Cladistics Cladistics 26 (2010) 456–481 10.1111/j.1096-0031.2009.00298.x Phylogeny of the large extinct South American Canids (Mammalia, Carnivora, Canidae) using a ‘‘total evidence’’ approach Francisco J. Prevosti* División Mastozoologı´a, Museo Argentino de Ciencias Naturales ‘‘Bernardino Rivadavia’’—CONICET, Av. Angel Gallardo 470, C1405DJR Buenos Aires, Argentina Accepted 6 October 2009 Abstract South America currently possesses a high diversity of canids, comprising mainly small to medium-sized omnivorous species, but in the Pleistocene there were large hypercarnivorous taxa that were assigned to Protocyon spp., Theriodictis spp., Canis gezi, Canis nehringi and Canis dirus. These fossils have never been included in phylogenies based on quantitative cladistics, but handconstructed cladograms published in the 1980s included some of them in the South American canine clade and others in the Canis clade. In this work, the phylogenetic position of the large extinct South American canids was studied using a large sample of living and extinct canids, as well as different sources of characters (e.g. DNA and 133 osteological characters). The phylogenetic analysis corroborates the inclusion of Theriodictis and Protocyon in the ‘‘South American clade’’, where C. gezi is also included. In addition, the position of C. dirus as a highly derived Canis species is confirmed. The simultaneous analysis supports hypercarnivory having arisen at least three times in Caninae and once in the ‘‘South American clade’’. The combination of the phylogenetic analyses, the fossil record and divergence dates estimated in previous works suggests that at least three or four independent lineages of the ‘‘South American clade’’ invaded South America after the establishment of the Panama bridge around 3 million years ago, plus other events corresponding to the immigration of Urocyon and Canis dirus.  The Willi Hennig Society 2009. The family Canidae is a very conspicuous group of the continental vertebrate communities of all continents in the present and includes social hypercarnivorous forms (e.g. Lycaon and Canis lupus), omnivorous species (e.g. Cerdocyon thous) and other that are mainly insectivores (e.g. Otocyon megalotis) (Sillero Zubiri et al., 2004). Living canids comprise 13 genera and around 32 species, all included in the subfamily Caninae (Tedford et al., 1995; Wozencraft, 2005). The fossil record shows that the present diversity is a tiny portion of that of the past and that the family appeared in the late Eocene (37– 36 Myr bp) in North America (Wang, 1994; Wang et al., 1999, 2004a,b, 2008). The Caninae invaded the Old World in the Late Miocene (ca. 10 Myr bp) and arrived *Corresponding author: E-mail address: protocyon@hotmail.com  The Willi Hennig Society 2009 in South America in the Late Pliocene (ca. 4– 2.5 Myr bp) (Berta, 1989; see Flynn, 1991 regarding the date of this publication; Wang et al., 2004a,b, 2008; Prevosti, 2006; Prevosti et al., 2009). The migration of Caninae to South America generated a great diversity of species, represented by more than 10 living species, mainly small ⁄ medium omnivorous canids, and several extinct species that included large hypercarnivorous forms (Berta, 1987, 1989; Sillero Zubiri et al., 2004; Prevosti, 2006; Soibelzon and Prevosti, 2007). The large hypercarnivorous taxa were particularly well represented in the past (Van Valkenburgh, 1991), with two endemic genera (Theriodictis and Protocyon) and three species that have been included in Canis (Kraglievich, 1928; Berta, 1989; Berman, 1994; but see Prevosti, 2006). Recent morphological, molecular and combined phylogenetic analyses indicate that recent South American F.J. Prevosti / Cladistics 26 (2010) 456–481 canids (except Urocyon cinereoargenteus) constitute a monophyletic group, sister of the Canis clade (Canis + Cuon + Lycaon; Tedford et al., 1995; Wayne et al., 1997; Zrzavý and Řičánková, 2004; Bardeleben et al., 2005; Lindblad-Toh et al., 2005), but none of them analysed the position of the fossil taxa. The first analyses where Theriodictis, Protocyon and the South American species of Canis (C. dirus, C. nehringi, C. gezi) were included are those by Berta (1981, 1989); see also Berta, 1984, 1987), one of the first researchers who applied the cladistic methodology in the group (see also Tedford, 1976; Langguth, 1980). Berta used a manual cladistic approach, based on an unpublished manuscript by Tedford and Taylor (Berta, 1981; Tedford et al., 1995). She found that Theriodictis and Protocyon were part of the ‘‘South American clade’’, with Cerdocyon, Speothos, Dusicyon and Pseudalopex, and that C. dirus, C. nehringi, C. gezi and Chrysocyon brachyurus were part of the Canis clade (Berta, 1981, 1987, 1989). Molecular evidence and some North American fossils suggest that the immigration of canids to South America occurred through several independent events (Wayne et al., 1997; Wang et al., 2004a,b, 2008). This information suggests that Chrysocyon, Speothos and probably two lineages of South American foxes, plus Urocyon and Canis, invaded this continent after the elevation of the Panama bridge 4–2.5 Ma (DuqueCaro, 1990; Coates and Obando, 1996; Coates et al., 2004; Prevosti and Rincón, 2007). The assignment of fossils from the Pliocene of North America to South American genera is very important to this biogeographical discussion. In 1981, Torres Roldán and Ferusquı́a Villafranca (1981); see also Torres Roldán, 1980) described a new species of Cerdocyon (C. avius) for the early Blancan of Baja California (Mexico). Other specimens found in the Pliocene–Middle Pleistocene of the USA were described recently as new species of Cerdocyon, Chrysocyon and Theriodictis by Tedford et al. (2009; see also Berta, 1989; Wang et al., 2004a,b, 2008). These specimens are very fragmentary (e.g. an incomplete skull, a mandible, isolated teeth, or an incomplete postcranial skeleton) and have never been included in a cladistic analysis. The objectives of this contribution were: to analyse the phylogenetic position of the large fossil canids of South America using maximum parsimony (MP) and including a wide sampling of taxa and characters (mitochondrial and nuclear DNA, osteology and other sources); to test the phylogenetic position of the fossils from North America that have been assigned to South American taxa; to explore the congruence and phylogenetic signal of different sources of characters; and to discuss the North and South American interchange of canine lineages based on the results obtained and published information. 457 Materials and methods Taxon and character sampling I included all the recent South American canids (Speothos venaticus, Chrysocyon brachyurus, Cerdocyon thous, Atelocynus microtis, Dusicyon australis, Urocyon cinereoargenteus, Lycalopex griseus, L. gymnocercus, L. fulvipes, L. sechurae, L. culpaeus, L. vetulus), three species of Vulpes (V. vulpes, V. lagopus, V. zerda) and the living species of the genera Canis, Nyctereutes, Lycaon, Cuon and Otocyon. The South American fossil Caninae Theriodictis platensis, T. tarijensis, Protocyon troglodytes, P. scagliorum, Canis gezi, C. nehringi, C. dirus and Dusicyon avus, plus the North American C. armbrusteri, Eucyon davisi and Leptocyon vafer, were also included. Finally, two Hesperocyoninae (Hesperocyon gregarius and Mesocyon coryphaeus) and two Borophaginae (Archaeocyon leptodus and Phlaocyon leucosteus) were included as outgroups (see Wang, 1994; Wang et al., 1999 for the phylogenetic relationships of the three Canidae subfamilies), resulting in a total of 41 taxa. I coded 133 characters from the skull, dentition and postcranial skeleton (osteological partition; characters 0–132; Appendix 1; Appendix S2 in Supporting Information). Most came from the reanalysis of several publications (Gaspard, 1964; Stain, 1975; CluttonBrock et al., 1976; Berta, 1989; Munthe, 1989; Wang, 1993, 1994; Tedford et al., 1995; Wang et al., 1999), but new characters were also included. The anatomical terms and nomenclature used in this contribution are mainly those proposed by Evans (1993), with some modifications following several works on canid systematic and morphology (Hildebrand, 1954; Stain, 1975; Berta, 1989; Munthe, 1989; Wang, 1993, 1994; Tedford et al., 1995; Wang et al., 1999). In addition, I included 105 characters (characters 133–237; miscellaneous partition) of soft anatomy (fur, number of mammary glands, visceral morphology), behaviour, natural history and chromosomes as defined and coded by Zrzavý and Řičánková (2004); see also Prevosti, 2006; see Appendix S3 for the complete matrix). The following abbreviations are used for upper ⁄ lower incisors, premolars and molars, respectively: I ⁄ i, P ⁄ p, M ⁄ m. I downloaded sequences of 22 nuclear genes used in recent phylogenies by Wayne et al. (1997), Bardeleben et al. (2005) and Lindblad-Toh et al. (2005) from GenBank (http://www.ncbi.nlm.nih.gov): APOBE29S1 (characters 238–247); APOBE29S2 (characters 248– 254); BDNF (character 255); BRAC1S1 (characters 256–280); BRAC1S2 (characters 281–290); CH14 (characters 291–343); CH21 (characters 344–369); CH24 (characters 370–417); CHRNA1 (characters 418–424); CHST12 (characters 425–426); CKMOR1 (characters 427–429); CYPIA (characters 430–442); FES (characters 458 F.J. Prevosti / Cladistics 26 (2010) 456–481 443–462); FGFR (characters 463–486); GHR exon 9 (characters 487–489); GHR exon 10 (characters 490– 510); RAG1 (characters 511–530); TMEM20 (characters 531–575); TRSP(characters 576–617); VANGL2 (characters 618–624); VTN (characters 625–652); VWF (characters 653–667) (Appendix S3). Also, three partitions of mitochondrial DNA published by Wayne et al. (1997) and Bardeleben et al. (2005) were used: COI (characters 668–847); COII (characters 848–1030); cytochrome b (characters 1031–1248) (Appendix S4). Character analysis The characters and states were identified using the ‘‘tests’’ to define primary homologies (similarity and conjunction; Patterson, 1982; de Pinna, 1991; Brower and Schawaroch, 1996; Hawkins et al., 1997; Rieppel and Kearney, 2002; Freudenstein, 2005; De Laet, 2005). The characters were defined using a reductive approach (Hawkins et al., 1997; Lee and