Syst. Biol. 48(3):455–490, 1999 The Position of Cetacea Within Mammalia: Phylogenetic Analysis of Morphological Data from Extinct and Extant Taxa 1 MAUREEN A. O’LEARY1,3 AND JONATHAN H. GEISLER2 Department of Anatomical Sciences, State University of New York, Stony Brook, New York 11794-8081 , USA; E-mail: [email protected] 2 Department of Vertebrate Paleontology, American Museum of Natural History, New York, New York 10024-5192 , USA; E-mail: [email protected] Abstract.— Knowledge of the phylogenetic position of the order Cetacea (whales, dolphins, and porpoises) within Mammalia is of central importance to evolutionary biologists studying the transformations of biological form and function that accompanied the shift from fully terrestrial to fully aquatic life in this clade. Phylogenies based on molecular data and those based on morphological data both place cetaceans among ungulates but are incongruent in other respects. Morphologists argue that cetaceans are most closely related to mesonychians, an extinct group of terrestrial ungulates. They have disagreed, however, as to whether Perissodactyla (odd-toed ungulates) or Artiodactyla (even-toed ungulates) is the extant clade most closely related to Cetacea, and have long maintained that each of these orders is monophyletic. The great majority of molecule-based phylogenies show, by contrast, not only that artiodactyls are the closest extant relatives of Cetacea, but also that Artiodactyla is paraphyletic unless cetacean s are nested within it, often as the sister group of hippopotamids. We tested morphological evidence for several hypotheses concerning the sister taxon relationships of Cetacea in a maximum parsimony analysis of 123 morphological characters from 10 extant and 30 extinct taxa. We advocate treating certain multistate characters as ordered because such a procedure incorporates information about hierarchical morphological transformation. In all most-parsimonious trees, whether multistate characters are ordered or unordered, Artiodactyla is the extant sister taxon of Cetacea. With certain multistate characters ordered, the extinct clade Mesonychia (Mesonychidae + Hapalodectidae) is the sister taxon of Cetacea, and Artiodactyla is monophyletic. When all fossils are removed from the analysis, Artiodactyla is paraphyletic with Cetacea nested inside, indicating that inclusion of mesonychians and other extinct stem taxa in a phylogenetic analysis of the ungulate clade is integral to the recovery of artiodactyl monophyly. Phylogenies derived from molecular data alone may risk recovering inconsistent branches because of an inability to sample extinct clades, which by a conservative estimate, amount to 89% of the ingroup. Addition of data from recently described astragali attributed to cetaceans does not overturn artiodactyl monophyly. [Artiodactyla; astragalus; Cetacea; fossils; homoplasy; Mesonychia; morphology; phylogeny.] The importance of fossils in phylogeny reconstruction is most keenly appreciated when a clade is of great antiquity and has suffered numerous extinctions. Gauthier et al. (1988), Donoghue et al. (1989), and Novacek (1992a, 1994) have emphasized that under such circumstances, the importance of fossils in phylogeny reconstruction can outweigh the drawback of their incompleteness. These authors have shown that fossils often capture the primitive morphotype of a clade, because many fossil taxa have had less time to evolve homoplasies than have extant members of the same clades. Because fossils often have combinations of primitive and derived features not found in extant taxa, they can be critical for untangling 3 Address correspondence to Dr. Maureen A. O’Leary, Department of Anatomical Sciences, HSC T8 (040), SUNY at Stony Brook, Stony Brook, New York 11794–8081, USA. problems of long-branch attraction, where homoplasies masquerade as homologies. Depending on the sampling of characters and taxa, convergent similarities can have the potential to draw taxa together phylogenetically, as in the celebrated example of the paraphyletic clade “Haematothermia” that links birds and mammals but excludes crocodiles (Gauthier et al., 1988; Eernisse and Kluge, 1993). Mammalian systematists in particular continue to spar (Catzeis, 1993; Graur, 1993a, 1993b; Novacek, 1993) over the signicance of the often strongly differing tree topologies that emerge from phylogenetic analyses of data sets partitioned between molecular and morphological data or between extinct and extant taxa. The problem is exacerbated by different types of missing data: phylogenies based on molecular data cannot sample extinct taxa, and those based 455 456 SYSTEMATIC BIOLOGY on morphological data and drawing on fossil taxa cannot sample all morphological characters, even from the most exquisitely preserved fossils (Gauthier et al., 1988; Novacek, 1994; Simmons and Geisler, 1998). The part of the mammalian tree relating to the origin of Cetacea (whales, dolphins, and porpoises) has been the subject of particular controversy of late because of differing results from phylogenetic analyses based on molecular data and those based on morphological data. Both types of analyses support the hypothesis that cetaceans are nested within ungulates (Novacek, 1992b), but there is virtually no agreement as to the sister taxon of Cetacea or whether certain ungulate clades close to Cetacea are monophyletic. Without knowledge of the sister taxon of Cetacea, we cannot begin to understand the drastic anatomical and physiological transformations that occurred as terrestrial mammals returned to aquatic life and evolved into some of the most specialized known vertebrates. Initial phylogenetic analyses of morphological evidence bearing on the question of cetacean origins nothwithstanding (Prothero et al., 1988; Thewissen, 1994), a comprehensive phylogenetic analysis testing morphological evidence for the monophyly of various modern ungulate orders with respect to Cetacea has been conspicuously absent from the paleontological literature until recently (Geisler and O’Leary, 1997; Geisler and Luo, 1998). This is a serious shortcoming, because detailed morphological analyses are an integral part of any total evidence analysis (Kluge, 1989; and references therein) combining molecular and morphological data (also referred to as simultaneous analysis; Nixon and Carpenter, 1996). Lack of a comprehensive phylogenetic analysis of morphological data has persisted despite the suggestion by some molecular biologists that because of the “numerous extinctions along the stem lineage of cetaceans, most major insights into the transformation from a terrestrial ungulate to a fully aquatic cetacean will come from fossil taxa” (Gatesy et al., 1996:960) . Here we attempt to correct this shortcoming. VOL. 48 BACKGROUND ON CETACEAN ORIGINS Morphological Contributions Morphologists at one time viewed understanding the phylogenetic relationships of cetaceans to other orders of mammals as an intractable problem because the anatomy of cetaceans is so transformed relative to that of other mammals (e.g., Simpson, 1945). Within the last three decades, however, attributable in no small part to discoveries of new fossils, paleontologists have developed the hypothesis that an extinct order of carnivorous, hoofed mammals, the Mesonychia, is most closely related to Cetacea (Van Valen, 1966, 1968, 1969, 1978; McKenna, 1975; Prothero et al., 1988; Thewissen, 1994; Zhou et al., 1995; McKenna and Bell, 1997; Geisler and Luo, 1998; O’Leary, 1998a). Van Valen (1966) initially based this hypothesis on a variety of dental and cranial similarities. Mesonychians (Mesonychidae and Hapalodectidae [but not Andrewsarchus; see Van Valen, 1978; McKenna and Bell, 1997; O’Leary, 1998a]) , a group that some paleontologists argue is paraphyletic (Thewissen, 1994; Geisler and Luo, 1998), are known from the Early Tertiary of the Holarctic. Because they had hoofs, mesonychians have generally been classied among ungulates (Van Valen, 1966), but they differ from virtually all other ungulates in possessing laterally compressed, homodont lower dentitions that are strongly suggestive of a carnivorous diet (Szalay and Gould, 1966; Szalay, 1969a, 1969b; Zhou et al., 1992; O’Leary and Rose, 1995a, 1995b). McKenna (1975) classied Cetacea and Acreodi (Mesonychidae) together in the mirorder Cete within the grandorder Ungulata, thereby formalizing morphological arguments advanced by Van Valen (1966; see also McKenna and Bell, 1997). Paleontologists have not argued that mesonychians are nested among any of the following ungulate clades that have extant members: Artiodactyla, the even-toed hoofed mammals (ruminants, pigs, hippos, and camels); Perissodactyla, the oddtoed hoofed mammals (horses, rhinos and tapirs); or among the paenungulates: Proboscidea (elephants), Hyracoidea (hyraxes), and Sirenia (dugongs and manatees). 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA Initially, interpretations of morphological evidence for the extant clade most closely related to Cetacea tended to converge on Perissodactyla or were simply ambiguous. Among more recent treatments, Novacek (1982) placed cetaceans in an unresolved polytomy with Artiodactyla, Perissodactyla, and tentatively, Tubulidentata (aardvarks), with paenungulates as the next most closely related clade. A series of subsequent parsimony analyses of the mammalian orders tackled more precisely the morphological evidence for cetacean–ungulate relationships. On the basis of cranial traits, Novacek (1986) found cetaceans to be related to ungulates in the following manner: Artiodactyla (Cetacea (Perissodactyla (Hyracoids (Sirenians + Proboscideans))))–a result broadly consistent with information from patterns of variation in the stapedial artery (Wible, 1987). Novacek and Wyss (1986) and Novacek (1989) obtained similar results based on morphological information from several anatomical systems but introduced the possibility that Perissodactyla might be more closely related to Cetacea than is Artiodactyla. In a study including a variety of extinct “condylarths,” Prothero et al. (1988) argued that Cetacea was the sister group of Andrewsarchus and that, of the extant ungulate orders, Perissodactyla was most closely related to Cetacea. Their tree, however, may not be the most–parsimonious explanation of their data because these authors did not perform a parsimony analysis on their entire data set. Thus, phylogenetic analyses of morphological data initially tended to favor a close relationship between paenungulates , perissodactyls, and Cetacea but were unable to demonstrate this relationship with great certainty. In many of the studies described above, monophyly of Artiodactyla was assumed and not tested. The assumption that Artiodactyla is monophyletic originates primarily from the observation that artiodactyls share an ankle joint morphology known as the double-pulleyed astragalus (Schaeffer, 1947; Vaughan, 1986) (Fig. 1), a unique condition. This skeletal feature is one of the few characters that can be directly veried in 457 both extant and extinct forms, and its presence has been the primary criterion for membership in the order, particularly for fossil taxa. The signicance of the artiodactyl ankle is that it restricts the distal hindlimb to parasagittal motion and is thought to be less likely to dislocate as the animal engages in high–speed quadrupedal running across a terrestrial substrate (Schaeffer, 1947; O’Leary and Rose, 1995b). Despite the broad radiation of this group, the morphology of this joint has remained remarkably constant since its rst known appearance (Schaeffer, 1947) at the base of the Eocene ( ~ 55 million years ago). Figure 1 shows the astragali of various mammals associated with the basal ungulate radiation. Relatively unspecialized Early Tertiary mammals like Chriacus have an astragalus with a relatively at proximal end (trochlea) and a convex distal end (head). The astragalus of Phenacodus, a more derived ungulate, has a convex head but a grooved trochlea. Artiodactyls, perissodactyls, and some mesonychians also have a relatively grooved trochlea. These taxa differ, however, in the structure of the astragalar head, specically in its articular facet for the navicular, and each morphology is thought to be a convergent specialization for cursorial locomotion (O’Leary and Rose, 1995b). Both perissodactyls and mesonychians exhibit a saddle-shaped head (O’Leary and Rose, 1995b) but differ in that mesonychians have a distinct facet for the cuboid. Artiodactyls have a deeply grooved head with a more substantial cuboid articulation (e.g., Bunophorus [Fig. 1]; Schaeffer, 1947). Despite variations in body size and locomotor capabilities among artiodactyls, the morphology of the astragalus remains a diagnostic feature of the order (Schaeffer, 1947; Vaughan, 1986). Stem taxa outside of Artiodactyla exhibiting transitional morphologies leading to the primitive artiodactyl morphotype remain virtually unknown (Rose, 1987, 1996). A complete astragalus associated with other diagnostic material of a very primitive cetacean has never been described (Milinkovitch and Thewissen, 1997). For the archaic cetacean, Ambulocetus, the taxon for which we have the most detailed knowledge of skeletal 458 SYSTEMATIC BIOLOGY VOL. 48 FIGURE 1. Right astragali of Early Tertiary mammals associated with the basal ungulate radiation: the proximal end of the astragalus (trochlea); the articular facet for the distal tibia (character 109), towards the top of each drawing; the distal end of the astragalus (head); the articular facet for the navicular; and sometimes the cuboid (characters 104 and 106), towards the bottom of the page. Chriacus (AMNH-VP 92832) , an oxyclaenid arctocyonid (morphology resembles that of Arctocyon coded in this analysis); Phenacodus (AMNH-VP 15287), a stem taxon to extant perissodactyls; Dissacus (AMNH-VP 3359) and Pachyaena (AMNH 16154), mesonychians; Bunophorus (AMNH-VP 92847), an artiodactyl (morphology resembles that of Diacodexis coded in this analysis, representative of the artiodactyl double-pulleyed astragalus); Heptodon (AMNH-VP 95866), a perissodactyl; and Ambulocetus (H-GSP 18507) , an archaic cetacean (redrawn from Thewissen et al., 1996). The head and part of the trochlea of the Ambulocetus astragalus are not preserved in this specimen. Scale bar = 10 mm. anatomy, only the proximal astragalus is known (Fig. 1). It has a grooved trochlea that distinguishes it from generalized mammals such as Chriacus and Arctocyon, but the more diagnostic head is not preserved. Two partial astragali have recently been attributed to the archaic cetaceans Ambulocetus and Pakicetus on the basis of their grooved trochleae, large size, and faunal associations (Thewissen et al., 1998). The authors interpret these bones as possessing certain derived similarities that might link cetaceans to artiodactyls but not to mesonychians but qualify their argument by stating that these “cetacean” astragali do not have trochleated heads as in artiodactyls. Because these fossils are fragmentary and are not associated with diagnostic cetacean material, we continue to consider the morphology of the cetacean astragalus to be relatively poorly established. Three other relatively recent fossil discoveries also suggest the possibility of a close link between Cetacea and Artiodactyla. The hind foot of the archaic cetacean Basilosaurus (Gingerich et al., 1990) exhibits a paraxonic 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA condition (i.e., the axis of symmetry passes between the third and fourth digits), a feature consistent with a close phylogenetic relationship between cetaceans and both Mesonychia and Artiodactyla. The morphology of the incus of another archaic cetacean, Pakicetus (Thewissen and Hussain, 1993), has been argued to support a close relationship between Artiodactyla and Cetacea, as have features of the skeleton of a third archaic cetacean Ambulocetus (Thewissen et al., 1994, 1996), such as hoofs and a paraxonic foot, which reinforce a link with Mesonychidae as well. Thewissen (1994) conducted the rst parsimony analysis of morphological data bearing on the question of cetacean origins and identied several dental synapomorphies supporting a sister-taxon relationship between cetaceans and certain mesonychids. He found Mesonychia (Mesonychidae + Hapalodectidae) and Mesonychidae to be paraphyletic with respect to Cetacea and, in contrast to Thewissen and Hussain (1993), argued that the extant sister taxon to Cetacea was Perissodactyla, a hypothesis championed by Prothero (1993) as well. Thewissen’s (1994) analysis did not directly test monophyly of Artiodactyla because it included only one artiodactyl, Diacodexis. However, when Geisler and O’Leary (1997) and Geisler and Luo (1998) included several artiodactyls in parsimony analyses of morphological data, support emerged for a clade that included Artiodactyla, Cetacea, and Mesonychia to the exclusion of Perissodactyla, and for monophyly of Artiodactyla, Perissodactyla, Cetacea, and Mesonychidae. New dentitions of archaic cetaceans led O’Leary (1998a) to argue that Mesonychidae and Mesonychia are each monophyletic clades (in contrast to Prothero et al., 1988; Thewissen, 1994), the former being nested within the latter. Our work here represents a combination and expansion of these most recent phylogenetic analyses. Molecular Contributions Like morphological analyses, molecular analyses of cetacean origins have varied in their taxon and character sampling, methods of phylogenetic reconstruction, and 