar formula of Ictidopappus was reduced to the cusps are aligned in a trigonid-V, rather MI-2. However, the breaks at the poste- than anteroposteriorly in a straight line. rior ends of both the mandible of the type The talonid of P4 is very short, wide, and (USNM 9296) and the referred maxilla weakly cuspidate. The talonid in Ictidopap- (USNM 9295) do not preclude the possibility pus is largely formed by a strong, transverse, that M3 were actually present in this speci- posterior cingulum and lacks significant ba- men. If M3 were retained, Ictidopappus sining. This differs greatly from the talonid would be excluded from the Feliformia, and structure in other Carnivora, in which the would be more similar to the Caniformia in talonid is primitively longer, distinctly cus- this primitive retention. Alternatively, Ictid- pidate, and well basined. Even in carnivo- opappus may be a very derived, early cani- rans that reduce the talonid of P4, there is form that had already reduced the molar den- generally greater cusp development, partic- tition to Ml 2 ularly on the high buccal ridge, and better The morphology of the Ictidopappus P4 is basin development. very distinctive, and very different from the The outline of P4 is short and triangular, morphology of P4 in all other carnivorans. rather than the elongate oval or elongate The anterior accessory cusp is small, weakly rectangular outline found in other primitive elaborated, high on the tooth crown, far an- carnivorans. This suite of P4 characters in terolingual and connected to the main cusp Ictidopappus is unique among the Carniv- by a strong, uninterrupted ridge. This differs ora, and adds little information useful for de- greatly from the morphology in other feli- termining the phyletic relationships of this forms, as well as other Carnivora. The Feli- genus, or any other carnivoran. formia all have a small, but strong, anterior accessory cusp that is at the base of the tri- gonid of P4. The cusp in feliforms (and can- FAMILY VIVERRAVIDAE iforms) appears to be an elaboration of an WORTMAN AND MATTHEW, 1899 anterior, basal cingulum, and is separated TYPE GENUS: Viverravus Marsh, 1872. from the main cusp by a distinct notch or INCLUDED GENERA: Viverravus Marsh, wide valley. In contrast, the anterior acces- 1872; Simpsonictis Maclntyre, 1962. sory cusp in Ictidopappus is much higher on DISTRIBUTION: Middle Paleocene to ?late the trigonid, is much farther lingual, and is Eocene. a swelling on the anterior end of a ridge from DIAGNOSIS: Members of the Aeluroida and the main cusp, rather than a cingular cusp. Feliformia that differ from all other members There is a very slight lingual swelling of the of these groups in the following derived fea- enamel on the crown of P4, internal to the tures-the anterior accessory cusp on P4 is main cusp, that may represent a rudimentary far anterolingual and isolated from the main 40 AMERICAN MUSEUM NOVITATES NO. 2725 body of the trigonid by a narrow constriction two genera in a restricted concept of Wort- of the crown just posterior to the anterior man and Matthew's Viverravidae. accessory cusp; M2 talonid extremely elon- Simpsonictis tenuis is only known from a gate due to enlargement of the hypoconulid small cluster of lower dentitions collected and the posterior lobe of the talonid, hypo- from the San Juan Basin, New Mexico. Rig- conulid large, expanded and bulbous with a by (1980) named a new species of Simpson- distinct anterior facet for occlusion with the ictis, S. jaynannei, from the Paleocene posterior margin of M2; M1 hypoconulid Swain Quarry from northwestern Colorado. strong, posterocentrally located, and pro- The known dental material does provide jecting posteriorly from the talonid margin; enough information to permit a number of anterobuccal basal trigonid cingulum on M1 important phyletic conclusions to be drawn. weak; anterior trigonid notch on M2 reduced, Simpsonictis is very similar in overall mor- resulting in a continuous, low paralophid. phology to Viverravus acutus, particularly in A number of primitive Aeluroida and Fe- P4 and M2, which provided us with an indi- liformia characters are retained by the Vi- cation that Simpsonictis might be closely verravidae that are modified by various other allied to Viverravus rather than Protictis. feliforms, including: small, low anterior ac- Rigby (1980) mentions that personal cessory cusp on P4, separated from the main communication from Maclntyre suggested cusp by a broad, shallow concavity; narrow that Simpsonictis might be related to Vi- P4 talonid, shallow talonid basin, and low lin- verravus, although no explicit reasons for gual talonid ridge rimming the basin; molar allying the two genera were given in that pa- trigonids high, particularly the trigonid of per. The derived characters allying Simpson- M2; P3 without a protocone; P4 occlusal out- ictis with all other Aeluroida have been dis- line very irregular, not straight or arcuate; cussed previously, and the derived features cingulae poorly developed; bulla unossified; that further unite the genus with Viverravus postglenoid foramen large. as a family within the Aeluroida are consid- DISCUSSION: Following the erection of the ered below. Viverravidae by Wortman and Matthew The anterior accessory cusp on P4 in the (1899), usage of the family was abandoned Viverravidae is located at the base of the tri- by most workers. However, the concept of gonid and very far lingual of the main antero- a closely allied group of genera was retained posterior axis of the tooth. The cusp is small, by Matthew (1909, 1915) and Simpson (1935, as in primitive Feliformia, but its distinctly 1937, 1945) through use of a subfamily Vi- lingual position may be derived. There is a verravinae within the Miacidae. tight, narrow, transverse constriction of the We propose the resurrection of the family tooth crown just posterior to the anterior ac- Viverravidae, although we restrict the Vi- cessory cusp on an anterior lobe of the verravidae to include only the genera Vi- crown of P4, which distinctly separates the verravus and Simpsonictis. We agree with anterior accessory cusp from the main cusp. Rigby (1980) that Simpsonictis is a valid, dis- There is a definite lingual inflection of this tinct, diagnosable genus. Furthermore, anterior lobe that coincides with the antero- Simpsonictis is widely separated from the lingual positioning of the anterior accessory cluster of taxa allied in the genus Protictis cusp. and here grouped with Didymictis in the new The hypoconulid of M1 in the Viverravidae family Didymictidae. We support Mac- is very well developed, high and located pos- Intyre's (1962) original recognition of the dis- terocentrally on the tooth crown. The hy- tinct nature of the material included in Simp- poconulid also projects posteriorly from its sonictis tenuis, and because of the shared position at the posterocentral margin of the derived features between this genus and Vi- talonid. This results in a distinct, but trans- verravus (and the extremely close similarity versely narrow, tablike projection at the base of some aspects of the morphology between of the talonid, that appears to articulate with Simpsonictis and V. acutus) we ally these the anterior edge of the trigonid of M2 in a 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 41 lock-and-key type of arrangement. This un- all the trigonid cusps and of the trigonid usual hypoconulid morphology is found in height as a modification of the M2 trigonid as both Viverravus and Simpsonictis, but in no a flat grinding surface continuous with the other Feliformia, nor in the Caniformia, and talonid of M1. it appears to be a derived feature for the Vi- The Viverravidae differ strongly from the verravidae. However, in one outgroup tax- remainder of the Aeluroida in features of the on, Leptonysson, a similar projection of the basicranium. Based on the limited known hypoconulid occurs on both M1 and M2. The cranial material for the Viverravidae (known lack of this feature in all the other outgroups only in Viverravus, see Matthew, 1909, as examined, and its occurrence only on M1 Simpsonictis is represented only by dental in the Viverravidae, would seem to indi- elements), the family lacked any indications cate that it is a nonhomologous character of an ossified bulla and retained a large post- evolved independently in both groups, rather glenoid foramen. The bulla in Palaeogale than a primitive retention only in the Viver- and the Feloidea is fully ossified, in contrast, ravidae (and no other Carnivora). and preserved basicrania are unknown for All Viverravidae have a very weak antero- Ictidopappus. Didymictis also appears to buccal basal trigonid cingulum on M1,2. The lack an ossified bulla, based on an unde- presence and strong development of this cin- scribed, well-preserved skull of Didymictis gulum is primitive for all Carnivora, and is (AMNH 17030) from the Huerfano (see be- found in most of the Cretaceous outgroup low). As mentioned above (see Systemat- taxa. In primitive Carnivora, the cingulum is ics-Carnivora), an unossified bulla is prob- present as a low, basal elaboration of the ably primitive for the Carnivora, but the crown of M12 continuous along the anterior hazards of basing such a statement on the and buccal base of the trigonid. The Viver- limited material available for early, primitive ravidae reduce this cingulum to only a nar- members of the group are obvious. The ab- row, irregular, and sometimes discontinuous sence of an entotympanic bulla, or any in- rim at the base of the molar trigonids. Some dications of its presence in the surrounding of the Didymictidae elaborate the antero- bones of the basicranium, indicates that the buccal basal trigonid cingulum into a low, ear region in the Viverravidae remained wide shelf, whereas most other feliforms re- more primitive than those of the closely re- tain the primitive condition for this cingu- lated Palaeogale and the Feloidea. lum. Viverravus also retains the large postgle- The anterior trigonid notch on M2 is re- noid foramen primitively found in the Car- duced in size in the Viverravidae, and the nivora. The foramen is situated far postero- paraconid is relatively low. This results in a externally on the postglenoid process, at the continuous, low, horizontal paralophid on posterior margin of the ventral contribution M2 and a reduction in the depth of the car- of the squamosal to the postglenoid region nassiform notch. The primitive carnivoran (see fig. 7), and widely separated from the condition of a high M2 paraconid, a high par- posterior wall of the glenoid fossa. In the alophid with a deep vertical embrasure, and Caniformia, and other Mammalia, the post- a deep, sharp carnassiform notch found glenoid foramen is also large, but it is located primitively in the Didymictidae and other far anteromedially on the postglenoid pro- Feliformia is highly modified in the Viver- cess, directly posterior to the glenoid fossa, ravidae. Although the reduction of the para- and high on the posterior wall of the fossa. lophid is not as extreme in Simpsonictis as Even in Caniformia with a long basicranium, in Viverravus, it is greater than that ob- similar in length to that of Viverravus, served in other Aeluroida, and it appears to the foramen is far anterior and just lateral to be a derived feature uniting the two groups. the location of the fissura glaseri. Therefore, Some species of Didymictis also reduce the the posteroexternal position of the postgle- paraconid of M2 and paralophid, but this is noid foramen does not appear to be directly in conjunction with a reduction of the size of correlated with basicranial length. The po- 42 AMERICAN MUSEUM NOVITATES NO. 2725

FG

- GFO

- PGF

- PR

I IPP

0 5 10 o 5 o s Il A B C FIG. 7. Comparative basicrania of A) Vulpavus (AMNH 12626), B) Palaeogale (AMNH 25449), and C) Viverravus (AMNH 12621). Views of left, ventral surfaces of the basicrania; illustrations partially restored based on right side of skull, when specimen is damaged. Scales in millimeters. FG-fissura glaseri; GFO-glenoid fossa; OB-ossified bulla; OC-occipital condyle; PGF-postglenoid foramen; PP-paroccipital process; PR-promontorium. Hatching indicates foramina or areas covered by matrix.