Bryant, 1999; Strong and Lipscomb, 1999) focusing on their logical independency (Wilkinson, 1995a). This type of codification preserves the logic between the presence ⁄ absence of a structure and their attributes (e.g. colours, textures, morphology) (Fitzhugh, 2006). Some of the methodological problems that affect this approach (Maddison, 1993; Strong and Lipscomb, 1999) can be alleviated or corrected using the branch-collapsing rule number 3 (minimal branch length = 0; see Coddington and Scharff, 1994), the rule utilized in this work. Some authors have suggested recently that the reductive coding could be problematic in some cases where ‘‘intraorganismal homologies’’ are present (Harris et al., 2003). Of a total of 119 discrete osteological characters, only three sets could be coded as reductive or composite (character 24 vs. 25, 73 vs. 74–77 and 89 vs. 90), but some of them were not parsimoniously informative (characters 73–77, 90). Thus the choice of reductive or composite codification was not very relevant in this case. Several studies have suggested that developmental processes and functional roles could impose constraints in the morphology generating lack of independence between characters (Van Valkenburgh, 1991; Jernvall, 1995; Biknevicius and Van Valkenburgh, 1996; Polly, 1998; Zhao et al., 2000; Hlusko, 2004; Kangas et al., 2004; Goswami, 2006; Koepfli et al., 2007; Springer et al., 2007). Serial homologous structures (e.g. presence ⁄ absence of metaconule in M1 and M2) are only a special case. I did not discard this information, but instead I tried not to bias the character analysis with these suppositions (cf. Luckow and Bruneau, 1997). For that reason, the distribution of the states of the characters suspected of lack of independence due to a developmental or functional process were compared, and only if two or more of these characters had the same distribution, either they were combined or one of them was deleted (cf. Pol and Gasparini, 2009). Their optimizations were also compared a posteriori on trees obtained from different analyses. For these comparisons and other analyses of the characters, only unambiguous optimizations were considered in this work. Fourteen characters were coded as continuous (Goloboff et al., 2006) using the range of values observed as states for each taxon (Appendix 2). Continuous characters were indexes of linear measurements taken with a digital calliper from the specimens studied (Appendix 1; Prevosti, 2006). These ratios represented proportions (‘‘shape’’) of different structures, and were specially constructed in order to avoid allometric patterns. Several of the discrete characters used were in fact of continuous variation, but it was not possible to quantify them to be used as continuous. Multistate discrete characters were coded as nonadditive. The gaps of DNA characters were coded as a fifth state in the static searches (see below) to preserve their phylogenetic information. Some authors have shown that polymorphic characters have a phylogenetic signal and that they are useful to resolve phylogenies (Wiens and Servedio, 1997; Kornet and Turner, 1999; Wiens, 2000). In this work, polymorphisms were coded as such (coded as 0–1 if both states were represented by different specimens of a taxon; Campbell and Frost, 1993). With this method, the most parsimonious state is chosen from among the options for polymorphic character ⁄ taxa during the parsimony searches. On the other hand, the samples did not have an adequate size to use frequency of states as information (Wiens and Servedio, 1997; Wiens, 2000). Several characters and taxa had an elevated number of missing data, but they were not excluded from the analyses because, as shown by several authors (Kearney, 2002; Kearney and Clark, 2003; Wiens, 2003a,b, 2006; Wilkinson, 2003), these kinds of character ⁄ taxon could be useful in phylogenetic analyses. To explore the effect of missing data and detect ‘‘wild card’’ or unstable taxa, I utilized a strict reduced consensus (Wilkinson, 1995b, 2003). In the static approximation, the sequences were aligned using the program Dialign at BiBiServ (http:// bibiserv.techfak.uni-bielefeld.de/dialign; Morgenstern, 2004). The resulting alignments were adjusted manually, as conservatively as possible. After that, searches were performed with TNT 1.1 (Goloboff et al., 2008a). In addition, analyses using dynamic homologies (and consequently unaligned matrices), as implemented in POY (Varón et al., 2009), were also performed (vide infra). Phylogenetic analysis Maximum parsimony analyses were conducted using equal and implied weights (Goloboff, 1993). For the F.J. Prevosti / Cladistics 26 (2010) 456–481 implied weighting-based searches, a wide range of the concavity constant k was used (between 1 and 100). Implied weighting was used in the analysis of the osteological matrix and the simultaneous analysis. Homoplasy weighting tends not to be used in molecular analyses since Källersjö et al. (1999) showed the effects of eliminating ‘‘third position characters’’, which reduce the number of well supported clades and average jackknife resampling frequencies; but recent studies by Goloboff et al. (2008b) demonstrate that the use of implied weighting can improve the results (topology and support) in morphological matrixes as well as in molecular analyses, if the functions are well rescaled. The relative performance of implied related to equal weighting was not explored in combined analysis, but I explored how the use of different concavity values (strength of weighting against homoplasy) affects the results obtained with the complete data set. The heuristic searches were performed with 1000 series of random addition sequences (RAS), swapping the trees with tree bisection–reconnection (TBR), plus an additional rearrangement of all the most parsimonious trees found using TBR. Branch support was evaluated using symmetric resampling, which is not distorted by weighting the characters (Goloboff et al., 2003), and expressed by group frequencies (SR), and group differences (GC), which give an idea of the contradiction between the characters (Goloboff et al., 2003). Both measures of support were calculated by performing 1000 pseudoreplicates, each consisting of 10 RAS. For dynamic approximation with the direct optimization method (Wheeler, 1996, 2003) I used the software POY4 ver. 4.1.1 (Varón et al., 2009). Prior to the analyses, sequences were partitioned into regions in the conserved domains. This allows a more efficient heuristic analysis to be performed in POY, and in the case of missing fragments in the sequences, it helps avoiding wrong alignments due to the missing data. Searches were performed using the following costs for substitution, indels and gap opening: 1 : 1 : 1. In all cases, searches were performed by building 200 Wagner trees, swapping them retaining five trees per round and swapping trees 10% longer than the local optimum (‘‘threshold: 10’’), plus tree fusing and ratchet (commands fuse and perturb). During swapping, regions with few indels were treated as static characters using the command ‘‘auto_static_approx’’. The supports were estimated using the Bremer support index with the command calculate_support (Varón et al., 2009). Trees were rooted with H. gregarius in the morphological and simultaneous analyses and with U. cinereoargenteus in the DNA analyses—one of the most basal Canidae and living Caninae, respectively (Wang, 1994, Wang et al., 1999; Bardeleben et al., 2005; LindbladToh et al., 2005). The information present in the most 459 parsimonious trees was condensed using strict consensus trees (Bremer, 1990). In the static approximation I performed ‘‘partitioned’’ analyses to explore the signal of different types and sources of characters (continuous partition: characters 0–13; cranial: 14–53, 96–97; ‘‘postcranial’’: 98–132; osteological discrete: 14–132; osteological: 0–132; miscellaneous: 133–237; nuclear DNA: 238–667; mitochondrial DNA: 668–1248; DNA: 238–1248). Finally, this information was combined in a simultaneous analysis (Kluge, 1989; Nixon and Carpenter, 1996; Lecointre and Deleporte, 2004). The congruence of the trees obtained from these partitions was measured using three approaches: subtree pruning and regrafting (SPR) distances (Goloboff, 2007); nodes shared between strict consensuses (or the most parsimonious tree) of each partition; and the consensus fork index (CFI), which consists of the number of shared nodes divided by the number of possible nodes (Colless, 1980). For these comparisons, some nuclear genes (APOBE29S1, BRAC1S2, BNDF, CHRNA1, CHST12, CMKOR1, GHR exon 9, VANGL2) were combined in one partition because they had few informative characters (< 12). The Mann–Whitney test was used to evaluate differences of missing and polymorphic entries within taxa and characters, and SpearmanÕs R correlation index was used to test the relationship between percentage of polymorphic entries and number of specimens studied for each taxon (Zar, 1984). Results Characteristics of the osteological matrix The matrix obtained had 12.65% missing cells, mostly concentrated in the fossil taxa. With the exception of L. fulvipes and D. australis, the recent species had < 7% missing data (mean: 4.08%), while fossils had generally more than 10% (mean: 27.72%), although some possessed levels similar to those of recent species (C. dirus, T. platensis, D. avus, Hesperocyon: 3.76– 6.05%) (Table 1). The missing data were clearly not randomly distributed, and were caused mainly by the absence of known postcranial skeleton for several fossils and the living taxa L. fulvipes and D. australis. In those taxa, the level of missing entries was between 30 and 54%. Dental and cranial characters had similar percentages of missing data (2.18 and 4.24%, respectively), but postcranial characters possessed higher values (27.25%). These differences were significant under the Mann–Whitney test (cranial versus dental: U = 808, P = 0.012; cranial versus postcranial: U = 168.50, P < 0.000001; dental versus postcranial: U = 89, P < 0.000001). 