Falaeont. a/r., 37,53-79 (2001) PRELIMINARY PHYLOGENETIC ANALYSIS AND STRATIGRAPmC CONGRUENCE OF THE DICYNODONT ANOMODONTS (SYNAPSIDA: THERAPSIDA) by Kenneth D. Angielczyk Department 0./ Integrative Biology and Museum 0./ Paleontology, 1101 Valley Life Sciences Building, University 0./ California, Berkeley, CA 94720-4780 USA e-mail: etranger@socrates.berkeley.edu ABSTRACT A preliminary phylogenetic analysis of 20 well-known dicynodont taxa was conducted using modern cladistic methods . Many past phylogenetic hypotheses were corroborated, but others conflict with the results ofthis analysis . Most notably, Diictodon, Robertia, and Pristerodon are reconstructed in more basal positions than previously suggested, whereas Endothiodon and Chelydontops occupy a more crown ward position. These findings are consistent with novel evolutionary scenarios for characters such as the presence of postcanine teeth and anterior palatal ridges. The Relative Completeness Index and Gap Excess Ratio were used to examine the degree of fit between the most parsimonious cladograms of this study and the stratigraphic record of the dicynodonts . Although the results of this analysis suggest that the preferred cladogram is relatively consistent with stratigraphy, the presence of some ghost ranges and ghost lineages imply that the fossil record of dicynodonts is not as complete as is sometimes stated. These findings are important because there is a long tradition of intensive collecting in regions where dicynodont fossils are common; sections of several dicynodont lineages may not be preserved in these areas. KEYWORDS: Dicynodontia, Phylogeny, Stratigraphy, Anomodontia, Synapsida, Karoo Basin INTRODUCTION The dicynodont therapsids (sensu Hopson & Barghusen 1986; Modesto et at. 1999; Rybczynski 2000) represent a diverse radiation of specialized herbivorous non-mammalian synapsids that range in age from the Late Permian to Late Triassic and are known from every continent. The dicynodont fossil record from the Beaufort Group of South Africa is especially rich and well documented since the first description of therapsid fossils by Owen (1845). Together with a handful of more basal forms (e.g., Patranomodon, Venyukovia, Otsheria, Ulemica, Suminia) , dicynodonts form the Anomodontia (sensu Hopson & Barghusen 1986; Modesto et al. 1999; Rybczynski 2000), a clade of basal eutherapsids (Sidor & Hopson 1998). The exact placement of Anomodontia within Therapsida is still unclear (compare e.g., Gauthier eta!' 1988; Hopson & Barghusen 1986; Sidor & Hopson 1998), but this issue is not the focus of this paper. Despite the large number of specimens and the long history of work surrounding the group, the phylogenetic relationships of dicynodonts are still problematic. Many early workers, such as Robert Broom, concentrated on simply collecting and describing isolated, often poorly preserved or prepared skulls, and paid less attention to issues such as functional morphology, systematics, or even descriptions of post cranial skeletal elements. As a result, the group was extremely over-split (see Haughton & Brink's 1954 bibliographic database for the results of this tendency). Recent workers (e.g., Cluver & Hotton 1981; Cluver & King 1983; Keyser 1975,1993; Keyser & Cruickshank 1979; King 1988,1993; King & Rubidge 1993) have reduced the number of recognized taxa to a more manageable number, but the lineage status of most of these taxa has not been rigorously tested. Cluver & King (1983), King (1988; 1990), Cox (1998), andSurkov (2000) published cladistic analyses ofthe group, but the absence of a list of characters examined or character codings for the taxa included in these works makes it difficult to independently test their hypotheses. The analyses of Modesto eta!' (1999), Modesto & Rybczynski (2000), and Rybczynski (2000) are methodologically stronger, but the focus of these works has been on the relationships of non-dicynodont anomodonts and they include relatively few dicynodont taxa. Nevertheless, all of these analyses represent important first steps toward an understanding of dicynodont phylogeny. In the following study, I present a new, preliminary phylogenetic analysis of the dicynodonts. My primary goal is to examine the interrelationships of the better known Permian dicynodont taxa using up-to-date methods and technologies, such as the computer parsimony algorithm PAVP. Because considerable work remains to be conducted on most dicynodont taxa, the present analysis is not intended to be an exhaustive treatment of dicynodont phylogeny. Instead, it represents a rigorous test of the hypotheses of relationship and homology made by previous authors. Furthermore, I examine the results of this phylogenetic 54 analysis in the context of stratigraphy to determine how well the morphologically parsimonious cladograms fit the known fossil record. INSTITUTIONAL ABBREVIATIONS AM: Albany Museum, Grahamstown, South Africa AMNH: American Museum of Natural History, New York, USA BMNH: Natural History Museum, London, England BP: Bernard Price Institute for Palaeontological Research, Johannesburg, South Africa GSP: Geological Survey of South Africa, Pretoria, South Africa NM: PIN: ROZ: SAM: National Museum, Bloemfontein, South Africa Paleontological Institute, Moscow, Russian Federation Roy Oosthuizen Collection, South African Museum, Cape Town, South Africa South African Museum, Cape Town, South Africa TM: Transvaal Museum, Pretoria, South Africa UCMP: University of California Museum of Paleontology, Berkeley, USA METHODS Phylogenetic Analysis A matrix consisting of 20 anomodont taxa and 40 morphological characters was constructed and subjected to a maximum parsimony analysis using PAUP 4.0b4a (Swofford 2000). The characters used are listed and discussed in detail in Appendix 1, the matrix is shown in Appendix 2, and the sources of character codings are given in Appendix 3. Multistate characters were treated as unordered and all characters were equally weighted. Unknown and inapplicable characters were coded "?" (Strong & Lipscomb 2000). The heuristic search algorithm was used and 1 000 random addition sequence replicates were run to prevent the searches from becoming trapped in a local tree-length minimum (Maddison 1991). Support for the recovered clades was measured by decay analysis (Bremer 1988). Many of the characters in this analysis are based on proposed synapomorphies that have been used by King (1988,1990), Cox (1998), and Surkov (1998, 2000) to rdiagnose dicynodont clades or grades of organization \they have proposed or characters used in Modesto et af. (1999) and Rybczynski (2000). As noted above, although all of these authors used cladistic principles to construct their phylogenetic hypotheses, only the analyses of Modesto et af. (1999) and Rybczynski (2000) were based on discrete-state characters and constructed using a computer parsimony algorithm (e.g., PAUP). Accordingly, I have reinterpreted many of the characters so that they have discrete states that can be coded in a matrix. The character states and codings that I have used are based mainly on my observations of dicynodont specimens in various institutions in the United States, South Africa, and Russia. A detailed discussion of the character states and my reasons for coding different taxa as I have are provided in Appendix 1. Because this analysis includes many characters from past phylogenetic analyses of dicynodonts and newer methods than most, it represents a rigorous test of previous phylogenetic hypotheses. The skull has long been the primary focus of studies of dicynodonts, and this bias is reflected in collections of dicynodont material as well as studies of their morphology and systematics. Early dicynodont taxonomists placed much emphasis on skull proportions, patterns of facial and skull roof sutures, and the presence or absence of bones such as the postfrontal or septomaxilla (e.g., Broom 1932). I have avoided many of these characters because more recent authors (e.g., Cluver & Hotton, 1981; Cluver & King, 1983; Keyser 1975; King 1993; Rubidge 1984; Toerien 1953; Tollman eta!' 1980) have shown that they are highly variable within dicynodont genera, species, and presumed age classes. They are also susceptible to alteration by taphonomic proce,sses such as weathering or plastic deformation. Since the seminal work of Toerien (1953), much greater attention has been paid to the structure and functional relationships of the palate and mandible of dicynodonts because these regions are rich sources of potentially informative characters that show less individual variation. Other potentially informative regions ofthe skull, such as the braincase, have not been thoroughly examined to determine if they preserve a useful phylogenetic signa~. The majority of the cranial characters I have used in this analysis are features of the palate and mandible, although some deal with other regions of the skull. So much attention has been paid to the dicynodont skull, including the recent focus on the feeding system, that other sources of phylogenetic characters have remained almost completely unexplored in the dicynodont literature. Despite well over a century of collection and description, the postcranial skeleton of dicynodonts remains poorly known; only scattered descriptions of its osteology or functional morphology are available (e.g., Boonstra 1966; Broom 1905; Camp & Welles 1956; Cluver 1978; Cox 1959, 1972; DeFauw 1986; King 1981a, 1981b, 1985; Olson & Byrne 1938; Pearson 1924; Rubidge eta!' 1994; Walter 1986; Watson 1960). This lack of published information is unfortunate because these works and others that have attempted to include postcranial characters in systematic studies (Camp 1956; King 1988; Surkov 1998) strongly suggest that dicynodont postcrania preserve a valuable and informative phylogenetic signal. The consideration of postcranial characters has enabled more parsimopious explanations of patterns of evolution to be discovered in broader studies of synapsids (Hopson 1995; Rowe & van den Heever 1986). The five postcranial characters that I have included appear to be phylogenetically informative and are easily coded even for poorly preserved or prepared specimens. Eighteen dicynodont genera form the in group and two non-dicynodont anomodonts (Patranomodon and Otsheria) are used as outgroups. Because the monophyly of the Anomodontia is accepted (Hopson, 1991 ; Hopson & Barghusen 1986; Modesto et a!. 1999; Rybczynski 2000; Sidor & Hopson 1998) and I am concerned only with relationships among dicynodonts, I did not include any non-anomodont outgroups in this analysis. Patranomodon is a useful outgroup because it is a basal anomodont (Modesto et al. 1999; Rubidge & Hopson 1990, 1996; Rybczynski 2000) represented by a very well-preserved skull and associated postcranial elements that retain many primitive anomodont characters. Otsheriais a venyukovioid anomodont (sensu Modesto et a!. 1999; Rybczynski 2000), and represents a slightly more advanced grade of anomodont evolution. Anomocephalus, the most basal known anomodont (Modesto & Rubidge 2000; Modesto et a!. 1999), was not used as an outgroup because the only known specimen (BP/1/5582) does not preserve many of the characters I have used. The in group taxa in this analysis are known primarily from the Late Permian of Africa, with the exceptions of Lystrosaurus, Myosaurus, and Kannemeyeria~ which appear in the Triassic. The focus of this analysis is the pattern of relationships among the Permian dicynodont taxa, and I have tried to include as many of the well­ known genera as possible. The Triassic genera are mainly intended to serve as place-holders, and the results of this and other studies (Cluver & King 1983; Cox 1965, 1998; Cox &Li 1983; Keyser & Cruickshank 1979; King 1988, 1990) suggest that most Triassic dicynodonts are members of the monophyletic clade including Dicynodon, Lystrosaurus, and Kanne­ meyeria. Future analyses that include additional Triassic taxa can be used to test this assumption. All included genera have been recently described, redescribed, or revised in the literature, and I have had the opportunity to personally examine specimens