sition of the postglenoid foramen in Viver- gale (see also the section on Palaeogale, be- ravus appears to be very different from, and low). much more derived than, the primitive po- We agree with Maclntyre's (1966) criti- sition for the Carnivora. In the Feloidea, the cism of Simpson's (1937) idea that Ictido- postglenoid foramen is extremely reduced in pappus might be ancestral to Viverravus, as size, or absent. When present in feloids, this Ictidopappus is certainly a very unusual and foramen is located somewhat anteromedial- derived form. However, as discussed in the ly. It is unclear whether the foramen even section on Ictidopappus, we do consider this passes any significant circulatory vessel, or genus to be related to Viverravus and Simp- whether it just serves as a nutrient foramen sonictis (the Viverravidae), for both of these in those feloids that do retain it. Palaeogale taxa are members of the Aeluroida and are has a large postglenoid foramen in a position phyletically distant from the other early Ter- very similar to that found in Viverravus. tiary Feliformia. The characters distinguish- Didymictis (based on AMNH 17030) has a ing Ictidopappus from the Viverravidae large postglenoid foramen that is located far were already considered in the systematic anteromedially, just posterior and external to discussion of that genus, and in the earlier the fissura glaseri. The posteroexternal lo- parts of this discussion detailing the derived cation of the postglenoid foramen is certainly features uniting the Viverravidae. Ictidopap- derived over the primitive position for the pus is here excluded from the Viverravidae, Carnivora, and it probably represents a syn- but included within the Aeluroida, because apomorphy for the Viverravidae and Palaeo- it lacks the derived features allying Viver- 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 43 ravus and Simpsonictis and it has the derived the anteroposterior alignment of the cristid characters of the infraorder. Ictidopappus is along the buccal margin of the tooth. The not more closely related to Viverravus and talonid becomes trenchant in Viverravus, Simpsonictis than to any other members of due to the emphasis on the large, anteropos- the Aeluroida. terior ridge formed by the cristid obliqua. Viverravus has a number of important fea- Some species of Viverravus parallel ad- tures that are more advanced than those vanced Caniformia in the development of a found in Simpsonictis. P3 is higher than P4; distinct hypocone cusp-. This cusp is an elab- metaconid height is lower than the paraconid oration of the well-developed posterolin- on M1, and the metaconid is directly lateral gual cingulum present at the base of the pro- to the protoconid; the cristid obliqua on M1l2 tocone in all Viverravus. is aligned more directly anteroposteriorly, In many ways Simpsonictis remains prim- and the sulcus obliquus becomes more trans- itive with respect to Viverravus. Simpson- versely narrow; and some species develop a ictis lacks all the derived features mentioned distinct hypocone as an elaboration of the above for Viverravus, but there are a number postcingulum. of advanced features that characterize the The enlargement of P3 to a greater overall genus. Two of the most distinctive charac- size than P4, in Viverravus, is unusual and ters of the lower dentition of Simpsonictis derived for the Carnivora. All other early are the moderately large anterior accessory Tertiary Caniformia and Feliformia, except cusp on P4 and the development of high, Procynodictis and Daphoenictis, have P3 sharp, piercing cusps on all the cheek teeth. much shorter in length and lower in height The development of an enlarged P4 anterior than P4. In Viverravus, P3 increases in length accessory cusp may be derived for Simpson- (although it generally remains shorter than ictis, and in some ways it parallels the en- P4), and the height of the main cusp on P3 is largement of this cusp in Protictis. However, always greater than on P4. the cusp in Simpsonictis is never so high or The metaconid of M1 is reduced in all Vi- large as in Protictis, nor is it elaborated into verravus, and the paraconid is relatively en- an anteroposteriorly aligned, laterally com- larged, particularly in more derived species. pressed, robust blade as in Protictis. Fur- The combination of the open-V on the tri- thermore, P4 lacks the deep notch between gonid of M1 in all Aeluroida, enlargement of the anterior accessory cusp and the main the paraconid, and reduction of the metaco- cusp found in all the Didymictidae. For these nid results in an enlarged and elaborated, reasons, and because of the many derived more anteroposteriorly aligned, lower car- features supporting the phyletic separation nassial blade. Shear on the paralophid and of Simpsonictis and Protictis, it is assumed the carnassial notch of M1 is thereby further that the moderately large P4 anterior acces- emphasized, and shear on the posterior face sory cusp in Simpsonictis is a feature of the trigonid characteristic of the Didy- uniquely derived within the Aeluroida. mictidae is reduced. The metaconid is farther Simpsonictis is also the only member of anterior on the trigonid, and is located di- the Feliformia to modify its dentition for rectly lateral to the protoconid, rather than piercing (and presumably insectivorous hy- posterolingual to it, as in other Carnivora. percarnivory). All the cusps are high, thin, Any posterior shear occurring on the trigonid sharp and conical, rather than the more ro- of M1 is therefore oriented more transverse- ly, rather than obliquely. bust and shorter cusps found primitively in In all Viverravus the cristid obliqua on the Feliformia. Maclntyre (1962, 1966) pro- M1_2 is aligned parallel to the anteroposte- vided a detailed discussion of this aspect of rior axis of the tooth, rather than following the dentition of Simpsonictis, and it is suf- the primitive oblique course across the tal- ficient to point out here that this modification onid. The sulcus obliquus is transversely of the cusp morphology is very derived for narrow and deep, presumably as a result of the Carnivora, is unique to Simpsonictis, 44 AMERICAN MUSEUM NOVITATES NO. 2725 and it serves to distinctly define Simpsonic- Palaeogale has a number of features that tis as a monophyletic taxon. appear to be derived in common with all oth- PALAEOGALE VON MEYER, 1846 er Feliformia. A moderately large parastyle cusp on P4 is present in Palaeogale, although DISCUSSION: Simpson (1946) provided the its size is variable in different species of the most recent review of the morphology and genus. The size of the parastyle varies from relationships of Palaeogale in his discussion small to moderately enlarged, but in all cases on Palaeogale, Plesiogale and Plesictis. In the parastyle is an elaboration of the anterior this study Simpson synonymized the Amer- basal cingulum on P4 and is not a modifica- ican genus Bunaelurus with Palaeogale. tion of the anterior paracone ridge, as in the Previous workers had used the presence of Nimravidae. Those species of Palaeogale M2 in Bunaelurus to distinguish this genus that have a large parastyle do enlarge the from Palaeogale, which lacks M2. Simpson cusp to a similar degree to that found in all did not consider this single character differ- other Feliformia. ence to be generically significant, and he in- As pointed out by Simpson (1946, p. 12), cluded a lengthy discussion of his reasons for there is a narrow, deep cleft on P4 separating concluding that the presence or absence of the paracone portion from the metastylar M2 was insufficient grounds for separating part of the elongate blade. This is the same these two genera (see also the discussion in as the derived condition of the metastyle this paper, sections on Plesiomiacis and notch in the Feliformia, and is not similar to Quercygale). Palaeogale, Plesiogale, and the notch exhibited by the Caniformia. Fur- Plesictis were all believed to be "the earliest thermore, all the living Mustelidae lose the definitely recognizable mustelids .. carnassial notch on P4, altogether, and only (Simpson, 1946, p. 1), based on the dentition a broad, shallow concavity on the metastylar and cranial foramina of these forms. In his blade remains in these forms. redefinition of Palaeogale and Plesiogale, All Palaeogale lack M3 and M2 is reduced Simpson (1946, pp. 12-13) pointed out a in size, or is absent. Both of these are de- number of features that distinguish the two rived morphologies that are found in the Fe- genera, and that also support the assignment liformia. However, advanced caniforms of Palaeogale to the Feliformia rather than (such as the Mustelidae) also lose M3, and the mustelid Caniformia. strongly reduce, or lose, M2. Depending on Hunt (1974b) briefly mentioned Palaeo- the determination of the phyletic affinities of gale, which he believed was very similar to Palaeogale based on the distribution of oth- Viverravus. There was no discussion of the er derived characters, as well as this one, basis for allying Palaeogale with Viverrav- one could consider the genus a member of us, other than the suggestion that the two either the Feliformia or of a derived branch exhibited many morphological similarities. of the Caniformia. The lack of many derived The precise phyletic position of Palaeo- mustelid characters in Palaeogale, as well as gale is debatable. It appears that much of the the presence of derived feliform features, reason Simpson considered it a mustelid was lead us to hypothesize a phyletic position its supposed close alliance with Plesiogale within the Feliformia for this genus. Palae- and Plesictis, which may well be early mem- ogale would be more advanced than most bers of the Mustelidae. However, Simpson of the primitive feliforms, as it has reduced did not provide any clear documentation or M2 to a minute size, or even lost it alto- solid evidence for his assignment of Palaeo- gether, in contrast to the larger and more gale to the Mustelidae, and we believe that functionally important M2 of early feliforms. many of the characters presented in his di- The lower dentition of Palaeogale has a agnosis, as well as others considered during number of features that place the genus with- this investigation, support an alternative in the Feliformia, in general, and specifically placement of the genus in the Feliformia. within the Aeluroida. M2 is very reduced in 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 45 overall size, but the tooth remains functional elaborate, continuous lingual cingulum that and has an elongate oval outline, rather than is derived for the Caniformia. This lack of a the short squared or rectangular outline char- lingual cingulum does not necessarily ally acteristic of the Caniformia. It is difficult to Palaeogale with the Feliformia, as it is only discern the standard molar cusp pattern on a retention of a primitive character, but it M2 of Palaeogale, but the length of the tal- does indicate that Palaeogale may not oc- onid portion does seem to be equal to or cupy a position within the Caniformia, as greater than the length of the trigonid. This here defined. The Nimravidae also lack the would be a derived feature common to both modified cingulum, and it is conceivable that Palaeogale and the other Feliformia. M2s of Palaeogale may be allied to this family the other "mustelids" discussed by Simp- (based on other characters, see below). son, Plesiogale and Plesictis, are not de- The protocone of P4 in Palaeogale is rel- scribed as having the derived, feliform, elon- atively small and is closer in relative size to gate oval outline mentioned for Palaeogale. the reduced protocone of caniforms than to Palaeogale also has the open trigonid-V that of feliforms. on M1 characteristic of all the Aeluroida. The If Palaeogale is a member of the Felifor- paraconid is relatively large, far anterior on mia, the precise phyletic position of the ge- the trigonid, and located more buccally (clos- nus must then be determined. Palaeogale er to the midline) than the paraconid in prim- does not have any of the derived features itive feliforms and Carnivora. The carnassial that are diagnostic for the Didymictidae, but notch between the paraconid and protoconid it does possess the characters that are de- is deep and the paralophid is well developed rived for the Aeluroida. The open-V trigonid and aligned more anteroposteriorly than in on M1 is present, as discussed above, and some feliforms (e.g., the Didymictidae). An- the absence of conules on the upper molars terobuccal shear on the paralophid is greatly is also characteristic of Palaeogale. Is Pa- emphasized and posterior shear on the pro- laeogale, then, more closely related to Vi- tolophid is very reduced, as in all Aeluroida. verravus, as suggested by Hunt (1974b), or This results in an open-V and anteroposte- to the Feloidea, within the Aeluroida? riorly oriented paralophid on the trigonid of Although Hunt (1974b, p. 1039) stated that M1. Palaeogale further modifies the trigonid the ancestry of Palaeogale could be traced of M1 by losing the metaconid entirely. Many directly to the closely related Viverravus other Aeluroida (particularly within the Fe- minutus, he gave no direct evidence for this loidea) reduce the metaconid of M1 but none and provided only a brief discussion of his have totally eliminated the metaconid as a reasons for allying the taxa. Hunt refers to distinct cusp. Some advanced Nimravidae the similarities in the basicranial structure also lose the metaconid of M1, but since this and a "complex of cranial features" (1974b, only occurs in the more derived members of p. 1039), but the reader may only indirectly the family, it may be hypothesized that the infer the precise nature of these characters. loss of the metaconid occurred independent- Palaeogale and Viverravus have a large ly in Palaeogale and only some nimravids. postglenoid foramen that is located far pos- The upper molars in Palaeogale are more teroexternally on the postglenoid process of similar to feliforms than caniforms in their the squamosal. The large size of the foramen lack of a continuous lingual cingulum around is primitive for all Eutheria, but the position the base of the protocone. Even though the of the foramen is certainly derived within the upper molars of Palaeogale are reduced in Carnivora. It seems likely, then, that the size, particularly in the lingual portion of the posteroexternal position of the postglenoid teeth, there is no indication of any develop- foramen is a derived feature uniting Palaeo- ment of a single, strong lingual cingulum that gale with the Viverravidae. This would agree entirely encircles the protocone. In contrast, with the relationship between the two genera Plesiogale and Plesictis do seem to have the proposed by Hunt. As no other Carnivora, 46 AMERICAN MUSEUM NOVITATES NO. 2725 or Eutheria that we have observed, exhibit other known member of the Aeluroida with this derived morphology of the postglenoid a well-preserved basicranium, Viverravus, foramen we feel that this character unites lacks any indication of an ossified bulla. If Palaeogale and the Viverravidae. the lack of an ossified bulla is primitive for Palaeogale and Viverravus share the de- the Carnivora and the Feliformia, the pres- rived basicranial morphology of extreme re- ence of an ossified bulla in Palaeogale and duction of the paroccipital and supraoccipital the Feloidea would indicate a closer phylo- processes (see fig. 7). Other Carnivora (e.g., genetic relationship between these taxa than Feloidea, Vulpavus, other Caniformia, etc.) between Palaeogale and any other Aeluroi- have these processes moderately well devel- da. However, Palaeogale lacks the ectotym- oped, as do the Creodonta and members of panic septum in the bulla found in all mem- other eutherian outgroup taxa (see van der bers of the Feloidea. This indicates that the Klaauw, 1931, pp. 95-96 for a detailed dis- two groups might share only the ossification cussion of the paroccipital morphology in the of the bulla, and Palaeogale remains more Creodonta and the primitive Carnivora). Ac- primitive in the lack of any development of cording to Matthew (1910b; as cited in van the ectotympanic into an internal septum in der Klaauw, 1931, p. 95): "the paroccipital the bulla. process is short and backwardly directed, The large size of the postglenoid foramen ... a primitive eutherian character common also excludes Palaeogale from a position to nearly all Eocene mammals." The great within the Feloidea, but it does not preclude reduction in size of these processes, and the a closest sister group relationship between resultant similarities in the morphology of Palaeogale and the Feloidea. these areas of the basicranium are derived The only other reasonable alternative hy- features that would closely relate Viverravus pothesis of relationship for Palaeogale, and Palaeogale, to the exclusion of any oth- would be a close affinity with the Nimravi- er Carnivora for which basicrania are dae. Palaeogale has the reduced upper and known. lower dental formulae characteristic of the Hunt may also be referring to basicranial nimravids. The Nimravidae reduce the molar length and overall similarity of the basicra- dentition to MI-2, however, Palaeogale nium of the two genera in his suggestion of variably reduces it to M1-2 2 their close relationship. The basicrania of The bulla is totally ossified in Palaeogale Palaeogale and Viverravus are generally and some Nimravidae. However, total ossifi- similar, but most of these features are prim- cation of the bulla in the Nimravidae is only itive retentions in both. Without an explicit known for certain in the advanced nimravid analysis and discussion of the polarity of the Barbourofelis. Most nimravids, such as Di- supposed basicranial features allying the two nictis and Nimravus, never preserve a total- genera, it is impossible to evaluate their util- ly ossified bulla; the bulla is always broken ity in determining the precise phyletic rela- and fragmented ventrally. Hunt (1974b) be- tionship of Palaeogale and Viverravus. Such lieved that this was just an artifact of pres- an analysis is not undertaken in this project. ervation, and that all nimravids would have An alternative hypothesis of relationship had a completely ossified bulla, in life. How- for Palaeogale within the Aeluroida would ever, it is just as reasonable to assume that be a close alliance with the Feloidea. Pa- the bulla was only partially ossified in the laeogale may share the derived development primitive Nimravidae, and that the preser- of an ossified, bipartite bulla with the feloids. vation observed in the large collections of The bulla is known to be totally ossified in these forms is indicative of the true condition Palaeogale, although there is no ontogenetic of bullar ossification in the living animals. evidence as to the percentage of the bulla Both the Nimravidae and Palaeogale retain contributed by either the ecto- or ento-tym- the primitive condition of a lack of any true panic components. All feloids also possess septum within the bulla. a completely ossified bulla, whereas the only Palaeogale and the Nimravidae share a 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 47 similar M' morphology, but some of these ravidae and the remainder of the Caniformia, features are also shared by the feliforms, Ic- as the earliest nimravid, Palaeogale, would tidopappus and the Feloidea, and may just retain a relatively large parastyle. be convergent adaptations to hypercarnivo- The Nimravidae also lack the derived, slit- ry. The general tooth outline is similar, in like, P4 metastyle notch found in Palaeogale that there is a large parastyle wing, very and the Feliformia, and their postglenoid fo- small metastylar wing, robust buccal margin, ramen remains in the primitive anteromedial anteroposteriorly shortened lingual margin, position on the postglenoid process, and is and transverse elongation of M1. The proto- not located far posteroexternally as in Pa- cone is the only internal cusp, and the lingual laeogale and Viverravus. cingulum is extremely weak, or absent, and The primitive lack of a teres process of the is not continuous around the base of the pro- scapula in Palaeogale excludes this genus tocone. The size of MI is also very small, in from a close phyletic relationship to the Nim- comparison to P4 size and the size of M' in ravidae and other Caniformia (see below). primitive Carnivora. The lack of continuity It seems most parsimonious, at this time, of the cingulum around the base of the pro- to include Palaeogale within the Feliformia, tocone is a primitive retention and cannot be as the closest sister group to the Viverravi- considered as indicative of special relation- dae. We prefer to group Palaeogale and the ship between Palaeogale and the Nimravi- Viverravidae in the superfamily Viverravoi- dae. Reduction in the size and the similarity dea Wortman and Matthew, 1899.8 However, in shape, of M1 may be derived for these two the character contradictions and uncertain- taxa, but similar reduction and morphology ties discussed above, indicate that this is an also occurs in some Feliformia. equivocal hypothesis of relationship readily The metaconid of M1 is absent in Palaeo- subject to modifiction based on rigorous test- gale and in some Nimravidae, whereas other ing of the hypothesis and more detailed analy- nimravids retain a small M1 metaconid. In ses of character distributions between the this case, either Palaeogale is more closely taxa considered above. related to some advanced nimravids, than all the nimravids are to each other, or the loss of the metaconid on M1 occurred indepen- dently in both groups. If the latter is the case, then the loss of the metaconid in Palaeogale SUPERFAMILY FELOIDEA HAY, 1930 and some Nimravidae cannot be considered INCLUDED TAXA: Hyaenidae, Felidae, and a derived character between the two taxa. Viverridae. All Nimravidae lack a true, basal cingulum DISTRIBUTION: Late Eocene to Recent. parastyle on P4, whereas at least some DIAGNOSIS: Aeluroida that differ from all species of Palaeogale exhibit a large para- other members of the Aeluroida in the de- style. If Palaeogale is considered a close rel- velopment of the following derived fea- ative of the Nimravidae, within the Canifor- tures-ossified, bipartite bulla with a large mia, then the species with a relatively large ectotympanic septum; great reduction or loss parastyle must be among the most primitive Caniformia, as they would have retained the 8 Article 36 of the International Code of Zoological primitive eutherian condition of a moderate- Nomenclature states that a name proposed for a taxon ly well-developed P4 parastyle. The more ad- in any category in the family-group (tribe, subfamily, vanced "species" of Palaeogale that reduce family, superfamily) has priority in the naming of any the parastyle would then be more re- other category in the family-group. Therefore, the erec- closely tion of a family Viverravidae by Wortman and Matthew lated to the Nimravidae, which lose the cusp in 1899 established nomenclatural priority for all other entirely, and the genus Palaeogale would be taxonomic ranks in this family-group. Our use of the paraphyletic. Furthermore, the reduction superfamily Viverravoidea is simply recognition of the and loss of the parastyle of P4 would have to existence of a taxon implicit in the prior naming of the be independently derived in both the Nim- family Viverravidae. 48 AMERICAN MUSEUM NOVITATES NO. 2725 of the postglenoid foramen (i.e., effective the inflected edges of the ectotympanic and loss of venous drainage of the cranium via the caudal entotympanic. Hyaenids appear this foramen in all Feloidea); scaphoid and to have a septum with only a tympanic con- lunar united into a single ossification, the tribution, and would therefore lack a true scapho-lunar; paroccipital process broadly septum bullae, according to Hunt. Nonethe- based and closely appressed to the posterior less, all feloids have a complete, large ecto- wall of the bulla; development of an intra- tympanic septum that divides the bulla into cranial arterial rete in the orbital region, in- two separate chambers. This contrasts ternal carotid artery reduced in size, or ves- strongly with the internal construction of the tigial. bulla in both Palaeogale and all Caniformia DISCUSSION: The Feloidea is a monophy- with an ossified bulla. Palaeogale has a com- letic taxon that includes all the living and fos- pletely ossified bulla (although without early sil members of the families Hyaenidae, Vi- ontogenetic material, the relative propor- verridae, and Felidae, of Simpson (1945) but tions of the ecto- and ento-tympanic contri- excludes the Nimravidae as used in this pa- butions are indeterminable), but lacks any per. An alliance of the felids, hyaenids, and indication of the development of an internal viverrids results in a grouping of taxa with septum, either ectotympanic or entotympan- a varied suite of dental, skeletal, and soft ic. Within the Caniformia, members of the anatomical modifications. Such an assem- Canidae develop a small caudal entotympan- blage of taxa that have evolved numerous ic septum, but it is in a different topographic autapomorphies in the various familial lin- position than in the Feloidea, it is never so eages represents only the most basic hypoth- large as in the Feloidea, nor is there any ec- esis of relationships for these groups. The totympanic component to the septum such characters used in the diagnosis above dis- as that found in all the Feloidea. tinguish the Feloidea from all other Aeluroi- The postglenoid foramen is extremely re- da and Feliformia, and indicate the close duced in size, or absent, in all Feloidea. phyletic relationship of the three included Those feloids that do retain a small foramen families. We do not, however, attempt to re- on the postglenoid process may have it var- solve this trichotomy in any further detail, iably developed on either side of the skull of as we recognize that the relationships of the a single individual. However, when the fo- diversity of taxa included within the Feloidea ramen is present, it is positioned far antero- must be considered with much closer scru- medially on the basicranium. This is similar tiny than is possible, or desirable, in this pa- to the location of the large postglenoid fo- per. Various aspects of the phylogeny of the ramen of the Caniformia, Creodonta and oth- three living feliform families are currently er eutherians, but differs from the extreme under study by numerous other workers. posteroexternal position found in the long- An important feature uniting all the Feloi- skulled Viverravus, and the slightly more dea is the development of a complete, large, moderate posteroexternal position exhibited ectotympanic septum within a totally ossi- by Palaeogale, with a shorter basicranium fied, bipartite (ectotympanic and entotym- (see previous section). panic) bulla. All Feloidea share the derived It appears, therefore, that the loss of ve- condition of a large entotympanic (especially nous drainage of the head through the post- the caudal entotympanic) component, and a glenoid foramen is a derived feature that smaller ectotympanic component forming unites the Feloidea. Venous drainage of the the bulla. The composition and morphology cranium and inferior petrosal sinus would be of the septum varies within the Feloidea, but captured by another vessel, and at that point all the members of the superfamily have a there would not be a significant vein or ve- large, primary ectotympanic component of nous circulation passing through the postgle- the septum. Hunt (1974a) showed that both noid foramen. This would result in a great felids and viverrids (except Nandinia) have reduction of the size or total loss, of the post- a true, bilaminar septum bullae formed by glenoid foramen. All other Carnivora, except 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 49 some aquatic carnivorans and some Nimrav- idae, retain the large, primitive postglenoid 'C, foramen and the passage of significant ve- nous circulation. The Feloidea are derived in the develop- ment of a single scapholunar ossification. Fusion of the independent scaphoid and lu- 7a-d 7a-d nar bones occurs a number of times in the I X 742a-f Eutheria, but it may be hypothesized to have -e \ a-e 11 a-e 2a-f occurred only once in the phylogenetic his- tory of the Feliformia. All the primitive fe- ossified liforms for which feet are known have inde- bulla pendent ossification of the scaphoid and lunar, as do many of the primitive caniforms. This feature was used by most of the early fused workers to ally all the primitive Carnivora in scapholunar the Miacidae. A separate scaphoid and lunar may therefore be assumed to be the primitive condition for Carnivora. More advanced Caniformia also evolve a single scapholunar FIG. 8. Hypothesis of monophyly for the but from the "Fissipeda." Traditional phylogeny in which the bone, probably independently Feloidea and Canoidea (of Simpson) are most Feloidea, as the more primitive members of closely related to each other, to the exclusion of both suborders have a separate scaphoid and the "Miacoidea," based on the derived characters lunar. The hypothesis that a scapholunar was "ossified bulla" and "fused scapholunar bone." evolved twice within the Carnivora, rather The numbers and letters are character contradic- than only once, is based on parsimony ar- tions and refer to the derived features summarized guments; fewer assumptions of parallel evo- in the Phylogeny of the Carnivora section. Com- lution of derived features are required if the pare this phylogeny to the more parsimonious, "miacids" are included in the Caniformia alternative arrangement proposed in figure 4 and and Feliformia, than if the later Tertiary car- the Phylogeny of the Carnivora section. Some nivorans are considered a "miacines" refers to those "miacines" that have monophyletic a fused scapholunar-Vassacyon, Procynodictis, group from which most "miacids" are ex- and some Vulpavus; other "miacines" are those cluded (see fig. 8). that have independent scaphoid and lunar bones. All Feloidea have a laterally expanded, broadly based paroccipital process that is very closely appressed to the posterior wall of the bulla. The process is curved and cup- an intracranial arterial (anastomatic branch shaped and forms a cradle that extends an- of the external carotid) rete in the orbital re- teriorly along the posterior and ventral walls gion, and may reduce, or even lose entirely, of the bulla (see Van der Klaauw, 1931, pp. the median internal carotid artery. The me- 100-101). This cupping of the bulla by the dian internal carotid artery is absent in all paroccipital and enlargement of the process, Hyaenidae; greatly reduced, or vestigial, in is derived over the condition found in out- the Felidae; greatly reduced or absent in group taxa such as the Creodonta, Canifor- Nandinia, and less reduced in all other Vi- mia, and Palaeogale, and it appears to be a verridae (Hunt, 1974a, p. 47). good synapomorphy for the Feloidea. A combination of the primitive Aeluroida Davis and Story (1943) and Hunt (1974a) features and the derived characters for the provided description and discussion of the Aeluroida and the Feloidea may be assumed internal carotid circulation in Carnivora. to represent the primitive morphotype (for Evidence from these studies indicates that the features considered here) for all the Fe- all Feloidea are derived in that they develop loidea. All the families within the Feloidea 50 AMERICAN MUSEUM NOVITATES NO. 2725 have evolved divergent and characteristic a number of taxa that have previously been specializations. considered members of the Feloidea or Vi- verravidae appear to be members of the Can- SUBORDER CANIFORMIA KRETZOI, 1943 iformia, as is discussed later. INCLUDED TAXA: Uintacyon, Miacis, Pa- The inclusion of the early Tertiary "mi- laearctonyx, Tapocyon, Vulpavus, Vassa- acines" within the Caniformia results in a cyon, Oodectes, Quercygale, Procynodictis bipartite division of the Carnivora into two (including "Plesiomiacis"), Nimravidae, monophyletic taxa that include all the known and all other taxa included in Tedford's early Tertiary, late Tertiary, and living car- (1976) Caniformia. nivorans. Such an arrangement provides a DISTRIBUTION: Early Eocene to Recent. vertical hypothesis of relationship for all the DIAGNOSIS: Carnivorans that differ from Carnivora and may serve to bridge the phy- all other Carnivora in the development of the logenetic gap between the archaic and the following derived features-loss of the cal- modern carnivoran families lamented by caneal fibular facet in the pes; great reduc- Tedford (1976, p. 364). This paper, however, tion or absence of the parastyle on P4; re- does not attempt to determine the phyletic duction in size of the protocone of P4; lingual relationships of the "miacines" in any more cingulum on all upper molars complete detail than to assign them to the Caniformia. around the base of the protocone; large teres Determination of the precise relationships of major process on the scapula (in the taxa for the various "miacine" genera would be an which this region of the scapula is pre- important and useful study and would pro- served). vide insight into the early evolution of the Members of the Caniformia primitively re- caniform Carnivora. tain the following carnivoran characteristics: Matthew (1909) pointed out that the "mia- metastyle blade of P4 elongate with a broad, cine miacids" differed from the "viverra- shallow carnassial notch; M2 talonid shorter vines" in the loss of the calcaneal fibular fac- than the trigonid; M1-3 present, M2 large et in the pes. This character is a derived and well developed; hypoconid on M1_3 ro- feature uniting the "miacines" and all the bust, forming the largest talonid cusp; M1l2 rest of the Caniformia. Matthew's state- talonid basins moderately wide; M1l2 trigon- ments about the "miacine" foot were based ids high, paraconid < metaconid < proto- largely on articulated skeletal material as- conid size, trigonid cusps form a relatively signable to Vulpavus, as postcranial ele- closed-V; P4 anterior accessory cusp small ments definitely attributable to other "mi- and separated from the main cusp by a broad acine" genera were unknown. However, valley; P4 talonid long and with two posterior based on information from the Vulpavus accessory cusps; conules present on all up- specimen illustrated by Matthew (1909, per molars; independent ossification of the AMNH 12626, type of Vulpavus profectus), scaphoid, lunar, and centrale; bulla unossi- and on specimens assigned to Miacis, Uin- fied. tacyon, and Hesperocyon, primitive cani- DIscUSSION: The Caniformia represent the forms had already lost the calcaneal fibular subordinal division of the Carnivora comple- facet primitively retained in the Feliformia, mentary to the Feliformia. As originally con- Creodonta, other eutherians, and the mar- structed by Kretzoi (1943, 1945) the Canifor- supials. A sampling of living Caniformia mia included all the members of Flower's (Taxidea, Ursus, Canis, and Mephitis) in- Arctoidea and Cynoidea, the Miacidae of dicates that these taxa, and presumably all Matthew, and the aeluroid Hyaenidae and the Caniformia, have lost an articular contact Viverridae. Tedford (1976) restricted usage between the fibula and the calcaneum, and of the Caniformia to the Canoidea and Pin- no longer retain a large dorsal facet on the nipedia of Simpson (1945). We follow Ted- calcaneum for articulation with the fibula. ford's usage, but we also include the "miac- Some of the caniforms (e.g., some Nimrav- ine miacids" within the Caniformia. Further, idae, Mephitis) even develop a lateral pro- 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 51 jection on the astragalus at the anteroexter- iforms elaborate on this condition and de- nal base of the astragalar fibular facet that velop a lingual cingulum that completely acts as a mechanical stop to prevent the fib- encircles the base of the protocone, although ula from moving too far anterior, and that in some Caniformia (e.g., Oodectes) the con- effectively eliminates any possibility of the nection of the posterior and anterior cingulae fibula articulating ventrally with the calca- is only very weakly developed. In contrast, neum. the Feliformia retain the primitive separation As early as 1899 (Wortman and Matthew, and distinction of two cingular expansions on 1899) the degree of development of the para- the lingual margin of the molar crowns. The style of P4 was used to distinguish two major morphology exhibited by the feliforms is also groups within the early Tertiary Carnivora. found in the outgroup taxa, which maintain All Caniformia greatly reduce or lose the cin- separate postcingulae and precingulae. Most gular parastyle cusp on P4. Most of the can- of the Caniformia develop a distinct hypo- iforms lose the cusp and reduce the basal cone cusp as an expansion of the posterior cingulum early in their phylogeny, and those portion of the lingual cingulum, as do some that do retain the parastyle have reduced it Feliformia (e.g., some Viverravus). How- to a very small swelling or cuspule on the ever, in the Feliformia the hypocone arises anteroexternal portion of the basal cingulum. from an isolated postcingulum that is not The presence of a small parastyle is primitive continuous with the precingulum, and is for the Eutheria, but extreme reduction of therefore a feature derived independently the cusp, and in some cases the supporting from, and in a different manner than, the hy- cingulum as well, is a derived feature uniting pocone in the Caniformia. All caniforms de- all the Caniformia. velop the continuous lingual cingulum, The protocone of P4 is reduced in size in whereas the development of a hypocone is all Caniformia. The primitive condition for a derived feature for only some taxa within all eutherians is a large, lingual protocone on the Caniformia. P4, as is maintained by the Feliformia. Re- Tedford (1976, p. 372) considered the de- duction of the protocone is less extreme in velopment of a teres process and postscap- some of the early Tertiary caniforms than in ular fossa on the scapula to be derived for later members of the suborder, but in all the Arctoidea. He also discussed the mor- cases the cusp is lower in height and occu- phology of the scapula in "pinnipeds" and pies a smaller area of the surface of the arctoids and concluded that the Otarioidea crown of P4 than is primitively found in the were most closely related to the Ursoidea, Eutheria and Feliformia. Moderate reduction and the Phocidae were most closely related of the protocone in some primitive caniforms to the Mustelidae within the Arctoidea. De- makes it difficult to distinguish between tails of the scapular morphology indicated those taxa and members of the Feliformia that the otarioids and phocids are very dif- with a relatively small protocone, on the ba- ferently derived in their shoulder girdles sis of this feature alone. Nevertheless, the even though both groups are aquatically protocone is relatively reduced in all cani- adapted. Davis (1949) provided a detailed forms, and when analyzed in conjunction discussion of the shoulder architecture of a with other characters this derived feature number of living carnivorans. This study may be used to unite the members of the concluded that the teres major process was Caniformia. relatively large to very large in procyonids The Caniformia are derived in the devel- and ursids, and absent in canids and felids opment of a continuous lingual cingulum on (no "pinnipeds" were considered). the upper molars (Tedford, 1980, personal Our study indicates that the development comm.). Primitive Carnivora have both a of a large teres major process is more com- precingulum and a postcingulum at the base mon in the Carnivora than was originally in- of the molar protocones, without a connec- dicated by Davis (1949) and Tedford (1976), tion around the lingual base of the cusp. Can- and that this feature may be derived for the 52 AMERICAN MUSEUM NOVITATES NO. 2725