459 ultimately in their conclusions. Some initial parsimony analyses of amino acid sequences (Goodman et al., 1985; Miyamoto and Goodman, 1986) or of a combination of amino acid sequences, immunodiffusion, and morphology (Shoshani, 1986) generally revealed no greater resolution than that Cetacea formed a clade with Artiodactyla and Perissodactyla. Evidence supporting a close relationship between Perissodactyla and Cetacea to the exclusion of Artiodactyla emerged from parsimony analyses of amino acid sequences of eye lens proteins (De Jong, 1985; McKenna, 1992), and from analysis of a combination of nuclear gene sequences and amino acid sequences (Stanhope et al., 1993). This last result has been among the minority of results based on molecular data. The molecule-based hypothesis that Artiodactyla is the closest living relative of Cetacea was rst put forth on the basis of serological precipitin tests (Boyden and Gemeroy, 1950). Maximum parsimony analyses of amino acid sequences of a and b hemoglobin chains (Shoshani et al., 1985; Czelusniak et al., 1990b), pancreatic ribonucleases (Beintema et al., 1986, 1988), and a combination of eight different types of amino acid sequences (Czelusniak et al., 1990a) also showed that Cetacea grouped more closely with artiodactyls than with perissodactyls. Several molecular analyses, however, began to recover phylogenies with Cetacea nested within Artiodactyla, a highly controversial result from a morphological perspective, given the reknowned monophyly of Artiodactyla. Czelusniak et al. (1990b) recovered a clade of cetaceans and pecorans (deer, sheep, giraffe, and their close relatives) to the exclusion of other artiodactyls in maximum parsimony analyses of amino acid sequences. They qualied their result as likely to be “phylogenetically incorrect” (Czelusniak et al., 1990b:614) because the new topology differed from established ideas of relationships and was based only on a small subset of all molecular evidence. Using a combination of amino acid and nucleotide sequences for 18 artiodactyls (including Hippopotamus, camels, suids, and pecorans, a fair representation of taxonomic 460 SYSTEMATIC BIOLOGY variation within Artiodactyla), Czelusniak et al. (1990a) found some evidence for artiodactyl paraphyly, but equally parsimonious trees also supported artiodactyl monophyly. Their cautious interpretations of trees supporting artiodactyl paraphyly were not followed by other researchers examining nucleotide sequence data. Parsimony analyses of nucleotide sequence data of the mitochondrial cytochrome b gene (Árnason et al., 1991) and of the 12S and 16S ribosomal genes (Milinkovitch et al., 1993) recovered a sistertaxon relationship between Artiodactyla and Cetacea to the exclusion of Perissodactyla, based on relatively small samples of artiodactyls (1 to 3 taxa). Increasing the artiodactyl sample to 11 taxa for mitochondrial cytochrome b gene sequences, Irwin et al. (1991) recovered a paraphyletic Artiodactyla with Cetacea nested inside as the sister taxon of camels. A sister taxon relationship between Cetacea and ruminants resulted from parsimony analyses of combined nucleotide sequences of both cytochrome b and cytochrome c oxidase subunit II (Honeycutt et al., 1995) and received mixed support from nuclear gene sequences (Stanhope et al., 1996). Graur and Higgins (1994) analyzed 11 nuclear-encoded protein sequences and ve mitochondrial DNA sequences in four taxa, using maximum likelihood (with the assumption of constant rates of substitution [Hasegawa and Adachi, 1996]), neighbor joining, and maximum parsimony, and found that Cetacea was the sister group of the cow to the exclusion of pigs and camels. They advocated revisions to mammalian ordinal-level taxa to formalize their result, despite its genesis from relatively poor taxonomic sampling. Others criticized the four-taxon method of Graur and Higgins (1994), recognizing that it risks recovering “robust, but false relationships” (Philippe and Douzery, 1994:149; see also Adachi and Hasegawa, 1996) because of such confounding factors as long-branch attraction. Authors of other nucleotide sequence analyses that had recovered artiodactyl paraphyly (Queralt et al., 1995; Smith et al., 1996) qualied their results as tentative, sensitive to such variables as outgroup VOL. 48 choice and ingroup sampling, and possibly reective of the unique evolutionary history of a particular gene but not of the clades in question. Hasegawa and Adachi (1996) reexamined the data sets of both Graur and Higgins (1994) and Irwin and Árnason (1994), combined with hemoglobin sequences, using a maximum-likelihood analysis with a rate heterogenous model for nucleotide substitution. They found greatly reduced statistical support for artiodactyl paraphyly but still achieved that result. This suggested to them that artiodactyl monophyly is a viable alternative hypothesis that could not be dismissed until more genes had been examined. None of the above nucleotide sequence analyses included hippopotamids in the sample of artiodactyls, and none explicitly tested the effect of this group of artiodactyls on tree topology. Once hippopotamids were included with 11 other artiodactyls in a parsimony analysis of nucleotide and amino acid sequences of the cytochrome b gene, hippopotamids formed a clade with Cetacea to the exclusion of other artiodactyls (Irwin and Árnason, 1994). Gatesy et al. (1996) found a similar result from parsimony analysis of milk-protein gene sequences, concluding that the next outgroup to the hippocetacean clade was ruminant artiodactyls, then pigs and peccaries, and nally camels. Increasing the number of genes and the number of taxa examined (Gatesy, 1997) produced similar results, and Montgelard et al. (1997), using cytochrome b and 12s rRNA sequences, also found support for an Ancodonta (Hexaprotodon + Hippopotamus) + Cetacea clade to the exclusion of other artiodactyls. By contrast, the hippo-cetacean clade had not been recovered in an earlier parsimony analysis of amino acid sequences (Czelusniak et al., 1990a). Finally, Shimamura et al. (1997) found evidence for artiodactyl paraphyly on the basis of retroposons: nucleotide sequences that have been inserted into a genome at particular loci (Li, 1997). As implemented by Shimamura et al. (1997), the homology statement for retroposons is their presence or absence in a particular position in the genome. The results of these authors demon- 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA strated a close relationship between Hippopotamus, ruminants, and cetaceans, to the exclusion of the camel pig and several outgroup taxa. On the basis of these characters alone, Shimamura et al. (1997:669) asserted “we believe that recent molecular data will lead to the reinterpretation by paleontologists of many fossil records of Artiodactyla to match our conclusions.” Furthermore, Milinkovitch and Thewissen (1997) claimed that retroposons are “noise-free,” arguing that it is improbable that retropositional elements inserted themselves independently into orthologous positions in the genomes of different taxa. No phylogenetic data, however, are known a priori to be free of homoplasy (Wiley et al., 1991). Thus, molecular papers arguing for artiodactyl paraphyly are numerous. Different molecular analyses have found perissodactyls, hippopotamids, ruminants, and camels each to be most closely related to cetaceans. A hippopotamid–ruminant– cetacean grouping describes the most common nding of many of these studies, notably those with the densest character and taxon sampling. These unconventional topologies recovered from molecular analyses have implications not only for cetacean phylogeny but also for the monophyly of clades within Artiodactyla. A previous phylogenetic analysis of morphological characters divided Artiodactyla into two clades, each with extant and extinct members (Gentry and Hooker, 1988): Selenodontia, consisting of various ruminating artiodactyls (including Tragulidae, Cervidae, Girafdae, Antilocapridae, and Tylopoda and fossil relatives); and Bunodontia (including Suidae, Tayassuidae, and Hippopotamidae and fossil relatives). Many of the molecular topologies, however, support paraphyly of the clade including Suidae, Tayassuidae, and Hippopotamidae, challenging the hypothesis that these taxa are more closely related to each other than any is to either camels or ruminants (Matthew, 1929; Pickford, 1983; Gentry and Hooker, 1988). An analysis of cytochrome b and 12s rRNA sequences, combined with morphological data for extant taxa only (Montgelard et al., 1998), and aimed at testing the monophyly 461 of the clade that includes Suidae, Tayassuidae, and Hippopotamidae, reinforced the notion that conicting signals are present in morphological and molecular data sets. This analysis did not include cetaceans, thereby making it difcult to interpret the results in the context of many other molecular analyses. At the same time, monophyly of ruminants and camels, to the exclusion of pigs, peccaries, and hippos, has been argued on the basis of suites of morphological characters (Webb and Taylor, 1980; Vaughan, 1986; Gentry and Hooker, 1988; Langer, 1988) but is disrupted in many of the molecular phylogenies. The hypothesis that Artiodactyla is paraphyletic with respect to Cetacea contradicts traditional morphological ideas of artiodactyl monophyly but merits explicit testing in the wake of the numerous molecular studies supporting artiodactyl paraphyly. Furthermore, the morphological hypothesis of artiodactyl monophyly has been formulated without knowledge of the morphology of the astragalus of a primitive cetacean—the skeletal character argued to be among the most important for determining membership within Artiodactyla (Luckett and Hong, 1998). Because it is paraphyly, not polyphyly, of Artiodactyla that is in question, lack of evidence about the cetacean astragalus invites the query: on what basis do morphologists know that whales are not highly derived artiodactyls? We therefore investigated the morphological evidence for the following four questions: (1) Is Artiodactyla monophyletic with respect to Cetacea; (2) what is the sister taxon of Cetacea; (3) does exclusion of fossil taxa from the phylogenetic analysis result in a pruned version of the tree based on extant taxa alone or a different tree; and (4) what impact does the morphology of the astragalus of a primitive cetacean have on tree topology? MATERIALS AND METHODS The Signicance of a Morphological Data Partition When Most of the Ingroup Is Extinct We subscribe fully to the notion that a total evidence analysis (i.e., one based on com- 462 VOL. 48 SYSTEMATIC BIOLOGY bined molecular and morphological data) is an important step in the overall investigation of the position of Cetacea within Mammalia and that it is preferable to a taxonomic congruence approach for reasons outlined by Kluge (1989). It also has been argued in the overall discussion of the signicance of total evidence that system-based data partitions (i.e., molecules, osteology, behavior, or various subsets of these partitions) are not at present derived from an understanding of biological processes but are instead based on intuition (Kluge and Wolf, 1993). Total evidence analyses combining molecular data for a few extant taxa with morphological data for extant taxa and a large number of extinct taxa, however, face an operational obstacle imposed by extinction. With > 99% of all organisms extinct (Novacek and Wheeler, 1992), can such total evidence analyses be considered robust when there are no molecular, soft tissue, or behavioral data for the vast majority of life? Because of the potential importance of numerous extinct taxa for reconstructing the phylogeny of Cetacea, taxa that are known only from morphology and not from molecular biology, we believe it is also important to investigate and present the morphological (predominantly osteological) signal independently. In such a data partition, extinct taxa, which can far outnumber extant taxa in an ingroup, can be maximally inuential on tree topology. The vast majority of ingroup taxa relevant to cetacean phylogeny are extinct. Molecular and morphological contributions to the problem of the phylogenetic position of Cetacea are congruent in nesting Cetacea within ungulates, or the grandorder Ungulata (McKenna and Bell, 1997). Ungulata (sensu McKenna and Bell, 1997) is diverse and includes several extinct clades besides Artiodactyla, Cete (which includes Cetacea), and Perissodactyla (Table 1). In this study (see below) ve orders comprise the ingroup: Artiodactyla, Perissodactyla, Procreodi, Cete, and “Condylarthra.” Most (89%) of the genera in these orders are extinct (McKenna and Bell, 1997). Maximally, a phylogenetic study based on extant taxa alone can access no more than 11% of genera in the ingroup. At present, larger moleculebased phylogenies (Gatesy, 1997) have sampled only 1% of ingroup genera. Even the present study, which includes 37 ingroup taxa, almost triple that of Gatesy (1997), uses only 3% of relevant genera. Since the aim of this analysis is to recover deep splits between mammalian orders such as Artiodactyla, Perissodactyla, and Cetacea, sampling exhaustively within the ingroup may not be essential if the most primitive members of a clade can be sampled. It is not, however, likely that extant mammals are primitive members of the clades in question, because both molecular and morphological estimates indicate that the splits between mammalian orders are ancient (Novacek, 1992b; Kumar and Hedges, 1998), and their extant members have had much time to evolve homoplasies. TABLE 1. Estimate (derived from McKenna and Bell, 1997) of the extinction in the ingroup in this analysis and other recent molecule-based phylogenetic analyses of the position of Cetacea within Mammalia (e.g., Gatesy, 1997, and references therein). Shown are the number of extinct and extant genera from ve orders that make up the ingroup. This estimate of the number of ingroup taxa is conservative because the ingroup may include all genera within Ungulata (see McKenna and Bell, 1997) (i.e., Tubulidenatata, Arctostylopida, Litopterna, Notoungulata, Astrapotheria, Xenungulata, Pyrotheria, and Urantotheria) and because it is highly unlikely that the fossil record has preserved all ingroup taxa. Taxon Artiodactyla Perissodactyla Cete “Condylarthra” Procreodi Total Extinct genera Extant genera 589 (86%) 236 (97.5%) 243 (86%) 59 (100%) 29 (100%) 94 (14%) 6 (2.5%) 39 (14%) 0 0 1156 (89%) 139 (11%) The problem of looking at only a few taxa, particularly only a few extant mammalian taxa, is that because mammalian orders have ancient splits (long terminal branches) separated by short internodes (Novacek et al., 1998), sampling of the extant taxa alone risks reinforcing an inconsistent branching pattern attributable to long-branch attraction (Felsenstein, 1978). Adding fossils to a data 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA matrix has been argued to be one way to expose and overturn a long-branch attraction problem (Gauthier et al., 1988). The study of Gauthier et al., 1988, and others of seed plants (Doyle and Donoghue, 1987) and eutherians (Novacek, 1992a), all show that parsimony analysis of a matrix of morphological data for extant taxa alone results in a tree that is incongruent with results of a parsimony analysis of a more complete matrix (extant + extinct taxa) of the same characters. Interestingly, none of these empirical analyses has shown that the tree based on extant taxa alone was simply a pruned version of the tree based on the more complete matrix. The assumption that the tree resulting from more taxa is the most robust is implicit in this methodology and is logical, because there is no justication for excluding ingroup taxa. Each study performed experiments in which fossil taxa were deleted and parsimony analyses run to investigate changes to the tree topology (Donoghue et al., 1989). Drawing on independent evidence from simulations where the true tree is known, Huelsenbeck (1991) recognized that such deletion/restoration experiments are one of the few means of testing whether fossils have a signicant effect on tree topology. Inclusion of fossil taxa is particularly important when studying the relationships of taxa separated by large stretches of time because fossils aid in the identication of homoplasies that might otherwise go unrecognized (Doyle and Donoghue, 1987; Huelsenbeck, 1991). Essentially, fossils may break up long branches. Fossil taxa are particularly likely to break up long branches in the radiation of cetaceans and other ungulates. Morphological and molecular evidence argues that the splits among Cetacea, Artiodactyla, and Perissodactyla are > 50 million years old. A fossil-based estimate for the divergence of Cetacea from Artiodactyla is early Paleocene or very Late Cretaceous (~ 65 million years ago; Gingerich and Uhen, 1998); for Perissodactyla and Artiodactyla, Late Cretaceous (Novacek, 1992b; but see also Novacek et al., 1998). A molecule-based estimate for the divergence of Cetacea from Artiodactyla is ~ 58 million years ago, and 463 that for Cetacea + Artiodactyla from Perissodactyla is ~ 83 million years ago (Kumar and Hedges, 1998). Thus, by all estimates of divergence, the split under consideration is ancient and hence vulnerable to long-branch attraction problems. Mesonychians, which range from early Paleocene through late Eocene (McKenna and Bell, 1997), are fossils that meet two phylogenetically signicant criteria (Huelsenbeck, 1991): They are relatively complete, and they have a time of appearance relatively close to that of Cetacea. Many of the diverse array of exinct artiodactyls, perissodactyls, and “condylarths” known fulll similar criteria. Other studies have reinforced the importance of taxonomic sampling. Wheeler (1992) demonstrated on the basis of computer simulations that accuracy of cladograms is more affected by number of taxa included than by model of evolution, number of characters (specically, length of nucleotide sequences), or rate of evolution. Although Kim (1996) argued that adding taxa does not always increase phylogenetic accuracy, Graybeal (1998) recognized that many of the taxa added in his study did not break up long branches. She found that phylogenetic accuracy improved as the number of taxa increased, even if simultaneously the number of characters decreased. Importantly, the taxa added specically broke up long branches, and she emphasized that it is the addition of such taxa that improves phylogenetic accuracy, a result consistent with paleontological studies. We follow the methods of Doyle and Donoghue (1987) and Gauthier et al. (1988) and use deletion/restoration experiments with fossil taxa to examine the effect of extinction on tree topology. Indeed, investigation of the morphological (again, primarily osteological) signal alone is one of the few ways to maximize the empirical effect of extinct clades on tree topology. If tree topology changes on addition of morphological data from fossil taxa, one hypothesis explaining this result is that it would also happen if molecular data were accessible for all extinct ingroup taxa. Obviously, we cannot test this hypothesis empirically because of missing data. However, the operational reality faced 464 SYSTEMATIC BIOLOGY by total evidence analyses is that extinction creates a data partition that can inuence results and obscure phylogenetic signals. Because osteology can be studied across extinct and extant organisms, its signal alone deserves consideration, for its potential to offer an important perspective on relationships not apparent from neontological data. Data Collection Scoring all extant and extinct ungulate genera, particularly Artiodactyla, Perissodactyla, Cetacea, Mesonychia, and archaic ungulates, was an unrealistic point of departure because, as noted above, some of these taxa are highly diverse. From among the clades mentioned above we chose 37 genera to form the ingroup (Appendix 1), primarily following principles outlined by Hillis (1998:5 [methods 3 and 4]), and assumed that each of these genera is monophyletic on the basis of the morphological similarity of the species within it. The ingroup consists primarily of fossil genera known from relatively complete specimens, chosen both because we believe they are representative of forms close to the basal morphotype of various extant ungulate clades and because they are representative of the diversity of ungulate clades argued to be closest to Cetacea (Gentry and Hooker, 1988; Prothero et al., 1988; McKenna and Bell, 1997). We did not sample extensively within paenungulates because molecular and morphological phylogenies described above do not argue for a close relationship between Artiodactyla, Cetacea, and paenungulates . We did, however, include several taxa that have been argued on the basis of cladistic analysis (Thewissen and Domning, 1992) to be among the most primitive members of the paenungulate clade (i.e., Meniscotherium, Phenacodus; see Appendix 1, “archaic ungulates”). Other paenungulates are assumed to be nested among the taxa sampled. The artiodactyl sample includes the oldest member of this order, Diacodexis (Rose, 1982); an entelodont, Archaeotherium; an anthracothere, “Elomeryx”; an oreodontoid, Agriochoerus; a camelid, Poebrotherium; and several extant forms: Sus (pig), VOL. 48 Hippopotamus, Hexaprotodon (= Choeropsis, pygmy hippotamus), Ovis (sheep), Tragulus (chevrotain), and Camelus (camel) (McKenna and Bell, 1997). The taxonomy of anthracotheres is much in need of revision, and we place “Elomeryx” in quotes because we have also used specimens attributed to Bothriodon (Ancodus) to score the morphology of “Elomeryx”; these taxa appear to be very similar, are from similar deposits, and may be synonyms. Our sample of perissodactyls includes Equus (horse), an extinct equid, Hyracotherium, and an extinct tapiroid, Heptodon. All relatively complete genera of mesonychians are included in our analysis, as well as a number of extinct cetaceans—Pakicetus, Ambulocetus, and Remingtonocetus, among others—and two extant cetaceans: odontocete Tursiops (bottle-nosed dolphin) and a mysticete Balaenoptera (rorqual whale). Several relevant taxa, collectively described as “archaic ungulates” (see below and Appendix 1) or “condylarths,” were also included. One genus for which the assumption of monophyly remains controversial is Pakicetus. We follow Thewissen and Hussain (1998) as to which specimens constitute Pakicetus. These authors caution that some of these identications remain poorly substantiated. Where possible, original specimens were examined, but character coding was supplemented from the literature when original specimens were unavailable (Appendix 1). Ordinal-level relationships of mammals are incompletely resolved (Novacek, 1992b), making outgroup choice ambiguous in a study of this kind examining variation across more than four orders. In Novacek’s (1992b) tree, the node below our ingroup (Artiodactyla, Perissodactyla, and Cetacea) is “Condylarthra,” a paraphyletic assemblage of primitive, herbivorous-omnivorous, placental mammals loosely allied with ungulates (Carroll, 1988). Although “Condylarthra” is not specically designated as an outgroup in our analysis, our tree contains a number of “condylarths”: Arctocyon, Hyposodus, Meniscotherium, Phenacodus, triisodontines, 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA and mesonychians (Carroll, 1988), which effectively polarize characters within Artiodactyla, Perissodactyla, and Cetacea. The sister taxon of “Condylarthra” is an unresolved polytomy consisting of archontans, glires, carnivorans, and insectivorans (Novacek, 1992b). We chose one outgroup from this assemblage, the insectivoran Leptictis, and two others at still lower nodes on the mammalian tree: the primitive eutherian Asioryctes and the extant marsupial Didelphis. We decided against using Tubulidentata as an outgroup, despite the hypothesized proximity of this order to our ingroup (Novacek, 1992b), because the highly derived dentition of tubulidentates would be useless for polarizing dental characters in the ingroup. Our matrix heavily emphasizes osteological characters. We scored 123 morphological characters; 33 basicranial, 9 other cranial, 45 dental, 29 postcranial, and 7 soft morphological characters (Appendix 2). The amount of missing data for a given taxon ranged from 2% to 75% (average 29%; Appendix 3). Most (88, or 72%) of the characters were binary, and 35 (28%) were multistate. Multistate characters were treated in two ways: (1) all unordered, or (2) 24 of the 35 multistate characters ordered (20% of the total number of characters, 68% of multistate characters); these characters are specied in Appendices 2 and 3. Ordering was sequential as follows: 0 « 1 « 2 « 3. Multistate characters were not ordered if we did not hypothesize a hierarchical transformation for that character. Treatment of multistate characters as ordered or unordered constitutes an assumption about evolutionary process (Hauser and Presch, 1991; Wilkinson, 1992; Barriel and Tassy, 1993; Slowinski, 1993) that must be faced a priori in cladistic analysis. Unordered multistate characters assume that one character state can transform directly into any other character state without passing through an intermediate character state. Ordered multistate characters assume that transformations do pass through such an intermediate stage. Wilkinson (1992) argued that ordering certain characters is a logical extension of Hennig’s auxiliary principle 465 and “explains the similarity between a subset of the character states in terms of synapomorphy” (Wilkinson, 1992:380). Slowinski (1993) demonstrated that neither ordering nor unordering necessarily increases taxonomic congruence, one measure of phylogenetic accuracy. He concluded that either way of treating multistate characters is valid (Slowinski, 1993:163) and proposed ordering as favorable if a transformation series describes morphoclinal variation. The characters treated as ordered here capture hierarchical morphological change, and we argue that treating these characters as unordered would constitute a loss of information important for reconstructing this phylogeny. Certain characters discussed as important for substantiating the monophyly of Artiodactyla were not included in our study if we could not score them in a consistent fashion (e.g., the relative expansion of the pars facialis of the lacrimal bone). We excluded one soft morphological character, relative elongation of the blastocyst (Thewissen, 1994; Geisler and Luo, 1998), because it has not been described in sufcient detail for many of the extant taxa in question. The hypocone of artiodactyls has been described as nonhomologous to the hypocones of various of the ungulates (Prothero et al., 1988). This cusp is thought to develop from the metaconule in Artiodactyla and from the lingual cingulum in many other ungulates (Hunter and Jernvall, 1995). Instead of assuming a priori that the hypocones are nonhomologous, we scored both the metaconule and the hypocone on the basis of position on the tooth: the hypocone being the cusp in the distolingual corner of the upper molar, and the metaconule being the cusp positioned more labially or anterolabially on a diagonal line between the metacone and the protocone. Outgroup taxa do not necessarily have state 0 for every character (Appendix 3) for two reasons: (1) we use several genera rather than a hypothetical ancestor as an outgroup, and there is morphological variation among the outgroups; and (2) certain ordered characters (e.g., character 46) branch in two directions from the hypothesized primitive state, which is most easily coded by desig- 466 VOL. 48 SYSTEMATIC BIOLOGY nating the primitive state as 1 and the derived states as 0 and 2. Character Independence An assumption of parsimony analysis is that the characters used in a phylogenetic analysis (homology statements) are independent. As discussed by Kluge and Wolf (1993) independence in phylogenetic analysis must be addressed in two forms: (1) logical independence of characters, and (2) concurrent origin of the characters. The most egregious violation of the assumption of independence is to score redundant characters, that is, those that fail the test of logical independence. Farris (1983:20) describes an example of two characters that are not logically independent as scoring “tarsal segments and twice that number” separately. We strive not to include duplications of this nature in our data matrix. The second criterion for independence, concurrent origin, is much more difcult to verify (Kluge and Wolf, 1993), particularly for fossil taxa. Rejection of the hypothesis that two synapomorphies evolved independently requires knowledge of tokogenetic relationships that “even our best-supported phylogenetic propositions do not provide” (Kluge and Wolf, 1993:192). Numerous biological processes (e.g., genetic, selective, developmental) may contribute to the non-independence of different characters. However, as Kluge and Wolf (1993:192) contend, “the actual processes responsible for nonindependent evolution cannot be read from even the most detailed patterns of organism relationship.” We interpret this to mean that similar distribution of character states in two or more characters is not necessarily indicative of the nonindependence of those characters. Hence, to dismiss covarying characters without exquisite knowledge of tokogenetic relationships risks an a priori dismissal of phylogenetic signal (character congruence). We, therefore, include all logically independent characters in our analysis, even two characters that appear to covary exactly (i.e., characters 81 and 82). Exact covariance does not seem sufcient reason to combine two characters in the matrix. These characters cannot be scored for all taxa in the matrix, and thus it is quite possible that the exact covariance may disappear as our knowledge of the fossil record improves. Secondly, we are studying taxa whose origin is separated by millions of years, making it very likely that we have not discovered, and have not scored, all relevant forms. Although in certain cases characters may appear to have been acquired at the same node (i.e., covary exactly), this may be a byproduct of extinction, fossilization, and the scale of the analysis, not evidence of the nonindependence of the synapomorphies. This paper contains references to basicranial, cranial, dental, postcranial, and soft morphological data partitions. These partitions are, however, simply constructs to organize information and are not meant at present to convey any knowledge of biologically based character-relatedness. Phylogenetic Analyses The matrix was compiled in MacClade 3.07 (Maddison and Maddison, 1992) and analyzed by using either the heuristic or branch-and-bound search algorithms in PAUP 3.1.1 (Swofford, 1993), depending on the size of the character-taxon matrix in a particular run. Once trees were found in PAUP, character distributions were analyzed in MacClade. Because of the large size of the data set, a branch-and-bound analysis in PAUP was possible only in runs that excluded all fossils (runs 3 and 4; see below). All other searches were heuristic with the following settings: addition sequence random; number of replications = 1,000; treebisection and reconnection (TBR) branchswapping performed; MULPARS option in effect; steepest descent option not in effect; branches having maximum length 0 collapsed to yield polytomies; trees unrooted; uninformative characters excluded; and multistate taxa interpreted as polymorphism. Trees were rooted with Didelphis as outgroup. All tree lengths reported are calculated in PAUP 3.1.1 (Swofford, 1993). The character-taxon matrix was subject to maximum parsimony analysis under the following parameters: run 1, certain multistate characters ordered (Appendices 2 and 3); 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA run 2, all multistate characters unordered; run 3, extant taxa only, and certain multistate characters ordered; run 4, extant taxa only, and all multistate characters unordered; run 5, certain multistate characters ordered, and evidence from Thewissen et al. (1998) included; run 6, all multistate characters unordered, and evidence from Thewissen et al. (1998) included. Runs 3 and 4 with extant taxa only were constructed to mimic taxon sampling available in molecular analyses, thereby testing the effect of fossils on tree topology. On the basis of available data, we tested the hypothesis that the morphology of the cetacean astragalus is central to establishing artiodactyl monophyly (Milinkovitch and Thewissen, 1997; Luckett and Hong, 1998) by adding in runs 5 and 6 new data on what are described as astragali of primitive cetaceans (Thewissen et al., 1998). The characters described by Thewissen et al. (1998) are found on two specimens, neither of which is associated with diagnostic cetacean material. For this reason, we treat these data with particular caution. These specimens potentially provide data on three characters that are missing data in our initial matrix: the navicular facet (character 104), the sustentacular facet (character 105), and the lateral process (character 107). (The cetacean conditions for the astragalar canal [character 103] and the proximal astragalus [character 109], are known for Ambulocetus [Thewissen et al., 1996].) On the basis of the new bones, we coded character 104 as a new state 3, at (in Pakicetus); character 105 as state 0, narrow (in Pakicetus); and character 107 as state 1, absent (in Ambulocetus). The bone described by Thewissen et al. (1998) appears to introduce new variability to character 105 not orignally reected in our coding regime. Where our matrix emphasizes width of this character, Thewissen et al. (1998) emphasize length. This character is very likely in need of more detailed description in the future to accommodate the variation in the ingroup. We calculated decay values (Bremer, 1988) for nodes in the consensus trees from runs 1 and 2 by conducting heuristic searches (with 100 replications, using TBR branchswapping and random addition) in PAUP 467 3.1.1 (Swofford, 1993) and using constraint trees to nd trees that do not contain a particular clade in question. The difference in length between these trees and our most-parsimonious tree equals the decay index for each node. We emphasize that the decay values are estimates because all searches are heuristic. Estimated bootstrap values (Felsenstein, 1985) were calculated in PAUP* 4.0b1 (Swofford, 1998) for the unordered and ordered trees by using heuristic searches (1,000 replications) with simple addition and TBR branch-swapping, and including groups compatible with the 50% majority rule consensus. Optimization of characters was performed by using both ACCTRAN and DELTRAN options in PAUP 3.1.1. Finally, to compare the topology found in one of the more densely sampled molecule-based analyses (Gatesy, 1997) with the morphology-based