that can be referred cqnfidently to these taxa (Appendix 3). Although the taxonomy of the included genera is not the focus of this paper, my treatment of Eodicynodon, Chelydontops, and Tropldostoma necessitates some explanation. Two described species currently are assigned to the genus Eodicynodon, Eodicynodon oosthuizeni (Barry 1974a) and Eodicynodon oelofteni (Rubidge 1990b). Although the taxa share several diagnostic features, a number of notable differences also exist, including the presence or absence of a canine tusk, the size and arrangement of the postcanine teeth, and the morphology of the lower jaw (Rubidge 1990b). Also, many more specimens of E. oosthuizenihave been collected (over 20 well-preserved skulls versus the holotype and a second specimen (NM QR3003) that is probably referable to E. oelofteni, both of which are poorly preserved; Rubidge, 1990a; personal observation) and its cranial and postcranial osteology have been much more thoroughly described (Barry 1974b; Cluver & King 1983; Rubidge 1984,1985, 1990a; Rubidge eta!' 1994). Because of these discrepancies, I have considered only E. oosthuizem: and my character state codings for Eodicynodon should be considered valid only for this species. Cluver (1975) described the taxon Chelydontops altldentalis based on two specimens (SAM 11558 and 55 SAM 12259) discovered in the Tapinocephalus Assemblage Zone. Recently, Cox (1998) has proposed synomymizing C. altldentalis with the genus Prodicynodon, creating the new combination Prodicynodon altidentalis. Although the holotype (SAM 11558) and referred (SAM 12259) specimens of Chelydontops are not completely preserved, they do show a distinct, diagnostic suite of features, including a relatively wide intertemporal bar, a dome­ like pineal boss, a shelf-like area lateral to the upper postcanine teeth, a large, leaf-shaped palatal exposure of the palatine, a distinctly developed coronoid eminence, and a median palatal ridge that has a flattened, expanded, diamond-shaped area anteriorly. The holotypes of both Prodicynodon pearstonensis (AM 2551) and P. beaufortensis (AMNH 5509) are very fragmentary and preserve only the snout and anterior portion of the lower jaw. The palate is not fully exposed in either specimen, and the details of the intertemporal region are also unknown. Because both specimens preserve almost no information regarding their possible affinities with Chelydontops altldentalis, I believe it is premature to refer this taxon to the genus Prodicynodon. I agree with Keyser (1993) that both P. pearstonensis and P. beaufortensis should be regarded as nomina dubia until more informative material is discovered. The codings for Chelydontops in this analysis are based on my observations of the two SAM specimens. The genus Tropldostoma is problematic because although there is considerable variation in the width of the intertemporal region in the taxon, most of the specimens referred to it are otherwise very similar. Keyser (1973) largely ignored this variation in his revision of Tropldostoma, but Cluver & King (1983) favored placing specimens with narrow intertemporal regions in Tropldostoma while retaining the genus Cteniosaurus for specimens with wider intertemporal regions. The holotype specimen for Tropidostoma (BMNH R868) has a relatively narrow intertemporal region, with the parietals exposed in a median depression between the postorbitals, which partially overlap them. In my observations, I have found that this pattern is relati vel y constant, although the depth of the depression, degree of overlap, and overall width of the intertemporal bar can be quite variable. These features also are highly susceptible to alteration by plastic deformation (for example, the holotype of T. dunni (BMNH R866) has a narrower intertemporal region than the holotype of T. microtrema (BMNH R868), but the former specimen has clearly been laterally compressed). For these reasons, as well as the fact that there is almost no other morphological difference between Cteniosaurus and Tropldostoma, I have followed Keyser (1973) and treat them as synonyms. Stratigraphic Analysis The biostratigraphic relationships of dicynodonts, especially those of the Karoo Basin of South Africa, have a long history of study (e.g., Broom 1906; Keyser 56 & Smith 1977-1978; Kitching 1977; Lucas 1996, 1998; Rubidge 1995a; Seeley 1892; Watson 1914) and frequently have been used to add a temporal dimension to phylogenetic and adaptational hypotheses about the group. Because of this tradition, and the fact that the Beaufort Group and its eight assemblage zones have been the subject of recent scrutiny (Rubidge 1995a), dicynodonts are an ideal group with which to examine the fit between stratigraphy and phylogeny. A number of methods have been proposed recently to measure the fit between stratigraphic data and cladograms (e.g., Benton & Storr, 1994; Gauthier etal 1988; Huelsenbeck 1994; Norell & Novacek 1992a, 1992b; Wills, 1999). There has also been debate over whether and how one should include stratigraphic data in phylogenetic analyses (e.g., Clyde & Fisher 1997; Fisher 1992,1994,1997; Fox etall997, 1999; Norell & Novacek 1997; Rieppell997; Smith 2000; Thewissen 1992). In this analysis the Relative Completeness Index (RCI) metric of Benton & Storrs (1994) and the Gap Excess Ratio (GER, Wills, 1999) were used to examine the nature of the gaps in the fossil record required by the morphologically parsimonious cladograms. A cladogram that implies fewer or shorter gaps in the fossil record is a more highly corroborated hypothesis than an equally parsimonious cladogram that requires longer or more numerous gaps. I calculated the RCI and GER for all of the most parsimonious cladograms of this analysis. In addition, these metrics were calculated for cladograms up to four steps longer and a cladogram based on the topology of King (1988, 1990), but only including the taxa I examined (Figure lc), to determine if any of these hypotheses fit the known fossil record better than the morphologically most parsimonious cladograms. The Basic program "Ghosts" (version 2.4, 1000 random replicates, polytomies resolved as the worst case; Wills 1999) also was used to determine whether the most parsimonious cladograms of this analysis and the cladogram of King (1988, 1990) fit the fossil record significantly better than random. Stratigraphic data were taken from Rubidge (1995a) and King (1988). Because all taxa in the analysis except Otsheria are known from southern Africa, potential biases caused by different preservation rates or biostratigraphic correlation difficulties should be minimal. Taxa known to occur in only part of an assemblage zone were not treated differently than taxa found throughout an assemblage zone. This assumption simplifies data analysis, but reduces resolution. To calculate the RCI (Benton & Storrs 1994; see also Benton & Hitchin 1996; Benton & Simms 1995; Benton & Storrs 1996; Hitchin & Benton 1997a) Minimum Implied Gaps (MIGs; see Benton 1994; Norell 1993; Norell & Novacek 1992a, 1992b; Smith & Littlewood 1994; Storrs 1993; Weishampel & Heinrich 1992) were assessed for the most parsimonious cladograms as well as the cladogram based on King's (1988, 1990) topology. This is easily done by coding a stratigraphic character (Appendices 1, 2) in MacClade 3.08 (Maddison & Maddison 1999; see also Fisher 1992) and noting the number of steps added. The Simple Range Length (SRL; Storrs, 1993) for each taxon was measured by counting the number of assemblage zones in which that taxon appears. These values were then substituted into the equation given in Benton & Storrs (1996) and the equation was solved for RCI. Because the RCI takes into account time duration and missing time, not just the relative ranks of clades, it has been suggested as an estimate of the completeness of the fossil record of the group in question implied by a phylogenetic hypothesis (Hitchin & Benton 1997a; although see Wagner 2000). The RCI is known to be sensiti ve to the choice of taxa and magnitude of time examined (Benton & Storrs 1994; Hitchin & Benton 1997a). It may also be affected by number of taxa and clade asymmetry (Siddall 1996; 1997; but see Benton et al 1999; Hitchin & Benton 1997a, 1997b). However, these biases should not be a problem in this analysis because the identities and number of included taxa, as well as the magnitude of time in consideration, are the same for all of the cladograms in question. Although there are slight differences in asymmetry among the cladograms, they are unlikely to be large enough to affect the results seriously. The randomization procedure implemented in "Ghosts" also helps to control for cladogram size and balance biases when estimating the significance of the RCI (Wills 1999). . The Gap Excess Ratio (GER) has been proposed by Wills (1999) as a means to examine the fit of a cladogram to stratigraphy by examining the amount of ghost ranges required. The GER represents the excess ghost range above and beyond the minimum possible value for a set of stratigraphic data, expressed as a fraction of the total range of ghost values possible for that data (Wills 1999). To calculate the GER, the minimum and maximum number of implied gaps possible for the stratigraphic data set used here were calculated. These values were then substituted into equation 3 of Wills (1999) with the MIG (see above) for each of the most parsimonious cladograms of this analysis, the cladograms several steps longer, and the cladogram of King (1988, 1990). The GER appears not to be biased by the number of taxa included in the cladograms being compared (Benton et al1999; Wills 1999), butthis potential source of error is not a concern because all of the cladograms included in this analysis have the same number of taxa. It can be affected by symmetry differences among examined cladograms (Benton et al 1999), but variation in symmetry among the cladograms in this analysis;s minimal and should not cause undue bias. As with the RCI, the randomization procedure of "Ghosts" helps to control for cladogram size and shape biases when calculating the significance ofthe GER (Wills 1999). RESULTS Phylogenetic analysis The parsimony analysis recovered a single most parsimonious cladogram with a length of 125 steps, a consistency index of 0.57, and a retention index of 0.67 (Figure la). The topological results of the analysis are presented in Figure 1. For comparison, a cladogram based on the topology of King (1988,1990.), but including only the taxa and characters examined here (Figure lc), has a length of 136 steps, a consistency index of 0..53, and a retention index of 0..59. Tree decay analysis shows that most of the hypothesized clades in this phylogeny are relatively weakly supported. At 126 to 127 steps, the major clades of dicynodonts are resolved, but most resolution within these clades is lost (Figure 1 b). The observed loss of resolution is likely due to the generally short branch lengths within the major clades. At 128 steps, the ingroup is resolved from the outgroup, but no branches within the ingroup are resolved. All resolution is lost at 133 steps. A large number of character state changes take place on the branch separating the ingroup from the outgroup, resulting in the strong decay support for that node. The topological results of this analysis are generally compatible with those of King (1988, 1990.), although some differences do exist. Most notably, Robertla and Dlidodon, and Pristerodon are reconstructed in a more basal position, whereas Endothiodon and Chelydontops appear in a more crownward position. Although clades including the same taxa as the Cryptodontinae and Emydopidae of King (1988, 1990.) are present in the most parsimonious cladogram of this analysis, the relationships within these clades are slightly different (Figure la, Nodes I, L; Figure lc, Nodes C, E). The results are also compatible with those of Modesto et al. (1999), Modesto & Rybczynski (20.0.0.), Rybczynski (20.0.0.) and Surkov (20.0.0.), although there are notable differences in taxon sampling. Character state transformations were optimized on the most parsimonious cladogram using the Delayed Transformation (DELTRAN) algorithm of MacClade 3.a8a (Maddison & Maddison, 1999) to arrive at the following diagnoses of the recovered clades. The clade including Eodicynodon, Kannemeyena, and all descendants of their most recent common ancestor (= Dicynodontia sensu Modesto et al. 1999; Rybczynski 20.0.0.; Figure la, Node A) is diagnosed by the absence of premaxillary teeth (Character 2, State 1), the presence of a caniniform process (6, 1), a posterior median ridge with an expanded, flattened anterior area on the palatal . surface of the premaxilla (8, 1), a lateral squamosal fossa for the origination of the lateral branch of the M. Adductor MandibulaeExternus (21,1), a rounded, bulbous surface of the palatal exposure of the palatine suggesting a keratinized covering (22, 1), a single, median nasal boss (23, 1), a dorsolateral notch in the squamosal (32, 1), a relatively long interpterygoid vacuity that does not reach the level ofthe palatal exposure of the palatines (33,1), and a lateral palatal fenestra located at the level of the anterior portion of the palatal exposure of the palatines (35, 1). It is important to note that some of these characters (e.g., 21,1) would likely diagnose more inclusive clades if more non-dicynodont anomodonts were included in the analysis. 