Caniformia. Within the Carnivora Oodectes, separating the teres process from the infra- Vulpavus, the Amphicyonidae, Nimravidae, spinous fossa. and Arctoidea all possess a large teres pro- Most other mammals lack an enlarged cess on the scapula. The Canidae, Palaeo- teres process and development of a second- gale, and all Feloidea lack a teres process ary scapular spine. The primitive condition (the scapula is unknown in any early Tertiary for the Theria is a scapula with a relatively Feliformia); in these forms the teres major straight axillary border, no teres major pro- origin is marked only by a scar on the axil- cess and a single scapular spine separating lary border near the vertebral angle (Davis, the infraspinous fossa from the supraspinous 1949, p. 285). fossa. Anteaters and armadillos, however, The preserved scapulae of Vulpavus parallel some Carnivora in the development (AMNH 11498) and Palaeogale (AC 22- of a huge teres process and secondary scap- 2012) are both broken, but enough of the ular spine, presumably as an adaptation for infraspinous fossa and axillary border are digging (see Davis, 1949). present to indicate the presence or absence The wide distribution of the teres major of the teres process in each. Palaeogale process within the primitive and derived agrees well with all other known feliforms in members of the Caniformia, and its absence possessing a scapula with a straight axillary in most other mammals, leads us to conclude border and no indication of a teres process. that this is a derived feature for the Canifor- Vulpavus, however, has a moderately large mia. All known Feliformia lack any indica- teres major process that is similar in size to tion of a teres major process or a secondary the process found in many procyonids. A scapular spine. The Canidae, however, par- small, vertical rib separates the infraspinous allel the Feliformia in the lack of a teres ma- fossa from the teres process in Vulpavus. jor process; presumably this is a secondary Oodectes also has a large teres process and loss within the canid caniforms. a small bony rib separating the teres process The Caniformia retain all the primitive car- from the infraspinous fossa (see Matthew, nivoran features mentioned in the diagnosis 1909, p. 378). of the suborder. These characters have been Members of the Amphicyonidae (Canifor- discussed in earlier sections of this paper, mia) have a very large teres process that is and are not considered further here. The list separated from the infraspinous fossa by a of characters primitive for the Caniformia is large secondary scapular spine (see Olsen, provided in the diagnosis as an aid in char- 1960, p. 8, fig. 6, and pl. 1; Peterson, 1910, acterizing members of the suborder, and as pp. 231-232, pls. 78, 83 and 84; Bergounioux a hypothetical morphotype to assist in the and Crouzel, 1973, fig. 47). This resembles determination of character polarities within the scapular morphology of the Ursidae, the Caniformia. Otariidae, and Phocidae within the Arcto- A number of taxa that were previously al- idea. The Procyonidae (Arctoidea) have a lied with taxa we place in the Feliformia have smaller teres process, but they lack a sec- been excluded from the suborder because of ondary scapular spine. Within the Procyon- their lack of the derived features uniting the idae the teres major process is generally Feliformia. These non-feliform taxa appear "relatively large and plate-like" (Davis, to belong in the Caniformia, instead, as they 1949, p. 293) as in Bassariscus, Nasua, and share the derived characters of the cani- Procyon. forms. Included within this group of contro- The Nimravidae (here considered Canifor- versial taxa are Oodectes, Plesiomiacis, mia, see below) have a teres major process Quercygale and the Nimravidae. All of these that is very broad and large in Hoplophoneus taxa are more thoroughly discussed below. (see Osborn, 1910, fig. 87 and Hough, 1950, pls. 27-29) and Dinictis (AMNH 38805, and FAMILY NIMRAVIDAE COPE, 1880b see Matthew, 1901b, pl. 38). In both of these DIscuSSION: The distinction of two major genera there is a well-developed bony rib groups in the Felidae has been established 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 53 by numerous workers (for example, Cope, PACN 1880b; Teilhard de Chardin, 1915; Piveteau, PAC 1961; de Beaumont, 1965). We recognize the nimravids or "paleofelids" as a separate group of carnivorans having nothing to do directly with the phylogeny of the true cats (or neofelids). The "paleofelids" are not members of the Felidae, they are excluded 11111111i11 from the Feloidea (Felidae, Viverridae, and 0 5 10 Hyaenidae), nor can they even be considered A members of the Feliformia, as constructed in this paper. In addition, the distinction of the Nimravinae and Hoplophoneinae as two dis- tinct, recognizable groups of "paleofelids" by de Beaumont (1964) appears to be valid here, and the two lineages are considered FIFII' subfamilies of the family Nimravidae. We 0 5 follow Tedford's (1978) resurrection of the B family Nimravidae to include both the Nim- ravinae and the Hoplophoneinae. Martin (1980) and Baskin (1981) also recognize the Nimravidae, and include in this family de Beaumont's Nimravinae and Hoplophonei- METB nae. The "paleofelids," which include Dinic- tis, Pogonodon, Nimravus, and Dinaelurus in the Nimravinae, and Hoplophoneus, Eu- 111111 I smilus, Sansanosmilus, and Barbourofelis in 0 5 10 the Hoplophoneinae, have generally been C considered by previous workers to be mem- FIG. 9. Comparison of P4 morphology in the bers of, or close allies to, the Felidae. Anal- Nimravidae and the Feloidea. A) Barbourofelis ysis of dental characters, and the subsequent (AMNH 61895), B) Hoplophoneus (AMNH construction of a cladogram based on these 69417), and C) Crocuta (AMNH Comparative data, indicate that the Nimravidae are not Anatomy Collection 27765). All views are lingual, members of the Feliformia clade. All nim- scales in millimeters. PACN-notch in paracone ravids lack the diagnostic, well-developed ridge; all other abbreviations as in figure 1. parastyle on P4 and the slitlike form of the metastyle notch on P4 that are found in all these two structures are not homologous (see feliforms. fig. 9). The condition in Barbourofelis seems The parastyle is very poorly developed or to be an elaboration of the anterior paracone absent in all nimravids except Barbourofelis ridge and groove found in Nimravus and in which a large "parastyle" is developed in Hoplophoneus, which results in a superficial much the same manner as in various sabre- similarity to the large, true parastyle incor- toothed neofelids. Some nimravids (e.g., porated into the shearing blade of Smilodon. Nimravus) have a deep groove high on the Furthermore, the structure of the broad, strong ridge running anteriorly from the apex deep, rounded groove in Barbourofelis is of the paracone, resulting in a cusplike struc- distinctly different from the very narrow, ture just anterior to the groove. This struc- sharp, slitlike notch between the parastyle ture is positioned much higher on the flank and paracone in sabretoothed neofelids. Bar- of the paracone than would be a true cingular bourofelis and other nimravids also lack the parastyle cusp, and it appears obvious that deep, sharp notch in the metastyle blade; 54 AMERICAN MUSEUM NOVITATES NO. 2725 they have a shallower, broader, grooved convergence in the dentition. Hough (1953) notch that does not extend dorsally as a slit proposed a superfamily Cynofeloidea to in- in the crown. clude the Canidae, the Felidae and the Nim- The Nimravidae also have the weak, an- ravinae (considered a subfamily of the Feli- teriorly located protocone on P4 that is char- dae). This was based on a belief that the acteristic of the Caniformia. Although the nimravines represented morphological inter- protocone is most reduced in the more de- mediates between the Felidae and Canidae, rived Nimravidae (e.g., Barbourofelis), this in which the dentition resembled that of the cusp in all nimravids is much smaller than neofelid "sabretooths" and the basicranium that found in the Feliformia or the eutherian was much more similar to the Canidae. Any outgroup taxa (see fig. 3). Therefore, the similarities between the Felidae and other Nimravidae differ from the Feliformia in the Aeluroidea were assumed to be the result of reduction of the protocone of P4, and instead convergence. It may also be inferred, here, appear to share this derived feature with oth- that if the ancestral felid morphology was er members of the Caniformia. represented by the nimravines (as indicated The Nimravidae lack the calcaneal fibular in the Hough, 1953 phylogeny), then neofe- facet and the articulation between the fibula lids lacking smilodonty secondarily evolved and the calcaneum. Representative tarsi of a less derived dental morphology. Just such the nimravid genera Dinictis, Nimravus, Po- a phylogeny and morphocline polarity was gonodon, Hoplophoneus, and Barbourofelis more explicitly presented by Matthew lack the calcaneal fibular facet and any in- (19lOa) in his discussion of the phylogeny of dication of a facet on the fibula for articula- the Felidae. tion with the calcaneum. The presence of a Hough's (1953) alliance of the Nimravinae large calcaneal fibular facet is primitive for with the Canidae, or at least the Caniformia the Carnivora (see discussion in the Carniv- of this paper, is here considered valid. The ora section, above). A large facet is found in use of this group as a morphological the primitive feliform Didymictis and a "bridge" between the Canidae and true Fe- smaller facet is present in many living Vi- lidae is unacceptable, however, as it is based verridae. The loss of this calcaneal fibular solely on dental characteristics that are con- facet is surely derived, and it appears to be vergent, demands secondary reversal and a shared derived feature uniting the Nimrav- simplification of dental morphology within idae with all other Caniformia. some Felidae, and requires ad hoc assump- The presence of a large teres major pro- tions of convergence in dental and basicran- cess separated from the infraspinous fossa ial features between the Felidae and other by a bony ridge, in all Nimravidae for which Feloidea. a scapula is known, is a derived feature unit- If the Nimravidae (including Hoplopho- ing the Nimravidae with the Caniformia. neinae) are hypothesized as Feliformia, their This feature is found throughout primitive lack of a fully ossified, true septum bullae and derived Caniformia taxa, is only rarely and the presence of a postglenoid foramen observed in non-carnivoran mammals (see excludes them from a phyletic position with- discussion in Caniformia section), but is nev- in the Feloidea, represented by the trichot- er present in any Feliformia. This derived omy of the Felidae, Hyaenidae and Viverri- scapular morphology supports exclusion of dae at Node 10 of the cladogram (fig. 4). The the Nimravidae from the Feliformia and basicranial foramina, and the inferred circu- placement of the family within the Canifor- latory pattern, further support the exclusion mia, instead. of the Nimravidae from the Feloidea. Nim- Cranial evidence strongly supports the ravid basicrania are generally more canoid in separation of the Nimravidae from the Fe- arrangement and are strikingly different from loidea, and further corroborates the hypoth- the derived pattern found in true felids. esis that they are members of the Caniformia Hough's (1953) extensive analysis of the that resemble certain felids only through auditory region in fossil Felidae (which in- 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 55 cluded taxa presently considered "paleofe- canal, the close connection or confluence of lids" Hoplophoneus, Eusmilus, Dinictis, the condyloid foramen and the foramen la- and Nimravus and the neofelid genus Smi- cerum posterius, and the extreme reduction lodon, which was considered a "paleofelid") in size of the carotid canal. presented evidence that the ear region in Work presently in progress by N. Neff and "paleofelids" was distinctly canoid in struc- R. Tedford on the basicrania and carotid cir- ture and pattern. Hough stated that the bulla culation of the Nimravidae should provide is never completely ossified and it lacks any important new information relevant to the trace of a true septum bullae (although Bar- determination of the phyletic relationships of bourofelis, a derived hoplophoneine un- this group. known to Hough, does have a completely The striking morphological differences be- ossified bulla and internal bony struts, but no tween the basicrania of the Nimravidae and septum bullae). the true Felidae, and the lack of any derived Baskin (1981) discusses the "septum bul- felid characters in the Nimravidae, precludes lae" in the Nimravidae, as well as its role in a close relationship between the two groups. the classification of the Carnivora. Baskin Since the Nimravidae are not members of the cites work by Clark and Guensburg, and Felidae, any dental features in common (hy- Brunet, as well as his own observations, doc- percarnivory, smilodonty, etc.) must have umenting the presence of what they consider been evolved independently and convergent- a septum in the bullae of Dinictis and Hop- ly in the two lineages. If indeed the Nimrav- lophoneus. These workers have not convinc- idae are convergent with the Felidae, then ingly supported the assertion that there is a that similarity may have evolved at any point complete septum in the bulla of any nimrav- in the phyletic lineage of the Carnivora. No id. We do not believe that there is a true longer must we consider the nimravids as septum in the bulla of any Nimravidae that close relatives of the cats. Instead, the mor- would be homologous with the septum in phological evidence from the basicrania and either the Canidae or the Feloidea. Martin the dentition strongly indicates an associa- (1980, p. 152) also states that "No septum tion of the Nimravidae with the Caniformia, bullae occurs in any paleofelid, including rather than with the Aeluroida. We advocate Barbourofelis." The development of a bony removal of the Nimravidae from the Felifor- outgrowth in the bullae of the Nimravidae mia, altogether, and alternatively hypothe- may correspond in topographic position to size a relationship with the Caniformia. All the complete septum found in the Canidae, previous workers have considered the nim- and its presence and precise location of de- ravids as ancestors, or very close relatives, velopment within the bulla may be a derived of the Feloidea. We believe that the lack of feature uniting the Nimravidae and the Can- any of the derived features of the Feloidea idae. or Feliformia excludes the Nimravidae from The basicranial foramina were shown a phyletic position within these groups. The (Hough, 1953, p. 99) to be distinctly different Nimravidae are a very specialized group rep- from the pattern and numbers found in the resenting a divergent lineage within the Can- feloids, and to much more closely resemble iformia. A determination of the precise phy- the canoid condition. Some of these char- letic position of the Nimravidae within the acters, however, may be primitive for all the Caniformia awaits a more detailed analysis Carnivora. An alisphenoid canal, large post- of this group. glenoid foramen, large condyloid foramen separate from the foramen lacerum posteri- us, and a large carotid canal are all found in the Nimravidae and many Caniformia, but are Oodectes Wortman, 1901 absent or modified differently in all Feloidea. The living and most fossil Felidae are ex- DISCUSSION: Oodectes was proposed by tremely different in the lack of an alisphenoid Wortman (1901) as a new genus within the 56 AMERICAN MUSEUM NOVITATES NO. 2725