results here, we used a backbone constraint tree in PAUP* 4.061 to search for the most-parsimonious morphological tree under this molecular topological constraint. This exercise provided an estimate of whether or not forcing the topology generated on the basis of molecular data alone substantially increased the length of the tree based on morphology alone. This experiment was performed on the matrix with ordered characters only. RESULTS Run 1, in which 20% of the total number of characters (69% of multistate characters) were ordered (Fig. 2, Appendices 2 and 3), produced two most-parsimonious trees of 536 steps each, the strict consensus of which supported monophyly of each the following clades: Artiodactyla, Perissodactyla, Mesonychia, Mesonychidae, and Cetacea. The sister taxon of Cetacea is Mesonychia (Hapalodectidae and Mesonychidae), and the sister group to the cetaceanmesonychyian clade is the triisodontine arctocyonid, Andrewsarchus. Among clades with extant members, Artiodactyla is more closely related to the mesonychian-cetacean clade than is Perissodactyla. Within Artiodactyla, Sus, the hippopotamids, the entelodont, and the anthracothere are more 468 SYSTEMATIC BIOLOGY VOL. 48 FIGURE 2. Strict consensus of two most-parsimonious trees of 536 steps each from run 1 (CI = 0.384, RI = 0.696, HI = 0.698, RC = 0.267), 20% of total characters ordered (68% of multistate characters) (Appendices 2 and 3). Artiodactyla (A), Cetacea (C), Mesonychidae (D), Mesonychia (M), and Perissodactyla (P) are each monophyletic. Table 2 describes synapomorphies found with ACCTRAN and DELTRAN optimization at nodes marked with letters. Extant taxa = bold, extinct taxa = y. Numbers above branches are estimated decay values; numbers below branches are estimated bootstrap support. closely related to each other than any is to the ruminant clade, which consists of Ovis, Tragulus, Camelus, and Poebrotherium. Meniscotherium, Phenacodus, and Hyopsodus form a monophyletic clade with Perissodactyla. The ingroup overall is not monophyletic; the “condylarth” Arctocyon falls outside of a clade that includes the insectivoran Lep- 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA tictis. The data are not strong enough to unite the two triisodontine arctocyonids Andrewsarchus and Eoconodon as monophyletic, or even to draw Eoconodon into the ingroup, where it is hypothesized to belong based on current taxonomy (McKenna and Bell, 1997). Estimated decay indices indicate that the monophyly of Artiodactyla is more strongly supported than either the sister taxon relationship between Mesonychia and Cetacea or the clade Artiodactyla + (Andrewsarchus + (Mesonychia + Cetacea))). Run 2 (all characters unordered; Fig. 3) resulted in 36 most-parsimonious trees of 505 steps each, the strict consensus of which supports monophyly of each of the following clades: Artiodactyla, Perissodactyla, and Cetacea. The sister taxon of Cetacea is now the mesonychian Hapalodectes, the polarity within Mesonychidae has reversed in comparison with Figure 2, and both Mesonychia and Mesonychidae are paraphyletic with respect to Cetacea. Two results occur in all most-parsimonious trees, as in run 1: Poebrotherium and Camelus form a clade with the ruminants (Tragulus and Ovis), and Hyopsodus, Phenacodus, and Meniscotherium form a clade with Perissodactyla (Fig. 2). The tree topology within Artiodactyla is relatively similar to that found in run 1, except that the oreodontoid Agriochoerus joins the ruminant clade. Removing all fossils from the charactertaxon matrix and treating certain characters as ordered (run 3; Fig. 4a-c) recovers three most-parsimonious trees of 173 steps each, the strict consensus of which is a paraphyletic Artiodactyla with internal relationships of the ingroup poorly resolved. The three most-parsimonious trees (Fig. 4) support a cetacean sister group relationship with Ovis or with Ovis + (Camelus + Tragulus). Running extant taxa only with all multistate characters unordered (run 4; Fig. 4d-e) resulted in two most-parsimonious trees of 150 steps each, again with a paraphyletic Artiodactyla and Cetacea nested inside as the sister taxon of either hippopotamids or hippopotamids + Sus. The backbone constraint tree (Fig. 5b), which forced the most-parsimonious solution for the morphological matrix congruent 469 with one topology generated on the basis of molecules (Gatesy, 1997), was 551 steps, 15 steps longer than the most-parsimonious tree that was based on our morphological matrix with certain multistate characters ordered (Fig. 2). Under this constraint, mesonychians are still the sister taxon of Cetacea, and this entire clade falls inside Artiodactyla. Optimizations Synapomorphies for the tree found in run 1 (certain characters ordered) calculated with both ACCTRAN and DELTRAN optimization algorithms (Table 2) are described below. Such synapomorphies describe the strongest support for particular nodes because they persist whether parallelism or reversals are favored. The clade including Perissodactyla + (Artiodactyla + (Andrewsarchus + (Mesonychia + Cetacea))) is supported by four such synapomorphies (Table 2; Fig. 2: node a), which come from basicranial, dental, and postcranial partitions of the data matrix. Artiodactyla + (Andrewsarchus + (Mesonychia + Cetacea)) is supported by 12 synapomorphies (Fig. 2: node b) from basicranial, cranial, postcranial, and soft-morphological partitons of the data matrix. Mesonychia + Cetacea, however, is supported by only two synapomorphies (Fig. 2: node d). This is due in large part to the fragmentary nature of Andrewsarchus, the outgroup to this clade, which makes polarity calculations ambiguous. Cetacea is one of the more strongly supported nodes with 15 synapomorphies from basicranial, cranial, and dental data (Fig. 2: node e). Mesonychia (Hapalodectes + Mesonychidae) (Fig. 2: node f), by contrast, is supported by three dental synapomorphies, and Mesonychidae by four synapomorphies, from basicranial, cranial, and dental data partitions (Fig. 2: node g). Artiodactyla (Fig. 2: node j) is supported by 8 synapomorphies from basicranial, dental, and postcranial partitions of the data matrix. Perissodactyla, and clades that include stem taxa leading to it, are well-supported nodes with several synapomorphies each (Table 2). 470 SYSTEMATIC BIOLOGY VOL. 48 FIGURE 3. Strict consensus of 36 most-parsimonious trees, 505 steps each (CI = 0.406, RI = 0.694, HI = 0.679, RC = 0.282), all characters unordered (run 2). Artiodactyla (A), Cetacea (C), and Perissodactyla (P) are each monophyletic; Mesonychidae (“D”), and Mesonychia (“M”) are paraphyletic. Numbers above branches represent estimated decay values, numbers below branches represent estimated bootstrap support, branches without bootstrap values are nodes recovered in < 5% of bootstrap replicates . Extant taxa are in bold; extinct taxa = y. Unequivocal synapomorphies supporting artiodactyl monophyly in all 36 trees are the following: 27. Alisphenoid canal, present (1) ® absent (0); 32. Post-temporal canal, present (0) ® absent (1); 86. dp4 , resembles M1 (0) ® six-cusped (1); 92. Entepicondyle, wide (0) ® narrow (1); 93. Entepicondylar foramen, present (0) ® absent (1); 96. Proximal radius, two fossae (1) ® three fossae (2); 105. Sustentaculum, narrow (0) ® wide (1); 107. Lateral process of astragalus, present (0) ® absent (1); 109. Proximal astragalus, at (0) ® deeply grooved (2). The unequivocal synapomorphies supporting Hapalodectes + Cetacea are these: 37. Lacrimal tubercle, present (1) ® absent (0); 60. M3 , absent (3) ® equal to M2 in size (1); 83. Reentrant grooves, distal (2) ® proximal (0); and 95. Olecranon process, deep (1) ® shallow (0). 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA 471 Because removal of extinct taxa from the data matrix changed the topology of the tree (compare Figs. 2 and 4a-c, or Figs. 3 and 4d-e), we examined which of the characters specically supporting the sister taxon relationship with Cetacea were acting as homologies in the cladograms of extant taxa alone (Fig. 4) but were homoplasies when the fossils were included. (Figs. 2 and 3; Table 2). These characters, listed in the caption to Figure 4, came from the cranial, postcranial, or soft-morphological data partitions. Addition of Astragalar Data from Thewissen et al. (1998) When the data from Thewissen et al. (1998) are incorporated into the matrix (runs 5 and 6), cetaceans are scored as having one character that was shown above (Table 2) to be a synapomorphy for Artiodactyla: character 107, lateral process of the astragalus absent. Adding these characters to the matrix with certain multistate characters ordered (run 5) results in 2 trees of 539 steps each with a topology no different from that of Figure 2. Adding these characters to the matrix with all multistate characters unordered results in a concensus tree no different from that found in Figure 3; that is, the new data do not change the toplogies of the trees. D ISCUSSION Results of these parsimony analyses show that on the basis of 123 morphological characters (116 osteological and 7 soft morFIGURE 4. Results of parsimony analyses with all extinct taxa removed from the data matrix (Appendix 3). (a-c) The three most-parsimonious trees of 173 steps each (CI = 0.618, RI = 0.590, HI = 0.422, RC = 0.365), based on the matrix with 20% of multistate characters ordered (Appendices 2 and 3) (run 3); Cetacea is the sister taxon of Ovis in (a) and (b) or Ovis + (Camelus + Tragulus) in (c). Synapomorphies supporting the sister taxon relationship between Cetacea and other taxa found with both ACCTRAN and DELTRAN algorithms are as follows: Tree a-1. Subarcuate fossa, present (0) ® absent (1); 15. Articulation of ectotympanic to squamosal, broad (0) ® absent (3); and 16. Exoccipital-ectotympanic contact, present (1) ® absent (0); tree b-1 and 15; tree c-26. Foramen ovale, medial, posterior wall formed by petrosal (2) ® anterior, posterior wall formed by alisphenoid (0); 94. Length of olecranon process, long (1) ® short (0); 113. Second metatarsal, unreduced (0) ® highly reduced (2); and 114. Fifth metatarsal, unreduced (0) ® highly reduced (2). (d and e) The two most-parsimonious trees of 150 steps each (run 4) (CI = 0.660, RI = 0.617, HI = 0.387, RC = 0.407), based on the matrix with all multistate characters unordered. Synapomorphies supporting Cetacea as the sister taxon of Sus + (Hippopotamus + Hexaprotodon) (tree d) are these: 7. Mastoid, exposed (0) ® not exposed (1); 23. Postglenoid foramen, present, enclosed by squamosal (0) ® absent (2). The following support Cetacea as the sister taxon of Hippopotamus + Hexaprotodon (tree e): 1. Subarcuate fossa: present (0) ® absent (1); 15. Articulation of ectotympanic to squamosal: broad (0) ® absent (3); 121. Hair: abundant (0) ® absent (1); and 122. Sebaceous glands: present (0) ® absent (1). 472 SYSTEMATIC BIOLOGY FIGURE 5. Comparison of backbone constraint tree to molecule-based tree, both addressing the question of artiodactyl monophyly. (a) pruned tree from Gatesy (1997) used as the constraint; (b) strict consensus of nine most-parsimonious trees of 551 steps each resulting from a heuristic search using the tree from (a) as a backbone constraint and the matrix with certain multistate characters ordered. Extant taxa are in bold. Cetacea (C), Mesonychidae (D), Mesonychia (M), and Perissodactyla (P) are each monophyletic, “A” = Artiodactyla here is monophyletic only if Cetacea are included. VOL. 48 phological) from 10 extant and 30 extinct taxa, Artiodactyla is more closely related to Cetacea than is Perissodactyla. This conclusion is supported when certain multistate characters (20% of the total data set) are treated as ordered, when all multistate characters are treated as unordered, and when we exclude fossils from the analysis. We believe that results of runs with certain multistate characters ordered better reect hierarchical morphoclinal transformations and relationships between different character states (Wilkinson, 1992; Slowinski, 1993), that are important to reconstructing phylogeny. With certain characters ordered, the sister taxon of Cetacea is a monophyletic Mesonychia, as has been argued on the basis of dental characters alone (O’Leary, 1998a). The triisodontine arctocyonid Andrewsarchus is the stem taxon to this cetacean-mesonychian clade, and a monophyletic Artiodactyla is the sister taxon to that clade (Fig. 2). This tree also supports traditional groupings within Artiodactyla (Gentry and Hooker, 1988): a ruminantcamel clade and a hippopotamid-suid clade, each with extinct relatives. That Cetacea is more closely related to Artiodactyla than to Perissodactyla corroborates the conclusions of virtually all molecule-based phylogenetic analyses of this question noted in the introduction, as well as several exclusively morphological descriptions or analyses (Gingerich et al., 1990; Novacek, 1992b; Thewissen and Hussain, 1993; Geisler and O’Leary, 1997; Geisler and Luo, 1998), but is incongruent with conclusions of several other morphological analyses (Novacek and Wyss, 1986; Wible, 1987; Prothero et al., 1988; Novacek, 1989; Prothero, 1993; Thewissen, 1994). When all multistate characters are treated as unordered, the strict consensus of 36 most-parsimonious trees also has a monophyletic Artiodactyla, but Mesonychia, Mesonychidae, and triisodontine arctocyonids are paraphyletic. Other nodes at which this tree is congruent with that from the run with certain multistate characters ordered are these: monophyly of Perissodactyla, the order of stem taxa leading to Perissodactyla (Meniscotherium, Phenacodus, Hyopsodus), a clade including 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA 473 TABLE 2. Character optimizations for particular nodes in the two most-parsimonious trees from run 2 (certain multistate characters ordered, Fig. 2). Character state transformations were calculated by using both ACCTRAN and DELTRAN algorithms in PAUP 3.1.1. Complete character state descriptions are given in Appendix 2. Indicated are transformations found with (A) ACCTRAN only, (D) DELTRAN only, or (A/D) ACCTRAN and DELTRAN (in bold). Node a b c Character A 2. Tegmen tympani A/D 11. Ectotympanic A/D 60. M3 equal to M2 A 95. Olecranon fossa A/D 97. Distal radius A/D 115. Distal phalanges A/D 16. Exoccipital-ectotympanic contact A/D 17. Sigmoid process A 20. Squamosal part of external auditory meatus A/D 37. Lacrimal tubercle A 56. M1 -M2 hypocone A 58. Paraconule on M2 A 59. Metaconule on M2 A/D 92. Entepicondyle A 93. Entepicondylar foramen A/D 99. Manus A/D 104. Navicular facet A/D 108. Pes A/D 112. First metatarsal A/D 118. Lumen A/D 119. Stomach lumen A/D 120. Cavernous tissue of penis A/D 123. Lung bronchi D 20. Squamosal part of external auditory meatus A 22. Internal carotid foramen A/D 23. Post-glenoid foramen A 25. Position of foramen for ramus superior of stapedial artery A 26. Foramen ovale A 30. Foramen rotundum A 31. Mastoid foramen A 33. Preglenoid process A 52. M2 metacone D 56. M1 -M2 hypocone A/D 57. Trigon basin A 71. M2 paraconid A 73. M3 paraconid A 78. Protoconid A 83. Reentrant grooves A 94. Olecranon process A 95. Olecranon fossa A 98. Centrale A 116. Distal phalanges A 121. Hair A 122. Sebaceous glands Transformation (0) Uninated ® (1) inated (0) Ring ! (1) bulla (2) Smaller than M2 ! (1) equal (0) Shallow ® (1) deep (0) Single, concave fossa ! (1) scaphoid and lunate fossae (0) Curved ! (1) straight (0) Absent ! (1) present (0) Absent ! (1) present (0) Absent ® (1) deep groove (1) Present ! (0) absent (1) Present ® (0) absent (1) Present ® (0) absent (1) Present ® (0) absent (0) Wide ! (1) narrow (0) Present ® (1) absent (0) Mesaxonic ! (1) paraxonic (0) Convex ! (1) saddle-shaped (0) Mesaxonic ! (1) paraxonic. (1) Reduced ! (2) highly reduced (0) Unilocular ! (1) plurilocular (0) Absent ! (1) present (0) Abundant ! (1) sparse (0) Two ! (1) three (0) Absent ® (1) deep groove (0) Absent ® (1) present (0) Enclosed by squamosal ! (1) medial to petrosal/squamosal suture (0) In petrosal/squamosal suture ® (1) anterolateral to epitympanic recess (1) Medial ® (0) anterior (0) Absent ® (1) present (0) Present ® (1) absent (0) Absent ® (1) present (0) Subequal to paracone ® (1) half the size of paracone (1) Present ® (0) absent (0) Broad ! (1) narrow (0) Present ® (1) absent (0) Present ® (1) absent (0) Subequal to height of talonid ® (1) twice height of talonid (1) Absent ® (0) proximal (0) Short ® (1) long (1) Deep ® (0) shallow (1) Absent ® (0) present (0) Compressed ® (1) broad (0) Abundant ® (1) absent (0) Present ® (1) absent (continued on next page) 474 TABLE 2. Node d e VOL. 48 SYSTEMATIC BIOLOGY Character D 2. Tegmen tymapani D 31. Mastoid foramen A 45. Embrasure pits D 52. M2 metacone A 54. Lingual cingulum on M2 A/D 57. Trigon basin D 73. M3 paraconid D 83. Reentrant grooves A/D 84. Talonid basins A/D 85. M3 hypoconulid D 94. Olecranon process D 116. Distal phalanges A 3. Anterior process of petrosal A 4. Tensor tympani fossa D 4. Tensor tympani fossa A 5. Sulcus for internal caroid artery A/D 6. Proximal stapedial artery sulcus A/D 7. Mastoid process external exposure D 9. Facial nerve sulcus A 9. Facial nerve sulcus A/D 12. Pachyosteosclerotic involucrum A/D 15. Articulation of ectotympanic bulla to squamosal A/D 24. Foramen for ramus superior of stapedial artery D 26. Foramen ovale A/D 27. Alisphenoid canal A 33. Preglenoid process A 35. Postorbital bar A 44. Premaxillae A/D 48. P4 protocone A/D 49. P4 paracone A 52. M2 metacone A/D 67. Lingual cingulid A/D 68. M1 