460 F.J. Prevosti / Cladistics 26 (2010) 456–481 Table 1 Percentage of missing entries per taxa and discrete osteological partition (number of characters per partition in parentheses) Taxa Type* Continuous (14) Dental (42) Cranial (42) Postcranial (35) Archaeocyon leptodus Atelocynus microtis Canis adustus Canis armbrusteri Canis aureus Canis dirus Canis gezi Canis latrans Canis lupus Canis mesomelas Canis nehringi Canis rufus Canis simensis Cerdocyon thous Chrysocyon brachyurus Cuon alpinus Dusicyon australis Dusicyon avus Eucyon davisi Hesperocyon Leptocyon vafer Lycalopex culpaeus Lycalopex fulvipes Lycalopex griseus Lycalopex gymnocercus Lycalopex sechurae Lycalopex vetulus Lycaon pictus Mesocyon coryphaeus Nyctereutes procyonoides Otocyon megalotis Phlaocyon leucosteus Protocyon scagliorum Protocyon troglodytes Speothos venaticus Theriodictis platensis Theriodictis tarijensis Urocyon cinereoargenteus Vulpes lagopus Vulpes vulpes Vulpes zerda Total f r r f r f f r r r f r r r r r r f f f f r r r r r r r f r r f f f r f f r r r r 57.143 0 0 21.429 0 0 50.000 0 0 0 21.429 14.286 0 0 0 0 57.143 0 28.571 0 50.000 0 7.143 0 0 0 0 0 42.857 0 0 7.143 57.143 57.143 0 0 92.857 0 0 0 0 13.763 7.143 0 0 11.905 0 0 7.143 0 0 0 19.048 0 0 0 0 2.381 7.143 0 11.905 0 21.429 0 9.524 0 0 0 0 2.381 2.381 0 0 14.286 19.048 7.143 16.667 0 19.048 0 0 0 0 2.178 16.667 0 0 0 0 2.381 35.714 0 0 0 19.048 0 0 0 0 2.381 9.524 14.286 19.048 9.524 26.190 0 9.524 0 0 0 0 0 11.905 0 4.762 19.048 28.571 21.429 0 4.762 71.429 7.143 4.762 4.762 4.762 4.239 20 0 5.714 100 2.857 2.857 91.429 0 0 2.857 100 2.857 5.714 0 2.857 2.857 100 5.714 100 8.571 100 2.857 100 2.857 2.857 0 20 0 45.714 0 2.857 20 100 51.429 0 8.571 100 0 2.857 0 2.857 27.247 Total (133) 18.797 0 1.504 32.331 0.752 1.504 42.857 0 0 0.752 40.602 2.256 1.504 0 0.752 2.256 37.594 6.015 39.098 5.263 46.617 0.752 33.083 0.752 0.752 0 5.263 0.752 21.053 0 2.256 16.541 47.368 28.571 5.263 3.759 64.662 2.256 2.256 1.504 2.256 12.653 *r, Living species; f, fossil taxa. The discrete morphological matrix had 8.98% polymorphic entries, a value that varied from 0 to 19.33% within each taxon (Table 2). The percentage of polymorphic cells of each taxon was positively correlated with the number of scored specimens (SpearmanÕs R = 0.78, P < 0.000001). By type of character, dental characters had more polymorphisms (mean = 15.04%), followed by cranial (mean = 8.30%), and finally postcranial characters (mean = 2.51%). These differences were significant under the Mann–Whitney test (P < 0.001). Partitioned analysis The analysis of the ‘‘continuous’’ partition resulted in one MP tree of 14.325 steps under equal weights, in which Caninae and most of the recognized clades and genera were not recovered. The larger species (e.g. Canis spp., Lycaon, Theriodictis spp., Protocyon spp., Cuon alpinus, Chrysocyon brachyurus) were contained in a large pectinate clade, while species with a medium body size (e.g. D. avus and L. culpaeus) fell at its base. The branch supports of this tree were very low, with < 40 SR or 20 GC in most nodes. With the ‘‘cranial’’ partition (discrete and continuous characters) and equal weights, 15 trees of 128.779 steps were found (Fig. 1). Caninae, Canini (including Nyctereutes sensu Tedford et al., 1995) and Vulpes were natural groups, while the ‘‘South American clade’’ and the Canis clade were paraphyletic. Lycalopex and Dusicyon formed a clade, but neither was monophyletic 461 F.J. Prevosti / Cladistics 26 (2010) 456–481 Table 2 Percentage of polymorphic entries per taxa and discrete osteological partition (number of characters per partition in parentheses) Taxa Dental (42) Cranial (42) Postcranial (35) Total (119) Archaeocyon leptodus Atelocynus microtis Canis adustus Canis aureus Canis dirus Canis gezi Canis latrans Canis lupus Canis mesomelas Canis nehringi Canis rufus Canis simensis Canis armbrusteri Cerdocyon thous Chrysocyon brachyurus Cuon alpinus Dusicyon australis Dusicyon avus Eucyon davisi Hesperocyon Leptocyon vafer Lycalopex culpaeus Lycalopex fulvipes Lycalopex griseus Lycalopex gymnocercus Lycalopex sechurae Lycalopex vetulus Lycaon pictus Mesocyon coryphaeus Nyctereutes procyonoides Otocyon megalotis Phlaocyon leucosteus Protocyon scagliorum Protocyon troglodytes Speothos venaticus Theriodictis platensis Theriodictis tarijensis Urocyon cinereoargenteus Vulpes lagopus Vulpes vulpes Vulpes zerda Total 16.667 23.810 11.905 23.810 35.714 0 23.810 28.571 14.286 0 7.143 9.524 16.667 28.571 19.048 14.286 0 16.667 0 14.286 0 23.810 0 33.333 33.333 23.810 26.190 21.429 9.524 9.524 7.143 4.762 0 16.667 4.762 19.048 7.143 21.429 9.524 23.810 16.667 15.041 2.381 9.524 9.524 19.048 11.905 0 21.429 19.048 16.667 0 14.286 2.381 14.286 9.524 9.524 9.524 0 2.381 0 7.143 0 11.905 0 14.286 16.667 14.286 16.667 21.429 0 16.667 7.143 2.381 0 0 9.524 9.524 0 7.143 9.524 4.762 0 8.304 2.857 5.714 5.714 5.714 2.857 0 11.429 8.571 0 0 0 0 0 2.857 11.429 0 0 0 0 0 0 5.714 0 5.714 5.714 2.857 0 2.857 0 2.857 0 0 0 0 5.714 5.714 0 5.714 0 2.857 0 2.509 7.563 13.445 9.244 16.807 17.647 0 19.328 19.328 10.924 0 7.563 4.202 10.924 14.286 13.445 8.403 0 6.723 0 7.563 0 14.286 0 18.487 19.328 14.286 15.126 15.966 3.361 10.084 5.042 2.521 0 5.882 6.723 11.765 2.521 11.765 6.723 10.924 5.882 8.977 because L. culpaeus was nested in Dusicyon. Cerdocyon (Nyctereutes (Speothos + Atelocynus)) and C. gezi + Protocyon + Theriodictis were the successive sister groups of the Canis clade. Finally, C. dirus + C. nehringi occupied a terminal position in the Canis clade as the sister group of C. lupus. Branch supports were generally low, and only Caninae, Canini, Urocyon + Otocyon, C. gezi + Protocyon + Theriodictis and C. dirus + C. nehringi + C. lupus had SR and GC values above 70 ⁄ 60. The consensus tree of the 100 trees of 108.129 steps found with the ‘‘dental’’ partition under equal weights was scarcely resolved, and most of the previously recognized clades were not monophyletic (Fig. 2). A large polytomy grouped most of the species of Lyca- lopex, Dusicyon, Vulpes, Cerdocyon, Atelocynus, and the clades Phlaocyon + Urocyon + Otocyon + Chrysocyon and Canis + Eucyon + Lycaon + Cuon + Speothos + Theriodictis + Protocyon. In the latter group, the Canis species with large body size occurred in terminal position as successive sisters of a highly carnivorous group (Lycaon + Cuon + Speothos + Theriodictis + Protocyon). Lycaon + Cuon was the sister group of Speothos + Theriodictis + Protocyon, C. gezi was the sister species of T. platensis, and C. dirus was the sister species of C. nehringi. The clades with branch supports above 60 or 70 were E. davisi + C. adustus, Lycaon + Cuon + Speothos + Theriodictis + Protocyon + C. gezi, Lycaon + Cuon and the more basal groups of the tree. 462 F.J. Prevosti / Cladistics 26 (2010) 456–481 Fig. 1. Strict consensus of the 15 trees of 128.779 steps found with the cranial partition under equal weights. Numbers below branches correspond to support values (group frequencies (SR) ⁄ group differences (GC)). Taxa in grey are the ‘‘South American clade’’ (excluding Nyctereutes), those in bold are the large South American fossils. Fig. 2. Strict consensus of the 100 trees of 108.129 steps found with the dental partition under equal weights. References as in Fig. 1. With the ‘‘postcranial’’ partition, 1502 trees of 66.192 steps were obtained. They resulted in a very polytomic consensus tree where Caninae was monophyletic with strong support (SR ⁄ GC: 98), but lacked internal resolution. The complete ‘‘osteological’’ matrix (continuous plus discrete characters) with equal weights gave four most parsimonious trees of 350.552 steps (Fig. 3). In the consensus tree, Caninae (if we exclude Leptocyon), Canini, Vulpini and Vulpes were the only monophyletic F.J. Prevosti / Cladistics 26 (2010) 456–481 genera. The ‘‘South American clade’’ was polyphyletic, but most species of Lycalopex and D. australis formed a natural group. The species of Canis and Eucyon formed a pectinated clade with D. avus, Chrysocyon and Atelocynus as successive sister taxa, and larger and more carnivorous species as a terminal node. Lycaon, Cuon, Theriodictis, Protocyon and Speothos were included in this clade as a terminal node representing a group of large and highly carnivorous taxa. Lycaon + Cuon was the sister group of Theriodictis + Protocyon + Speothos + C. gezi. Canis nehringi was joined to C. lupus and C. dirus, while Theriodictis was not monophyletic and T. platensis was the sister species of C. gezi. Finally, C. thous formed a natural group with Nyctereutes, and D. australis was grouped with L. culpaeus. With implied weights using k = 100–65, I obtained one of the topologies obtained with equal weights, but the node 41 was resolved and T. tarijensis was the most basal species, followed by Speothos and a paraphyletic Protocyon. With k = 64–27 the only difference observed was that L. fulvipes and L. vetulus were sister species. With k = 26–25 more changes occurred: Cerdocyon + Nyctereutes was the sister group of the large Canis clade, Vulpes was paraphyletic, Leptocyon was included in Vulpini, C. armbrusteri was joined to C. rufus and C. latrans, and C. mesomelas was placed in a more terminal node. But more drastic modifications occurred using k = 24–10 or lower values (four topologies in total): Speothos, Nyctereutes, Atelocynus and Cerdocyon were placed as successive basal species of Caninae (i.e. Canini is paraphyletic), Chrysocyon was 463 the sister group of C. gezi + Theriodictis + Protocyon and the clade Canis + Eucyon + Lycaon + Cuon. C. gezi was the sister taxon of T. platensis; in addition, Protocyon + T. tarijensis was the terminal node of the clade. Dusicyon avus was placed in a polytomy with P. culpaeus and D. australis using k = 9–1. Some changes were observed in the basal part of the clade of Canis + Eucyon + Lycaon + Cuon, and C. rufus was grouped with C. armbrusteri with k = 3–1. The clade of Dusicyon and most Lycalopex species was placed as sister taxa of the group L. fulvipes + L. vetulus + Leptocyon + Vulpini, Lycaon + Cuon was displaced to the base of the ‘‘Canis clade’’, and C. nehringi became the sister species of C. dirus under k = 1. Most of the nodes of these topologies had low or very low supports, with the exception of the most basal ones, Lycaon + Cuon, Urocyon + Otocyon and C. lupus + C. nehringi + C. dirus, which presented SR ⁄ GC between 60 and 70. The strict consensus of the four MP trees (2931 steps) obtained from the DNA matrix greatly resembled recently published trees, where Canini (excluding Nyctereutes), the ‘‘South American clade’’, Vulpes and Lycalopex were monophyletic groups (Bardeleben et al., 2005; Lindblad-Toh et al., 2005). Otocyon was the sister taxon of Vulpes, and Nyctereutes the sister taxon of the remaining Caninae. Canis rufus and L. fulvipes changed their position among the most parsimonious trees, generating two polytomies also containing C. latrans, C. rufus, C. lupus and L. culpaeus, L. griseus, L. gymnocercus, respectively. Most nodes had support values of Fig. 3. Strict consensus of the four trees of 350.552 steps found with the osteological partition under equal weights. Grey numbers above branches are node numbers. Other references as in Fig. 1. 