Viverravidae, but was later moved by Mat- from China9 is very similar to Oodectes thew (1909) to a position within the Miaci- in both size and morphology. This genus nae. Simpson (1945) followed Matthew's also appears to belong in the Caniformia. work and included Oodectes within the Mia- It has a short M2 talonid, very high M2 cinae. trigonid, M2 hypoconid larger than the hypo- The precise systematic position of Oo- conulid, and the parastyle of P4 is a weak, dectes is debatable, but we include it within low elaboration of the cingulum and not the Caniformia. The presence of three mo- well developed as a cusp; all of these primi- lars excludes Oiodectes from the Feliformia, tive characters are variably modified in the as here defined, but this primitive retention Feliformia. Furthermore, Xinyuictis has the does not necessarily ally it with the Canifor- reduced P4 protocone and lingual upper mia. Qodectes has a moderately well-devel- molar cingulum continuous around the base oped basal P4 parastyle, but it is not as en- of the protocone that are diagnostic for the larged as in other primitive feliforms, nor as Caniformia. The presence or absence of M3 reduced as in some caniforms. This mor- in Xinyuictis is not known from the presently phology may represent a retention of the available material, but in most other features primitive carnivoran (and eutherian) condi- the dentition of this genus closely resembles tion, the most primitive elaboration of the Oodectes, and the two may be closely relat- feliform parastyle, or the least reduction of ed taxa. the parastyle within the Caniformia lineage. MacIntyre and Guthrie (1979) have re- The lack of a hypocone in Oodectes is irrel- cently expanded the hypodigm of Oodectes evant, as not all caniforms develop molar by synonymizing Paeneprolimnocyon with hypocones. The lingual molar cingulae are Qodectes. All the material of the type moderately well-developed, and may be only species, P. amissadomus was included with- weakly continuous around the base of the in the only species of Qodectes, 0. herpes- protocone. Again, this represents the primi- toides. tive morphology and does not ally the genus OQidectes appears to lack most of the de- with the Feliformia, but would make it a very rived features of both the Caniformia and the primitive caniform if it were included within Feliformia. Due to the lack of any definitive the Caniformia. However, as mentioned synapomorphies between Oodectes and the above, other Caniformia have only a very Feliformia, and the presence of a reduced P4 weak development of the continuous lingual protocone, large teres process on the scap- cingulum, and these forms do not appear to ula, and a weakly continuous lingual cingu- be significantly more advanced than 06- lum on the molars of Oodectes, we consider dectes in this trait. Oodectes is more similar the genus a member of the Caniformia. to the Caniformia in the presence of a re- duced P4 protocone and in the shallowness "Plesiomiacis" (Stock, 1935) of the carnassial notch on P4. The talonid of DIscuSSION: Plesiomiacis Stock, 1935, M2 is short in Qodectes, and is never as en- originally described as a subgenus of Viver- larged as in the Feliformia. The hypoconulid ravus, was recognized by Simpson (1945) as is smaller than the hypoconid on M2, as in a distinct genus in the "Viverravinae." In the Caniformia and outgroup taxa, and not this context Plesiomiacis extended the tem- expanded in size as in the Feliformia. Pres- poral range of the "Viverravinae" into the ence or absence of the fibulo-calcanear facet late Uintan. The phyletic position of Ple- cannot be determined in Oodectes. siomiacis progressus Stock and The scapula of Oodectes has the large proposed by teres process and secondary scapular rib that is a derived feature found in all Caniformia 9 Casts of the genus are housed in the American Mu- (see discussion in Caniformia section, above). seum of Natural History and Museum of Paleontology, Xinyuictis tenuis Zheng, Tong and Ji, 1975, Michigan. 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 57