paraconid A/D 70. M2 paraconid A 72. M3 paraconid A/D 74. M1 metaconid A/D 75. M2 metaconid A 76. M3 metaconid A/D 87. Elongate shearing facets A 91. Scapular spine A 96. Proximal radius (Continued) Transformation (0) Uninated ® (1) inated (0) Present ® (1) absent (0) Absent ® (1) present (0) Sub-equal to paracone ® (1) half size of paracone (0) Present ® (1) absent (1) Narrow ! (2) very narrow (0) Present ® (1) absent (1) Absent ® (0) proximal (0) Basined ! (1) reduced (0) Long ! (2) absent (0) Short ® (1) long (0) Compressed ® (1) broad (0) Absent ® (1) present (1) Circular pit, no groove ® (2) circular pit, deep groove (0) Elongate fossa ® (2) circular pit, deep groove (0) Present ® (1) absent (0) Present ! (1) absent (0) Present ! (1) absent (0) Absent ® (2) anterior wall = mastoid process (3) Anterior wall formed by meatal tube ® (2) anterior wall formed by mastoid process (0) Absent ! (1) present (0) Broad ! (1) circular facet (0) Present ! (1) absent (1) Medial to glenoid fossa, posterior wall = alisphenoid ® (0) anterior to glenoid fossa, posterior wall formed by alisphenoid (1) Present ! (0) absent (1) Present ® (0) absent (0) Absent ® (1) present, almost complete (0) Short ® (1) elongate (0) Present ! (1) absent (0) Equal to height of M1 paracone ! (1) twice height of M1 paracone (1) Half the size of the paracone ® (2) highly reduced (0) Absent ! (1) present (0) Present ! (1) absent (0) Present ! (1) absent (0) Present ® (1) absent (0) Present ! (1) absent (0) Present ! (1) absent (0) Present ® (1) absent (0) Absent ! (1) present (0) Acromion overhangs glenoid ® (2) acromion does not overhang glenoid (1) Two fossae ® (0) one fossa (continued on next page) 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA TABLE 2. Node f g Character A 14. Median furrow of tympanic bulla A 19. Ectotympanic part of meatal tube D 30. Foramen rotundum D 33. Preglenoid process A 41. Mandibular condyle height A/D 50. P4 metacone A/D 53. M1 parastyle D 54. Lingual cingulum A/D 69. M1 paracristid position D 71. M2 paracristid position A 90. Number of sacrals A 93. Entepicondylar foramen D 4. Tensor tympani fossa A 9. Facial nerve sulcus A 22. Internal carotid foramen A/D 23. Post-glenoid foramen D 25. Foramen for ramus superior of stapedial artery D 26. Foramen ovale h i A/D 29. Posterior opening of alisphenoid canal A 37. Lacrimal tubercle D 41. Height of mandibular condyle A 45. Embrasure pits A/D 60. M3 size A/D 83. Reentrant grooves A/D 95. Olecranon fossa D 2. Tegmen tympani A 14. Median furrow of tympanic bulla A 41. Condyle D 58. Paraconule on M2 D 59. Metaconule on M2 A/D 65. P3 metaconid A/D 66. P4 metaconid A/D 77. M1 -M2 hypoconulid D 78. Protoconid D 90. Number of sacral vertebrae D 95. Olecranon fossa A/D 106. Astragalar-cuboid contact A 4. Fossa for tensor tympani D 5. Transpromontorial sulcus for internal carotid artery D 6. Proximal stapedial artery D 53. M1 parastyle A/D 79. Protoloph A 90. Number of sacral vertebrae A 91. Scapular spine A/D 92. Entepicondyle A/D 93. Entepicondylar foramen A/D 103. Astragalar canal A/D 104. Navicular facet of astragalus 475 (Continued) Transformation (1) Median notch ® (0) absent (1) Short ® (2) long (0) Absent ® (1) present (0) Absent ® (1) present (1) Even with dentition ® (0) below dentition (0) Absent ! (1) present (0) Absent ! (2) strong (0) Absent ® (1) present (0) Lingual ! (1) anterior (0) Lingual ® (1) anterior (2) Four ® (1) two or three (1) Absent ® (0) present (0) Elongate fossa ® (1) circular pit (3) Anterior wall formed by meatal tube ® (1) anterior wall formed by squamosal (1) Present ® (0) absent (1) Present ! (2) absent (0) In petrosal/squamosal suture ® (1) anterolateral to epitympanic recess (1) Medial to glenoid fossa, posterior wall formed by alisphenoid ® (0) anterior to glenoid fossa (0) Separated from foramen ovale ! (1) in a recess with the foramen ovale (0) Absent ® (1) present (1) Even with dentition ® (0) below dentition (1) Present ® (0) absent (1) Equal to M2 ! (2) small (0) Proximal ! (2) distal (0) Shallow ! (1) deep (0) Uninated ® (1) inated (1) Median notch ® (2) anteroposterior furrow (1) Even with dentition ® (2) superior to dentition (0) Absent ® (1) present (0) Absent ® (1) present (0) Absent ! (1) present (0) Absent ! (1) present (0) Absent ! (1) present (1) Twice height of talonid ! (0) subequal to height of talonid (1) Two or three ® (2) four (0) Shallow ® (1) deep (1) Present ! (0) absent (0) Elongate fossa ® (1) circular pit (0) Present ® (1) absent (0) Present ® (1) absent (0) Absent ® (1) weak (0) Absent ! (1) present (2) Four ® (3) ve or six (0) Large acromion ® (1) small acromion (0) Wide ! (1) narrow (0) Present ! (1) absent (0) Present ! (1) absent (0) Convex ! (1) saddle-shaped (continued on next page) 476 TABLE 2. Node j VOL. 48 SYSTEMATIC BIOLOGY (Continued) Character A/D 109. Proximal astragalus A/D 112. First metatarsal A/D 114. Fifth metatarsal A 2. Tegmen tympani D 4. Tensor tympani fossa A/D 27. Alisphenoid canal A/D 32. Post-temporal canal A/D 36. Frontal and maxillary contact D 78. Protoconid A/D 86. dp4 A 91. Acromion D 93. Entepicondylar foramen D 95. Olecranon fossa A/D 96. Proximal radius D 98. Centrale A/D 101. Third trochanter A/D 104. Navicular facet of astragalus A/D 105. Sustentacular width A 106. Astragalar-cuboid contact A/D 107. Lateral process of astragalus A 109. Proximal astragalus camels and ruminants, and one including hippopotamids, Sus, and fossil relatives. Milinkovitch and Thewissen (1997) argued that determination of the sister group of Cetacea hinges strongly on the recovery of an astragalus of a primitive cetacean (e.g., Pakicetus, Ambulocetus). Two fragmentary astragali argued to be cetaceans on the basis of faunal data have recently been described (Thewissen et al., 1998). These bones are not, however, associated with diagnostic cetacean material, and we treat their impact on the phylogeny of Cetacea with some caution. The new fossils are extremely fragmentary, and we have concluded that these bones introduce new and possibly autapomorphic character states not previously recognized. Addition of these new fossils does not change the tree topology or the number of trees found whether multistate characters are treated as ordered or unordered. These bones have been described as exhibiting more similarities to artiodactyls than to mesonychians (Thewissen et al., 1998); nonetheless, the invariance of our result may Transformation (1) Grooved ! (2) deeply grooved (1) Reduced ! (2) highly reduced (1) Reduced ! (2) highly reduced (1) Inated ® (0) uninated (0) Elongate fossa ® (1) circular pit (1) Present ! (0) absent (0) Present ! (1) absent (0) Absent ! (1) present (1) Twice height of talonid ® (0) subequal to height of talonid (0) Resembles M1 ! (1) six-cusped (0) Large ® (1) small (0) Present ® (1) absent (0) Shallow ® (1) deep (1) Two fossae ! (2) three fossae (0) Present ® (1) absent (0) Present ! (1) highly reduced (1) Saddle-shaped ! (2) trochleated (0) Narrow ! (1) wide (1) Small ® (2) large (0) Present ! (1) absent (0) Flat ® (2) deeply grooved be related to the fact that artiodactyl monophyly is supported by basicranial, dental, and postcranial synapomorphies, not simply synapomorphies based on ankle morphology. As found in other phylogenetic analyses (Gauthier et al., 1988), removing fossils from an analysis does not simply result in a pruned version of the tree generated from the matrix with extinct and extant taxa, but instead results in a tree with a different topology. When all extinct taxa are removed from the character-taxon matrix, mimicking the taxon sampling of many molecular analyses, we consistently recover artiodactyl paraphyly with Cetacea as the sister taxon of Ovis, a clade including Ovis + (Camelus + Tragulus), hippopotamids, or hippopotamids + Sus. Characters supporting these sister taxon relationships with Cetacea come from a variety of anatomical systems and are interpreted as homoplasies in the more complete matrix that includes extant and extinct taxa. Interestingly, artiodactyl paraphyly is also the result found in most 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA molecule-based analyses of the phylogenetic position of cetaceans, studies that cannot sample extinct taxa. The topology supported in the molecule-based analysis of Gatesy (1997) would increase the length of the most-parsimonious tree based on our matrix by 15 steps, a substantial number in a tree based on only 123 characters. Given a conservative estimate that 89% of the ingroup genera relevant to the phylogenetic position of Cetacea are extinct, we propose that analysis of a morphology-based, largely osteological, phylogenetic signal is particularly important, because extinct taxa can be maximally effective in such an analysis. We cannot test the effect of most molecular data on tree topology, because they are not available for mesonychians, triisodontine arctocyonids, other “condylarths,” lepticitids, basal artiodactyls, or basal perissodactyls. These extinct taxa may have even more information on branching sequences and character transformation than do extant taxa, because by virtue of their antiquity the extinct taxa have had less time to evolve homoplasies. The stem taxa that we have added here, mesonychians—“condylarths,” triisodontines, and basal artiodactyls and perissodactyls—introduce so much data on character transformation that they overturn nodes supported in most-parsimonious trees that are based on extant taxa alone (Figs. 2–4). These stem taxa simultaneously introduce substantial homoplasy: The consistency index (CI) for the trees based on extant taxa alone (Fig. 4a-c) is 0.618, whereas that for the more complete matrix (extant + extinct) taxa (Fig. 2) is 0.391. Adding ingroup taxa to a phylogenetic problem, as we have done, typically increases homoplasy (Sanderson and Donoghue, 1989), something that may affect the bootstrap and decay values of the nal tree. Strict comparisons between decay and bootstrap values found in this study and those in moleculebased studies are not a means of assessing relative robustness of the different phylogenetic hypotheses they support because the trees are derived from different characters and taxa. Values presented here for the decay and bootstrap are lower than 477 those found in a molecule-based analysis of Gatesy (1997), which supports a sister taxon relationship between hippopotamids and cetaceans, not artiodactyl monophyly. It has been shown elsewhere that addition of an extra ingroup taxon with homoplasies to a parsimony analysis can reduce the stability of the ingroup (i.e., make it decay faster) (Novacek, 1991). However, if that taxon is part of the ingroup, it simultaneously contributes to an increase in understanding of character transformation and relationships. Given such effects, Novacek (1991:347) argued that “cladistic results are not uniformly amenable to evaluation by simple measures of the stability of their components.” Such measures include the bootstrap and decay indices. Novacek (1991) argued further that addition of stem taxa may elucidate character transformation at the price of destabilizing certain nodes. In an important effort to synthesize information on differing molecular and morphological hypotheses of cetacean origins, Gatesy et al. (1996) proposed three hypotheses of relationship incorporating extinct and extant taxa: (1) cetaceans and mesonychians are derived artiodactyls; (2) cetaceans are derived artiodactyls and the sister group of hippopotamids, and mesonychians are a separate clade outside of Artiodactyla; and (3) Artiodactyla is monophyletic and mesonychians are the sister taxon of cetaceans. On the basis of all the morphological evidence presented here, neither mesonychians nor cetaceans are derived artiodactyls (hypotheses 1 and 2), making hypothesis 3 the best explanation of the morphological data. Hypothesis 3 renders as homoplasies the six morphological and behavioral synapomorphies of hippopotamids and cetaceans proposed by Gatesy et al. (1996). These include absence of hair, aquatic habitat, and underwater vocalization. These results suggest future avenues of investigation likely to impact the phylogenetic position of Cetacea. The triisodontine arctocyonid Andrewsarchus does not have the close relationship to Cetacea proposed by Prothero et al. (1988), but instead falls outside of a mesonychian-cetacean 478 VOL. 48 SYSTEMATIC BIOLOGY clade largely on the basis of dental evidence, as argued by O’Leary (1998a) . Andrewsarchus forms part of a paraphyletic grouping with Eoconodon, but expanded investigation of the triisodontine arctocyonids, a relatively primitive and poorly known group, is necessary before the paraphyly of this group should be recognized by taxonomic changes. In particular, knowledge of such features as the morphology of the distal phalanges of Hapalodectes and Eoconodon will be important for establishing their positions within ungulates. The result that the insectivoran Leptictis, an outgroup taxon, falls closer to our ingroup than does the primitive “condylarth” Arctocyon also warrants further testing and may be an artifact of including only one insectivoran. It is not clear that the hypocone in all members of Artiodactyla is derived from the metaconule, because both the hypocone and the metaconule are present together in the clade comprising Archaeotherium, “Elomeryx,” and Sus (Appendix 3). Phylogenetic analyses focusing on the relationships within Cetacea (Uhen, 1996) do not support the positions of Basilosaurus and Dorudon found in this analysis but instead place Dorudon closer to extant cetaceans. Because the aim of this study is to resolve more basal nodes in the ungulate tree, we suspect that the addition of more characters or more basilosaurid cetaceans to this matrix may result in the topology supported in Uhen (1996). Results based on the data set with certain multistate characters ordered (Fig. 2) require that cetaceans diverged by the middle Paleocene (Torrejonian), the time of appearance of the oldest mesonychian, Dissacus, a result also supported by other phylogenetic analyses (Geisler and Luo, 1998; O’Leary, 1998a). This tree topology species the existence of a cetacean ghost lineage ~ 10 million years long that remains to be discovered, something also suggested by the large number of synapomorphies uniting cetaceans (Table 2; Fig. 2: node e). This estimate of the time of origin of Cetacea is greater than likelihoodbased estimates (Gingerich and Uhen, 1998), which place the origin of Cetacea in the Eocene, much closer to the rst appearance of archaeocetes in the fossil record. As emphasized by the count of extinct and extant taxa relevant to the question of the position of Cetacea within Mammalia (Table 1), this analysis only begins to sample the diverse array of taxa likely to impact the position of Cetacea within Mammalia. Besides numerous extinct artiodactyls and perissodactyls, 12 other mesonychian genera and ve other triisodontine genera, known from fragmentary remains, are likely to affect the relationships presented here (McKenna and Bell, 1997). Furthermore, located either within the ungulate clade or as stem taxa to it are 65 other “condylarth” genera (McKenna and Bell, 1997), many of which are also poorly known and have not been scored here. Other critical taxa include the pakicetid Nalacetus, which has been argued to have the most primitive dental morphology of any known cetacean (O’Leary, 1998a; Thewissen and Hussain, 1998); Chriacus (Rose, 1996), which exhibits modications of the postcranial skeleton that suggest a link between arctocyonid “condylarths” and basal artiodactyls; Microclaenodon, a triisodontine arctocyonid, which is the rst outgroup to Mesonychia on the basis of dental evidence (O’Leary, 1998a), but which is known from little other material; and Sinonyx (Zhou et al., 1995), a mesonychian from the Paleocene of Asia, which is well-preserved postcranially but remains undescribed. Discovery and description of material of these key fossils will be crucial for testing the stability of the tree topologies presented here. ACKNOWLEDGMENTS We acknowledge the generosity of many colleagues who provided helpful discussion and access to specimens, including undescribed fossil material: P. D. Gingerich, R. E. Heinrich, E. Heizmann, L. T. Holbrook, S. G. Lucas, W. P. Luckett, D. Lunde, Z. Luo, R. D. E. MacPhee, P. J. Makovicky, M. C. McKenna, M. J. Novacek, K. D. Rose, A. E. Sanders, N. B. Simmons, and J. G. M. Thewissen. We are particularly grateful to J. G. M. Thewissen for unpublished data on Ambulocetus, to M. D. Uhen for assisting with the coding of Dorudon, to L. B. Nash for preparing Figure 1, and to C. Heezy who provided data on the postorbital bars of several 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA ungulates. M. J. Novacek, C. F. Ross, and D. M. Sabatini generously allowed us to conduct hours of phylogenetic analysis on their computers. R. J. Asher, J. Gatesy, G. J. P. Naylor, M. J. Novacek, C. F. Ross, N. B. Simmons, J. B. Slowinski, and W. C. Wheeler read and greatly improved earlier versions of the manuscript. 479 homologous sequences. Methods Enzymol. 183:601– 615. DAUDT , W. 1898. Beitrage zur Kenntnis des Urogenitalapparates der Cetaceen. Jena. Z. Naturwiss. 32:231– 312. DE J ONG, W. W. 1985. Superordinal afnities of Rodentia studied by sequence analysis of eye lens proteins. Pages 211–226 in Evolutionary relationships among rodents: A multidisciplinar y analysis REFERENCES (W. P. Luckett and J.-L. Hartenberger, eds.). Plenum, ADACHI, J., AND M. HASEGAWA . 