57 The clade including Robertla, Kannemeyena, and all descendants of their most recent common ancestor (Figure la, Node B) is diagnosed by the presence of fused premaxillae (3,1), upper postcanine teeth located medially, with the more posterior teeth approaching the lateral margin of the maxillae (4, 1), lower postcanine teeth located on a medial shelf or swelling (10, 1), and a reduced transverse flange of the pterygoid (37, 1). In addition, this clade is ambiguously diagnosed by the presence of a symphyseal region of the lower jaw with an upturned margin and a scooped-out depression on its posterior surface (18,l). The clade including Dlidodon, Robertla, and all descendants of their most recent common ancestor (= Robertiidae sensu King 1988, 1990.; Figure 1 a, Node C; Figure lc, Node A) is diagnosed by the presence of a caniniform process with a notch anterior to it (6, 2), paired anterior ridges that converge posteriorly on the palatal surface of the premaxilla (7, 1), contact between the anterior portion of the squamosal and the maxilla (34, 1), the presence of an ectepicondylar foramen (38, 1), and the presence of a cleithrum (39, 1). In addition, this clade is ambiguously diagnosed by the presence of an elongate dentary table bounded medially by a tall, thin, dorsally-convex blade (15, 2), and a relatively small lateral dentary shelf (17,1). The presence of paired, anterior palatal ridges that converge posteriorly may diagnose a more inclusive clade, but the testing of this hypothesis must wait until future analyses that include currently undescribed specimens collected in the lower Tapinocephalus Assemblage Zone of South Africa (Rubidge, personal communication, 20.0.0.; also see Appendix 1). The clade including Endothiodon, Kannemeyena, and all descendants of their most recent common ancestor (Figure la, Node D) is diagnosed by the presence of a relatively flat surface of the premaxillary secondary palate lateral to the posterior median palatal ridge (9,2), the presence of a posterior dentary sulcus (16,1), and a relatively smooth palatal surface of the palatine with fine pitting suggestive of a keratinized covering (22, 2). The clade including Endothiodon, Chelydontops, and all descendants of their most recent common ancestor (= Endothiodontoidea sensu King 1988, 1990.; Figure la, Node E; Figure lc, Node B) is diagnosed by the presence of a shelf-like area lateral to the upper post­ canine teeth (5, 1), paired nasal bosses located near the dorsal margin of the external nares (23, 2), and a strongly developed bony boss around the pineal foramen (26, 1). The clade including Pristerodon, Kannemeyena, and all descendants of their most recent common ancestor (Figure la, Node F) is diagnosed by contact between the palatine and premaxilla (27, 1), and contact between the anterior portion of the squamosal and the maxilla (34,1). The clade including Kingona, Kannemeyeria, and all descendants of their most recent common ancestor (Figure la, Node G) is diagnosed by the absence of upper postcanine teeth (4, 3), a posterior median palatal 58 ridge that lacks an expanded, flattened anterior area (8, 2), the absence of lower post canine teeth (10, 2), a mid­ ventral vomerine plate that is of constant width (12, 1), and an expanded femoral head that encroaches on the anterior surface of the bone (31,1). The clade including Kingoria, Myosaurus, and all descendants of their most recent common ancestor (Figure la, Node H) is diagnosed by the presence of a postcaniniform keel (1,1), an embayment of the palatal rim anterior to the caniniform process (14, 1), the absence of a dentary table (15, 0), and a shovel-shaped symphyseal region of the lower jaw (18, 3). This clade is also ambiguously diagnosed by the presence of a well­ developed lateral dentary shelf (17,3), although this feature may characterize a more inclusive clade. The clade including Emydops, Myosaurus, and all K N A Patranomodon Otsheria Eodicynodon Robertia c Diictodon Endothiodon E Chelydontops Pristerodon Emydops Cistecephalus Myosaurus Kingoria Tropidostoma Oudenodon Rhachiocephalus Pelanomodon o Aulacephalodon Dicynodon p Lystrosaurus Q Kannemeyeria c descendants of their most recent common ancestor (= Emydopidae sensu King, 1988, 1990, although see above; Figure 1 a, Node I; Figure 1 c, Node C) is diagnosed by the presence of groove-like depressions lateral to the posterior median palatal ridge (9, 1), a narrow, blade-like mid-ventral vomerine plate (13, 1), a squamosal with a relatively straight contour in occipital view (32, 0), and the presence of an ectepicondylar foramen on the humerus (38, 1). The clade including Myosaurus, Cistecepha/us, and all descendants of their most recent common ancestor (Figure la, NodeJ; also seeCluver, 1974b) is diagnosed by the presence of a foramen on the palatal surface of the palatine (24, 1) and a partial closing-off of the snout by the anterior margins of the orbits (25, 1). E The clade including Tropidostoma, Kannemeyeria, B Patranomodon Otsheria Eodicynodon Robertia A D(ictodon Emydops Myosaurus Cistecephalus Kingoria Pristerodon Tropidostoma Oudenodon Rhachiocephalus Pelanomodon F Aulacephalodon Dicynodon Lystrosaurus G Kannemeyeria Endothiodon B Chelydontops Patranomodon Otsheria Eodicynodon Robertia Diictodon Endothiodon Chelydontops Pristerodon Emydops Cistecephalus Myosaurus Kingoria Tropidostoma Oudenodon Rhachiocephalus Pelanomodon Aulacephalodon Dicynodon Lystrosaurus +2 Kannemeyeria Figure 1: A: Single most parsimonious cladograrn of this analysis (length 125 steps, CI = 0.57, RI = 0.67). The lettered nodes are discussed in the text. B: Cladogram showing the results of the decay analysis. Numbers indicate the number of steps beyond the most parsimonious cladogram required for a given node to collapse. Unnumbered nodes collapse at one step beyond the most parsimonious cladogram. C: Cladogram based on the topology of King (1988 , 1990), but including only the taxa examined in this analysis. Lettered nodes are higher taxa given in King (A = Robertiidae, B = Endothiodontoidea, C = Emydopidae, D = Dicynodontidae, E = Cryptodontinae, F = Aulacephalodontinae, G = Kannemeyeriinae). 59 TABLEt Results of the stratigraphic analysis. SMIG is the sum of Minimum Implied Gaps for each cladogram examined (= the number of steps added by the stratigraphic character in MacClade). RCI is the Relative Completeness Index for each cladogram examined. GER is the Gap Excess Ratio for each cladogram examined. The cladograms from the decay analysis have been broken down by their length, given in parentheses. Because more than one cladogram exists at each of the lengths in the decay analysis, the range ofSMIG, RCI, and GER values is given. Cladogram SMIG Preferred Cladogram 14 King (1988, 1990) 18 Decay Results (126) 13 - 18 Decay Results (127) 12 - 28 Decay Results (128) 11 - 30 Decay Results (129) 11 - 31 and all descendants of their most recent common ancestor (= Dicynodontidae sensu King, 1988, 1990; Figure la, Node K; Figure Ie, Node D) is diagnosed by the presence of paired anterior ridges that do not converge posteriorly (7, 2), a narrow, blade-like mid-ventral vomerine plate (13, 1), a palatal surface of the palatine that is highly rugose and pitted, suggesting a keratinized covering (22, 3), paired nasal bosses near the dorsal or posterodorsal margin of the external nares (23, 2), an expanded humeral head that encroaches on the dorsal surface of the bone (30, 1), and the presence of five sacral vertebrae (36, 2). In addition, this clade is ambiguously diagnosed by an elongate dentary table that is bounded by low ridges (15, 3) and a relatively small lateral dentary shelf (17, 1), although both of these features may characterize a more inclusive clade. The clade including Tropidostoma, Rhachiocephalus, and all descendants of their most recent common ancestor (= Cryptodontinae sensu King, 1988, 1990, although see above; Figure la, Node L; Figure Ie, Node E) is diagnosed by the presence of a postcaniniform crest (28, 1). The clade including Tropidostoma, Oudenodon, and all descendants of their most recent common ancestor (Figure la, Node M) is diagnosed by parietals that are exposed in a groove or depression between the postorbitals, which partially overlap them (20, 1) and a relatively long interpterygoid vacuity that reaches the level of the palatal exposure of the palatines (33, 2). The clade including Pelanomodon, Kannemeyeria, and all descendants of their most recent common ancestor (Figure 1 a, Node N) is diagnosed by the presence of a labial fossa (19, 1). The clade including Pelanomodon, Aulacephalodon, and all descendants of their most recent common ancestor (= Aulacephalodontinae sensu King, 1988, 1990; Figure la, Node 0; Figure Ie, Node F) is diagnosed by a relatively wide intertemporal region in which the postorbitals are steeply placed on the lateral sides of the skull and concave laterally (20, 2) and the presence of a transverse ridge across the snout at the level of the prefrontals (40, 1). The clade including Dicynodon, Kannemeyeria, and all descendants of their most recent common ancestor (Figure 1 a, Node P) is diagnosed by a moderately rugose PALAEONTOlOGIA AFRICANA VOl31 2001 - E RCI GER 0.632 0.863 0.526 0.784 0.658 - 0.526 0.882 - 0.784 0.684 - 0.263 0.902 - 0.588 0.711 - 0.211 0.922 - 0.549 0.711 - 0.184 0.922 - 0.529 and pitted palatal surface of the palatine that is suggestive of the presence of a keratinized covering (22, 4) and a relatively short interpterygoid vacuity that does not reach the level of the palatal surface of the palatines (33, 0). The clade including Lystrosaurus, Kannemeyeria, and all descendants of their most recent common ancestor (Kannemeyeriinae sensu King, 1988, 1990; Figure la, Node Q; Figure Ie, Node G) is diagnosed by a wedge­ shaped symphyseal region ofthe lower jaw (18, 4) and six sacral vertebrae (36, 3). Stratigraphic Analysis The results of the stratigraphic analysis are summarized in Table 1. The most parsimonious cladogram has aMIG value of 14, an RCI value of 0.632 and a GER value of 0.863. The RCI value is above the mean value of 0.498 found by Hitchin & Benton (1997) for continental tetrapods, and comparable to the mean of o. 601 reported by Benton eta!' (1999) for five non-mammalian synapsid cladograms. The GER values are notably higher than the average of 0.654 for the same five non-mammalian synapsid cladograms and the average of 0.765 for continental tetrapods (Benton eta!' 1999). The c1adogram of King (1988, 1990) has a MIG value of 18, an RCI value of 0.526, and GER value of 0.784. Both the most parsimonious cladogram of this analysis and the c1adogram of King were found to fit the fossil record significantly better than random (p < 0.001). As noted above, MIG, RCI, and GER values were also calculated for c1adograms up to four steps longer than the most parsimonious c1adogram (i.e., up to 129 steps) to