Simpson is erroneous. Plesiomiacis is not a position within the "Miacinae," nor did it closely related to Viverravus, nor can it be necessarily require inclusion within the considered a "viverravine" feliform; in- "Viverravinae." This was based on the syn- stead, it is definitely a "miacine" caniform. onymy of Quercygale (a supposed "viver- Plesiomiacis has the reduced P4 protocone ravine" lacking M3) with Tapocyon (a "mi- and parastyle, and continuous lingual cin- acine" that retained M3). Therefore, Van gulum on the upper molars that are diagnos- Valen assumed that the lack of M3 was not tic for the Caniformia, and lacks all the fea- confined exclusively to the "Viverravinae," tures diagnostic for the Feliformia (except and that advanced species of true "Miaci- for a presumed analogous loss of M33). nae" could lose M3 in parallel. This is cor- Plesiomiacis is extremely similar morpho- rect, and Plesiomiacis represents another il- logically to the caniform Procynodictis, dif- lustration of this situation. Careful and fering only in very minor details of the teeth rigorous analysis of character distributions and in the absence of M3 (and presumably in these taxa, indicates that they can all be M3, also). The premolars are very similar in assigned to the Caniformia; both Plesiom- both genera, as are the molar trigonids. The iacis and Quercygale lack the defining syn- molar talonids are slightly different, but both apomorphies of the Feliformia and possess genera have a trenchant cristid obliqua and those that characterize the Caniformia, in- a moderately developed lingual basin on the stead. Even if both Plesiomiacis and Quer- talonids. Plesiomiacis and Procynodictis cygale were considered generically distinct share a distinctive M1 morphology in which from Procynodictis and Tapocyon, respec- the parastyle wing is long and narrow, para- tively, they may still be considered cani- cone is on the anteriormost edge of the tooth forms that are more derived than most other and widely separated from the metacone, lin- early Tertiary caniforms in the loss of the gual portion of the tooth is slightly to mod- primitively retained M3. Simpson (1946) pro- erately more narrow than the buccal part, vides a similar discussion of the reasons for hypocone is moderately well developed as a not assigning generic value to differences in distinct cusp, all the cingulae (especially the the presence or absence of a very reduced anterior cingulum at the base of a straight tooth between otherwise very similar taxa. anterior margin) are reduced, conules (par- In any case, the occurrence of "Plesiomi- ticularly the metaconule) are reduced or ab- acis" does not extend the temporal range of sent, and the paraconule and metaconule the 'Viverravinae" (primitive Feliformia) cristae are also reduced. P4 and M2 in both into the late Uintan of North America. "Ple- genera are typically caniform in the very siomiacis" is a member of the Caniformia, short M2 talonid length, P4 protocone re- included in the genus Procynodictis, and is duced and far forward, and presence of a not closely related to the members of the very small basal parastyle. M3 may already Feliformia. be absent in Procynodictis (although this cannot be determined with certainty), and M3 is very reduced. The characters cited by Stock (1935) to distinguish Plesiomiacis pro- Quercygale Kretzoi, 1945 gressus from Viverravus are all shared with DISCUSSION: Both Van Valen (1967) and Procynodictis. Therefore, although P. pro- de Beaumont (1965) recognized the resem- gressus is probably a species distinct from blance of Quercygale ("Humbertia" of de Procynodictis vulpiceps, it is doubtful that Beaumont, 1965) to Tapocyon. Van Valen P. progressus should be considered generi- believed that Quercygale was either synon- cally distinct solely on the basis of the loss ymous with or should be considered a sub- of M3. genus of the "miacine" Tapocyon, whereas Van Valen (1967) first recognized that the de Beaumont retained "Humbertia" in the absence of M3 did not exclude a species from "Viverravinae." We believe that Quercy- 58 AMERICAN MUSEUM NOVITATES NO. 2725 gale is not closely related to the feliforms bladelike P4 anterior accessory cusp; open with which the genus was previously classi- M1 trigonid-V; great reduction of M2; strong fied within the "Viverravinae," that it is def- lingual cingulum on M1; and large, "hook"- initely a member of the Caniformia, and that shaped blade on the parastyle of M1. the genus may be closely related to Tapo- cyon and Uintacyon. PHYLOGENY OF THE CARNIVORA Quercygale has the very reduced P4 para- style and protocone, and a weak, but contin- The following list of characters abstracts uous, lingual cingulum on the upper molars, the features that are assumed to be derived which are diagnostic features for the Cani- for the members of each taxon. The nodes formia. Furthermore, Quercygale lacks all of are those given in the phylogeny of figure 4. the derived features for the Feliformia, ex- Node 1: Order Carnivora cept for the loss of M3. Quercygale retains A) P4/M1 modified as principal carnassial primitive carnivoran features such as a short, teeth, lacking a migratory locus for the unexpanded M2 talonid; M2 hypoconid larger carnassial than the hypoconulid; P4 metastyle notch B) P4 protocone located anterolingually, only moderately deep and narrow; a large, far forward of the paracone anteromedially located postglenoid foramen; C) extremely well-developed, elongate P4 and an unossified bulla. This suite of char- metastyle wing with a well-developed acters indicates that Quercygale is certainly metastyle blade and carnassial notch a caniform, although in many features it is D) well-developed, elongate P4 talonid more derived than many other early Tertiary with at least two posterior accessory caniforms, and parallels some other cani- cusps forms in the loss of M3. E) M232-3 reduced in size Quercygale is morphologically similar not F) processus hyoideus formed by a ven- only to Tapocyon, as was recognized by oth- tromedial prolongation of the squa- er workers, but to Uintacyon as well. Quer- mosal. cygale appears to be closely related to either Node 2: Suborder Feliformia one, or both, of these genera. Quercygale A) P4 parastyle enlarged resembles Uintacyon in the general outline B) very deep, narrow, slitlike carnassial of P4, M1, and P4; the position of the talonid notch in the metastyle blade of P4 cusps on the buccal margin of P4, and the C) M33 lost development of a large, lingual basin; the D) M2 reduced in size, simple in mor- closed-V formed between the paralophid and phology protolophid of M1; and many features of M1, E) M2 talonid extremely elongate; tooth such as weak lingual cingulum, lack of meta- with elongate oval outline conule, weak parastyle "hook" and strong F) M2 hypoconulid larger than, or equal development of the metastyle, etc. Alterna- to, hypoconid size. tively, Quercygale resembles Tapocyon in Node 3: Infraorder Didymictida (Family the development of an M1 metaconid > Didymictidae) paraconid size; anteroposterior alignment of A) M2 paraconid extremely, and metaco- the cristid obliqua of M1; outline of M1; and nid moderately, reduced; paraconid development of a pronounced flange at the much smaller than or subequal to anterior end of the mandible. Some of these metaconid size resemblances may only be shared primitive B) well-developed notch between anterior features and would not be indicative of any accessory cusp and main cusp on P4 special relationship between the various C) P4 talonid broad taxa. Also, in many features Tapocyon ap- D) canines multigrooved. pears to be more derived than both Uinta- Node 4: Protictis cyon and Quercygale. These include the de- A) anterior accessory cusp on P4 enlarged velopment of a relatively large, robust, and very high 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 59