1996. InstabilNew York. ity of quartet analyses of molecular sequence DONOGHUE, M. J., J. A. DOYLE , J. GAUTHIER , A. G. KLUGE , data by the maximum likelihood method: The AND T. ROWE. 1989. The importance of fossils in Cetacea/Artiodactyla relationships. Mol. Phylophylogeny reconstruction. Annu. Rev. Ecol. Syst. genet. Evol. 6:72–76. 20:431–460. ÁRNASON , Ú., A. GULLBERG , AND B. WIDEGREN . 1991. The DOYLE , J. A., AND M. J. DONOGHUE. 1987. The imporcomplete nucleotide sequence of the mitochondrial tance of fossils in elucidatin g seed plant phylogeny DNA of the n whale, Balaenoptera physalus. J. Mol. and macroevolution . Rev. Palaeobot. Palynol. 50:63– Evol. 33:556–568. 95. BARNES , L. G. 1984. Whales, dolphins and porpoises: EERNISSE , D. J., AND KLUGE , A. G. 1993. Taxonomic Origin and evolution of the Cetacea. Pages 139–154 congruence versus total evidence and amniote phyin Mammals. Notes for a short course, Studies in Gelogeny inferred from fossils, molecules and morpholology 8(1-4) (P. D. Gingerich and C. E. Badgely, eds.). ogy. Mol. Biol. Evol. 10:1170–1195. Department of Geological Sciences, Univ. Tennessee, FARRIS, J. S. 1983. The logical basis of phylogenetic Knoxville. analysis. Pages 1–36 in Advances in Cladistics, Vol. BARRIEL, V., AND P. TASSY . 1993. Characters, observa2 (N.I. Platnick and V.A. Funk, eds.). Columbia Unitions and steps: Comment on Lipscomb’s “Parsiversity Press. mony, homology and the analysis of multistate charFELSENSTEIN , J. 1978. Cases in which parsimony or acters.” Cladistics 9:223–232. compatibility methods will be positively misleading. BEINTEMA , J. J., W. M. FITCH , AND A. CARSANA . Syst. Zool. 27:401–410. 1986. Molecular evolution of pancreatic-type riFELSENSTEIN , J. 1985. Condence limits on phylogebonucleases. Mol. Biol. Evol. 3:262–275. nies: An approach using the bootstrap. Evolution BEINTEMA , J. J., C. SCHULLER , M. IRIE, AND A. CARSANA . 39:783–791. 1988. Molecular evolution of the ribonuclease superFRAAS, E. 1904. Neue Zeuglodonten aus dem unteren family. Prog. Biophys. Mol. Biol. 51:165–192. Mitteleoc än vom Mokattam bei Cairo. Geol. PalaeonBOYDEN , A., AND D. GEMEROY . 1950. The relative potol. Abh. 6:199–220. sition of Cetacea among the orders of Mammalia as GASKIN , D. E. 1978. Form and function in the digesindicated by precipitin tests. Zoologica 35:145–151. tive tract and associated organs in Cetacea, with a BREMER , J. L. 1904. On the lung of the opossum. Am. consideration of metabolic rates and specic energy J. Anat. 3:67–73. budgets. Oceanogr. Mar. Biol. Ann. Rev. 16:313–345. BREMER , K. 1988. The limits of amino acid sequence GATESY , J. 1997. More DNA support for a Cetacea/ data in angiosperm phylogenetic reconstruction. Hippopotamidae clade: The blood-clotting protein Evolution 42:795–803. gene gamma-brinogen. Mol. Biol. Evol. 14:537–543. CARROLL , R. L. 1988. Vertebrate paleontology and G ATESY , J., C. HAYASHI , M. A. CRONIN , AND P. evolution. W. H. Freeman, New York. ARCTANDER . 1996. Evidence from milk casein genes CATZEFLIS , F. 1993. Mammalian phylogeny: Morpholthat cetaceans are close relatives of hippopotamid arogy and molecules. Trends Ecol. Evol. 8:340–341. tiodactyls. Mol. Biol. Evol. 13:954–963. CHOW , M. 1959. A new arctocyonid from the upper G AUTHIER , J., A. G. KLUGE , AND T. ROWE . 1988. AmEocene of Lushih, Honan. Vertebr. Palasiat. 3:133– niote phylogeny and the importance of fossils. 138. Cladistics 4:105–209. CIFELLI , R. L. 1982. The petrosal structure of Hyopsodus GAZIN , C. L. 1965. A study of the early Tertiary condywith respect to that of some other ungulates, and its larthran mammal Meniscotherium. Smithson. Misc. phylogenetic implications . J. Paleontol. 56:795–805. Collect. 149:1–98. CZELUSNIAK , J., M. GOODMAN , B. F. KOOP, D. O. TAGLE , J. SHOSHANI, G. BRAUNITZER , T. K. KLEINSCHIMDT , W. GAZIN , C. L. 1968. A study of the Eocene condylarthran mammal Hyopsodus. Smithson. Misc. ColW. DE J ONG, AND G. MATSUDA . 1990a. Perspectives lect. 153:1–90. from amino acid and nucleotide sequences on cladisGEISLER , J. H., AND Z. LUO. 1996. The petrosal and intic relationships among higher taxa of Eutheria. ner ear of Herpetocetus sp. (Mammalia, Cetacea) and Pages 545–572 in Current mammalogy, volume 2 (H. their implications for the phylogeny and hearing of H. Genoways, ed.). Plenum, New York. archaic myticetes. J. Paleont. 70:1045–1066. CZELUSNIAK , J., M. GOODMAN , N. D. MONCRIEF, AND S. GEISLER , J. H., AND Z. LUO . 1998. Relationships of M. KEHOE. 1990b. Maximum parsimony approach to construction of evolutionary trees from aligned Cetacea to terrestrial ungulates and the evolution 480 SYSTEMATIC BIOLOGY of cranial vasculature in Cete. Pages 163–212 in The emergence of whales (J. G. M. Thewissen, ed.). Plenum, New York. GEISLER , J. H., AND M. A. O’LEARY . 1997. A phylogeny of Cetacea, Artiodactyla, Perissodactyla, and archiac ungulates: The morphological evidence. J. Vertebr. Paleontol. 17-(Suppl. to No. 3):48A. GEISLER , J. H., A. E. SANDERS , AND Z. LUO. 1996. A new protocetid cetacean from the Eocene of South Carolina, U.S.A.: Phylogenetic and biogeographic implications . Paleontol. Soc. Spec. Publ. 8:139. GENTRY, A. W., AND J. J. HOOKER. 1988. The phylogeny of the Artiodactyla. Pages 235–272 in The phylogeny and classication of the tetrapods, volume 2. Mammals Syst. Assoc. Spec. Vol. 35B (M. J. Benton, ed.). Clarendon Press, Oxford, England. GINGERICH, P. D., M. ARIF, AND W. C. CLYDE . 1995. New archaeocetes (Mammalia, Cetacea) from the middle Eocene Domanda Formation of the Sulaiman Range; Punjab (Pakistan). Contrib. Mus. Paleontol. Univ. Mich. 29:291–230. GINGERICH, P. D., AND D. E. RUSSELL . 1981. Pakicetus inachus, a new archaeocete (Mammalia, Cetacea) from the early-middle Eocene Kuldana Formation of Kohat (Pakistan). Contrib. Mus. Paleontol. Univ. Mich. 25:235–246. GINGERICH, P. D., AND D. E. RUSSELL . 1990. Dentition of the early Eocene Pakicetus (Mammalia, Cetacea). Contrib. Mus. Palentol. Univ. Mich. 28:1–20. GINGERICH, P. D., B. H. SMITH , AND E. L. SIMONS. 1990. Hind limbs of Eocene Basilosaurus: Evidence of feet in whales. Science 249:154–157. GINGERICH, P. D., AND M. D. UHEN . 1998. Likelihood estimation of the time of divergence of Cetacea and Artiodactyla. Paleontol. Electron. 2:1–47. GOODMAN , M., J. CZELUSNIAK , AND J. E. BEEBER . 1985. Phylogeny of primates and other eutherian orders: A cladistic analysis using amino acid and nucleotide sequence data. Cladistics 1:171–185. GRAUR , D. 1993a. Molecular phylogeny and the higher classication of eutherian mammals. Trends Ecol. Evol. 8:141–147. GRAUR , D. 1993b. Reply from D. Graur. Trends Ecol. Evol. 8:341–342. GRAUR , D., AND D. G. HIGGINS. 1994. Molecular evidence for the inclusion of cetaceans within the order Artiodactyla. Mol. Biol. Evol. 11:357–364. GRAYBEAL , A. 1998. Is it better to add taxa or characters to a difcult phylogenetic problem? Syst. Biol. 47:9–17. GUTHRIE , D. A. 1967. The mammalian fauna of the Lysite Member, Wind River Formation (early Eocene), of Wyoming. Mem. South. Calif. Acad. Sci. 5:1–53. HASEGAWA , M., AND J. ADACHI . 1996. Phylogenetic position of cetaceans relative to artiodactyls: Reanalysis of mitochondrial and nuclear sequences. Mol. Biol. Evol. 13:710–717. HAUSER , D. L., AND W. PRESCH . 1991. The effect of ordered characters on phylogenetic reconstruction. Cladistics 7:243–265. VOL. 48 HILLIS , D. M. 1998. Taxonomic sampling, phylogenetic accuracy, and investigator bias. Syst. Biol. 47:3–8. HONEYCUTT , R. L., M. A. NEDBAL , R. M. ADKINS, AND L. L. J ANECEK . 1995. Mammalian mitochondrial DNA evolution: A comparison of the cytochrome b and cytochrome c oxidase II genes. J. Mol. Evol. 40:260–272. HUELSENBECK , J. P. 1991. When are fossils better than extant taxa in phylogenetic analysis? Syst. Biol. 40:458–469. HUNTER, J. P., AND J. JERNVALL . 1995. The hypocone as a key innovation in mammalian evolution. Proc. Natl. Acad. Sci. USA. 92:10718–10722. IRWIN, D. M., AND Ú. ÁRNASON . 1994. Cytochrome b gene of marine mammals: Phylogeny and evolution. J. Mammal. Evol. 2:37–55. IRWIN, D. M., T. D. KOCHER, AND A. C. WILSON . 1991. Evolution of the cytochrome b gene of mammals. J. Mol. Evol. 32:128–144. KELLOGG, A. R. 1936. A review of the Archaeoceti. Carnegie Inst. Washington Publ. 482:1–366. KIELAN -J AWOROWSKA, Z. 1977. Evolution of the therian mammals in the Late Cretaceous of Asia. Part II. Postcranial skeleton in Kennalestes and Asioryctes. Palaeontol. Pol. 37:65–83. KIELAN -J AWOROWSKA, Z. 1981. Evolution of the therian mammals in the Late Cretaceous of Asia. Part. IV. Skull structure in Kennalestes and Asioryctes. Palaeontol. Pol. 42:25–78. KIM , J. 1996. General inconsistency conditions for maximum parsimony: Effects of branch lengths and increasing numbers of taxa. Syst. Biol. 45:363–374. KITTS, D. B. 1956. American Hyracotherium (Perissodactyla, Equidae). Bull. Am. Mus. Nat. Hist. 110:1– 60. KLUGE, A. G. 1989. A concern for evidence and a phylogenetic hypothesis of relationships among Epicrates (Boidae, Serpentes). Syst. Zool. 38:7–25. KLUGE, A. G., AND A. J. WOLF. 1993. Cladistics: What’s in a word? Cladistics 9:183–199. KUMAR, S., AND B. HEDGES . 1998. A molecular timescale for vertebrate evolution. Nature 392:917– 920. KUMAR, K., AND A. SAHNI . 1986. Remingtonocetus harudiensis, new combination, a middle Eocene archaeocete (Mammalia, Cetacea) from Western Kutch, India. J. Vertebr. Paleontol. 6:326–349. LANGER , P. 1988. The mammalian herbivore stomach: Comparative anatomy, function and evolution. Fischer, Stuttgart. LI, W.-H. 1997. Molecular evolution. Sinauer, Sunderland, Massachusetts. LUCKETT , W. P., AND N. HONG. 1998. Phylogenetic relationships between the orders Artiodactyla and Cetacea: A combined assessment of morphological and molecular evidence. J. Mammal. Evol. 5:127–182. LUO , Z., AND K. MARSH . 1996. Petrosal (periotic) and inner ear of a Pliocene kogiine whale (Kogiinae, Odontoceti): Implications on realionships and hearing evolution of toothed whales. J. Vertebr. Paleontol. 16:328–248. MACFADDEN , B. J. 1992. Fossil horses. Cambridge Univ. Press, New York. 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA MACLEOD , N., AND K. D. ROSE. 1993. Inferring locomoter behavior in Paleogene mammals via eigenshape analysis. Am. J. Sci. 293-A:300–355. MACPHEE, R. D. E. 1981. Auditory regions of primates and eutherian insectivores: Morphology, ontogeny, and charcater analysis. Contrib. Primatol. 18:1–282. MACPHEE, R. D. E. 1994. Morphology, adaptations , and relationships of Plesiorycteropus, and a diagnosis of a new order of eutherian mammals. Bull. Am. Mus. Nat. Hist. 220:1–214. MADDISON , W. P., AND D. R. MADDISON . 1992. MacClade: Analysis of phylogeny and character evolution. Version 3. Sinauer, Sunderland, Massachusetts. MATTHEW , W. D. 1897. A revision of the Puerco Fauna. Bull. Am. Mus. Nat. Hist. 9:59–110. MATTHEW , W. D. 1929. Reclassication of the artiodactyl families. Bull. Geol. Soc. Am. 40:403–408. MATTHEW , W. D. 1937. Paleocene faunas of the San Juan Basin, New Mexico. Trans. Am. Philos. Soc. 30:1– 510. MCCRADY, E., JR. 1940. The development and fate of the urogenital sinus in the opossum, Didelphy virginiana. J. Morphol. 66:131–154. MCKENNA , M. C. 1975. Toward a phylogenetic classication of the Mammalia. Pages 21–46 in Phylogeny of the primates (W. P. Luckett and F. S. Szalay, eds.). Plenum, New York. MCKENNA , M. C. 1992. The alpha crystallin A chain of the eye lens and mammalian phylogeny. Ann. Zool. Fenn. 28:349–360. MCKENNA , M. C., AND S. K. BELL. 1997. Classication of mammals above the species level. Columbia Univ. Press, New York. MILINKOVITCH , M. C., G. ORTI, AND A. MEYER . 1993. Revised phylogeny of whales suggested by mitochondrial ribosomal DNA sequences. Nature 361:346–348. MILINKOVITCH , M. C., AND J. G. M. THEWISSEN . 1997. Even-toed ngerprints on whale ancestry. Nature 388:622–624. MIYAMOTO , M. M., AND M. GOODMAN . 1986. Biomolecular systematics of eutherian mammals: Phylogenetic patterns and classication. Syst. Zool. 35:230– 240. MONTGELARD , C., F. M. CATZEFLIS , AND E. DOUZERY . 1997. Phylogenetic relationships of artiodactyls and cetaceans as deduced from the comparison of cytochrome b and 12S RNA mitochondrial sequences. Mol. Biol. Evol. 14:550–559. MONTGELARD , C., S. DUCROQ, AND E. DOUZERY. 1998. What is a suiforme (Artiodactyla)? Mol. Phylog. Evol. 9:528–532. MULLER , O. 1898. Untersuchungen uber die Veranderungen, welche die Respirationsorgane der Saugetiere durch die anpassung an das Leben im Wasser erlitten haben. Jena. Z. Naturwiss. 32:95–230. NIXON , K. C., AND J. M. CARPENTER . 1996. On simultaneous analysis. Cladistics 12:221–241. NOVACEK , M. J. 1977. Aspects of the problem of variation, origin and evolution of the eutherian bulla. Mammal Rev. 7:131–149. NOVACEK , M. J. 1980. Cranioskeletal features in tupaiids and selected Eutheria as phylogenetic evidence. 481 Pages 35–93 in Comparative biology and evolutionary relationships of the tree shrews (W. P. Luckett, ed.). Plenum, New York. NOVACEK , M. J. 1982. Information for molecular studies from anatomical and fossil evidence on higher eutherian phylogeny. Pages 3–41 in Macromolecular sequences in systematic and evolutionary biology (M. Goodman, ed.). Plenum, New York. NOVACEK , M. J. 1986. The skull of leptictid insectivorans and the higher-level classication of eutherian mammals. Bull. Am. Mus. Nat. Hist. 183:1–111. NOVACEK , M. J. 1989. Higher mammal phylogeny: The morphological/molecular synthesis. Pages 421–435 in The hierarchy of life, Nobel Symposium 70 (R. Fernholm, K. Bremer, and H. Jornvall, eds.). Elsevier Science, Amsterdam. NOVACEK , M. J. 1991. “All tree histograms” and the evaluation of cladistic evidence: Some ambiguities. Cladistics 7:345–349. NOVACEK , M. J. 1992a. Fossils, topologies, missing data, and the higher-level phylogeny of eutherian mammals. Syst. Biol. 41:58–73. NOVACEK , M. J. 1992b. Mammalian phylogeny: Shaking the tree. Nature 356:121–125. NOVACEK , M. J. 1993. Mammalian phylogeny: Morphology and molecules. Trends Ecol. Evol. 8:339–340. NOVACEK , M. J. 1994. Morphological and molecular inroads to phylogeny. Pages 85–130 in Interpreting the hierarchy of nature (L. Grande and O. Rieppel, eds.). Academic Press, San Diego. NOVACEK , M. J., K. GAO , M. A. NORRELL , AND G. ROUGIER . 1998. Ghost lineages, phylogeny, and ranges of selected vertebrate lineages across the K/T boundary. J. Vertebr. Paleontol. 18(Suppl. to No. 3):67A. NOVACEK , M. J., AND Q. D. WHEELER . 1992. Extinct taxa: Accounting for 99.999...% of the earth’s biota. Pages 1–16 in Extinction and phylogeny (M. J. Novacek and Q. D. Wheeler, eds.). Columbia Univ. Press, New York. NOVACEK , M. J., AND A. R. WYSS. 1986. Higher-level relationships of the eutherian orders: Morphological evidence. Cladistics 2:257–287. O’LEARY, M. A. 1998a. Phylogenetic and morphometric reassessment of the dental evidence for a mesonychian and cetacean clade. Pages 133–161 in The emergence of whales (J. G. M. Thewissen, ed.). Plenum, New York. O’LEARY, M. A. 1998b. Morphology of the humerus of Hapalodectes (Mammalia, Mesonychia). Am. Mus. Novit. 3242:1–6. O’LEARY, M. A., AND K. D. ROSE . 1995a. New mesonychian dentitions from the Paleocene and Eocene of the Bighorn Basin, Wyoming. Ann. Carnegie Mus. 64:147–172. O’LEARY, M. A., AND K. D. ROSE . 1995b. Postcranial skeleton of the early Eocene mesonychid Pachyaena (Mammalia: Mesonychia). J. Vertebr. Paleontol. 15:401–430. PETERSON , O. A. 1931. New mesonychids from the Uinta. Ann. Carnegie Mus. 20:333–339. PHILIPPE, H., AND E. DOUZERY . 1994. The pitfalls of molecular phylogeny based on four species, as illus- 482 SYSTEMATIC BIOLOGY trated by the Cetacea/Artiodactyla relationships. J. Mammal. Evol. 2:133–152. PICKFORD, M. 1983. On the orgins of Hippopotamidae together with descriptions of two new species, a new genus and a new subfamily from the Miocene of Kenya. Geobios 16:193–217. PROTHERO , D. R. 1993. Ungulate phylogeny: Molecular vs. morphological evidence. Pages 173–181 in Mammal phylogeny, volume 2. Placentals (F. S. Szalay and M. J. Novacek, eds.). Springer-Verlag, New York. PROTHERO , D. R., E. M. MANNING, AND M. FISCHER . 1988. The phylogeny of the ungulates. Pages 201–234 in The phylogeny and classication of the tetrapods, volume 2. Mammals. Syst. Assoc. Spec. Vol. 35B (M. J. Benton, ed.). Clarendon, Oxford, England. QUERALT , R., R. ADROER, R. OLIVA , R. J. WINKFEIN, J. D. RETIEF , AND G. H. DIXON . 1995. Evolution of protamine P1 genes in mammals. J. Mol. Evol. 40:601– 607. RADINSKY, L. B. 1965. Evolution of the tapiroid skeleton from Heptodon to Tapirus. Bull. Mus. Comp. Zool. 34:1–106. ROSE, K. D. 1982. Skeleton of Diacodexis, oldest known artiodactyl. Science 216:621–623. ROSE, K. D. 1985. Comparative osteology of North American dichobunid artiodactyls. J. Paleontol. 59:1203–1226. ROSE, K. D. 1987. Climbing adaptation s in the early Eocene mammal Chriacus and the origin of Artiodactyla. Science 236:314–316. ROSE, K. 1990. Postcranial skeletal remains and adaptations in early Eocene mammals from the Willwood Formation, Bighorn Basin, Wyoming. Geol. Soc. Am. Spec. Pap. 243:107–133. ROSE, K. D. 1996. On the origin of Artiodactyla. Proc. Natl. Acad. Sci. USA 93:1705–1709. ROSE, K. D., AND M. A. O’LEARY. 1995. The manus of Pachyaena gigantea (Mammalia: Mesonychia). J. Vertebr. Paleontol. 15:855–859. RUSSELL , D. E. 1964. Les mammiféres Paleocenes d’Europe. Mem. Mus. Natl. Hist. Nat. Ser. C (Paris) 13:1–324. RUSSELL , D. E., J. G. M. THEWISSEN , AND D. SIGOGNEAU RUSSELL . 