determine if any of these topologies fit the fossil record better than the most parsimonious c1adogram. MIG values for these c1adograms range from 11 to 31, RCI values range from 0.711 to 0.184, and GER values range from 0.922 to 0.529. Thus, although some of these topologies fit the known stratigraphic record of dicynodonts better than the morphologically most parsimonious c1adogram, most require more and/or longer gaps. Also, the c1adograms that are closest in length to the most parsimonious cladogram have metric scores that are less variable and closer to those of the most parsimonious cladogram than those that are longer (i.e., the 126 step cladograms have MIG values of 13-18, while the 129 step cladograms have MIG values of 60 11-31). The cladograms that do have a closer fit to the fossil record have topologies that are generally congruent with that of the most parsimonious cladogram. Many of the major clades of dicynodonts are resolved in these cladograms (e.g., Figure la, nodes C, H, K, P, and Q), but the relationships among the taxa in these clades varies. Also, the positions of Endothiodon, Chefydontops and Pristerodon tend to be variable in these cladograms. DISCUSSION The topologies of the cladograms recovered in this analysis differ in some respects from those published by past authors, but are similar in others. Because this analysis represents a test of these previous hypotheses of homology and relationship, cases of congruence can be interpreted as corroboration, whereas cases of incongruence represent falsification. However, similarities between the results presented here and those of past workers also may reflect the fact that many of the characters examined are based on features that they used to construct their trees. In addition, the large number of palatal and jaw characters included could be misleading if these areas are prone to homoplasy related to feeding methods and dietary preferences. Future analyses that include a greater diversity of characters should be able to test this phylogenetic signal. In the following discussion, I have limited my comparisons mainly to the cladogram of King (1988, 1990) because only thatcladogram and the cladogramofCluver & King (1983), which has a nearly identical topology, include taxon sampling comparable to that undertaken in this analysis. Although the topological differences are discussed in some detail, it is important to note that the results ofthis analysis are largely compatible with those of Cluver & King (1983) and King (1988, 1990), corroborating many of their results. The preferred cladogram of this analysis reconstructs the clade including Robertia, Diictodon, and their most recent common ancestor (= Robertiidae of King 1988, 1990) in a relati vely basal position, just one node above Eodicynodon (Figure la, Node B). King (1988, 1990) favored a more crown ward position for this clade, nesting it within the clade including Kingoria, Myosaurus, and all descendants of their most recent common ancestor (Figure 1c). Much of the available evidence is equivocal when comparing these two alternatives. For example, the postcranial skeleton of Robertia and DtiCtodon retain several plesiomorphic character states, such as relatively weakly developed humeral and femoral heads (DeFauw 1986; King 1981; Surkov 1998). The postcrania that have been described for Kingoriaand Cistecephalus(Cluver 1978, DeFauw 1986, King 1985) suggest that at least some members of that clade possess more derived postcranial skeletons (although my observations of two undescribed Emydops humeri, SAM K5974, SAM K1 0009, suggest that taxon retained a more primitive humeral morphology). However, members of both clades possess an ectepicondy lar foramen and an ossified cleithrum, features that are otherwise rare in dicynodonts and suggestive of close relationship. Differing interpretations of morphological features are also an issue. King (1988, 1990) considered the notch anterior to the caniniform process of Diictodon and Robertia to be homologous with the embayment of the palatal rim found in taxa such as Emydops and Myosaurus. I have treated these features as separate characters; a notched caniniform process is considered to be a type of caniniform process (Character 6, State 2), whereas the embayment is an unrelated feature (Character 14). My justification for this decision is bas,ed on the fact that in my observations of specimens of DtiCtodon and Robertia, I found little evidence of the pronounced lateral bowing of the palatal rim that forms the embayment in taxa such as Emydopsor Myosaurus. Also, there is not a depression on the medial surface of the palatal rim near the caniniform process as in Kingoria. Instead, the notch seems to be formed by a modification of the anterior edge of the caniniform process, such that the edge originates medial to the palatal rim, as opposed to being contiguous with it. Thus the notch likely represents a transformation of the caniniform process, and therefore a notched caniniform process must be considered a type of (i.e., homologous with) caniniform process. This interpretation of these characters is more consistent with the hypothesis that Diictodon and Robertia are not closely related to taxa such as Kingoria or Emydops. The more basal placement of Diictodon and Robertta also better fits the known fossil record of dicynodonts (Figure 2) and more parsimoniously explains the evolution of characters such as the posterior dentary sulcus, postcaniniform keel, and the morphology of the dentary symphysis. King (1988, 1990) also favored a more basal position for Endothiodon and Chelydontops than I have presented here (Figure la, Node D; Figure 1c, Node B) In some ways (e.g., presence of premaxillary teeth, relatively short premaxillary secondary palate, long, narrow maxillae that lack a caniniform process, midventral plate of vomers with an expanded area posterior to the junction with the premaxilla) the cranial morphology of Endothiodonis highly suggestive of this taxon being part of a very basal dicynodont lineage, perhaps even more basal than Eodicynodon. However, several other features (e.g., relatively narrow intertemporal region, jugal modified for the possible origination of a masseter-like muscle; Cox 1964; Ray 2000), reduced transverse flange of the anterior pterygoid process, quadrate that allows propalinal sliding of the jaw, presence of a posterior dentary sulcus) as well as its stratigraphic occurrence in the Pristerognathus to Cistecephalus Assemblage Zones of South Africa (Rubidge 1995a) suggest that Endothtodon is actually a more advanced, highly autapomorphic dicynodont.lfthe hypothesis that Chefydontops is the sister taxon of Endothtodon is correct (supported in this analysis by two synapomorphies; see Results and below), then the more crownward position of these taxa becomes more logical. Chelydontops is known from the Tapinocephalus Assemblage Zone of South Africa and possesses two synapomorphies (a shelf lateral to the maxillary teeth and a dome-like pineal boss) that unite it with Endothiodon in this analysis (I do not consider Chelydontops to possess premaxillary teeth; see Appendix 1, Character 2). In most other respects, Chelydontops is much more similar to "typical" dicynodonts than to Endothiodon. In particular, the relatively wide intertemporal region, possible evidence for the secondary loss of the anterior palatal ridges (Appendix I, Character 7), and the morphology of the palatines, posterior median palatal ridge, and lower jaw give the two known specimens of Chelydontops a very Pristerodon-like appearance. This superficial similarity may indicate that the lineages including Endothiodon and Chelydontops (Figure I a, Node E) and Pristerodon and Kannemeyeria (Figure la, Node F) evolved from a Pnsterodon-like common ancestor, supporting the more crownward placement of Endothiodon and Chelydontopsproposed here. In addition, this hypothesis suggests that some apparently plesiomorphic features of Endothiodon, such as premaxillary teeth and the unusual morphology of the premaxilla and maxilla, are actually reversals, and that Chelydontops may be a useful model for many features of a hypothetical ancestor of Endothiodon (also see Cox 1998). The more basal position of Pristerodon suggested in this analysis (Figure la, node F) is supported by the retention of several plesiomorphic character states in that taxon, including weakly developed anterior palatal ridges that converge with an expanded area of the posterior median ridge, a relatively wide intertemporal region, a relatively small dentary table, a relatively large lateral dentary shelf, the presence of maxillary and dentary teeth, and the relatively smooth but finely pitted palatal surface of the palatines. King (1988, 1990) favored a more crownward placement (Figure 1c) based on the modification of the posterior dentary sulcus into a "deep, thin-walled sulcus" (King 1988, p. 71). Although I have included only the presence or absence of a dentary sulcus in the data matrix presented here, in my personal observations I have found the dentary sulcus of Pristerodon to more closely resemble that of Chelydontops or Endothiodon than that of taxa such as Tropidostoma. I consider these observations to be consistent with the more basal position suggested by the preferred cladogram, and future analyses that take into account the detailed morphology of the dentary sulcus (not just its presence or absence) can further test this hypothesis. The placement of Pnsterodon suggested by this analysis also has important implications for one other character, the presence of paired anterior ridges that converge posteriorly. When this character is optimized on the preferred cladogram of this analysis, it appears to have evolved independently in Pristerodon and the common ancestor of Robertia and Dti"ctodon. However, this reconstruction is likely to be an artifact of the taxa included in the analysis. In addition to Diictodon, 61 Robertia, and Pnsterodon, several specimens collected in the lower Tapinocephalus Assemblage Zone (e.g., BP/l/5580, BP/l/5589, NM QR3145, NM QR3505; Rubidge, personal communication, 2000) also possess this arrangement of the palatal ridges. In all cases the ridges converge with an expanded, flattened area of the posterior median ridge. In some specimens (e.g., NM QR3145, NM QR 3505) the expanded area of the posterior median ridge is Y -shaped, while in others (e.g., BP/1I5589, Diictodon, Pnsterodon) the expanded area is narrower and V -shaped. Although not included in this analysis, the morphology and stratigraphic occurrence of these specimens suggest that they likely would fall on the branch between Eodicynodon and the clade including Robertia and Dtictodon in the cladogram presented here (Modesto & Rubidge pers. comm.). If this hypothesis is correct, then it is possible that convergent anterior ridges evolved only once, and were later lost in the clades including Chelydontops and Endothiodon (Figure la, Node E) and Kingoria and Myosaurus (Figure la, Node H). These hypothesized losses are corroborated by the fact that both Chelydontops and Emydops possess posterior palatal ridges with an expanded anterior area and an anterior palatal morphology suggestive of being derived from an ancestor that possessed ridges (Appendix 1, Character 7). The parallel anterior ridges in taxa such as Troptdostoma or Dicynodon likely represent a modification of the convergent ridge morphology. The taxa Emydops, Myosaurus, and Cistecephalus are reconstructed as forming a clade (Figure 1 a, Node I) that has the same membership as the Emydopidae of King (1988, 1990; Figure 1c, Node C). However, the relationships among the taxa in the clade presented here are different than those of King; instead of Emydops, Ctstecephalus is reconstructed as the sister taxon of Myosaurus. This pattern of relationship is supported by two synapomorphies in this analysis (the presence of a foramen on the palatal surface of the palatine, and posteromedially extended anterior orbital margins that partially close off the back of the snout). Both of the proposed synapomorphies are quite distinctive, and I have only observed these features in Ctstecephalus, Myosaurus, and Cistecephaloides (which Cluver (1974a) proposed was closely related to Ctstecephalus; a palatal foramen has also been reported in Kawingasaurus, another potential relative of Ctstecephalus (Cox, 1972); also see Appendix 1, Characters 24, 25). In addition, the synapomorphies King (1988, 1990) proposed to support a close relationship between Myosaurus and Emydops to the exclusion of Ctstecephalus (reduced upper teeth, long, straight anterior pterygoid rami, and a shortened basicranial region) also can be found in Ctstecephalus, and are thus diagnostic of a more inclusi ve clade. Cluver (197 4b) also noted the possibility of a close relationship between Cistecephalus and Myosaurus, and suggested that both taxa could be derived from a Myosaurotdes-like ancestor. 