B) lingual margin of P4 anteroposteriorly B) M2 talonid extremely elongate; hypo- straight conulid large and bulbous with anterior C) P3 develops a weak, but distinct, pro- occlusal facet tocone cusp anterior to the paracone C) anterobuccal basal trigonid cingulum- apex on M1 weak D) P4 main cusp situated on buccal tooth D) M1 hypoconulid strong, posterocen- margin trally located, and projecting poste- E) main cusp on P4 leaf-shaped in lateral riorly from talonid margin outline. E) anterior trigonid notch on M2 reduced, Node 5: Protictis (Bryanictis) resulting in a continuous, low paralo- A) P4 anterior accessory cusp tremen- phid. dously enlarged and modified for Node 10: Superfamily Feloidea shearing A) ossified, bipartite bulla with a large ec- B) expansion and very deep basining of totympanic septum lingual portion of P4 talonid B) great reduction or loss of the postgle- C) large, well-developed P3 protocone noid foramen D) all P4 cusps very strongly laterally C) scaphoid and lunar united in a single compressed scapholunar bone E) alignment of P4 cusps in a straight, D) paroccipital process broadly based and anteroposterior line on buccal tooth closely appressed to the posterior wall margin; buccal tooth margin antero- of the bulla posteriorly straight. E) development of an intracranial arterial Node 6: Unnamed taxon (P. (Protictis) and rete in the orbital region, internal ca- P. (Protictoides)) rotid artery reduced in size or vestigial. A) buccal margin of P4 straight Node 11: Suborder Caniformia B) P4 cusps compressed and trenchant A) loss of calcaneal fibular facet in the pes C) P4 parastyle extremely large, higher B) great reduction or absence of the para- relative to protocone style on P4 D) deep second posterior notch between C) reduction in size of the P4 protocone enlarged first and second posterior ac- D) lingual cingulum on upper molars com- cessory cusps on P4. plete around the protocone base Node 7: Infraorder Aeluroida E) large teres major process on the scap- A) conules absent on upper molars ula. B) M1 paraconid and metaconid far sepa- rated, resulting in an open-V between the paralophid and protolophid on the trigonid CHARACTER CONTRADICTIONS C) loss of the second posterior accessory cusp on P4; talonid simpler The following list is a summary of the dis- D) reduction of molar cingulae. tribution of characters that contradict the Node 8: Superfamily Viverravoidea phylogenetic arrangement proposed above: A) paroccipital process extremely re- 1) reduction of P4 talonid to one posterior duced accessory cusp-Didymictis, Aeluroida, B) supraoccipital process extremely re- some early and late Tertiary Caniformia. duced 2) ossified, single scapholunar bone-Fe- C) postglenoid foramen located far loidea, some Vulpavus (see Matthew, 1909), posteroexternally on the postglenoid Vassacyon (Matthew, 1915, p. 40), Procy- process of the squamosal. nodictis (Wortman and Matthew, 1899, p. Node 9: Family Viverravidae 122), and late Tertiary Caniformia. A) P4 anterior accessory cusp far antero- 3) completely ossified bulla-Palaeogale, lingual on crown Feloidea, late Tertiary Caniformia. 60 AMERICAN MUSEUM NOVITATES NO. 2725