1983. A new dichobunid artiodactyl (Mammalia) from the Eocene of North-West Pakistan, Part II. Cranial osteology. Proc. K. Ned. Akad. Wet., Ser. B 86:285–299. SANDERSON , M. J., AND M. J. DONOGHUE . 1989. Patterns of variation in levels of homoplasy. Evolution 43:1781–1795. SCHAEFFER , B. 1947. The origin of a mammalian ordinal character. Evolution 2:164–175. SCOTT, W. B. 1888. On some new and little known creodonts. J. Acad. Nat. Sci. Phila. 9:155–185. SCOTT, W. B. 1894. The structure and relationships of Ancodus. J. Acad. Nat. Sci. Phila. 9:461–497. SHIMAMURA , M., H. YASUA , K. OHSHMIA, H. ABE, H. KATO, T. KISHIRO , M. GOTOS, I. MUNECHIKAI , AND N. OKADA. 1997. Molecular evidence from retroposons that whales form a clade within even-toed ungulates. Nature 388:666–670. VOL. 48 SHOSHANI, J. 1986. Mammalian phylogeny: Comparison of morphological and molecular results. Mol. Biol. Evol. 3:222–242. SHOSHANI, J., M. GOODMAN , J. CZELUSNIAK , AND G. BRAUNITZER . 1985. A phylogeny of Rodentia and other eutherian orders: Parsimony analysis utilizing amino acid sequences of alpha and beta hemoglobin chains. Pages 191–210 in Evolutionary relationships among rodents: A multidisciplinar y approach (W. P. Luckett and J.-L. Hartenberger, eds.). Plenum, New York. SIMMONS , N. B., AND J. H. GEISLER . 1998. Phylogenetic relationships of Icaronycteris, Archaeonycteris, Hassianycteris, and Palaeochiropteryx to extant bat lineages, with comments on the evolution of echolocation and foraging strategies in Microchiroptera. Bull. Am. Mus. Nat. Hist. 235:1–182. SIMPSON , G. G. 1945. The principles of classication and a classication of mammals. Bull. Am. Mus. Nat. Hist. 85:1–350. SISSON , S., AND J. D. GROSSMAN . 1953. The anatomy of the domestic animals. W. B. Saunders, Philadelphia. SLIJPER, E. J. 1936. Die Cetaceen VergleichendAnatomische und Systematisch. Capita Zool. 7:1– 590. SLOWINSKI , J. B. 1993. “Unordered” versus “ordered” characters. Syst. Biol. 42:155–165. SMITH , M. R., M. S. SHIVJI, V. G. WADDELL , AND M. J. STANHOPE . 1996. Phylogenetic evidence from the IRBP gene for the paraphyly of toothed whales, with mixed support for Cetacea as a suborder of Artiodactyla. Mol. Biol. Evol. 13:918–922. SMUTS , M. M. S. AND A. J. BEZUIDENHOUT . 1987. Anatomy of the dromedary. Clarendon Press, Oxford. SOKOLOV, V. E. 1982. Mammal skin. Univ. California Press, Berkeley. STANHOPE , M. J., W. J. BAILEY, J. CZELUSNIAK , M. GOODMAN , J.-S. SI, J. NICKERSON , J. G. SGOUROS, G. A. M. SINGER, AND T. K. KLEINSCHMIDT . 1993. A molecular view of primate supraordinal relationships from both nucleotide and amino acid sequences. Pages 251–292 in Primates and their relatives in phylogenetic perspective (R. D. E. MacPhee, ed.). Plenum, New York. STANHOPE , M. J., M. R. SMITH , V. G. WADDELL , C. A. PORTER, M. S. SHIVJI, AND M. GOODMAN . 1996. Mammalian evolution and the interphotoreceptor retinoid binding protein (IRPB) gene: Convincing evidence for several superordinal clades. J. Mol. Evol. 43:83– 92. SWOFFORD, D. L. 1993. PAUP: Phylogenetic analysis using parsimony, version 3.1.1. Illinois Natural History Survey, Champaign. SWOFFORD, D. L. 1998. PAUP* 4.0b1: Phylogenetic Analysis Using Parsimony (* and Other Methods). Sinauer Associates, Sunderland, Massachusetts. SZALAY , F. S. 1969a. The Hapalodectinae and a phylogeny of the Mesonychidae (Mammalia, Condylarthra). Am. Mus. Novit. 2361:1–26. SZALAY , F. S. 1969b. Origin and evolution of function of the mesonychid condylarth feeding mechanism. Evolution 23:703–720. 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA SZALAY, F. S., AND S. J. GOULD . 1966. Asiatic Mesonychidae (Mammalia, Condylarthra). Bull. Am. Mus. Nat. Hist. 132:128–173. THEWISSEN , J. G. M. 1990. Evolution of Paleocene and Eocene Phenacodontidae (Mammalia, Condylarthra). Univ. Mich. Pap. Paleontol. 29:1–107. THEWISSEN , J. G. M. 1994. Phylogenetic aspects of cetacean origins: A morphological perspective. J. Mammal. Evol. 2:157–184. THEWISSEN , J. G. M., AND D. P. DOMNING . 1992. The role of phenacodontids in the origin of the modern orders of ungulate mammals. J. Vertebr. Paleontol. 12:494– 504. THEWISSEN , J. G. M., AND S. T. HUSSAIN . 1990. Postcranial osteology of the most primitive artiodactyl Diacodexis pakistanensis (Dichobunidae). Anat. Histol. Embryol. 19:37–48. THEWISSEN , J. G. M., AND S. T. HUSSAIN . 1993. Origin of underwater hearing in whales. Nature 361:444–445. THEWISSEN , J. G. M., AND S. T. HUSSAIN . 1998. Systematic review of the Pakicetidae, early and middle Eocene Cetacea (Mammalia) from Pakistan and India. Ann. Carnegie. Mus. 34:220–238. THEWISSEN , J. G. M., S. T. HUSSAIN , AND M. ARIF. 1994. Fossil evidence for the origin of aquatic locomotion in archaeocete whales. Science 263:210–212. THEWISSEN , J. G. M., S. I. MADAR, AND S. T. HUSSAIN . 1996. Ambulocetus natans, an Eocene cetacean (Mammalia) from Pakistan. Cour. Forschungsinst. Senckenb. 191:1–86. THEWISSEN , J. G. M., S. I. MADAR, AND S. T. HUSSAIN . 1998. Whale ankles and evolutionary relationships. Nature. 395:452. THEWISSEN , J. G. M., D. E. RUSSELL , P. D. GINGERICH, AND S. T. HUSSAIN . 1983. A new dichobunid artiodactyl (Mammalia) from the Eocene of North-West Pakistan, dentition and classication. Proc. K. Ned. Akad. Wet. Ser. B 86:153–180. TING, S., AND C. LI. 1987. The skull of Hapalodectes (Acreodi, Mammalia), with notes on some Chinese Paleocene mesonychids. Vertebr. Palasiat. 25:161– 186. UHEN , M. D. 1996. Dorudon atrox (Mammalia, Cetacea): Form, function, and phylogenetic relationships of an archaeocete from the late middle Eocene of Egypt. Ph.D. thesis, Univ. Michigan, Ann Arbor. UHEN , M. D. 1998. Middle to late Eocene basilosaurines and dorudontines. Pages 29–61 in The emergence of whales (J. G. M. Thewissen, ed.). Plenum, New York. VAN VALEN , L. 1966. The Deltatheridia, a new order of mammals. Bull. Am. Mus. Nat. Hist. 132:1–126. VAN VALEN , L. 1968. Monophyly or diphyly in the origin of whales. Evolution 23:37–41. VAN VALEN , L. 1969. The multiple origins of the placental carnivores. Evolution 23:118–130. VAN VALEN , L. 1978. The beginning of the age of mammals. Evol. Theory 4:45–80. VAUGHAN , T. A. 1986. Mammalogy. Harcourt, Brace, Jovanovich, Fort Worth, Texas. WEBB , S. B., AND B. E. TAYLOR. 1980. Phylogeny of hornless ruminants and a description of the cranium of Archaeomeryx. Bull. Am. Mus. Nat. Hist. 167:1–157. 483 WEST , R. M. 1981. Geology and paleontology of the Bridger Formation, Southern Green River Basin, Southwestern Wyoming. Part 5. Harpagolestes macrocephalus and comments on structure, function and diversity of middle Eocene to early Oligocene large mesonychids. Contrib. Biol. Geol. Milwauk. Publ. Mus. 43:1–17. WHEELER , W. C. 1992. Extinction, sampling and molecular phylogenetics. Pages 205–215 in Extinction and phylogeny (M. J. Novacek and Q. D. Wheeler, eds.). Columbia Univ. Press, New York. WIBLE, J. R. 1987. The eutherian stapedial artery: Character analysis and implication s for superordinal relationships. Zool. J. Linn. Soc. 91:107–135. WIBLE, J. R. 1990. Petrosals of Late Cretaceous marsupials from North America, and a cladistic analysis of the petrosal in therian mammals. J. Vertebr. Paleontol. 10:183–205. WILEY, E. O., D. SIEGEL-CAUSEY , D. R. BROOKS, AND V.A. FUNK. 1991. The Compleat Cladist. Univ. Kansas Mus. Nat. Hist. Spec. Publ. 19:1–158. WILKINSON , M. 1992. Ordered versus unordered characters. Cladistics 8:375–385. WILLIAMSON , T. E., AND S. G. LUCAS . 1992. Meniscotherium (Mammalia, “Condylarthra”) from the Paleocene/Eocene of western North America. N. M. Mus. Nat. Hist. Sci., Bull. 1:1–75. WORTMAN , J. L. 1901. Studies of Eocene Mammalia in the Marsh Collection, Peabody Museum. Am. J. Sci. 11:1–90. ZHOU, X., AND P. D. GINGERICH. 1991. New species of Hapalodectes (Mammalia, Mesonychia) from the early Wasatchian, early Eocene of northwestern Wyoming. Contrib. Mus. Paleontol. Univ. Mich. 28:215–220. ZHOU, X., W. J. SANDERS , AND P. D. GINGERICH. 1992. Functional and behavioral implication s of verterbral structure in Pachyaena ossifraga (Mammalia, Mesonychia). Contrib. Mus. Paleontol. Univ. Mich. 28:289–319. ZHOU, X., R. ZHAI , P. D. GINGERICH, AND L. CHEN . 1995. Skull of a new mesonychid (Mammalia, Mesonychia) from the late Paleocene of China. J. Vertebr. Paleontol. 15:387–400. Received 29 April 1998; accepted 14 August 1998 Associate Editor: G. Naylor APPENDIX 1. SPECIMENS AND LITERATURE REFERENCES USED TO C OMPILE THE C HARACTER -T AXON MATRIX FOR PARSIMONY ANALYSIS Data on the deciduous fourth premolar throughout matrix are from Luckett and Hong (1998). AMNH-M = Department of Mammalogy, AMNH-VP, Department of Vertebrate Paleontology, American Museum of Natural History, New York; ChM PV = Charleston Museum, Charleston, South Carolina; GSP-UM = Geological Survey of Pakistan/University of Michigan, Ann Arbor; H-GSP = Howard University/Geological Survey of Pakistan, Washington, D.C.; IVPP = Institute of Vertebrate Paleontology and Paleoanthropology, Bei- 484 VOL. 48 SYSTEMATIC BIOLOGY jing, China; MCZ = Museum of Comparative Zoology, Harvard University, Cambridge, Massachusetts; NMMNH = New Mexico Museum of Natural History, Albuquerque; USGS = Johns Hopkins University/United States Geological Survey (now at the Smithsonian Institution, Washington, D.C.); USNM = National Museum of Natural History, Smithsonian Institution, Washington, D.C.; YPM = Yale Peabody Museum, New Haven, Connecticut; YPM-PU, Princeton University collection (now at Yale Peabody Museum). Archaic Ungulates Andrewsarchus.—AMNH-VP 20135; Chow (1959). Arctocyon.—AMNH-VP 55900 (cast) , 55901 (cast), 55902; Russell (1964). Eoconodon.—AMNH-VP 764, 774, 3177, 3181, 3187, 3280, 4052, 16329, 16341; Matthew (1897, 1937). Hyopsodus.—AMNH-VP 1, 39, 1717, 10977, 10979, 11330, 11349, 11350, 11363, 11393, 11415, 11899; Gazin (1968); Thewissen and Domning (1992). Meniscotherium.—AMNH-VP 2560, 4414, 4426, 4434, 4447, 48083, 48126, 48129, 48555; Gazin (1965); Cifelli (1982); Williamson and Lucas (1992). Phenacodus.—AMNH-VP, 2961, 4370, 4378, 4403, 15262, 15266, 15268, 15271, 15275, 15279, 15284, 15286, 117195; Thewissen (1990); O’Leary and Rose (1995b). Artiodactyla Agriochoerus.—AMNH-VP 685, 1349, 1355, 7407, 7409, 9808, 9811, 38843, 38932, 95324, 95332, 99275. Archaeotherium.—AMNH-VP 39127, 39455. Camelus.—AMNH-M 2911, 14109, 35463, 35563, 69405, 80227, 90433; Smuts and Bezuidenhout (1987); Langer (1988 [based on condition in Llama]). Diacodexis.—AMNH-VP 16141; USGS 2352; Rose (1982, 1985); Russell et al. (1983); Thewissen et al. (1983); Thewissen and Hussain (1990); Thewissen (1994). “Elomeryx”-(includes specimens attributed to Bothriodon).—AMNH-VP 579, 582, 583, 1242, 1245, 1259, 1483, 10041, 12461, 39015, 101668; Scott (1894). Hexaprotodon.—AMNH-M 2423, 54265, 81899, 89626, 146848, 146849, 148452, 185383, 202423; Vaughan (1986); Langer (1988); Sokolov (1982). Hippopotamus.—AMNH-M 15898, 24282, 24284, 24285, 24289, 53773, 54248, 80813, 81856, 130247, 176118; Vaughan (1986); Langer (1988); Sokolov (1982). Ovis.—AMNH-M 6231, 6239, 10074, 14515, 15584, 35520, 53598, 88702, 100012, 100072, 146547; Sisson and Grossman (1953); Langer (1988); Sokolov (1982). Poebrotherium.—AMNH-VP 8955, 39085, 42240, 42248, 42249, 42261, 42272, 42276, 42281, 42284, 42292, 47003, 47008, 47052, 47103, 47182, 47284, 47317, 47324, 47333, 47077, 47907, 63701, 63704, 63712, 63713, 97103. Sus.—AMNH-M 20871, 69422, 69442, 100260, 235190, 235192, 236144, 238325, 238331; Sisson and Grossman (1953); Langer (1988); Sokolov (1982). Tragulus.—AMNH-M 10101, 14137, 14139, 34252, 32645, 53602, 53609, 60759, 90101, 90193, 102091, 102176, 106552, 240913; Langer (1988); Sokolov (1982). Balaenoptera.—AMNH-M 28274, 84870, 148407, 219212, 219220; Daudt (1898); Muller (1898); Sokolov (1982); Gaskin (1978). Basilosaurus.—AMNH-VP 14381, 129577 (cast); GSPUM 97507; Kellogg (1936). Cross Whale.—ChM PV 5401; Geisler et al. (1996). Dorudon.—Uhen (1996, 1998). Pakicetus.—GSP-UM 084 (cast); H-GSP 96231, 96386, 96431, 18410, 18570, 96505; Gingerich and Russell (1981, 1990); Thewissen and Hussain (1993); Thewissen (1994); Thewissen and Hussain (1998). Protocetus.—Fraas (1904). Remingtonocetus.—GSP-UM 3009, 3054, 3057; Kumar and Sahni (1986); Gingerich et al. (1995). Tursiops.—AMNH-M 120920, 184930, 212554; Daudt (1898 [based on condition in other odontocetes]); Muller (1898 [based on condition in other odontocetes]); Langer (1988). Mesonychidae Ankalagon.—AMNH-VP 776, 777, 2454; NMMNH 16309. Dissacus.—AMNH-VP 3356, 3357, 3359, 3360, 3361, 15996, 39276, 55925 (cast); IVPP 4266 (cast); YPM-PU 16159; O’Leary and Rose (1995a, 1995b). Harpagolestes.—AMNH-VP 1878, 1892, 1945, 2308, 26267, 26300, 26301; USNM 14708, 14915; Wortman (1901); Peterson (1931); West (1981); Zhou et al. (1995). Mongolestes.—AMNH-VP 26064, 26065. Mongolonyx.—AMNH-VP 26661, 26662. Mesonyx.—AMNH-VP 1716, 5021, 11552, 12641, 12643 (cast), 19204, 122122; Scott (1888); O’Leary and Rose (1995b). Pachyaena.—AMNH-VP 72, 75, 1522, 2823, 2959, 4262, 15224, 15228, 15728, 15730, 16154; USGS 7185; YPM 50000; YPM-PU 14708; O’Leary and Rose (1995a, 1995b); Rose and O’Leary (1995). Sinonyx.—VPP V10760 (skull and lower dentition); Zhou et al. (1995). Synoplotherium.—AMNH-VP 19203, 122122; Wortman (1901). Hapalodectidae Hapalodectes.—AMNH-VP 78, 12781, 14748 (cast), 128561, 17558, 20172, 80802 (cast); IVPP 5254 (cast); USGS 275 (cast), 9628, 10293 (cast); Guthrie (1967); Szalay (1969a); Ting and Li (1987); Zhou and Gingerich (1991); O’Leary (1998b). Perissodactyla Equus.—AMNH-M 1244C, KMM 99; AMNH-VP 113 (VP), LA 118; Sisson and Grossman (1953); Vaughan (1986); Langer (1988). Heptodon.—AMNH-VP 294, 4855, 4858, 14865, 14884, 16861, 141881; Radinsky (1965). Hyracotherium.—AMNH-VP 55267, 55268, 55269, 96274, 96277, 96284, 96298, 129209; Kitts (1956); MacFadden (1992); Rose (1990); Thewissen and Domning (1992). Cetacea Ambulocetus.—H-GSP 18507; Thewissen et al. (1996); Thewissen pers. comm. (pelvis data). Outgroups Asioryctes.—Kielan-Jaworowska (1977, 1981). 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA Didelphis.—AMNH-M 7310, 14335, 70089, 180358, 217701, 219212, 219220, 240517, 242660; Bremer (1904); McCrady (1940); Vaughan (1986); Langer (1988). Leptictis.—AMNH-VP 1413a, 5346, 38920, 39444, 90256, 96766; MCZ 19678; USNM 336367; Novacek (1980, 1986). APPENDIX 2. CHARACTERS AND CHARACTER STATES FOR PARSIMONY ANALYSIS Multistate characters treated as ordered are specied below. The character-taxon matrix is presented in Appendix 3. Basicranial Characters 1. Subarcuate fossa.—Present (0); absent (1) (Novacek, 1986). 2. Shape of tegmen tympani (ordered).—Uninated, forms thin lamina lateral to facial nerve canal (0); inated, forms barrel-shaped ossication lateral to the facial nerve canal (1); hyperinated, transverse width greater than or equal to width of promontorium (2) (modied from Cifelli, 1982; Geisler and Luo, 1998). 3. Anterior process of petrosal.—Absent (0); present, anterior edge of tegmen tympani extending far anterior to edge of pars cochlearis (1) (Luo and Marsh, 1996; Geisler and Luo, 1998). 4. Fossa for tensor tympani muscle.—Shallow, anteroposteriorly elongate fossa (0); circular pit, no groove (1); circular pit with deep tubular anterior groove (2); long narrow groove between tegmen tympani and promontorium (3) (Luo and Marsh, 1996; Geisler and Luo, 1998). 5. Transpromontorial sulcus for internal carotid artery.— Present, forms anteroposterior groove on promontorium, medial to fenestrae rotunda and ovalis (0); absent (1) (Cifelli, 1982; Thewissen and Domning, 1992). 6. Sulcus on promontorium for proximal stapedial artery.— Present, forms a groove that branches from the transpromontorial sulcus anteromedial to the fenestra rotunda and extends to the medial edge of the fenestra vestibuli (0); absent (1) (Cifelli, 1982; Wible, 1987; Thewissen and Domning, 1992). 7. Mastoid process of petrosal.—Exposed externally on posterior face of braincase as a triangle between lambdoidal crest of the squamosal dorsolaterally, the exoccipital ventrally, and the supraoccipital medially (0); not exposed posteriorly, lambdoidal crest of squamosal in continuous contact with exoccipital and supraoccipita l (1). 8. Stylomastoid foramen.—Complete, ectotympanic contacts tympanohyal laterally and petrosal medially, in some cases ectotympanic separated from petrosal by a narrow (< 1 mm) ssure (0); incomplete, ectotympanic does not contact the petrosal either anterior or posterior to the fenestra rotunda, medial side of foramen open (1) (modied from Geisler and Luo, 1998). 9. Facial nerve sulcus distal to stylomastoid foramen.— Absent (0); present, anterior wall of sulcus formed by squamosal (1); present, anterior wall formed by 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 485 mastoid process of petrosal (2); anterior wall formed by meatal tube of ectotympanic (3) (modied from Geisler and Luo, 1996, 1998). Some taxa coded as polymorphic for this character are not truly polymorphic, and the character state is simply ambiguous, but certain states can be eliminated from the list of possibilities. Articulation of pars cochlearis with basisphenoid/basioccipital.—Present (0); absent (1) (Thewissen and Domning, 1992). Ectotympanic.—Simple ring, no bulla formation (0); medial edge expanded into bulla (1) (derived from Novacek, 1977; MacPhee, 1981; Geisler and Luo, 1998). Pachyosteosclerotic involucrum of bulla.—Absent (0); present (1) (Thewissen, 1994). Lateral furrow of tympanic bulla.