464 F.J. Prevosti / Cladistics 26 (2010) 456–481 more than 60, with the exception of the nodes: Otocyon + Vulpes, Otocyon + Vulpes + Canini and Lycaon + Cuon + the derived species of Canis. Simultaneous analysis The ‘‘total evidence’’ analyses, under equal and implied weighting (k = 100–72), gave as a result one tree of 3657.222 steps (Fig. 4), where the arrangement of the living canids was highly congruent with the DNA tree. Leptocyon was excluded from Caninae (node 43), while Eucyon was the most basal taxon of the Canis clade. Canis dirus was the sister species of C. lupus, and C. armbrusteri the sister of the latter clade. Oddly, C. nehringi fell as the sister taxon of Lycacon + Cuon. Theriodictis, Protocyon and C. gezi occurred in the Chrysocyon–Speothos group inside the ‘‘South American clade’’. Consequently, C. gezi was the sister taxon of Speothos + Protocyon + Theriodictis, while Protocyon and Theriodictis were a natural group. This last genus was paraphyletic because T. tarijensis was clustered with the Protocyon species. Finally, D. avus and D. australis were successive sisters of Lycalopex. Six other topologies were recovered using different concavity values during the implied weighting searches. In the topologies obtained with k = 71–2, the main changes were in the placement of Nectereutes and Otocyon, which became the successive sisters of the rest of the Caninae. Dusicyon avus and D. australis were a clade (under k = 37–2), L. fulvipes were the sister species of L. culpaeus (under k = 11–2) and Leptocyon was a cluster inside Vulpes (under k = 6–2). With the concavity values k = 11–1, C. nehringi was the sister taxon of C. dirus, while with k = 6–1, C. armbrusteri formed a clade with C. aureus + C. rufus. Theriodictis tarijensis was the sister species of P. troglodytes with concavity values k = 11–4 and of P. scagliarum with k = 3–1. With concavity value k = 1, a large modification was observed: Canini and the ‘‘South American clade’’ were paraphyletic because Speothos became the most basal Caninae, followed by the clade Chrysocyon + C. gezi + Theriodictis + Protocyon, the South American foxes clade, the vulpines plus Leptocyon and the Canis clade. Canini, the ‘‘South American clade’’ and several other nodes (e.g. 45, 48, 63, 75, 76) had moderate supports, but the others possessed low SR ⁄ GC values (Fig. 4). Dynamic homology approach The analysis of the DNA partition with POY gave a tree nearly identical (cost: 4040) to the one obtained with the static approximation, but in which the terminal polytomies of Canis and Lycalopex were resolved. In Canis, C. lupus was the sister taxon of C. latrans and Fig. 4. Most parsimonious tree (3657.222 steps) found with the simultaneous analysis, under equal weights and implied weighting with k = 100–72. References as in Figs 1 and 3. F.J. Prevosti / Cladistics 26 (2010) 456–481 C. rufus of C. aureus, while in Lycalopex, L. fulvipes and L. gymnocercus were successive sisters of the group L. culpaeus + L. griseus. The simultaneous analysis performed with POY gave one tree (cost: 4422) similar to the one found in the static approximation (Fig. 5). Vulpini was also paraphyletic with Nyctereutes as the most basal Caninae, but here Urocyon and Otocyon was a clade sister to Canini. Leptocyon was included in the Vulpes clade, E. davisi as the sister taxa of C. adustus and L. sechurae in a polytomy with D. avus and D. australis. Canis nehringi was the sister taxa of C. dirus, as in some morphological trees (Figs 1 and 2). 465 other comparisons). Rescaling the SPR distances by the number of possible movements (Goloboff, 2007; Goloboff et al., 2008a) gave results similar to those obtained with other measurements (data not shown). It must be noted that the congruence between and within the ‘‘types’’ of characters presented large variation, with overlapping of ranges and even 25–75% quartiles, thus implying that some osteological trees or partitions had the same level of congruence with the DNA partitions when comparing with the congruence observed between different nuclear or mitochondrial genes. Congruence between partitions Discussion The three measurements used for analysing the congruence gave similar results. Comparisons within nuclear genes, mitochondrial genes, and osteological partitions showed higher levels of congruence than comparisons between the different kinds of partition. Mitochondrial vs. nuclear genes showed less congruence, and the lowest levels of congruence were between osteological and miscellaneous vs. other kinds of characters (Fig. 6). The SPR indicated a low congruence between the osteological partitions, but this was partly because these partitions had more taxa (41 vs. ca. 24 in Pattern and effect of missing data and polymorphism The osteological matrix showed moderate levels of missing data that were not randomly distributed, matching the most common pattern of distribution (Wheeler, 1992; Wiens, 2003a; Prevosti and Chemisquy, 2009). As usually stated, fossils have more missing data than living species (Wilkinson, 1995b), but it must be said that there is a wide superposition in the level of missing data between these groups, and that some living Fig. 5. Most parsimonious tree (fit: 4422) found with the dynamic homology approach and osteological plus DNA characters. Numbers below branches correspond to Bremer support values. 466 F.J. Prevosti / Cladistics 26 (2010) 456–481 Fig. 6. Congruence intrapartition and among partitions. (a) Box plots of consensus fork index (CFI) values; (b) box plots of subtree pruning and regrafting (SPR) distances; osteo: osteological partition; misc: miscellaneous partition; mit: mitochondrial genes; nuc: nuclear genes. species had higher levels of incompleteness than some fossils. With the complete matrix, the percentage of missing data was far greater but still not random. Despite the large percentage of missing data in some taxa, the wild card bias was detected only in the analysis of some partitions of the discrete osteological characters, and was very localized (data not shown). The inclusion of continuous characters eliminated this problem, in part by the introduction of more information, but also probably because with this kind of character the trees are more resolved due to their fractional length (it is less probable to obtain several trees with the same length; Ramı́rez, 2003). Another negative effect of missing entries was lowering of the supports of some nodes. For example, excluding Leptocyon from the consensus estimation during resampling in the simultaneous analysis increased the supports of Caninae from 56 ⁄ 22 (SR ⁄ GC) to 97 ⁄ 96. Excluding C. gezi, the Theriodictis + Protocyon clade (node 78 of Fig. 4) changed supports from 27 ⁄ )11 to 56 ⁄ 44, and Speothos plus the former clade (node 74 of Fig. 4) changed supports from 45 ⁄ 6 to 82 ⁄ 75. The same was observed when C. nehringi was excluded: most nodes of the Canis clade (53, 54, 59, 60) obtained values above 70, but this appeared to be related also to the presence of polymorphisms in C. dirus. The high levels of polymorphic entries contrast with the intraspecific variation detected in previous studies (Berta, 1989; Tedford et al., 1995), but agree with published information about osteological variation in recent canids (Szuma, 2002, 2003, 2004, 2008a,b,c; Prevosti, 2006). As stated in previous works (Campbell and Frost, 1993), the number of polymorphic cells is positively related to the number of specimens scored by taxa, which could explain the difference observed with previous studies. The different percentage of polymorphisms observed between cranial, dental, and postcranial characters could be related in part to the number of specimens scored in each partition. This explains why postcranial characters, which have the lowest number of scored specimens, have low polymorphisms, but not the highest number of polymorphisms of the dental characters, compared with the cranial partition, as both have a similar number of specimens scored. Probably the highest number of polymorphisms of the dental characters could be related to developmental factors (Szuma, 2002). Polymorphisms and missing data apparently biased the position of C. nehringi in the simultaneous analyses with equal and implied weights with k = 100–7. This species and C. dirus are nearly identical (Berta, 1989) and have even been considered synonyms, although this has not been formally published (Prevosti, 2006). According to these two works, these species formed a monophyletic group (Berta, 1989; Prevosti, 2006), a result also found in the present contribution in several searches using the osteological partition and in the simultaneous analysis based on the dynamic approach (see above; Fig. 5). In fact, the only differences between the scored characters of both species are that C. nehringi has several missing entries (all the postcranial data set and some dental and cranial characters) and that C. dirus has several polymorphic entries. This clade is recovered if character number 20 is coded as derived (orbital fissure and optic canal in a common pit, which has been interpreted previously as a synapomorphy of these species; Berta, 1989; Prevosti, 2006) in C. dirus, which is the most common state in this species. A