62 Troptdostoma, Oudenodon, and Rhachiocephalus form a clade (Figure la, Node L) that has the same membership as the Cryptodontinae of King (1988, 1990; Figure la, Node E), but here again the relationships of the taxa within the clade are different. In this case, Oudenodon and Tropidostoma are reconstructed as being more closely related to each other than either is to Rhachiocephalus, whereas King favored a sister group relationship between Oudenodon and Rhachiocephalus. King (1988, 1990), used the loss of teeth as a synapomorphy to unite Rhachiocephalus and Oudenodon and exclude Troptdostoma. Although Tropidostomadoes possess upper and lower postcanine teeth (some specimens also possess tusks), there are several other features that strongly suggest a close relationship between Troptdostoma and Oudenodon. In this analysis, two synapomorphies unite Oudenodon and Tropidostoma, the presence of a relatively long interpterygoid vacuity that reaches the level of the anterior portion of the palatines and a temporal region in which the postorbitals partially overlap the parietals, with the parietals exposed in a median groove or depression. King (1988, 1990) used the latter feature to diagnose her Tropidostomini, but in my observations this character state also characterizes nearly all Oudenodon specimens, suggesting it is diagnostic for a more inclusive clade. The intertemporal region of Rhachiocephalusis distinctly different because it is narrower and the postorbitals nearly completely overlap the parietals. In addition, the detailed morphology of the nasal bosses also supports the topology presented here. All three taxa possess paired nasal bosses, but the bosses of Tropidostoma and Oudenodon are less elongate and usually centered approximately over the center of the external narial openings. The bosses of Rhachiocephalus tend to be more elongate and ridge­ like, and usually are located posterodorsal to the narial opening. Rhachiocephalus also tends to be larger than either Oudenodon or Troptdostoma, and it possesses a distinct pineal boss, which the other taxa lack. However, it is important to note that the results of the decay analysis show that only one additional morphological step is required to make Rhachiocephalus the sister taxon of Oudenodon. The evolutionary history of the dentition of dicynodonts implied by the preferred cladogram of this analysis also deserves mention because it requires gains and losses of teeth that initially seem somewhat counter-intuitive. Basal anomodonts possess laterally placed teeth on the premaxilla, maxilla, and dentary, whereas many advanced dicynodonts have completely replaced their teeth with a keratinized beak. Several other dicynodonts possess a keratinized beak but also retain some teeth. Some of these taxa (e.g., Robertia, Emydops, Pristerodon, Tropidostoma) have been considered basal members of toothless clades that document the process of tooth loss (Cox 1998; Watson 1948). However, the preferred cladogram of this analysis implies that the presence of teeth in some of these taxa represent reversals to a toothed state from a toothless ancestor. The cladogram presented here suggests that premaxillary teeth were lost in the common ancestor of Eodicynodon and Kannemeyeria (Figure la, Node A), but a reversal later occurred in Endothiodon. This reversal may characterize the lineage including both Endothiodon and Chelydontops (Figure la, Node D), depending on how the palate of Chelydontops is interpreted (Appendix 1, Character 2). I do not consider Chelydontops to possess premaxillary teeth, but the presence of teeth very close to the maxilla/premaxilla suture in that taxon supports the hypothesis that a secondary anterior lengthening of the upper tooth row occurred in this clade. Maxillary teeth were independently lost in Diictodon and the common ancestor of Kingoria and Kannemeyeria (Figure la, Node G). The loss in Diictodon is corroborated by an undescribed, toothed Diictodon-like specimen from the lower Tapinocephalus Assemblage Zone (Rubidge pers . comm., 2000). The loss in the ancestor of Kannemeyeria and Kingoria is more interesting because it implies reversals to a toothed state in Emydops and Troptdostoma. However, these reversals are also required in some optimizations of this character on the cladogram of King (1988, 1990), as is an additional reversal to a toothed state in Robertia. The implied reversal in Robertia is not well supported because the undescribed, toothed Diictodon-like specimen stro~gly suggests that the common ancestor of Diictodon and Robertta was also toothed. A similar pattern is observed for the dentary teeth. The presence of dentary teeth is the basal character state for Anomodontia, and medially-placed dentary teeth characterize the clade including Robertta, Kannemeyeria, and all descendants of their most recent common ancestor (Figure la, Node B). The complete loss of teeth took place independently in Diictodon and the common ancestor of Kingona and Kannemeyeria (Figure la, Node G), and reversals to a toothed state occurred in Emydops and Tropidostoma. Some optimizations of this character on the cladogram of King (1988, 1990) also require a similar pattern of loss and reversal, as well as an additional reversal to a toothed state in Robertta. However, this reversal appears less likely because the undescribed, toothed Diictodon-like specimen strongly suggests that the common ancestor of Diictodon and Robertta possessed a toothed dentary. The patterns of tooth loss and gain in the skull and jaw implied by the preferred cladogram of this analysis suggest that some degree of plasticity existed in the developmental processes responsible for the formation of the dentition in dicynodonts. This perhaps is not surprising given that much of tooth development is the result of epithelium-mesenchyme interactions that can be altered easily by slight modifications in developmental sequences or the behavior of cell populations. In addition, it has been shown that chick epithelium retains the ability to participate in tooth formation (Kollar 1972; Kollar & Fisher 1980), despite coming from a lineage that has been toothless since at least the Early Tertiary. The 63 Age Assemblage Zone U Cynognathus ....... t/.l t/.l ~ S LystrosaurlJs Dicynodon ~ I I:::l .... ~ ~ s:! ~ ~ ~ s:! -% s:! ~ ~ -% s:! I ~ ~ -% ~ ~ s:! ~ s:! ~ {i s:! G '-> ;::s ~ .-.- ~ Q .- I:::l a Q s:! .... ~ ~ -% Ci 2- L- e O()~ ~ .s f .... ~k:I ~ s:! .~ ;::s -% :.... ~ ..... ~ ;::s I:::l ~ ..... ...s:! ..... I:::l e- I:::l s:! ...s:! ...s:! -% e- '-> e-~ '-> .... '-> . 9 ~ ...s:! ~ ..:: .~ I:::l '-> E I:::l ;::s ..g U ...s:! ~ ~ e::: ~ ~ '--- ..g Cistecephalus .... ~ ~ -c.r L- Z Tropidostoma ~ ....... :;8 0:::: ~ Pristerognathu 0.. .s:! ~ 1::: ~ ~ 2-...t:l ~ .... e::: s:! ..g ..Q ~ 6 s:! Tapinocephalus -% ~ Eodicynodon s:! s:! -% G .- ~ I:::l ~ ~ .... ~ :.... ~ ~ ~ s:! ...s:! T ~ ~ I ..... a ; 11 I I Figure 2: Cladogram based on that of Figure 1a showing the stratigraphic ranges of the included taxa. Solid bars indicate known ranges and open bars indicate ghost ranges implied by the cladogram. Thin lines indicate ghost lineages implied by the cladogram. Stratigraphic ranges and assemblage zones are based on Rubidge (l995a). Because Patranomodon and Otsheria are each known from a single specimen their stratigraphic ranges are approximate. The stratigraphic range of Che/ydontops also is not well constrained and therefore approximate. The range of Kingoria is an approximation based on data presented in King (1988). 64 losses and reversals required by this cladogram likely occurred over a much shorter period of time, possibly ten million years or less. It is important to note that slightly less parsimonious cladograms (two morphological steps longer than the most parsimonious cladogram, with the same degree of fit to stratigraphy) do not require the toothed states of Emydops and TroPldostoma to be reversals. The results of the stratigraphic analysis show both the preferred topology and the topology of King (1988, 1990) fit the known fossil record of dicynodonts relatively well. Both cladograms have a significantly better fit to stratigraphy than random, although the lower MIG and higher RCI and GER scores of the preferred cladogram are indicative of its greater consistency. Despite the good general fit, the preferred topology does imply range extensions for several taxa as well as the existence of a number of ghost lineages (Figure 2). This is an important finding because it suggests that the dicynodont fauna of South Africa is not as completely represented as is sometimes suggested (e.g., Cox 1998). It seems likely that some portions of the evolutionary history of the Karoo dicynodonts occurred in a different geographical location or environment where preservation and sampling rates were much lower. This scenario could explain some of the relatively long ghost ranges in Figure 2 (e.g., those associated with Kingoria, Myosaurus, and Kannemeyeda). Alternatively, some of the missing taxa could have been found, but remain unrecognized due to the taxonomic confusion surrounding dicynodonts. Although the RCI has been proposed as an explicit measure of the completeness of the fossil record of a group (Benton & Storrs 1994; Hitchin & Benton 1997a), I have not used it as such because it can be affected by factors such as the accuracy of the cladogram examined, sampling rates, and extinction (Paul 1998; Wagner 1998, 2000). The gaps implied by the preferred cladogram can be tested independently through the use of statistical methods such as those of Foote & Sepkoski (1996), and the results of such a test would be interesting given the productive fossil record of the Karoo Basin and the intensive collecting that has occurred there. The decay analysis shows that some of the slightly less morphologically parsimonious cladograms actually fit the fossil record better than the most parsimonious cladogram. In this paper I have given greater weight to the intrinsic features of the organisms (i.e., morphology) rather than properties partially controlled by extrinsic factors (i.e., stratigraphic occurrence). Thus I have not considered the topologies of the more stratigraphically parsimonious cladograms in detail. Howeve1:' further examination of these cladograms using stratocladistics might be a fruitful area for further research. ACKNOWLEDGEMENTS I wish to thank the curators and collections directors of the AM, AMNH, BP, GSP, NM, PIN, SAM, and TM for their hospitality and assistance during my visits to their institutions. P. A. Holroyd and K. Padian provided access to specimens at the UCMP. B. Rubidge and A. Kurkin provided invaluable advice and assistance in the planning and execution of my trips to South Africa and Russia, and I wish to thank them for their patience. B. Rubidge also allowed access to specimens on loan to him from the BMNH. M. A. Wills assisted in the stratigraphic analysis. A. Aronowsky, J.R. Hutchinson, G. M. King, D. R. Lindberg, B. D. Mishler, S. Modesto, K. Padian, J. F. Parham, and M. A. Wills all read and improved earlier versions of this manuscript. This research was supported by a Collections Usage Grant from the AMNH, a Grant-in-Aid-of­ Research from the Berkeley Chapter of Sigma Xi, and a grant from the Samuel P. Welles Fund of the UCMP. This is UCMP contribution 1736. REFERENCES BARGHUSEN, H. R. 1976. Notes on the adductor jaw musculature of Venjukovia, a primitive anomodont therapsid from the Permian of the U.S.S.R. Annals 0.1 the South African Museum 69, 249-260. BARRY, T. H. 1974a. Affinities and systematic position of the South African Lower Middle Permian dicynodonts (Therapsida: Dicynodontidae). In: Campbell, K. S. W., (ed.), Proceedings and Papers. 