4) enlargement of anterior accessory cusp named taxa when possible, and to maintain on P4-Protictis, Tapocyon. logical consistency between the phylogeny 5) M33 lost-Feliformia, and some derived and the classification. members of the Canidae, Procyonidae, Mus- telidae, Nimravidae, Quercygale, and "Plesi- Order Carnivora Bowdich, 1821 omiacis" (within the Caniformia). Suborder Feliformia Kretzoi, 1945 6) conules absent on upper molars-Ae- Infraorder Didymictida, new luroida, some late Tertiary Caniformia. Family Didymictidae, new 7) great reduction or loss of the postgle- Didymictis Cope, 1875 noid foramen-Feloidea, advanced Nimrav- Protictis (Matthew, 1937) idae. P. (Protictis) (Matthew, 1937) 8) lingual cingulum on upper molars com- P. (Protictoides), new plete around the base of the protocone- P. (Bryanictis) MacIntyre, 1966 Caniformia, some Didymictis. Infraorder Aeluroida, new 9) loss of the calcaneal fibular facet in the Superfamily Viverravoidea Wortman pes-Caniformia, some Feloidea. and Matthew, 1899 10) open trigonid-V on M1-Aeluroida, Palaeogale von Meyer, 1846 some Caniformia. Family Viverravidae Wortman and Characters 1-3 are found in all the later Matthew, 1899 Tertiary Carnivora, and 2-3 have tradition- Viverravus Marsh, 1872 ally been used to unite Simpson's Feloidea Simpsonictis Maclntyre, 1962 and Canoidea and to exclude the "Miacoi- Superfamily Feloidea Hay, 1930 dea." Of these, however, only the complete Infraorder Aeluroida, incertae sedis ossification of the bulla unambiguously sup- Ictidopappus 10 ports the traditional grouping of the Canoi- Suborder Caniformia Kretzoi, 1943 dea and Feloidea to the exclusion of the "miacids." All the characters other than the LITERATURE CITED ossified bulla have a variable distribution Baskin, Jon A. throughout both the early and later Tertiary 1981. Barbourofelis (Nimravidae) and Nim- Carnivora. The sporadic distribution of these ravides (Felidae), with a description of characters contradicts some of the relation- two new species from the late Miocene ships hypothesized in the phylogeny outlined of Florida. Jour. Mammal., vol. 62, no. above, but does not reasonably support the 1, pp. 122-139. traditional, or any other alternative, phylog- Berggren, William A., M. C. McKenna, J. Har- eny for the Carnivora. The contradictions denbol, and J. D. Obradovich 1978. Revised Paleogene polarity time scale. cited here highlight features that may have Jour. Geol., vol. 86, pp. 67-81. evolved convergently in various canivoran Bergounioux, F.-M., and F. C. Crouzel taxa, and suggest future areas of investiga- 1973. Amphicyon major Blainville du Mio- tion of the supposed homologies and distri- cene moyen de Sansan (Gers). Ann. Pa- butions of these morphological characters. leo., vol. 59, no. 1, pp. 3-52. 10 The incertae sedis placement of Ictidopappus with- CLASSIFICATION in the Aeluroida indicates that the genus shares all the derived features of the infraorder, but that the Viver- The following classification was construct- ravoidea and Feloidea do not share any features that to exclude Ictidopappus. Further, neither of the aeluroid ed be consistent with the phylogeny pre- superfamilies have uniquely derived features that are sented above and in figure 4. The classifica- known to exclude Ictidopappus from these taxa. There- tion does not precisely reflect the detail of fore, the precise phyletic position of Ictidopappus is relationships portrayed in the cladogram; in- uncertain; the genus may potentially be most closely stead, an attempt is made to minimize the related to either, or both, of the superfamilies, or even proliferation of new taxa, retain previously have a position within either superfamily. 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 61

Bowdich, Thomas E. vol. 31, no. 34, pp. 285-305, figs. 70-77, 1821. An analysis of the natural classifications tables 1-12. of Mammalia for the use of students and Davis, D. Dwight, and H. Elizabeth Story travellers. Paris, J. Smith, 115 + [31] 1943. The carotid circulation in the domestic pp., 16 pls. cat. Zool. Ser., Field Mus. Nat. Hist., Chow, Min-chen vol. 28, no. 1, pp. 1-47, figs. 1-9. 1975. Some carnivores from the Eocene of de Beaumont, Gerard China. Veet. Pal Asiatica, vol. 13, no. 3, 1964. Remarques sur la classification des Fe- pp. 165-168. (in Chinese, with English lidae. Ecologae geol. Helv., vol. 57, no. summary.) 2, pp. 837-845. Clemens, William A., Jr. 1965. Les Viverravinae (Carnivora, Miacidae) 1973. Fossil mammals of the type Lance For- de l'Eocene de la Suisse. Bull. Soc. mation, Wyoming. Part III. Eutheria Vaudoise Sci. Nat., vol. 69, pp. 133- and summary. Univ. of California Pub- 146. lications in Geol. Sciences, vol. 94, 1966. Les Miacinae (Carnivora, Miacidae) de Univ. Calif. Press, pp. i-vi, 1-102, 32 l'Eocene de la Suisse. Ibid., vol. 69, pp. figs., 10 tables. 273-285, 2 figs., 1 pl. Cope, Edward D. 1967. Observations sur les Herpestinae (Vi- 1875a. Systematic catalogue of Vertebrata of verridae, Carnivora) de l'Oligocene su- the Eocene of New Mexico, collected perieur avec quelques remarques sur in 1874. Report to the Engineer Dept., des Hyaenidae du Neogene. Premiere U.S. Army Geogr. Explor. Surv..... partie. Arch. Sci., Geneve, vol. 20, pp. 1-37. fasc. 1, pp. 79-107, 14 figs., 3 pls. 1875b. On the supposed Carnivora of the 1968. Note sur la region auditive de quelques Eocene of the Rocky Mountains. Proc. carnivores. Ibid., vol. 21, fasc. 2, pp. Acad. Nat. Sciences, Philadelphia, vol. 213-224, 1 fig. 27, pp. 444-448. Flower, William Henry 1877. Part II. Report upon the extinct Verte- 1869. On the value of the characters of the brata obtained in New Mexico by par- base of the cranium in the classification ties of the expedition of 1874. Chapters of the order Carnivora, and on the sys- XI-XII. Engineer Dept., U.S. Army, tematic position of Bassaris and other Report Upon United States Geographi- disputed forms. Proc. Zool. Soc. Lon- cal Surveys, pp. 1-370. don, pp. 4-37. 1880a. On the genera of the Creodonta. Proc. 1883. On the arrangement of the orders and Amer. Phil. Soc., vol. 19, pp. 76-82. familes of existing Mammalia. Ibid., pp. 1880b. On the extinct cats of America. Amer- 178-186. ican Naturalist, vol. 14, no. 12, pp. 833- Flower, William H., and Richard Lydekker 858. 1891. An introduction to the study of mam- 1882. Synopsis of the Vertebrata of the Puer- mals living and extinct. London, Adam co Eocene Epoch. Proc. Amer. Phil. and Charles Black, xvi + 763 pp. Soc., vol. 20, pp. 461-471. Gregory, William K., and Milo Hellman 1883. The Vertebrata of the Tertiary forma- 1939. On the evolution and major classifica- tions of the West. Book I. Report of the tion of the civets (Viverridae) and allied U.S. Geol. Surv. of the Territories, fossil and recent Carnivora: A phylo- Dept. Interior, F. V. Hayden, vol. III, genetic study of the skull and dentition. pp. 1-1009. Proceedings of the Amer. Phil. Soc., Cray, P. E. vol. 81, no. 3, pp. 309-392, 6 pls., 33 1973. Marsupialia, Insectivora, Primates, figs. Creodonta and Carnivora from the Hea- Hough, Jean R. don Beds (Upper Eocene) of Southern 1948. The auditory region in some members England. Bull. of the British Mus. (Nat. of the Procyonidae, Canidae, and Ur- Hist.), Geology, vol. 23, no. 1, pp. 1- sidae: Its significance in the phylogeny 102, 6 pls. of the Carnivora. Bull. Amer. Mus. Nat. Davis, D. Dwight Hist., vol. 92, art. 2, pp. 67-118, text 1949. The shoulder architecture of bears and figs. 1-11, pls. 9-15. other carnivores. Fieldiana: Zoology, 1950. The habits and adaptation of the Oli- 62 AMERICAN MUSEUM NOVITATES NO. 2725