similar situation could occur with C. gezi, which is very similar to T. platensis (Prevosti, 2001, 2006). Both formed a monophyletic group in most searches using the osteological partition, but occurred at the base of Speothos + Theriodictis + Protocyon in the simultaneous analyses. This happened because the ‘‘potential’’ synapomorphies of the former clade with T. platensis F.J. Prevosti / Cladistics 26 (2010) 456–481 are plesiomorphies in this tree, and the synapomorphic characters that linked Theriodictis and Protocyon were missing entries in C. gezi. Placing C. gezi at the base of Protocyon + Theriodictis or as the sister of T. platensis increased the length of the tree in only 1.477 and 1.416 steps, respectively. General outcomes for canid phylogeny The pattern obtained in these analyses is consistent with the published phylogenies (Wang, 1994; Tedford et al., 1995, 2009; Wayne et al., 1997; Wang et al., 1999, 2004a,b; Zrzavý and Řičánková, 2004; Bardeleben et al., 2005; Lindblad-Toh et al., 2005). For example, the living canids, Canini, the ‘‘Canis clade’’, and the ‘‘South American clade’’ were monophyletic. In the ‘‘Canis clade’’, C. adustus and C. mesomelas were the most basal group, rendering Canis as a paraphyletic taxa. Zrzavý and Řičánková (2004) suggested excluding these species from Canis and put them in Schaeffia and Lupulella, respectively, because in their analysis they were successive sisters of a Lycaon + Cuon + Canis s.s. Virtually nobody followed these suggestions, and a paraphyletic Canis has been used until now (Sillero Zubiri et al., 2004; Wilson and Reeder, 2005; Tedford et al., 2009). My results are in agreement with the fact that C. adustus and C. mesomelas must be excluded from Canis, but both could be included in the same genera. Schaeffia and Lupulella were described in the same paper by Hilzheimer (1906) and, following the International Code of Zoological Nomenclature (1999), the revisor could choose either name in this situation. The generic name of C. adustus and C. mesomelas must be established in the future by choosing among the available names, but that is beyond the scope of this work. The position of C. rufus as the sister species of C. aureus was in disagreement with previous works. In fact the status of this species is complex. Some authors considered it as a subspecies of C. lupus (Wilson and Reeder, 2005); a different species (Nowak, 1979, 2002; Wilson et al., 2009); or a hybrid between C. lupus and C. latrans (Roy et al., 1996). A very polymorphic osteological character (height of the femur head) grouped both species in the partitioned analysis based on the osteological matrix (Fig. 3) but not in the total evidence tree (Fig. 4), where substitutions of cytochrome b are the only synapomorphies for this clade. More data are necessary to corroborate the position of C. rufus. Another area of agreement is that E. davisi was placed as the sister taxa of the ‘‘Canis clade’’, except in the dynamic approach where it is the sister species of C. adustus, as stated by Tedford et al. (2009); and D. australis (and D. avus) formed a clade with Lycalopex spp., as previously found by Zrzavý and Řičánková (2004). 467 One interesting point is the position of the most basal living canids. In several analyses, Nyctereutes was the basal-most living species, but using equal weights (Fig. 4) and under different concavity values, changes in the position of Urocyon, Nyctereutes, Otocyon and Vulpes were observed. This variation could also be observed in the published phylogenies (Tedford et al., 1995, 2009; Wayne et al., 1997; Wang et al., 1999, 2004a,b; Zrzavý and Řičánková, 2004; Bardeleben et al., 2005; Lindblad-Toh et al., 2005). The basal position of Nyctereutes could be linked, in part, to the exclusion of Leptocyon from Caninae in several analyses (Figs 3 and 4). As suggested by one revisor (X. Wang), the hypocarnivorous borophagine Phlaocyon could be ‘‘attracted’’ inside Caninae due to the occurrence of omnivorous canines (e.g. Urocyon, Nyctereutes) at the base of this clade. If Phlaocyon is excluded, Vulpes, Nyctereutes, Otocyon and Urocyon were grouped in a monophyletic Vulpini and Leptocyon placed as the sister species of other Caninae, as previously established by Wang et al. (1999) and Tedford et al. (1995, 2009). Clearly, a denser sampling of fossil basal canines, Borophaginae and Hesperocyoninae, is needed to stabilize the position of the living Vulpini species. Phylogenetic relationships of the South American large extinct canids The trees obtained using the osteological matrix with equal weighting and implied weighting with k = 100–25 characters did not agree with previous hypotheses based on morphological characters published by Berta (1981, 1984, 1987, 1989) and Tedford et al. (1995), as the South American and Canis clades were not monophyletic in this osteological tree (Fig. 3). Chrysocyon was not the sister taxon of Canis as suggested by Berta (1987, 1989), but instead was associated with a clade composed of the species of the Canis clade, plus Atelocynus, Eucyon, Speothos, Theriodictis, Protocyon and the fossil species of Canis. However, there were several monophyletic groups in common with Tedford et al.Õs (1995) hypothesis, including Vulpini, Canini, Urocyon + Otocyon, Cuon + Lycaon, L. culpaeus + D. australis and Cerdocyon + Nyctereutes. The latter clade was also represented in BertaÕs trees. The South American fossil C. nehringi was placed in a polytomy with C. dirus and C. lupus, but was the sister taxon of C. dirus in the phylogeny of Berta (1989). Canis dirus and C. nehringi formed a monophyletic group only with k = 1. The other South American species of Canis, C. gezi, had a position very different from that published by Berta (1989), because it was placed as the sister group of T. platensis inside the monophyletic group of Speothos + Protocyon spp. + Theriodictis spp., corroborating my previous interpretation (Prevosti, 2001) based on anatomical studies. The placement of the hypercarniv- 468 F.J. Prevosti / Cladistics 26 (2010) 456–481 orous South American taxa Protocyon and Theriodoctis did not agree with BertaÕs tree either, as they were included as a terminal node in a large clade of species of the Canis clade and are not the sister group of D. australis as previously reported by Berta (1989). The clade of Theriodictis, Protocyon, Speothos and C. gezi was the sister group of other hypercarnivorous canids such as Lycaon + Cuon and more distant from the large-sized Canis species. This large clade of hypercarnivorous canids resembles a polyphyletic group composed of highly carnivore species that are currently placed in different families (Canidae, Amphicyonidae, Procyonidae), which in the past was considered a Canidae subfamily (Kraglievich, 1928, 1930; Matthew, 1930; Simpson, 1945; see also Wang, 1997; Tedrow et al., 1999; Peigné et al., 2005; Salesa et al., 2006). Lastly, Theriodictis and Protocyon were not monophyletic in the trees obtained and these topologies did not support the statement that Speothos is a ‘‘phyletic dwarf’’ of Protocyon (Wayne and OÕBrien, 1987), the statement that Protocyon is a descendant of Cerdocyon (Winge, 1895), the close relationship between Cuon and Protocyon (Kraglievich, 1928), or the inclusion of Theriodictis species in Canis (Kraglievich, 1928; Berman, 1994). Using stronger concavity values (k = 24–1) resulted in other topologies, but with a few exceptions they did not alter these statements. One important exception is that Canini became paraphyletic because Speothos, Nyctereutes, Atelocynus and Cerdocyon are successive sisters of other canines. Another exception is that Theriodictis + Protocyon + C. gezi were placed related to Chrysocyon outside the Canis clade, making the latter group monophyletic. Finally, Caninae is monophyletic in all these trees, but the North American Tertiary fossil Leptocyon is excluded from the subfamily in the topologies obtained with equal and implied weighting with k = 100–27. The simultaneous analyses changed this scenario, with Canis and the South American group recovered as monophyletic with the combined data set, and the topology of the recent taxa was highly congruent with the published studies based on DNA (Wayne et al., 1997; Bardeleben et al., 2005; Lindblad-Toh et al., 2005). Speothos and other South American taxa were successive sisters of other canines only under implied weighting with k = 1, thus making Canini and the South American group nonmonophyletic. In the simultaneous analysis, Protocyon and Theriodictis were placed with Speothos, as a sister group of Chrysocyon, corroborating its inclusion in the ‘‘South American clade’’ as inferred by Berta (1981, 1987, 1989) (Figs 4 and 6). The relationship between Protocyon and Theriodictis, but not their monophyly, was supported. Theriodictis tarijensis must be transferred to Protocyon to make these genera natural groups. Canis gezi was the sister species of Theriodictis + Protocyon + Speothos in these topologies. Using implied weighting with k = 6–1, C. nehringi was the sister group of C. dirus, but with different concavity values or equal weighting was placed as a sister taxon of Lycaon + Cuon. The extinct D. australis and D. avus were sister taxa of Lycalopex. Dusicyon was monophyletic only under implied weighting with k = 37–2. With k = 1, D. australis and D. avus formed a terminal node with L. culpaeus inside Lycalopex. The inclusion of the North American fossils C. avius in the simultaneous analysis did not support its relationship with Cerdocyon or other South American genera (Fig. 7). In a simultaneous analysis with equal weights it was placed as the sister taxa of Cerdocyon texanus, forming a clade with Urocyon clade (Fig. 7). Cerdocyon avius lacks the derived conditions present in other species of Cerdocyon, such as a squared angular process, while C. texanus has tall and acute premolars as in Urocyon. Tedford et al. (2009) also noted similarities between C. avius and C. texanus with the fossils Nyctereutes tingi and N. donnezani. The other two fossils, Chrysocyon nearcticus and Theriodictis? floridanus are very similar to the South American taxa and in fact are placed in the ‘‘South American