3rd fUGS Gondwana Symposium f973, 475-479. Canberra, Australian National University Press. BARRY, T. H .. 1974b. A new dicynodont ancestor from the Upper Ecca (lower Middle Permian) of South Africa. Annals 0.1 the SouthAfrican Museum 64, 117-136. BENTON, M. J. 1994. Palaeontological data and identifying mass extinctions. Trends in Ecology and Evolution 9, 181-185. BENTON, M. J. & HITCHIN, R. 1996. Testing the quality ofthe fossil record by groups and by major habitats. Historical Biology 12, 111- 157. BENTON, M. J., HITCHIN, R & WILLS, M. A. 1999. Assessing congruence between cladistic and stratigraphic data. Systematic Biology 48,581-596. BENTON, M. J. & SIMMS, M. J. 1995. Testing the marine and continental fossil records. Geology 23, 601-604. BENTON, M. J. & STORRS, G. W. 1994. Testing the quality of the fossil record: paleontological knowledge is improving. Geology22, 111- 114. BENTON, M. J. & STORRS, G. W. 1996. Diversity in the past: comparing cladistic phylogenies and stratigraphy. In: Hochberg, M. E., Colbert, J. & Barbault, R, (eds), Aspects 0.1 the Genesis and Maintenance 0.1 Biological Diversity, 19-40. New York, Oxford University Press. BOONSTRA, L. D. 1966. The girdles and limbs of the Dicynodontia of the Tapinocephalus zone. Annals 0.1 the South African Museum 50, 1-11. BRINKMAN, D. 1981. The structure and relationships of the dromasaurs (Reptilia: Therapsida). Breviora 465, 1-34. BREMER, K. 1988. The limits of amino acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42, 795-803. BROOM, R. 1905. On the structure and affinities of the endothiodont reptiles. Transactions 0.1 the South African Philosophical Society 15, 259-282. BROOM, R. 1906. On the Permian and Triassic faunas of South Africa. Geological Magazine 5, 29-30. BROOM, R. 1932. The Mammal-like Reptiles 0.1 South Africa and the Origin 0.1 Mammals. London, H. F. and G. Witherby. CAMP, C. L. 1956. Triassic dicynodont reptiles: part II. Triassic dicynodonts compared. Memoirs 0.1 the University 0.1 California 13,305-341. 65 CAMP, C. L. & WELLES, S. P. 1956. Triassic dicynodont reptiles : part I. The North American genus Placerias. Memoirs 0.1 the Univeristy 0.1 California 13, 255-304. CLUVER, M. A 1971. The cranial morphology of the dicynodont genus Lystrosaurus. Annals 0.1 the South A.frican Museum 56, 155-274. CLUVER, M. A 1974a. The skull and mandible of a new cistecephalid dicynodont. Annals 0.1 the South African Museum 64, 137-155. CLUVER, M. A 1974b. The cranial morphology ofthe LowerTriassic dicynodontMyosaurus gracillis. Annals 0.1 the SouthA.fricanMuseum 66,35-54. CLUVER, M. A. 1975. A new dicynodont from the TapinocephalusZone (Karoo System, Beaufort Series) of South Africa, with evidence of the jaw adductor musculature. Annals 0.1 the South A.frican Museum 67, 7-23. CLUVER, M. A 1978. The skeleton of the mammal-like reptile Cistecephalus with evidence for a fossorial mode of life. Annals 0.1 the South A.frican Museum 76, 213-246. CLUVER, M. A. & HOTTON, N. 1981 . The genera Dicynodon and Diictodon and their bearing on the classification of the Dicynodontia (Reptilia, Therapsida). Annals 0.1 the South A.frican Museum 83, 99-146. CLUVER, M. A. & KING, G. M. 1983. A reassessment of the relationships of the Permian Dicynodontia (Reptilia, Therapsida) and a new classification of dicynodonts. Annals 0.1 the South African Museum 91, 195-273. CLYDE, W. C. & FISHER, D. C. 1997. Comparing the fit of stratigraphic and morphologic data in phylogenetic analysis . Paleobiology 23, 1-19. COX, C. B. 1959. On the anatomy of a new dicynodont genus with evidence of the position of the tympanum. Proceedings 0.1 the Zoological Society 0.1 London 132, 321-367. COX, C. B. 1964. On the palate, dentition, and classification ofthe fossil reptile Endothiodon and related genera. AmericanMuseumNovitates 2171, 1-25. COX, C. B. 1965. New Triassic dicynodonts from South America, their origins and relationships. Philosophical Transactions 0.1 the Royal Society 0.1 London Series B248, 457-516. COX, C. B. 1972. A new digging dicynodont from the Upper Permian of Tanzania. In: I oysey, K. A & Kemp, T. S., (eds), Studies in Vertebrate Evolution, 173-190. Edinburgh, Oliver and Boyd. COX, C. B. 1998. The jaw function and adaptive radiation of the dicynodont mammal-like reptiles of the Karoo basin of South Africa. Zoological Journal 0.1 the Linnean Society 122, 349-384. COX, C. B. & LI, I.-I. 1983. A new dicynodont from East Africa and its classification. Palaeontology 26, 389-406. CROMPTON, A. W. & HOTTON, N. 1967. Functional morphology of the masticatory apparatus of two dicynodonts (Reptilia, Therapsida). Postilla 109, 1-51 . DEFAUW, S. L. 1986. The appendicular skeleton 0.1 A.frican dicynodonts. Unpublished Ph.D. Thesis, Wayne State University, Detroit. EWER, R F. 1961. The anatomy of Daptocephalus leoniceps (Owen) . Proceedings 0.1 the Zoological Society 0.1 London 136, 375-402. FISHER, D. C. 1992. Stratigraphic Parsimony. In: Maddison, W. P. & Maddison, D. R, MacClade (eds), Analysis 0.1 Phylogeny and Character Evolution, Version 3, 124-129. Sunderland, MA, Sinauer Associates. FISHER, D. C. 1994. Stratocladistics: morphological and temporal patterns and their relation to phylogenetic process. In: Grande, L. & Reippel, O. (eds), Interpreting the Hierarchy 0.1 Nature, 133-171. San Diego, Academic Press . FISHER, D. C. 1997. Stratocladistics and hypothesis choice. Journal 0.1 Vertebrate Paleontology 17 (supplement to 3), 46A. FOOTE, M. & SEPKOSKI, I. I. 1999. Absolute measures of the completeness of the fossil record. Nature 398,415-417. FOX, D. L., FISHER, D. C. & LEIGHTON, L. R. 1997. Comparison of stratocladistics and cladistics using simulated evolutionary histories. Journal 0.1 Vertebrate Paleontology 17 (supplement to 3), 47 A. FOX, D. L., FISHER, D. C. & LEIGHTON, L. R. 1999. Reconstructing phylogeny with and without temporal data. Science284, 1816-1819. GAUTHIER, I ., KLUGE, A. G. & ROWE, T. 1988. Amniote phylogeny and the importance of fossils. Cladistics 4, 105-210. GOLUBEV, V. K. 2000. The faunal assemblages of Permian terrestrial vertebrates from Eastern Europe. Paleontological Journal 34, S211-S224. HAUGHTON, S. H. & BRINK, A S. 1954. A bibliographic list of the Reptilia from the Karoo beds of Africa. PaleontologiaA.fricana 2, 1-187. HITCHIN, R. & BENTON, M . I . 1997a. Congruence between parsimony and stratigraphy: comparisons of three indices. Paleobiology 20-32. HITCHIN, R. & BENTON, M. I .1997b. Stratigraphic indices and tree balance. Systematic Biology 46,563-569. HOPSON, I. A. 1991. Systematics of the non-mammalian Synapsida and implications of evolution in synapsids. In: Schultze, H. -P. & Trueb, L., (eds), Origins 0.1 the Higher Groups 0.1 Tetrapods: Controversy and Consensus, 635-693. Ithaca, NY, Comstock Publishing Associates. HOPSON, I. A 1995. Patterns of evolution in the manus and pes of non-mammalian therapsids . Journal 0.1 Vertebrate Paleontology I5,615- 639. HOPSON, I. A& BARGHUSEN, H. R 1986. An analysis of therapsid relationships. In: Hotton, N., MacLean, P. D., Roth, I. I. , & Roth, E. c., (eds), The Ecology and Biology 0.1 Mammal-like Reptiles, 159-168. Washington, D. C , Smithsonian Institution Press. HOTTON, N. 1974. A new dicynodont (Reptilia, Therapsida) from CynognathusZone deposits of South Africa. Annals 0.1 the SouthA.frican Museum 64, 157-165. HUELSENBECK, I. P. 1994. Comparing the stratigraphic record to estimates of phy logeny. Paleobiology 20, 470-483. IVAKHNENKO, M. F. 1996. Primitive anomodonts, venyukoviids from the Late Permian of Eastern Europe. Paleontological Journa/30, 575-~82 KEYSER, A W. 1973. A re-evaluation of the genus Tropidostoma Seeley. Paleontologia Africana 16, 25-35 . KEYSER, A W. 1975. A re-evaluation of the cranial morphology and systematics of some tuskless Anomodontia. Memoirs 0.1 the Geological Survey 0.1 South A.frica 67, 1-110. KEYSER, A W. 1993. A re-evaluation of the smaller Endothiodontidae. Memoirs 0.1 the Geological Survey 0.1 South Africa 82, 1-53. KEYSER, A. W. & CRUICKSHANK, A. R I. 1979. The origins and classification of Triassic dicynodonts. Transactions 0.1 the Geological Society 0.1 South A.frica 82, 81-108. KEYSER, A. W & SMITH, R M. H. 1977-1978. Vertebrate biozonation of the Beaufort Group with special reference to the Western Karoo Basin. Annals 0.1 the Geological Survey 0.1 South A.frica 12, 1-35. KING, G. M. 1981a. The functional anatomy of a Permian dicynodont. Philosophical Transactions 0.1 the Royal Society 0.1 London Series B291,243-322. KING, G. M. 1981 b. The postcranial skeleton of Robertia broomiana, an early dicynodont (Reptilia, Therapsida) from the South African Karoo. Annals 0.1 the South A.frican Museum 84,203-231 . KING, G. M. 1985. The postcranial skeleton of Kingoria nowacki( von Huene) (Therapsida: Dicynodontia). Zoological Journal 0.1 the Linnean Society 84, 263-289. 66 KING, G. M. 1988. Anomodontia. Handbuch der Palaoherpetologie, 17C. Stuttgart, Gustav-Fischer Verlag. KING, G. M. 1990. The Dicynodonts: A Study in Palaeobiology. New York, Chapman and Hall. KING, G. M. 1993. How many species of Diictodon were there? Annals o./the SouthA./rican Museum 102, 303-325. KING, G. M. 1994. The early anomodont Venjukovia and the evolution of the anomodont skull. Journalo./Zoology, London 232,651-673. KING, G. M., RUBIDGE, B. S. & OELOFSEN, B. W. 1989. The evolution of the dicynodont feeding system. Zoological Journal o./the LinneanSociety96,185-21l. KING, G. M. & RUBIDGE, B . S. 1993. A taxonomic revision of small dicynodonts with postcanine teeth. ZoologicalJournal o./the Linnean Society 107, 131-154. KITCHING, J. W . 1977. The distribution of the Karoo vertebrate fauna. Memoirs o./the Bernard Price Institute./or PalaeontologicalResearch 1, 1-13l. KOLLAR, E. J. 1972. The development of the integument: spatial, temporal, and phylogenetic factors. American Zoologist 12, 125-135. KOLLAR, E. J. & FISHER, C. 1980. Tooth induction in chick epithelium: expression of quiescent genes for enamel synthesis. Science207, 993-994. LATIMER, E. M., GOW, C. E & RUBIDGE, B. S. 1995. Dentition and feeding niche of Endothiodon (Synapsida; Anomodontia). PalaeontologiaAfricana32,75-82. LUCAS, S. G. 1996. Vertebrate biochronology of the Mesozoic of China. Memoirs o./the Beijing Natural History Museum 55, 110-149. LUCAS, S. G. 1998. Placerias (Reptilia, Dicynodontia) from the Upper Triassic of the Newark Supergroup, North Carolina, USA, and its biochronological significance. Neues Jahrbuch./iir Geologie und Paliiontologie Monatshqie 1998,432-448. MADDISON, D. R. 1991. The discovery and importance of multiple islands of most-parsimonious trees. Systematic Zoology 40,315-328. MADDISON, W. P. & MADDISON D. R. 1999. MacClade: Analysis 0./ Phylogeny and Character Evolution, Version 3.08a. Sunderland, MA, Sinauer Associates. MODESTO, S. & RUBIDGE, B. S. 2000. A basal anomodont therapsid from the lower Beaufort Group, Upper Permian of South Africa. Journal o./Vertebrate Paleontology 20, 515-521. MODESTO, S. & RUBIDGE,B. S. & WELMAN,J. 1999. The most basal anomodonttherapsid and the primacy of Gondwana in the evolution of the anomodonts. Proceedings o./the Royal Society 0./ London Series B . 