gocene sabre tooth carnivore, Hoplo- ravine miacid (Mammalia, Carnivora). phoneus. U.S.G.S. Prof. Paper 221H, Amer. Mus. Novitates, no. 2118, pp. 1- pp. 125-135, pls. 26-31. 4, figs. 1-3. 1953. Auditory region in North American fos- 1966. The Miacidae (Mammalia, Carnivora) sil Felidae: Its significance in phyloge- * Part 1. The systematics of Ictidopap- ny. U.S.G.S. Professional Paper 243G, pus and Protictis. Bull. Amer. Mus. pp. 95-115. Nat. Hist., vol. 131, art. 2, pp. 115-210, Hunt, Robert M., Jr. figs. 1-21, pls. 1-20, tables 1-12. 1974a. The auditory bulla in Carnivora: An an- Maclntyre, Giles T. and Daniel A. Guthrie atomical basis for reappraisal of carni- 1979. Paenoprolimnocyon Guthrie, 1967, aju- vore evolution. Jour. Morph., vol. 143, nior synonym of Oodectes Wortman, no. 1, pp. 21-76, text figs. 1-3, pls. 1- 1901. Jour. Paleont., vol. 53, no. 4, pp. 13. 1034-1036. 1974b. Daphoenictis, a cat-like carnivore McKenna, Malcolm C. (Mammalia, Amphicyonidae) from the 1980. Late Cretaceous and early Tertiary ver- Oligocene of North America. Jour. Pa- tebrate paleontological reconnaissance, leont., vol. 48, no. 5, pp. 1030-1047. Togwotee Pass Area, Northwestern Kielan-Jaworowska, Zofia Wyoming. In Jacobs, Louis L. (ed.), 1977. Evolution of therian mammals in the Aspects of vertebrate history: Essays in Late Cretaceous of Asia. Part II. Post- honor of Edwin Harris Colbert. Flag- cranial skeleton in Kennalestes and staff, Mus. of Northern Arizona Press. Asioryctes. Paleont. Polonica, vol. 37, Marsh, Othniel C. pp. 65-83. 1872. Preliminary description of new Tertiary Klaauw, C. J., van der mammals. Part I. Amer. Jour. Sci., ser. 1931. The auditory bulla in some fossil mam- 3, vol. 4, pp. 122-128. mals. With a general introduction to this Martin, Larry D. region of the skull. Bull. Amer. Mus. 1980. Functional morphology and the evolu- Nat. Hist., vol. 52, pp. 1-352, figs. 1- tion of cats. Trans. Nebraska Acad. 18. Sci., vol. 8, pp. 141-153. Kretzoi, Miklos Matthew, William D. 1943. Kochitis centenii n.g. n.sp., ein alter- 1901a. Additional observations on the Creo- tumlicher Creodonte aus dem Oberoli- donta. Bull. Amer. Mus. Nat. Hist., gozan Siebenburgens. Foldtany Koz- vol. 14, pp. 1-38. lony, vol. 73, pp. 10-17, 190-195. 1901b. Fossil mammals of the Tertiary of 1945. Bemerkungen uiber das Raubtiersystem. Northeastern Colorado. Mem. Amer. Anns. Hist.-Nat. Mus. Natn. Hung., Mus. Nat. Hist., vol. 1, part 7, pp. 355- Budapest, vol. 38, pp. 59-83. 447, pls. 37-39. Lillegraven, Jason A. 1909. The Carnivora and Insectivora of the 1969. Latest Cretaceous mammals of upper Bridger basin, middle Eocene. Ibid., part of Edmonton Formation ofAlberta, vol. 9, pt. 6, pp. 289-567, figs. 1-118, Canada, and review of marsupial-pla- pls. 42-52. cental dichotomy in mammalian evolu- 1910a. The phylogeny of the Felidae. Bull. tion. Univ. Kansas Paleont. Contrib., Amer. Mus. Nat. Hist., vol. 28, pp. art. 50 (Vertebrata), 122 pp., 53 figs., 21 289-316. tables. 1910b. On the osteology and relationships of Love, J. David Paramys and the affinities of the Ischy- 1939. Geology along the southern margin of romyidae. Ibid., vol. 28, pp. 43-72. the Absaroka Range, Wyoming. Geol. 1915. A revision of the lower Eocene Wasatch Soc. Amer. Spec. Paper 20, 134 pp. and Wind River faunas. Part I.-Order MacFadden, Bruce J. Ferae (Carnivora), suborder Creodonta. 1980. Eocene perissodactyls from the type Ibid., vol. 34, pp. 4-103, figs. 1-87. section of the Tepee Trail Formation of 1937. Paleocene faunas of the San Juan basin, northwestern Wyoming. Contrib. to New Mexico. Trans. Amer. Phil. Soc., Geol., Univ. Wyoming., vol. 18, no. 2, new series, vol. 30, pp. i-vii, 1-510, pp. 135-143. figs. 1-85, pls. 1-65. MacIntyre, Giles T. Matthew, William D., and Walter Granger 1962. Simpsonictis, a new genus of viver- 1915. A revision of the lower Eocene Wasatch 1982 FLYNN AND GALIANO: PHYLOGENY OF EARLY TERTIARY CARNIVORA 63

and Wind River faunas. Bull. Amer. nivora. Evolution, vol. 8, pp. 166-171, Mus. Nat. Hist., vol. 34, art. 1, pp. 1- text figs. 1-7. 103, 87 text figs. Schlosser, Max Mellett, James 1888. Die Affen, Lemuren, Chiropteren, In- 1977. Paleobiology of North American sectivoren, Marsupialier, Creodonten Hyaenodon (Mammalia, Creodonta). und Carnivoren des Europaischen Ter- Contrib. to Vert. Evol., vol. 1, Hecht tiiirs. Beitr. z. Palaont. Geol. Oster- and Szalay (eds.), Basel, S. Karger. reich-Ungarns, II Theil, Bd. VII, pp. von Meyer, H. 225-386, pls. 6-9. 1846. No title. Neues Jahrb. Min., Geog., Simpson, George G. Geol., Petrefakten-Kunde, Jahrg. 1846, 1931. A new classification of mammals. Bull. pp. 462-476. Amer. Nat. Hist., vol. 59, pp. 259-293. Novacek, Michael J. 1935. New Paleocene mammals from the Fort 1977. Aspects of the problem of variation, or- Union of Montana. Proc. U.S. Natl. igin and evolution of the eutherian au- Mus., vol. 83, pp. 221-244. ditory bulla. Mammal. Rev. 1977, vol. 1937. The Fort Union of the Crazy Mountain 7, nos. 3 and 4, pp. 131-149. field, Montana, and its mammalian fau- 1980. Cranioskeletal features in tupaiids and nas. Bull. U.S. Natl. Mus., no. 169, pp. selected Eutheria as phylogenetic evi- i-x, 1-287, figs. 1-80, pls. 1-10. dence. In Luckett, W. P. (ed.), Com- 1945. The principles of classification and a parative biology and evolutionary rela- classification of mammals. Bull. Amer. tionships of tree shrews, Plenum Mus. Nat. Hist., vol. 85, pp. i-xvi, 1- Publishing Corp. 350. Olsen, Stanley J. 1946. Palaeogale and allied early mustelids. 1960. The fossil carnivore Amphicyon longi- Amer. Mus. Nat. Hist. Novitates, no. ramus from the Thomas Farm Miocene. 1320, pp. 1-14. Part II-Postcranial skeleton. Bull. Stock, Chester Mus. Comp. Zool., Harvard, vol. 123, 1934. New Creodonta from the Sespe Upper no. 1, pp. 1-44, 1 pl., figs. 1-21. Eocene, California. Proc. Natl. Acad. Osborn, Henry F. Sci., vol. 20, pp. 423-427, 1 pl. 1910. The age of mammals in Europe, Asia 1935. Plesiomiacis, a new Creodont from the and North America. New York, The Sespe Upper Eocene, California. Proc. Macmillan Co., xvii + 635 pp. Natl. Acad. Sci., vol. 21, no. 2, pp. 119- Peterson, Olaf August 122, 1 pl. 1910. Description of new carnivores from the Szalay, Frederick S. Miocene of Western Nebraska. Mem. 1969. Mixodectidae, Microsyopidae, and the Carnegie Mus., vol. 4, no. 5, pp. 205- insectivore-primate transition. Bull. 278, pls. 74-85, figs. 1-69. Amer. Mus. Nat. Hist., vol. 140, art. 4, 1919. Report upon the material discovered in pp. 193-330, figs. 1-28, pls. 17-57, ta- the Upper Eocene of the Uinta basin by bles 1-21. Earl Douglas in the years 1908-1909, Szalay, Frederick S., and Richard L. Decker and by O.A. Peterson in 1912. Ann. 1974. Origins, evolution, and function of the Carnegie Mus., vol. 12, pp. 40-168, tarsus in Late Cretaceous eutherians plates 34-47. and Paleocene primates, pp. 223-259. Piveteau, J. In Jenkins, F. A. (ed.), Primate loco- 1961. Carnivora. In Traite de paleontologie, motion. New York, Academic Press. vol. 6, no. 1, pp. 641-820, Paris. Tedford, Richard H. Rigby, J. Keith 1976. Relationship of pinnipeds to other car- 1980. Swain Quarry of the Fort Union For- nivores (Mammalia). Syst. Zool., vol. mation, Middle Paleocene (Torrejoni- 25, no. 4, pp. 363-374, 5 text figs., 1 an), Carbon County, Wyoming: Geolog- table. ic setting and mammalian fauna. Evol. 1978. History of dogs and cats: a view from Monographs 3, pp. 1-179, pls. 1-14, the fossil record. M23: 10 pp. in Nutri- figs. 1-13, tables 1-52. tion and management of dogs and cats, Russell, Loris S. Ralston Purina Co. 1954. Evidence of tooth structure on the re- Teilhard de Chardin, Pierre lationships of the early groups of Car- 1915. Les carnassiers des Phosphorites de 64 AMERICAN MUSEUM NOVITATES NO. 2725

Quercy. Annls. Paleont., Paris, vol. 9, Wortman, Jacob L. pp. 103-192, pls. 12-20. 1886. The comparative anatomy of the teeth Trouessart, Edouard Louis of the Vertebrata. In The American 1885. Catalogue des Mammiferes vivants et system of dentistry, pp. 351-504, pls. fossiles. Fasc. IV.-Carnivores i-vi, figs. 189-268. (Carnivora). Bull. Soc. Etudes Sci. An- 1901. Studies of Eocene Mammalia in the gers, vol. 15 pp. 1-108. Marsh collection, Peabody Museum. Van Valen, Leigh Amer. Jour. Sci., ser. 4, vol. 11, pp. 1965. Some European Proviverrini (Mamma- 333-348, figs. 1-6, pl. 5, pp. 437-450, lia, Deltatheridia). Paleontology, vol. 8, figs. 7-17, pl. 6; vol. 12, pp. 143-154, pt. 4, pp. 638-665. figs. 18-30, pp. 193-206, figs. 31-43, pp. 1966. Deltatheridia, a new order of mammals. 281-296, fig. 44, pls. 1-4, pp. 377-382, Bull. Amer. Mus. Nat. Hist., vol. 132, figs. 45-48, pp. 421-432, figs. 49-60, art. 1, pp. 1-176, pls. 1-8, tables 1-26. pls. 8-9. 1967. New Paleocene insectivores and insec- Wortman, Jacob L., and William D. Matthew tivore classification. Ibid., vol. 135, art. 1899. The ancestry of certain members of the 5, pp. 217-284, figs. 1-7, pls. 6-7, tables Canidae, the Viverridae, and Procyon- 1-7. idae. Bull. Amer. Mus. Nat. Hist., vol. 1969. The multiple origins of the placental 12, pp. 109-139, figs. 1-10, pl. 6. carnivores. Evolution, vol. 23, no. 1, Zheng, Jia-jian, Yong-sheng Tong, and Hong- pp. 118-130, figs. 1-11. xiang Ji Watrous, Larry E., and Quentin D. Wheeler 1975. Discovery of Miacidae (Carnivora) in 1981. The out-group comparison method of Yuanshui Basin, Kiangsi Province. Vert. character analysis. Syst. Zool., vol. 30, PalAsiatica, vol. 13, no. 2, pp. 96-104. no. 1, pp. 1-11. (In Chinese.)