clade’’ (Fig. 7). Theriodictis? floridanus was included in the Chrysocyon– Speothos clade, but as sister taxa of the ‘‘C.’’ gezi + Speothos + Theriodictis + Protocyon group, because it has a large trigonid in the m1, a strong subangular lobule and an increment of the height of the coronoid process. This species also has a deep angular process as in Theriodictis and Protocyon, but its m1 possesses a metaconid, and the entoconid is reduced to a cingulum. This combination of characters is not present in any specimen of the South American Theriodictis. Chrysocyon nearcticus was included in the ‘‘South American clade’’ but as part of the Lycalopex lineage (Fig. 7), joined to L. fulvipes by a lowering of the coronoid process of the mandible. Most of the scored characters for C. nearcticus were plesiomorphic in this phylogeny, thus its inclusion in Lycalopex could be an artefact. The inclusion of Theriodictis? floridanus and C. nearcticus in the ‘‘South American clade’’ supports the interpretation of Tedford et al. (2009) concerning the presence of the clade in North America during the Pliocene and Pleistocene, but the generic assignation of these fossils is not clear. Both taxa are represented only by the lower mandible and dentition, and it is possible that the similarities shared with the South American canids could be due to homoplasy. More complete fossils are needed to corroborate this hypothesis. Osteological synapomorphies In the tree shown in Fig. 4, the Caninae (node 43) are supported by elongation of the rostrum (character 0), F.J. Prevosti / Cladistics 26 (2010) 456–481 469 Fig. 7. Most parsimonious tree (3667.515 steps) found with the simultaneous analysis, under equal weights, including Cerdocyon avius, Cerdocyon texanus, Chrysocyon nearcticus, and Theriodictis? floridanus (white stars). References as in Figs 1 and 3. enlargement of the m1 (character 10), the orbit placed over the distal border of the P4 (character 36: 1), and several changes in postcranial skeleton (Appendix 2). Most of these postcranial derived states (e.g. character 114: 1, robustness of the ulna; 111: 1, loss of entoepicondylar foramen; 131: 1, reduction of the first metatarsal) could be related to more cursorial habits (Wang, 1993). Canini (excluding Nyctereutes; node 50 of Fig. 4) possess the following derived changes: further elongation of the rostrum (0); widening of the palate (1); reduction of the tympanic bulla (3); strengthening of the mandible body (5); further enlargement of the m1 (10); elongation of m1Õs trigonid (11); mastoid process as a knob-like process (21: 1); crest-like hypoconid in the m2 (93: 0); elongation of the fore limb related to the hind limb (12); elongation of the femur related to the tibia (13); greater trochanter and head of the femur sub-equal in height (122: 1). The Canis clade (node 57, Fig. 4) presented a widening of the palate (1), widening of the frontals (2), I3 with a strong mesiolingual cingulum (55: 1), low and robust principal cusps in the premolars (61: 3) and p3 with a distal accessory cusp (79: 1). The position of C. dirus as the sister species of C. lupus and a very derived species of the Canis clade is sustained by several derived states: enlargement of the I3 (7), further widening of the frontals (2), strengthening of the ramus of the mandible (5); higher coronoid process (6); elongation of lower carnassial (10); elongation of the fore legs (12); narrow and convex supraoccipital shield (17: 2); caudally extended inion (18: 1); caudally expanded frontal sinuses (25: 1); very caudally expanded vomer (40: 2); depressed tympanic bulla (46: 1); presence of a sharp medial constriction on the orbits (49: 1); ventrally placed superficial masseteric scar on the zygomatic bone (29: 1); P2 and P3 with distal accessory cusps (62: 1 and 63: 1, respectively); reduced protocone on the P4 (64: 1); large paracone in the M1 (70: 1); m1 with metastylid (88: 1); p4 with a second distal accessory cusp (80: 1); triangular sustentacular facet on the calcaneum (127: 0) (see Appendix 2 for other synapomorphies). Several of these synapomorphies (e.g. elongation of the m1, elongation of the trigonid of the m1, strengthening of the ramus of the mandible) have been interpreted as traits associated with a hypercarnivorous diet by several authors (Berta, 1989; Van Valkenburgh, 1991; Palmqvist et al., 1999; Prevosti and Palmqvist, 2001). 470 F.J. Prevosti / Cladistics 26 (2010) 456–481 The ‘‘South American clade’’ (node 49, Fig. 4) showed a wide scar of the superficial masseteric muscle on the zygomatic bone (4), a paraoccipital process with narrow base (16: 0), the absence of metaconule in the M1 (68: 0) and hypoconulid in the m1 (78: 0). The Crysocyon + Speothos + Theriodictis + Protocyon + C. gezi node (63 in Fig. 4) shared dental, cranial and postcranial synapomorphies including the elongation of the rostrum (0); the diminution of bulla size (3); the strengthening of the ramus of the mandible (5); the presence of caudally expanded frontal sinuses (25: 1); reduction of the labial cingulum in the M1 (70: 1); the fourth cervical vertebra with a straight lateral border (99: 2); and quadrangular sustentacular facet on the calcaneum (127: 2). Canis gezi, Speothos, Theriodictis and Protocyon showed derived states related to hypercarnivorous habits (widening of the palate (1), elongation of the trigonid of the m1 (11); reduction of the protocone on the P4 (64: 2) and hypocone in M1 (67: 1); large paracone in M1 (70: 1); reduction of mesolabial cingulum and metaconid in the m2 (91: 2; 92: 2), which suggests that the hypercarnivory arose only once in the ‘‘South American clade’’. Other derived states of this clade were: the paraoccipital process ventrocaudally directed (14: 1), the independent opening for the rotundum foramen and alisphenoid canal (37: 1) and a robust and sharp subangular lobule (50: 1). A reduced hypocone and metaconule of the M1 (69: 1; 72: 1) and an m1 without metaconid or hypoconulid (84: 1; 89: 1) supported the Speothos + Theriodictis + Protocyon clade (node 74, Fig. 4), while the anterior placement of the facial nerve related to the acoustic nerve in the dorsal face of the petrosal (43: 1), and the presence of a depressed capitulum in the humerus (105: 1) sustained the Theriodictis–Protocyon group (node 78, Fig. 4). Theriodictis tarijensis and Protocyon spp. (node 79, Fig. 4) shared the presence of a distal accessory cusp in the P3 (63: 1), a relict of a metacone in the M2 (74: 3) and a higher coronoid process. The trees of Figs 4 and 5 indicate that the hypercarnivory originated independently at least three times in Caninae, in the Lycaon–Cuon, C. armbrusteri + C. dirus + C. lupus and C. gezi + Speothos + Protocyon + Theriodictis clades. In the latter, Speothos presented a further specialization with a more reduced dentition (M2 and m3 were lost). The T. tarijensis + Protocyon clade also showed a more derived dentition with a relict of metacone in the M2. This shows that the occurrence of hypercarnivory is not related to only one modification in the cranial–dental anatomy, and that several changes took place in different branches, leading to the presence of mosaics of characters in several species. As seen in Appendix 2, most of these derived changes are not uncontradicted synapomorphies, and reversions and parallelisms happened in other branches, or these changes were hidden in the polymorphisms coded for several taxa. Several of these synapomorphies of the ‘‘South American clade’’ had been recognized previously by Berta (1981, 1984, 1987, 1989) and Tedford et al. (1995; 2009), but some of them were not recovered in this study. The presence of strongly arched zygomata (Berta, 1987, 1989) is a plesiomorphic state for Caninae (see character 28 and Appendix S3). The presence of a high angular process does not constitute a synapomorphy of this clade because C. brachyurus, Lycalopex and Dusicyon have the plesiomorphic state (a low process). Coding this character (52) as additive did not modify the tree, but the reconstruction for the base of the ‘‘South American clade’’ changed from low-hooked (state 0) to ambiguous (states 0–1) and a rectangular process (state 1) could have been a synapomorphy of this group only after an acctran optimization. The presence of an expanded fossa for the inferior branch of the medial pterigoid muscle in the mandible was not considered a synapomorphy here either. The presence of an expanded and rectangular fossa occurred in several Caninae and is not exclusive of the South American group. Cerdocyon thous and Speothos have a deeper fossa, but this is correlated with their deeper angular process. Nyctereutes procyonoides also has a fossa that is deeper than in other canines, but again its angular process is deeper (see comments about characters 51–52 in Appendix 3). This interpretation is more in agreement with the analysis of Tedford et al. (1995). The scoring of the coronoid process as a continuous character (character 6) did not support the ‘‘South American clade’’ having a low and long process. Information and congruence between character partitions The partitioned analyses indicate that the clade of the hypercarnivorous species, which includes Theriodictis, Protocyon, Speothos, Lycaon and several Canis species found with the osteological partition (vide supra, Fig. 3, node 54), was sustained mostly by dental characters (Fig. 2). In fact, the cranial partition alone did not recover this clade, and the trees obtained were more congruent with the DNA partition. Several of these characters were related to a reduction in the dentition (e.g. P4 protocone [64: 1], M1 hypocone [69: 1], m1 entoconid [88: 3]); an enlargement of the third upper incisors (7), lower carnassial (10) and its trigonid (11), M1 paracone (70: 1); and the strengthening of the upper canine (8). Most of those characters were interpreted as specializations to highly carnivore diets (Kraglievich, 1928; Berta, 1989; Van Valkenburgh, 1991; Prevosti and Palmqvist, 2001). Thus they could constitute a character complex related to adaptive or functional causes, and probably not really independent. If the weights of these characters are reduced to 1 ⁄ 2 of the weights of