266, 331-337. MODESTO, S. & RYBCZYNSKI, N. 2000. The amniote faunas of the Russian Permian: implications for Late Permian terrestrial vertebrate biogeography. In: Benton, M. J., Kurochkin, E. N., Shishkin, M. A., & Unwin, D. M ., (eds), TheAgeo./DinosaursinRussiaandMongolia, 17-34. Cambridge: Cambridge University Press. NORELL, M. A. 1993. Tree-based approaches to understanding history: comments on ranks, rules, and the quality of the fossil record. American Journal 0./ Science 293A, 407 -417. NORELL, M. A. & NOVACEK, M. J. 1992a. The fossil record and evolution. Comparing cladistic and paleontologic evidence for vertebrate history. Science 255,1690-1693. NORELL, M. A. & NOVACEK, M. J. 1992b. Congruence between superpositional and phylogenetic patterns: comparing cladistic patterns with fossil records. Cladistics 8, 319-337. NORELL, M. A. & NOVACEK, M . J. 1997. The ghost dance: a cladistic critique of stratigraphic approaches to paleobiology and phylogeny. Journal 0./ Vertebrate Paleontology 17 (supplement to 3), 67 A. OLSON, E. C. and BYRNE, F. 1938. Osteology of Aulacocephalodon peavoti. Journal o./Geology 46, 177-190. OWEN, R. 1845. Description of certain fossil crania discovered by A. G. Bain, Esq., in the sandstone rocks at the southeastern extremity of Africa, referable to different species of an extinct genus of Reptilia (Dicynodon), and indicative of a new tribe or suborder of Sauria. Transactions o./the Geological Society 0./ London 7,59-84. PAUL, C. R. C. 1998. Adequacy, completeness and the fossil record. In: Donovan, S. K. & Paul, C. R. c., (eds), TheAdequacyo./the Fossil Record, 1-22. New York, John Weily and Sons. PEARSON, H. S. 1924. A dicynodont reptile reconstructed. Proceedings 0./ the Zoological Society 0./ London 1924, 827 -854. RA Y, S. 2000. Endothiodont dicynodonts from the Late Permian Kundaram Formation, India. Palaeontology 43, 375-404. REIPPEL, O. 1997. Falsificationist versus verificationist approaches to history. Journal 0./ Vertebrate Paleontology 17 (supplement to 3) 71A. ROWE, T. & V AN DEN HEEVER, J. A. 1986. The hand of Anteosaurus magnificus (Dinocephalia: Therapsida) and it bearing on the origin of the mammalian manual phalangeal formula. SouthA./rican Journal 0./ Science 82,641-645 . RUBIDGE, B. S. 1984. The cranial morphology and palaeoenvironment of Eodicynodon Barry (Therapsida: Anomodontia). Navorsinge van die Nasionale Museum Bloenifontein 4,325-404. RUBIDGE, B. S. 1985. The first record of a complete snout of the primitive dicynodont Eodicynodon oosthuizeni, Barry 1974 (Therapsida, Dicynodontia. Navorsinge van die Nasionale Museum Bloem./ontein 4,501-512. RUBIDGE, B. S. 1990a. Redescription of the cranial morphology of Eodicynodon oosthut:zeni(Therapsida, Dicynodontia). Navorsinge van die Nasionale Museum Bloenifontein 7, 1-25. RUBIDGE, B. S. 1990b. The cranial morphology of a new species of Eodicynodon (Therapsida, Dicynodontia). Navorsinge van die Nasionale Museum Bloem./ontein 7,29-42.6 RUBIDGE, B. S. (Ed.). 1995a. Biostratigraphy of the Beaufort Group (Karoo Supergroup). South African Committee for Stratigraphy Biostratigraphic Series 1: 1-46. RUBIDGE, B. S. 1995b. Biostratigraphy of the Eodicynodon Assemblage Zone. In: Rubidge, B. S., (ed), Biostratigraphy o./the Beau./ort Group (Karoo Supergroup), 3-7. South African Committee for Stratigraphy Biostratigraphic Series l. RUBIDGE, B. S. & HOPSON, J. A. 1990. A new anomodont therapsid from South Africa and its bearing on the ancestry of Dicynodontia. South African Journal 0./ Science 86,43-45. RUBIDGE, B. S. & HOPSON, J . A. 1996. A primitive anomodont therapsid from the base of the Beaufort Group (Upper Permian) of South Africa. Zoological Journal o./the Linnean Society 117,115-139. RUBIDGE, B. S., JOHNSON, M. R., KITCHING, J. W., SMITH, R. M . H., KEYSER, A. W., & GROENEWALD, G. H. 1995. An introduction to the biozonation of the Beaufort Group. In: Rubidge, B. S., (ed), Biostratigraphy o./the Beau./ort Group (Karoo Supergroup), 1-2. South African Committee for Stratigraphy Biostratigraphic Series 1. RUBIDGE, B. S., KING, G. M. & HANCOX, P. J. 1994. The postcranial skeleton of the earliest dicynodont synapsid Eodicynodon from the Upper Permian of South Africa. Palaeontology 37, 397-408. RYBCZYNSKI, N. 2000. Cranial anatomy and phylogenetic position of Suminia getmanovi, a basal anomodont (Amniota: Therapsida) from the Late Permian of Eastern Europe. Zoological Journal 0./ the Linnean Sociery 130, 329-373. SEELEY, H. G. 1892. Researches on the structure, organization, and classification of the fossil Reptilia. Further observations on Pareiasaurus. Philosophical Transactions of the Royal Society of London Series B 183,311-370. SIDDALL, M. E. 1996. Stratigraphic consistency and the shape of things. Systematic Biology 45, 111-115. SIDDALL, M. E .. 1997. Stratigraphic indices and tree balance: a reply to Hitchin and Benton. Systematic Biology 46, 569-573. SIDOR, C. A. & HOPSON, J. A. 1998. Ghost lineages and "mammal ness": assessing the temporal pattern of character acquisition in the Synapsida. Paleobiology 24, 254-273. SMITH, A. 2000. Stratigraphy in phylogeny reconstruction. Journal of Paleontology 74,763-766. SMITH, A. & LITTLEWOOD, D. T. 1994. Paleontological data and molecular phylogenetic analysis. Paleobiology 20, 259-273. SMITH, R. M. H. & KEYSER, A. W. 1995. Biostratigraphy of the Tapinocephalus Assemblage Zone. In: Rubidge, B. S., (ed), Biostratigraphy of the Beaufort Group (Karoo Supergroup), 8-12. South African Committee for Stratigraphy Biostratigraphic Series 1. STORRS, G. W. 1993. The quality of the Triassic sauropterygian fossil record. Revue de Pateobiologie 7,217-228. STRONG, E. E. & LIPSCOMB, D. 2000. Character coding and inapplicable data. Cladistics 15,363-371. SURKOV, M. V. 1998. Morphological features of the postcranial skeleton in anomodonts reflecting the evolutionary development of the group. Paleontological Journal 32, 620-623. SURKOV, M. V. 2000. On the historical biogeography of Middle Triassic Anomodonts. Paleontological Journal 34, 84-88 . SWOFFORD, D. L. 2000. PAUP*. Phylogenetic Analysis Using Parsimony (*and Other Methods). Version 4.0b4a. Sunderland, Massachusetts, Sinauer Associates. THEWISSEN, J. G. M. 1992. Temporal data in phylogenetic systematics: an example from the mammalian fossil record. Journal of Paleontology66,1-8. TOERIEN, M. J. 1953. Evolution of the palate in some anomodonts and its classificatory significance. Palaeontologia Africana 1, 49-117. TOLLMAN, S. M., GRINE, F. E. & HAHN, B. D. 1980. Ontogeny and sexual dimorphism in Aulacephalodon (Reptilia, Anomodontia) . Annals of the South African Museum 81, 159-186. WAGNER, P. J. 1998. Phylogenetic analyses and the adequacy of the fossil record. In: Donovan, S. K. & Paul, C. R. C., (eds), TheAdequacy of the Fossil Record, 165-188. New York, John Weily and Sons. WAGNER, P. J. 2000. The quality of the fossil record and the accuracy of phylogenetic inferences about sampling and diversity. Systematic Biology 49, 65-86. WALTER, L. R. 1986. The limb posture ofkannemeyeriid dicynodonts: functional and ecological considerations. In: Padian, K. (ed), The Beginning of the Age of Dinosaurs, 89-98. New York, Cambridge University Press. WATSON, D. M. S. 1914. On the nomenclature of the South African pareiasaurians. Annals and Magazine of Natural History 14, 98-102. WATSON, D. M. S .. 1948. Dicynodon and its allies. Proceedings of the Zoological Society of London 188, 823-877 . WATSON, D. M. S. 1960. The anomodont skeleton. Transactions of the Zoolog ical Society of London 29, 131-208. WEISHAMPEL, D. B. & HEINRICH, R. E. 1992. Systematics of Hypsolophodontidae and basal Iguanodontia (Dinosauria: Ornithopoda). Historical Biology 6, 159-184. WILLS, M. A. 1999. Congruence between phylogeny and stratigraphy: randomization tests and the gap excess ratio. Systematic Biology48, 559-580. 68 APPENDIX! Description and discussion of characters used in phylogenetic analysis Morphologic Characters 1) Keel-like extension of the palatal rim posterior to the caniniform process present (1) or absent (0). A postcaniniform keel is found in Emydops, Kingoria, Cistecephalus, and Myosaurus. The holotypes of Kombuisia (BPI 115344) and Cistecephaloides (SAM 6243) also possess this feature, although these taxa have not been included in this analysis. The keel has a rounded ventral edge, a steep-sided, wall-like lateral side, bears teeth in some Emydops specimens, and is formed by a continuation of the lateral palatal rim posterior to the caniniform process. Its height is variable; Kingoria and Emydops tend to have taller keels whereas Cistecephalus and Myosaurus have lower, less distinct keels. The postcaniniform keel is distinct from the postcaniniform crest (Character 28) because the crest has a sharp ventral edge, a smoothly sloping lateral side, and, in well-preserved specimens (e.g., SAM K5227, Oudenodon), can be seen to arise from the posteromedial side of the caniniform process. The character has been coded as "1" for taxa that lack caniniform processes because it is not directly applicable to them. 2) Premaxillary teeth present and located laterally (0), medially (1), or absent (2). Most non-dicynodont anomodonts (sensu Modesto et aL, 1999; Rybczynski 2000) resemble the majority of synapsids by possessing premaxillary teeth that are located near the anterior or lateral margins of the palatal surface of the bone. Galeops may be an exception to this trend (Brinkman 1981), although undescribed specimens tentatively identified as Galeops in· the collections of the South African Museum (SAM 4005, SAM 12261) have clear evidence of premaxillary teeth located near , the margins of the palate. Endothiodon possesses 2 premaxillary teeth, but these are located posteromedially on the premaxillary secondary palate, well away from the margins of the premaxilla. Chefydontops also has been reported as having at least one tooth very near the premaxilla/maxilla suture (Cluver 1975), but my examination of the holotype (SAM 11558) and referred specimen (SAM 12259) has shown the suture to be poorly preserved in both cases. The most anterior tooth may touch the premaxilla/maxilla suture in the holotype, but it is not completely or nearly completely surrounded by the premaxilla as the premaxillary teeth are in Endothiodon, and as a result I have coded Chefydontops as lacking premaxillary teeth. All other dicynodont taxa included in this analysis clearly lack premaxillary teeth. 3) Premaxillae fused (1) or unfused (0). The premaxillae are unfused and sutured together in the majority of synapsid taxa. A suture between the premaxillae is also visible in non-dicynodont anomodonts and in well- preserved specimens of Eodicynodon (e.g., ROZ 1, NM QR 2905, NM QR2989, NM QR2990). The premaxillae are indistinguishably fused in all other described dicynodont taxa. 4) Upper postcanine teeth located near lateral margins of maxilla (0), located more medially, but with more posterior teeth often approaching the lateral margin of maxilla (1), located medially and with teeth a constant distance from the margin of the maxilla (2), or absent (3). The upper postcanine teeth are located near the lateral margin of the maxillae in non-dicynodont anomodonts, an arrangement similar to that of the majority of synapsid taxa. Eodicynodon is unique among dicynodonts in having postcanine teeth placed far laterally. In E oosthuizenithe teeth are in line with the midline of the canine tusk on the posterior surface of the caniniform process, while in E oeloftoni a tusk is absent and a row of teeth is present near the lateral margin of the palate (Rubidge, 1990b). A few specimens of E oosthuizeni also have one or two medially placed postcanine teeth (e.g., ROZ 9, ROZ 11, NM QR2905, NM QR2989), but the nature of this variation has not yet been examined in detail. Because of the relative rarity of these specimens and the distinct lateral placement of at least some of the postcanine teeth visible in most E oosthuizeni and E oeloftoni specimens, I have chosen to code Eodicynodon as having laterally placed teeth. In Pristerodon, Robertia, and Chefydontops the postcanine teeth are found in rows that are located on the medial part of the maxilla anteriorly, but approach the lateral margin ofthe palate more posteriorly. There are minor variations in this pattern (e.g., the "rows" are sometimes poorly formed in Robertla, the exact placement of the teeth is variable in Pristerodon; Keyser, 1993), but this variation does not appear to be of phylogenetic importance. I have given this coding to Emydops and Troptdostoma as well, although these taxa tend not to show the pattern as clearly. Their tooth rows are usually short (i.e., have few teeth) and located towards the posterior and lateral portion of the maxilla, although not as far laterally as in Eodicynodon. They often show a weak medial to lateral trend, and these teeth may be homologous with the more posterior teeth in the other taxa. If this hypothesis is accurate, then the coding used here should not be misleading. The postcanine teeth of Endothiodon are located in medially placed, laterally concave rows. Although curved, these tooth rows remain a constant distance from the lateral margins of the palate because they also are laterally concave. This morphology is unlike the condition in e.g. Pristerodon because there the tooth rows approach the lateral margin of the palate posteriorly. All other taxa included in this analysis lack postcanine teeth. 