—Absent (0); present, forms a groove on the lateral surface of the ectotympanic bulla anterior to the base of the sigmoid process or meatal tube (1) (Geisler and Luo, 1998). Median furrow of tympanic bulla for tympanohyal (ordered).—Absent (0); median notch on posterior rim of bulla (1); prominent anteroposteriorly oriented furrow splits the ventral surface of the bulla into medial and lateral halves (2) (Geisler and Luo, 1998). Articulation of ectotympanic bulla to squamosal (ordered).—Broad articulation with medial base of postglenoid process (0); circular facet on short entoglenoid process (1); contact with a crest of the entoglenoid process (2); contact absent (3) (Geisler and Luo, 1998). Contact between exoccipital and ectotympanic bulla.— Absent (0); present (1) (Geisler and Luo, 1998). Sigmoid process (= anterior crus of tympanic ring).— Absent (0); present, forms transverse plate that projects laterally from the anterior crus of the ectotympanic ring and forms the anterior wall of the external auditory meatus (1) (modied from Thewissen, 1994; Geisler and Luo, 1998). Sigmoid process shape.—Thin and transverse plate (0); broad and aring, base of the sigmoid process forms dorsoventral ridge on lateral surface of ectotympanic bulla (1) (Geisler and Luo, 1998). Ectotympanic part of the meatal tube (ordered).— Absent (0); present but short, length of tube < 30% maximum width of bulla (1); present and long, length > 60% maximum width of bulla (2) (Geisler and Luo, 1998). Squamosal portion of external auditory meatus.— Absent or shallow, depth < 25% of transverse length (0); deep groove, depth > 35% of transverse length (1). Basioccipital crests (falcate processes).—Absent (0); present, forming ventrolaterall y aring basioccipital processes (1) (derived from Barnes, 1984; modied from Thewissen, 1994). Internal carotid foramen.—Absent or conuent with the piriform fenestra (0); present at basisphenoid/basioccipital suture with lateral wall of foramen formed by both these bones and thus separated from the piriform fenestra (1) (Geisler and Luo, 1998). 486 VOL. 48 SYSTEMATIC BIOLOGY 23. Postglenoid foramen (ordered).—Present, enclosed by squamosal, immediately posterior to postglenoid process (0); present, medial to petrosal/squamosal suture (if bulla present, a secondary ventral opening between bulla and squamosal may form) (1); absent (2) (Novacek, 1986; Geisler and Luo, 1998). 24. Foramen for ramus superior of stapedial artery.— Present (0); absent (1) (modied from Novacek, 1986; Thewissen and Domning, 1992). 25. Position of foramen for ramus superior of stapedial artery.—In petrosal/squamosal suture on dorsolateral edge of epitympanic recesss (0); anterolateral to epitympanic recess (1) (modied from Geisler and Luo, 1998). Cannot be scored for taxa that lack a foramen for the ramus superior of the stapedial artery. 26. Foramen ovale.—Anterior to glenoid fossa, posterior wall formed by alisphenoid (0); medial to glenoid fossa, posterior wall formed by alisphenoid (1); medial to glenoid fossa, posterior wall formed by petrosal (2); posterior to glenoid fossa (3). 27. Alisphenoid canal (alar canal).—Absent (0); present (1) (Novacek, 1986; Thewissen and Domning, 1992). 28. Anterior opening of alisphenoid canal.—Within sphenorbital ssure, cannot be distinguished from sphenorbital ssure in lateral view (0); posterior to sphenorbital ssure (1) (Novacek, 1982; Thewissen and Domning, 1992). 29. Posterior opening of alisphenoid canal.—Wellseparated from foramen ovale (0); within a deep groove or recess with the foramen ovale (1) (derived from Zhou et al., 1995). 30. Foramen rotundum.—Absent, maxillary division of trigeminal nerve exits skull through sphenorbital ssure (0); present (1) (Novacek, 1986; Thewissen and Domning, 1992). 31. Mastoid foramen.—Present, skull in posterior view (0); absent (1) (denition of structure follows MacPhee, 1994). 32. Post-temporal canal (for arteria diploetica magna, also called pericranial foramen).—Present, occurs at petrosal/squamosal suture with skull in posterior view, the canal continuing within the petrosal/squamosal suture (0); absent (1) (Wible, 1990; MacPhee, 1994). 33. Preglenoid process.—Absent (0); present, forms transverse, ventrally projecting ridge at anterior edge of glenoid fossa (1) (modied from Thewissen, 1994). Cranial Characters 34. Supraorbital process.—Absent, region over orbit does not project laterally from sagittal plane (0); present, laterally elongate and tabular (1) (derived from Barnes, 1984). 35. Postorbital bar (ordered).—Absent (0); present and almost complete (1); present and complete (2). 36. Contact of frontal and maxilla in orbit.—Absent (0); present (1). 37. Lacrimal tubercle.—Absent (0); present, situated on anterior edge of orbit adjacent to the lacrimal foramen (1) (Novacek, 1986). 38. Palatine ssure.—Small (0); large (1). 39. Posterior margin of external nares (ordered).— Anterior to or over the canines, or at anterior edge of premaxilla if canine not present (0); between P1 and P2 (1); posterior to P2 (2) (Geisler and Luo, 1998). 40. Mandibular foramen.—Small, maximum height of opening # 25% the height of the mandible at M3 (0); large, continuous with a large posterior fossa, maximum height $ 50% the height of the mandible at M3 (1) (modied from Thewissen, 1994; Geisler and Luo, 1998). 41. Height of mandibular condyle relative to dentition (ordered).—Below level of the dentition (0); even with superior aspect of dentition (1); substantially superior to dentition (2). 42. Angle of dentary.—Distal end at same level as ventral edge of dentary below molars (0); forms distinct ange that projects posteroventrally well below ventral edge of dentary (1) (modied from Gentry and Hooker, 1988). Dental Characters 43. Upper incisors.—Present (0); absent (1). 44. Premaxillae.—Short with incisors arranged in transverse arc (0); elongate, incisors aligned longitudinally with intervening diastemata (1) (modied from Prothero et al., 1988; Thewissen, 1994). 45. Embrasure pits on palate.—Absent (0); present (1) (modied from Thewissen, 1994). 46. P1 (ordered).—Absent (0); present, one-rooted (1); present, two-rooted (2) (Zhou et al., 1995). 47. P3 roots.—Three (0); two (1) (Zhou et al., 1995). 48. P4 protocone.—Present (0); absent (1) (Thewissen, 1994). 49. P4 paracone.—Equal or subequal to height of paracone of M1 (0); greater than twice the height of M1 paracone (1) (Thewissen, 1994). 50. P4 metacone.—Absent (0); present (1) (Thewissen, 1994). 51. Stylar shelf.—Present (0); absent (1). 52. M2 metacone (ordered).—Distinct cusp, sub-equal to paracone (0); distinct cusp approximately half the size of the paracone (1); highly reduced, indistinct from paracone (2) (Zhou et al., 1995). 53. M1 parastyle (ordered).—Absent (0); weak (1); moderate to strong (2) (O’Leary, 1998a) . 54. Lingual cingulum on M2 .—Present (0); absent (1) (O’Leary, 1998a). 55. Ectocingula.—Present (0); absent (1) (O’Leary, 1998a). 56. M1 -M2 hypocone.—Absent (0); present (1) (Thewissen and Domning, 1992; O’Leary, 1998a). 57. Trigon basin (ordered).—Broad (0); somewhat narrow (1); very narrow (2) (modied from Thewissen, 1994; O’Leary, 1998). 58. Paraconule on M2 .—Absent (0); present (1) (O’Leary, 1998a). 59. Metaconule on M2 .—Absent (0); present (1) (O’Leary, 1998a). 60. M3 size (ordered).—Present, larger than M2 (0); present, approximately equal to M2 (1); present, small, maximum mesiodistal length than < 60% the length of M2 (2); absent (3) (modied from Zhou et al., 1995). 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA 487 61. Lower canines.—Approximately twice as large as in- 85. M3 hypoconulid.—Long, protrudes as separate discisors or larger (0); subequal in size to incisors, well tal lobe (0); short, does not protrude substantially separated from postcanine teeth (1). beyond rest of talonid (1); absent (2) (Thewissen, 1994). 62. Lower incisors.—Small and vertical (0); tusk-like (1). 63. Lower canine cross-section shape.—Oval (0); triangu- 86. dp4 .—Resembles M1 (0); six-cusped, additional neomorphic cusp on paracristid (1) (Gentry and lar (1) (Gentry and Hooker, 1988). Hooker, 1988; Luckett and Hong, 1998). 64. P1 .—Present (0); absent (1) (Zhou et al., 1995). 65. P3 metaconid.—Absent (0); present (1) (Thewissen 87. Elongate shearing facets on lower molars extending below the gum line.—Absent (0); present (1). and Domning, 1992). 66. P4 metaconid.—Absent (0); present (1) (Thewissen and Domning, 1992). Postcranial Characters 67. Lingual cingulid on lower molars.—Poorly dened or 88. Cervical vertebrae.—Long, length of centrum greater absent (0); continuous from mesial to distal extreme than or equal to the centra of the anterior thoracics (1) (O’Leary, 1998a). (0); short, length shorter than centra of anterior tho68. M1 paraconid.—Present (0); absent (1) (derived from racics (1); very long (2) (derived from Gingerich et O’Leary, 1998a) . al., 1995). 69. M1 paraconid/paracristid position.—Cusp lingual or 89. Articulation between sacral vertebrae and ilium of paracristid winding lingually (or both) (0); cusp pelvis (ordered).—Broad area of articulation beanterior or paracristid (or both) straight mesiodistween pelvis and S1 and sometimes S2 (0); narrow tally, sometimes poorly developed (1) (derived from articulation of pelvis with end of transverse process O’Leary, 1998a) . of S1 (1); articulation absent (2) (Geisler and Luo, 70. M2 paraconid.—Present (0); absent (1) (derived from 1998). O’Leary, 1998a) . 90. Number of sacral vertebrae (ordered).—One (0); two 71. M2 paraconid/paracristid position.—Cusp lingual or or three (1); four (2); ve or six (3); cannot be scored paracristid (or both) winding lingually (0); cusp anfor taxa that lack an articulation between the verteterior or paracristid (or both) straight mesiodistally, bral column and the ilium (Thewissen and Domnsometimes poorly developed (1) (O’Leary, 1998a) . ing, 1992; Gingerich et al., 1995). 72. M3 paraconid.—Present (0); absent (1) (O’Leary, 91. Scapular spine.—Bears large acromion pro1998a). cess that overhangs glenoid fossa (0); scapu73. M3 paraconid/paracristid position.—Cusp lingual or lar spine with acromion process small or abparacristid winding lingually (or both) (0); cusp ansent, does not encroach upon glenoid fossa (1); terior or paracristid (or both) straight mesiodistally, acromion process large, directed anteriorly and sometimes poorly developed (1) (O’Leary, 1998a) . does not encroach upon the glenoid fossa (2) 74. M1 metaconid.—Present, forms distinct cusp (0); ab(O’Leary and Rose, 1995b; Geisler and Luo, sent or occasionally present as swelling on lingual 1998). side of protoconid (1) (Thewissen, 1994; Zhou et al., 92. Entepicondyle of humerus.—Wide, $ 50% width of ul1995). nar and radial articular facets (0); narrow, # 25% 75. M2 metaconid.—Present, forms distinct cusp (0); abwidth of ulnar and radial articular facets (1) (desent or occasionally present as swelling on lingual rived from O’Leary and Rose, 1995b; Geisler and side of protoconid (1) (modied from Zhou et al., Luo, 1998). 1995). 93. Entepicondylar foramen.—Present (0); absent (1) 76. M3 metaconid.—Present, forms distinct cusp (0); ab(Thewissen and Domning, 1992). sent or occasionally present as swelling on lingual 94. Length of olecranon process.—Short, < 10% total ulside of protoconid (1) (modied from Zhou et al., nar length (0); long, > 20% total ulnar length (1) 1995). (derived from O’Leary and Rose, 1995b). 77. M1 -M2 hypoconulid.—Absent (0); present (1) (The- 95. Olecranon fossa.—Shallow (0); deep, sometimes perwisen, 1994). forate (1) (modied from O’Leary and Rose, 1995b). 78. Molar protoconid.—Subequal to height of talonid 96. Proximal end of radius.—Single fossa for trochlea and (0); closer to twice height of talonid or greater (1) capitulum of humerus (0); two fossae, one for the ca(O’Leary, 1998a). pitulum and one for the medial edge of trochlea (1); 79. Protoloph.—Absent (0); present (1). three fossae, same as (1) but with additional fossa 80. Metaloph.—Absent (0); present (1). for the lateral lip of the humeral articular surface (2) 81. Metalophid formation on anterior aspect of lower teeth (O’Leary and Rose, 1995b; Geisler and Luo, 1998). (between protoconid and metaconid).—Absent (0); 97. Distal articular surface of radius.—Single concave present (1). fossa (0); split into scaphoid and lunate fossae (1); 82. Hypolophid formation on posterior aspect of lower teeth convex to at, distinct facets (2) (derived from (between entoconid and hypoconid).—Absent (0); O’Leary and Rose, 1995b). present (1) (Thewissen and Domning, 1992). 98. Centrale.—Present (0); (1) absent (Thewissen, 1994). 83. Reentrant grooves.—Proximal (0); absent (1); distal 99. Manus.—Mesaxonic, axis of symmetry passes along (2) (Thewissen, 1994; O’Leary, 1998a). center of digit 3 (0); paraxonic, axis of symmetry 84. Molar talonid.—Basined, or slightly basined, passes between digits three and four (1). hypoconid and entoconid present (0); reduced (1) 100. Height of greater trochanter of femur (ordered).— (modied from Thewissen, 1994). Inferior to head of femur (0); approximately equal to 488 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. VOL. 48 SYSTEMATIC BIOLOGY head of femur (1); above head of femur (2) (derived from O’Leary and Rose, 1995b). Third trochanter of femur (ordered).—Present (0); highly reduced (1); absent (2) (derived from O’Leary and Rose, 1995b). Patellar articular surface on femur.—Wide (0); narrow (1). Astragalar canal.—Present (0); absent (1) (derived from Shoshani, 1986). Navicular facet of astragalus (ordered) (see Fig. 1).— Convex (0); saddle-shaped (1); trochleated (2) (modied from Thewissen and Domning, 1992; O’Leary and Rose, 1995b). Sustentacular width.—Narrow, # 50% width of the astragalus (0); wide, $ 70% width of the astragalus (1) (Geisler and Luo, 1998). Contact of distal astragalus with cuboid (ordered) (see Fig. 1).—Absent (0); present but small, contact # 30% of the width of the distal articular surface (1); present and large, contact $ 40% of the articular surface (2) (O’Leary and Rose, 1995b; Geisler and Luo, 1998). Lateral process of astragalus.—Present (0); absent (1) (Schaeffer, 1956). Pes (ordered).—Mesaxonic, axis of symmetry of foot passes along center of digit three (0); paraxonic, axis lies between digits three and four (1); mesaxonic, axis passes along center of digit four (2) (derived from Gingerich et al., 1990; Thewissen, 1994; our interpretation of this character differs from that of these previous authors). Proximal astragalus (ordered).—Nearly at to slightly concave (0); grooved, depth of trochlea < 25% width of trochlea (1); deeply grooved, depth of trochlea > 30% width of trochlea (2) (derived from Schaeffer, 1947; O’Leary and Rose, 1995b). Cuboid and navicular.—Unfused (0); fused (1) (Webb and Taylor, 1980). Proximal fusion of third and fourth metatarsals.— Absent (0); present (1). First metatarsal (ordered).—Unreduced, length > 50% length of third metatarsal (0); reduced, length < 50% length of third metatarsal (1); highly reduced in form of nodule or small splint or absent (2). 113. Second metatarsal (ordered).—Unreduced, length $ 50% length of third metatarsal (0); reduced, length # 50% length of third metatarsal (1); highly reduced in form of nodule, small splint or completely absent (2). 114. Fifth metatarsal (ordered).—Unreduced, length $ 50% length of third metatarsal (0); reduced, length # 50% length of third metatarsal (1); highly reduced in form of nodule or small splint or absent (2). 115. Ventral border of distal phalanges.—Curved inferiorly (0); straight (1) (derived from MacLeod and Rose, 1993). 116. Distal phalanges in dorsal view.—Phalanx compressed transversely (0); broad transversely, each phalanx bilaterally symmetrical with central anteroposterior axis (1); broad transversely, each phalanx asymmetrical (2) (derived from MacLeod and Rose, 1993). Soft Tissue Characters 117. Stomach epithelium.—Nonglandular (stratied squamous epithelium) (0); composite (nonglandular and glandular mucosa) (1); discoglandular (small patches of glandular epithelium surrounded by nonglandular mucosa) (2); glandular (mucosa with glands) (3) (derived from Langer, 1988). 118. Lumen.—Unilocular (stomach with simple chamber only) (0); plurilocular (lumen of stomach subdivided by folds into two or more chambers) (1) (derived from Langer, 1988). 119. Diverticulation of stomach lumen (blind tubes or sacs branching from the main gastric lumen).—Absent (0); present (1) (derived from Langer, 1988). 120. Cavernous tissue of penis.—Abundant (0); sparse (1) (derived from Slijper, 1936; Thewissen, 1994). 121. Hair.—Abundant to common on body (0); almost completely absent (1) (Gatesy, 1997). 122. Sebaceous glands.—Present (0); absent (1) (Gatesy, 1997). 123. Primary bronchi of lung.—Two (0); three (1) (Thewissen, 1994). 1999 O’LEARY AND GEISLER—PHYLOGENY OF CETACEA 489 APPENDIX 3. Data matrix of 123 characters for 40 taxa. ? = missing data; n = inapplicable character; character states described in Appendix 2. Polymorphic characters as follows: states 0 and 1 = 5; 0 and 2 = 6; 0 and 3 = 7; 1 and 2 = 8; 2 and 3 = 9; earliest appearance of the genus (McKenna and Bell, 1997) (E. = early; M. = middle; L. = late; Cret. = Cretaceous; Pal. = Paleocene; Eoc. = Eocene; Mioc. = Miocene; Pleist. = Pleistocene), and percent missing data listed at end of matrix. Characters in bold are treated as ordered in certain analyses, extant taxa in bold. The matrix is available at http://www.herbaria. harvard.edu/treebase/ under the accession number S387. (continued on next page) 490 SYSTEMATIC BIOLOGY APPENDIX 3. (Continued) VOL. 48
© Copyright 2026 Paperzz