5) Shelf-like area lateral to the upper postcanine teeth present (1) or absent (0). Endothiodon and Chelydontops possess a flattened or slightly concave, shelf-like area lateral to the upper postcanine teeth (Cluver 1975; Cox 1964). When well preserved, these areas have a somewhat rugose texture suggesting a keratinous covering (e.g., BP/I/1659, Endothiodon, SAM 12259, Chefydontops). In Chefydontops the shelf is raised slightly above the level of the more anterior portion of the secondary palate, whereas in Endothiodon it is closer to the level of the anterior portion of the palate. The shelf is relatively longer in Endothiodon, but this may be related to the relatively longer tooth rows present 69 in that taxon. The non-dicynodont anomodonts and other toothed dicynodonts included in this analysis do not possess a shelf lateral to the upper tooth rows. The toothless dicynodonts have been coded as "?" because the character is not directly applicable to them. 6) Caniniform process absent (0), present (1), present with notch anterior to it (2). In non-dicynodont anomodonts and Endothiodon the maxillary rim of the palate is not extended ventrally into a caniniform process. In his description of Chelydontops, Cluver (1975) reported that a caniniform process was absent. However, a caniniform process is present in SAM 12259, albeit weakly developed. The majority of dicynodonts (including Chelydontops) possess a caniniform process that is a smooth continuation of the palatal rim. The detailed morphology of the process in varies among taxa (compare e.g., Placerias, Kannemeyeria, Emydops, Oudenodon, and Cistecephalus). This variation is likely related to factors such as function and presence or absence of a canine tusk, and I have not distinguished the variants in my codings. In Dtictodon and Robertia a distinctive notch in the palatal rim is present so that the anterior edge of the caniniform process is set off from the palatal rim and meets the palatal surface of the maxilla medial to the rim. Well-preserved specimens of Eodicynodon posses a notch in the palatal rim anterior to the canine tusk, but this is not homologous with the condition in Diictodon or Robertta because the notch is within the caniniform process and the anterior edge of the process is not set off from the palatal rim. 7) Paired anterior ridges on premaxilla absent (0), present but converge posteriorly (l), or present and do not converge (2) . Non-dicynodont anomodonts and Eodicynodon do not possess paired anterior ridges on the palatal surface of the premaxilla. Several undescribed dicynodont specimens collected in the lower Tapinocephalus Assemblage Zone (e.g., NM QR3145, NM QR3505; Rubidge, personal communication, 2000) possess weakly-developed, paired anterior ridges that converge with a Y -shaped, expanded, flattened anterior portion of the median palatal ridge. A similar condition is found in Robertla, Dlictodon, and Pristerodon, except that the ridges are more weakly convergent and the expanded, flattened portion of the posterior median ridge is narrower and V -shaped. I did not distinguish between these morphologies in my coding of this character because the undescribed specimens were not included in this analysis. TroPldostoma, Oudenodon, Rhachiocephalus, Aulacephalodon, Dicynodon, Lystrosaurus, and Kannemeyena all possess well-developed anterior ridges that are roughly parallel, although some individual specimens of these taxa (e.g., SAM K1496, Lystrosaurus) show slight convergence of the ridges. A variably developed depression is often present between the ridges in these taxa. Chelydontops, Endothiodon, Emydops, Kingona, Cistecephalus, Myosaurus, and Pelanomodon lack distinct anterior palatal ridges. Some specimens of these taxa (e.g., BPIl/3858 and BP1l1l562, Kingoria; SAM 11060 and SAM K6623, Emydops; SAM 11558, Chelydontops) bear a depression at the front of the mouth that has slightly raised lateral edges that are suggestive of highly reduced ridges, indicating that the ridgeless condition in these and closely related taxa may represent a secondary loss. Pelanomodon also lacks anterior ridges (Cluver and King, 1983) and the anterior area of the premaxillary secondary palate of specimen GSP AF9183 is smooth and flat. Anterior palatal ridges are absent in Endothiodon, and the premaxillary secondary palate is unique in being highly vaulted (Cox, 1964; Latimer et aI., 1995; Ray, 2000). 8) Posterior median ridge on premaxilla absent (0), present with a flattened, expanded anterior area (1), or present without a flattened, expanded anterior area (2). Non-dicynodont anomodonts do not possess a posterior median ridge on the palatal surface of the premaxilla. A wide, thick median ridge that flattens and fans out anteriorly is present in Eodicynodon. The ridge is thinner and more blade-like in most other dicynodont taxa. Robertla, Dtictodon, Pristerodon, Emydops, and several undescribed specimens collected in the lower Tapinocephalus Assemblage Zone (e.g., NM QR3145, NM QR3505; Rubidge, personal communication, 2000) possess a thinner median ridge that flattens and expands into a slightly raised, Y -shaped or V -shaped area that converges with the anterior ridges in some cases (see Character 7). The median ridge of Chelydontops also flattens and expands anteriorly, but the expanded area is diamond-shaped. Most of the other taxa included in this analysis possess a thin median ridge that becomes lower anteriorly but lacks a distinct expanded area. The median ridge is relatively weakly developed in Kingona, and unusually modified so that it is T-shaped in cross-section in Myosaurus (Cluver, 1974b). For simplicity, I have distinguished only between taxa that possess an expanded anterior portion of the median ridge and those that do not when the ridge is present. This distinction should be appropriate for the level of taxon sampling attempted in this analysis, although further subdivision of this character may be desirable in future analyses that include a greater number of taxa. 9) Palatal surface of premaxilla with well-defined depressions with curved sides lateral to median ridge (0), with groove-like depressions that have straight sides and a rounded anterior end (1), or relatively flat with poorly defined or no depressions present (2). Non-dicynodont anomodonts do not possess a median palatal ridge, and those included in this analysis (Patranomodon and Otshena) have been coded as "?" because this character is not directly applicable to them. Eodicynodon possesses a distinct depression in the premaxillary secondary palate on each side of the posterior median ridge. The depressions are roughly oval-shaped with the long axis of the oval trending antero-posteriorly. The sides of the depressions are curved and the anterior and posterior ends are slightly pointed. Similar depressions are found in Diictodon and Robertla, although the depressions are slightly shallower in Robertia than in the other two taxa. Emydops, Cistecephalus, and Myosaurus possess more groove-like depressions lateral to the median ridge that have straight 70 edges, a rounded anterior end, and an open posterior end. Myosaurus is unique in having thin, channel-like depressions that are partially enclosed by the crossbar of the T -shaped median ridge located between the median ridge and the depressions in the palatal surface. Cluver (1974b) compared these channels to features found on the dorsal surface of -the premaxilla in Lystrosaurus that may have been associated with a vomeronasal organ. I have given Myosaurus the same coding as Emydops and Cistecepha!us because all three taxa possess groove-like palatal depressions, and the channels medial to them in Myosaurus seem to be unrelated, autapomorphic features. The rest of the taxa in this analysis have do not possess depressions set off from the rest of the palatal surface by distinct margins. Instead, the palatal surface lateral to the median ridge is flat or slightly concave and smoothly merges with the palatal rim. The exact morphology of the area seems to be variable both among and within these taxa and is susceptible to alteration by plastic deformation of the skull. I have included this series of morphologies under a single character state because they all can be distinguished easily from the other two states and the variability among them appears to be random. 10) Lower teeth present on dorsal surface of dentaries (0), present on a medial swelling or shelf (1), or absent (2). The lower teeth of all non-dicynodont anomodonts for which jaws are known are located on the dorsal surface of the dentaries. In Patranomodon and Suminia the teeth are located slightly medially on the dorsal surface, and the dentary forms a narrow rim just lateral to base of the tooth crowns. The dorsal surface of the jaw has been expanded laterally to form a wide, flat surface in U!emica, especially in large specimens (Ivakhnenko, 1996). The teeth are located on the medial side of this surface, but still clearly on the dorsal surface of the jaw. I have not distinguished between these variants in coding this character because only Patranomodon is included in this analysis. A lower jaw is not known for Otsheria, although Ivakhnenko (1996) noted that it may have been very similar to the jaw specimens assigned to Venyukovia. Eodicynodon also has teeth located towards the medial side of the dorsal dentary surface (Rubidge, 1984, 1990a, 1990b). The other toothed dicynodonts included in this analysis all have dentary teeth located on some type of medial shelf or swelling. Robertia and an undescribed, toothed, Dtictodon-like specimen collected in the lower Tapinocepha!us Assemblage Zone (BP/1/5589; Rubidge, personal communication, 2000) possess a row of teeth on the blade-like ridge that forms the medial boundary of dentary table. Endothiodon and Che!ydontops have teeth located on a swelling of the dentary ramus that is medial and slightly below the level of the posterior dentary sulcus, although the condition is more strongly developed in Endothiodon. Emydops and Pristerodon possess teeth on a raised swelling or shelf that is medial to the posterior dentary sulcus. In Troptdostoma the dentary teeth are located on a raised swelling that is medial to the junction of the dentary table and posterior dentary sulcus. I have not distinguished between these variations in coding this character because the differences between the variants are relatively minor. All other taxa in this analysis lack dentary teeth. I have coded Pe!anomodon as "?" because no jaw that is definitely referable to this taxon 'is currently known (Cluver and King, 1983). 11) Vomers fused (1) or unfused (0) . Most non-dicynodont anomodonts and Eodicynodon resemble many other basal synapsids in possessing a clear midline suture between the vomers. Although the mid-ventral plate of the vomers is poorly preserved in the only known specimen of Otsheria (PIN 1758/5), careful scrutiny has uncovered no indication of a midline suture. The vomers may also be fused for at least part of their length in Suminia (Rybczynski 2000). All other dicynodont taxa included in this analysis possess vomers that are indistinguishably fused. 12) Mid-ventral plate of vomers with an expanded area posterior to junction with premaxilla (0) or with out a notably expanded area posterior to junction with premaxilla (1). The palatal exposure of the mid-ventral plate of the vomers of Patranomodon, Eodicynodon, Robertia, an undescribed Diictodon-like specimen collected in the lower Tapinocepha!us Assemblage Zone (BP/1/55