the remaining characters in the osteological data set, a F.J. Prevosti / Cladistics 26 (2010) 456–481 topology very similar to the one obtained with the cranial partition (Fig. 1) is obtained, where the South American hypercarnivorous canids are placed outside the Canis clade. The addition of other sources of information (DNA in this case) or the use of implied weighting with some k values (24–1) also helped to break the ‘‘misleading’’ signal of the dental partition. On the other hand, the placement of Phlaocyon (a borophagine) in a clade with omnivorous canines was supported by dental traits that are correlated with hypocarnivorous diets. However, in this case the inclusion of cranial and postcranial partitions rescued the basal position of this genus related to the Caninae subfamily (Fig. 3). Related to that, the comparison of the congruence of dental, cranial and gene partitions with CFI and SPR distances showed that the cranial partition was more congruent with the DNA partition (with values similar to those observed between some genes) than with dental characters, or between dental and gene partitions (see above; Fig. 6). The ‘‘misleading’’ signal of the dental partition resembles the case studied by Naylor and Adams (2001) related to molecular and morphological phylogenetic analyses of Cetioartydactila, where the dental partition generated the main incongruence between molecular and morphological data sets. Other works have shown the homoplasy of dental and other morphological characters in the light of recent multigene phylogenies (Gaubert et al., 2005; Koepfli et al., 2007). Koepfli et al. (2007) dismissed a previous morphological analysis (Baskin, 2004) due to the presence of homoplasy and nonindependent characters, caused in part by the ‘‘atomization’’ of many dental traits during analysis of the characters. These authors argued that teeth are integrated structures, developmentally and functionally correlated, and based this argument on different works (Jernvall, 1995; Polly, 1998; Zhao et al., 2000; Kangas et al., 2004). I agree that developmental and functional studies (Van Valkenburgh, 1991; Jernvall, 1995; Biknevicius and Van Valkenburgh, 1996; Polly, 1998; Zhao et al., 2000; Hlusko, 2004; Kangas et al., 2004; Goswami, 2006) may lead to the existence of ‘‘character complexes’’, but this should not disqualify performing a sharp analysis of dentition or other morphological sources of information. However, it must be noted that several authors maintain that morphology is misleading due to the correlation of characters based on works about developmental patterns that were tested in only a few model species (Kangas et al., 2004), and argue that the analysis of these problems is simplistic (Koepfli et al., 2007; Springer et al., 2007, 2008). In fact, several characters that a priori could be seen as nonindependent were synapomorphies of different nodes (Appendix 2) and did not possess identical state distribution. A good example is the absence of metaconid (character 84) and entonconid in the m1 (character 85), two traits related to hypercarniv- 471 orous diets (Berta, 1989; Van Valkenburgh, 1991) that could be controlled by the same developmental processes. The entoconid is present and the metaconid is usually absent in T. platensis, while the first is absent and the second is present in L. pictus, showing the independence and the existence of mosaic evolution. Moreover, the conflation of these characters could lead to loss of phylogenetic information. A careful character analysis based on the search of homology and logical independence (Rieppel and Kearney, 2002) and the exploration of the distribution of states of putative correlated characters (see Materials and methods; Pol and Gasparini, 2009), without biasing the analysis with a priori assumptions of lack of independence, are needed properly to identify non-independent characters and to avoid losing information in the process. This contrasts with the suggestion of an abandonment of ‘‘atomistic’’ morphological character analyses claimed by some authors (e.g. Koepfli et al., 2007). The existence of ‘‘character complexes’’ and the lack of independence could be explored a posteriori through partitioned analysis and by exploring the optimization of characters, while their effect could be analysed with different approaches (e.g. downweighting or combining the correlated characters, deleting some characters, introducing missing entries in key taxa; Pol and Gasparini, 2009). Beyond the possible lack of independence, the presence of homoplasy is not a problem for morphology and ⁄ or parsimony analysis (cf. Farris, 1983), and differential character weighting could be used if it is considered that a homoplasic character is less reliable (Farris, 1969; Goloboff, 1993, 1997); however, the inclusion of different sources of morphological characters must be encouraged. The evaluation of the congruence based on SPR distances and CFI indicates that the congruence is lower between morphology and DNA than between each gene, but also that there is a wide overlapping and that some morphological partitions or trees have values similar to those observed between genes. This suggests that the lack of congruence between morphology and DNA is not a valid argument for discarding morphological information in cladistic analyses, at least for this study, as it is explicit or implicit in some recent works (Koepfli et al., 2007; Springer et al., 2007, 2008). Beyond that, there are good reasons not to discard evidence or avoid combining data sets because of the presence of incongruence (Kluge, 1989; Nixon and Carpenter, 1996; Gatesy et al., 1999; Jenner, 2004; Gatesy and Baker, 2005; Asher et al., 2008; for discussion of the beneficial contributions of morphology in cladistics analyses see Smith, 1998; Smith and Turner, 2005; Wiens, 2004). Canid immigration to South America Taking into account the phylogeny obtained with the simultaneous analysis, it is clear that different canid 472 F.J. Prevosti / Cladistics 26 (2010) 456–481 lineages invaded South America independently, as previously stated (Wayne et al., 1997; Prevosti, 2006). Using mitochondrial genes, Wayne et al. (1997) estimated the divergence of Speothos from Chrysocyon 6– 7 Ma; 2–3 Myr before the rising of the Panama Bridge. The first split in the South American foxes clade was estimated in 3–5 Ma by these authors, and could be synchronous with the rise of the Panama bridge. If these dates are accurate, the ‘‘South American clade’’ could be represented at least by three immigration events: one for Chrysocyon, another for the Theriodictis + Speothos + Protocyon lineage, and the third for the South American foxes (Cerdocyon + Atelocynus + Dusicyon + Lycalopex); or four if Cerdocyon has diverged prior to 4–2.5 Ma (Wayne et al., 1997). This is congruent with the presence of Cerdocyon, Chrysocyon and Theriodictis in the Pliocene–Pleistocene of North America, as suggested by some authors (Torres Roldán and Ferusquı́a Villafranca, 1981; Berta, 1989; Tedford et al., 2009), but, as discussed above, C. avius and C. texanus are not related to Cerdocyon. Thus the South American foxes (Lycalopex, Dusicyon, Atelocynus and Cerdocyon) are restricted to South America, but the inclusion of Theriodictis? floridanus and C. nearcticus in the ‘‘South American clade’’ in part supports this hypothesis. On the other hand, the molecular dates of divergence must be corroborated with other genes and by using other methods that do not use constant rates (e.g. ‘‘relaxed molecular clocks’’). Another alternative is that the ancestor of the ‘‘South American clade’’ invaded South America prior to this tectonic event that occurred ca. 4–2.5 Ma bp, as happened with the family Procyonidae (Marshall et al., 1979; Soibelzon and Prevosti, 2007). If this was the case, the origin and diversification of the ‘‘South American clade’’ could be interpreted as in situ radiation, as suggested by other authors. The South American fossil record does not support this latter hypothesis, and indicates that the oldest canid, a member of the foxes lineage, occurred at the Late Pliocene (ca. 3– 2.5 Myr bp), while the first fossils of the Speothos + Chrysocyon clade (C. gezi, Theriodictis and Protocyon) were recorded in the Lower–Middle Pleistocene. Finally, Speothos and C. brachyurus have their oldest fossils in the late Pleistocene (125 000–8000 yr bp) (see Berman, 1994; Cione and Tonni, 1995; Prevosti et al., 2005; Prevosti and Rincón, 2007; Soibelzon and Prevosti, 2007). As the fossil record has shown, the immigration of Urocyon and Canis occurred in the Late Pleistocene (25– 10 thousand years ago; Prevosti and Rincón, 2007) and could be explained by two independent events (I considered C. nehringi and C. dirus as synonyms, see above and Prevosti, 2006). Finally, in the Holocene, domestic dogs were introduced in South America by humans (Schwartz, 1997; Prates et al., 2009). Conclusions The combined analysis of the available information supports the monophyly of the South American canids (Atelocynus, Chrysocyon, Lycalopex, Dusicyon, Speothos and Cerdocyon) and the inclusion of the fossils Theriodictis spp., Protocyon spp. and C. gezi in this clade. These fossils form a clade with Chrysocyon and Speothos, but Protocyon and Theriodictis are paraphyletic because T. tarijensis is the sister species of P. troglodytes and P. scagliorum, and must thus be transferred to Protocyon. Canis dirus and C. nehringi are included as derived species in the Canis clade. The first is the sister taxon of C. lupus, but the position of the second is biased in the simultaneous analysis by the presence of missing and polymorphic entries. Beside this bias, the states scored for C. nehringi are identical to those observed in C. dirus, which supports the idea that both could be the same species. The North American fossils that were previously assigned to Cerdocyon are, in fact, related to Urocyon, while Theriodictis? floridanus and C. nearcticus were included in the ‘‘South American clade’’, supporting the hypothesis of Tedford et al. (2009). Hypercarnivory arose at least three times in Caninae: once in the ‘‘South American clade’’ (Speothos + C. gezi + Theriodictis + Protocyon monophyletic group), once in the Lycaon + Cuon clade and once in the derived species of Canis (e.g. C. dirus, C. lupus). The partitioned and simultaneous analyses show that the dental characters recovered a clade of hypercarnivorous canids composed of species of the Canis clade and the ‘‘South American clade’’, which is apparently supported by a complex of characters related to functional demands. The comparison of the congruence between partitions indicates that the level of congruence observed between the morphological and the DNA partition is lower than the congruence observed between genes, but several genes have values similar to those observed between DNA and morphology. If I considered that the presence of a certain level of incongruence is a valid argument to exclude some partitions from the combined analysis, I do not see any reason to exclude the osteological partition in the present study. The combination of the phylogenetic analyses, published dates of molecular divergence and the fossil record suggest that at least three or four independent lineages of the ‘‘South American clade’’ invaded South America after the rise of the Panama Bridge around 4–2.5 Ma, pointing to an out-of-South American origin for this clade. Urocyon and Canis dirus invaded this continent at least during the Late Pleistocene, and the domestic dog (Canis familiaris) was introduced by aborigines in the Holocene. F.J. Prevosti / Cladistics 26 (2010) 456–481 Acknowledgements Thanks to Amelia Chemisquy, who helped me in different stages of this work, especially with the molecular data, and for her moral and sentimental support. Thanks are also due to Christian Meister, Kim Aaris, Ascanio Rincón, Richard Tedford, and George Lyras for supplying casts of different canids. Richard Tedford gave me the opportunity to read an early draft of Tedford et al. (2009). Also to Norberto Gianini and Xiaoming Wang for their useful comments during the review process, to Dennis Stevenson and Ian Kitching for their help during with editorial management, and to Sergio Lucero for giving me a copy of LangguthÕs paper. Thanks to the curators of the collections I visited: A. Kramarz, J. Bonaparte, M. Reguero, S. Bargo, A. Dondas, O. Vaccaro, J. Powell, R. Barquez, D. Flores, M. Merino, D. Verzi, C. Vieytes, I. Olivares, C. Morgan, A. Rodrı́guez, D. Ibáñez, I. Ferrusquı́a Villafranca, J.L. Aguilar, C. De Muizon, A. Currant, R. Tedford, J. Flynn, B. MacFadden, R. Hulbert, B. Simpson, R. MacPhee, B. Patterson, A. Rincón, M. de Vivo, D. Dias Henriques, A. Salles, V. Pacheco, R. Salas Gismondi, J.L. Carrion, L. Albuja, P. Velazco, M. Mikaty del Castillo, J.N. Martı́nez, F. Bisbal, D. Romero, E. Massoia, J. Gadkin, J. Yañes, T. Amorosi, W. Joyce, D.B., M. Carrano, L.K. Gordon, M.-T. Schulenberg and W. Stanley. The American Museum of Natural History, Florida Museum of Natural History, Field Museum of Natural History, and CONICET gave me travel grants to visit different collections. The Consejo Nacional de Investigaciones Cientı́ficas y Técnicas (CONICET) and the Comisión de Investigaciones Cientı́ficas de la provincia de Buenos Aires (CIC) gave financial support. Thanks also to Adriana ‘‘La Gordi’’ Candela, Richad Fariña and Pablo Goloboff for comments made during the defence of my PhD thesis; to Richard Tedford, George Lyras, and Dima Ivanoff for discussions about different subjects in canid morphology and systematics; and to Victoria Gonzalez Eusevi for checking the English. References Asher, R.J., Geisler, J.H., Sánchez-Villagra, M.R., 2008. Morphology, palaeontology, and placental mammal phylogeny. Syst. Biol. 57, 311–317. Bardeleben, C., Moore, R.L., Wayne, R.K., 2005. A molecular phylogeny of the Canidae based in six nuclear loci. Mol. Phylogenet. Evol. 37, 815–831. Baskin, J.A., 2004. Bassariscus and Probassariscus (Mammalia, Carnivora, Procyonidae) from the early Barstovian (Middle Miocene). J. Vert. Paleontol. 24, 709–720. Berman, W.D., 1994. Los Carnı́voros Continentales (Mammalia, Carnivora) del Cenozoico en la Provincia de Buenos Aires. PhD Thesis, Universidad Nacional de La Plata, La Plata, Argentina. 473 Berta, A., 1981. Evolution of large canids in South America. Anais do II Congreso Latino-Americano de Paleontologı́a (Porto Alegre, 1981) 2, 835–845. Berta, A., 1984. The Pleistocene bush dog Speothos pacivorus (Canidae) from the Lagoa Santa Caves, Brazil. J. Mammal. 65, 549–559. Berta, A., 1987. Origin, diversification, and zoogeography of the South American Canidae. Fieldiana. Zool. 39, 455–471. Berta, A., 1989. Quaternary evolution and biogeography of the large South American Canidae (Mammalia: Carnivora). Univ. Calif. Publ. Geol. Sci. 132, 1–149. Biknevicius, A.R., Van Valkenburgh, B., 1996. Design for killing: craniodental adaptations of predators. In: Gittleman, J.L. (ed.), Carnivore Behavior, Ecology, and Evolution, Vol. 2. Cornell University Press, Ithaca, New York, pp. 393–428. Bremer, K., 1990. Combinable component consensus. Cladistics 6, 369–372. Brower, A.V.Z., Schawaroch, V., 1996. Three steps of homology assessment. Cladistics 12, 265–272. Campbell, J.A., Frost, D.R., 1993. Anguid lizards of the genus Abronia: revisionary notes, descriptions of four new species, phylogenetic analysis, and key. Bull. Am. Mus. Nat. Hist. 216, 1–121. Cione, A.L., Tonni, E.P., 1995. Chronostratigraphy and Ôlandmammal agesÕ in the Cenozoic of southern South America: principle, practices, and the ÔUquianÕ problem. J. Paleontol. 69, 135–159. Clutton-Brock, J., Corbet, G.B., Hills, M., 1976. A review of the family Canidae, with a classification by numerical methods. Bull. Brit. Mus. (Nat. Hist.) Zool. 29, 119–199. Coates, A.G., Obando, J.A., 1996. The geologic evidence of the Central American Isthmus. In: Jackson, B.C., Budd, A.F., Coates, A.G. (Eds.), Evolution and Environment in Tropical America. University of Chicago Press, Chicago, pp. 21–56. Coates, A.G., Collins, L.S., Aubry, M.-P., Berggren, W.A., 2004. The geology of the Darien, Panama, and the Miocene–Pliocene collision of the Panama arc with northwestern South America. Geol. Soc. Am. Bull. 116, 1327–1344. Coddington, J.A., Scharff, N., 1994. Problems with zero-length branches. Cladistics 10, 415–423. Colless, D.H., 1980. Congruence between morphometric and allozyme data for Menidia species: a reappraisal. Syst. Zool. 29, 288–299. De Laet, J.E., 2005. Parsimony and the problem of inapplicables in sequence data. In: Albert, V.A. (Ed.), Parsimony, Phylogeny, and Genomics. Oxford University Press, Oxford, pp. 81–116. Duque-Caro, H., 1990. Neogene stratigraphy, paleoceanography and paleobiogeography in northwest South America and the evolution of the Panama Seaway. Palaeogeogr. Palaeoclimatol. Palaeoecol. 77, 203–234. Evans, H.E., 1993. MillerÕs Anatomy of the Dog, 3rd edn. W. B. Saunders, Philadelphia. Farris, J.S., 1969. A successive approximations approach to character weighting. Syst. Zool. 18, 374–385. Farris, J.S., 1983. The logical basis of phylogenetic analysis. In: Platnick, N, Funk, V. (Eds.), Advances in Cladistics 2, Proceedings of the Second Meeting of the Willi Hennig Society. Columbia University Press, New York, pp. 7–36. Fitzhugh, K., 2006. The philosophical basis of character coding for the inference of phylogenetic hypothesis. Zool. Scr. 35, 261–286. Flynn, J.J., 1991. Review of ÔQuaternary evolution and biogeography of the large South American Canidae (Mammalia: Carnivora)Õ. Cladistics 7, 206–212. Freudenstein, J.V., 2005. Characters, states, and homology. Syst. Biol. 54, 965–973. Gaspard, M., 1964. La région de lÕangle mandibulaire chez les Canidae. Mammalia 28, 249–329. 474 F.J. Prevosti / Cladistics 26 (2010) 456–481 Gatesy, J., Baker, R., 2005. Hidden likelihood support in genomic data: can forty-five wrongs make a right? Syst. Biol. 54, 483–492. Gatesy, J., OÕGrady, P., Baker, R., 1999. Corroboration among data sets in simultaneous analysis: hidden support for phylogenetic relationships among higher level artiodactyl taxa. Cladistics 15, 271–313. Gaubert, P, Wozencraft, W.C., Cordeiro-Estrela, P., Veron, G., 2005. Mosaics of convergences and noise in morphological phylogenies: whatÕs in a viverrid-like carnivoran? Syst. Biol. 54, 865–894. Goloboff, P.A., 1993. Estimating character weights during tree search. Cladistics 9, 83–91. Goloboff, P.A., 1997. Self-weighted optimization: tree searches and character state reconstructions under implied transformation costs. Cladistics 13, 225–245. Goloboff, P.A., 2007. Calculating SPR distances between trees. Cladistics 23, 1–7. Goloboff, P.A., Farris, J.S., Källersjö, M., Oxelman, B., Ramı́rez, M.J., Szumik, C.A., 2003. Improvements to resampling measures of group support. Cladistics 19, 324–332. Goloboff, P.A., Mattoni, C.I., Quinteros, A.S., 2006. Continuous characters analyzed as such. Cladistics 22, 589–601. Goloboff, P.A., Farris, J.S., Nixon, K., 2008a. TNT: a free program for phylogenetic analysis. Cladistics 24, 774–786. Goloboff, P., Carpenter, J., Arias, J.S., Miranda Esquivel, D., 2008b. Weighting against homoplasy improves phylogenetic analysis of morphological data sets. Cladistics 24, 1–16. Goswami, A., 2006. Morphological integration in the carnivoran skull. Evolution 60, 169–183. Harris, S.R., Gower, D.J., Wilkinson, M., 2003. Intraorganismal homology, character construction, and the phylogeny of Aetosaurian Archosaurs (Reptilia, Diapsida). Syst. Biol. 52, 236–252. Hawkins, J.A., Hughes, C.E., Scotland, R.W., 1997. Primary homology assessment, character and character states. Cladistics 13, 275– 283. Hildebrand, M., 1954. Comparative morphology of the body skeleton in recent Canidae. Univ. Calif. Publ. Zool. Sci. 52, 399–496. Hilzheimer, M., 1906. Die geographische Verbreitung der Afrikanischen Grauschakale. Zool. Beobachter 47, 363–373. Hlusko, L.J., 2004. Integrating the genotype and the phenotype in hominid paleontology. Proc. Natl Acad. Sci. USA 101, 2653– 2657. Huxley, T.H., 1880. On the cranial and dental characters of the Canidae. Proc. Zool. Soc. Lond. 12, 238–288. ICZN, 1999. International Code of Zoological Nomenclature, 4th edn. International Trust for Zoological Nomenclature, London. Ivanoff, D.V., 2000. Origin of the septum in the canid auditory bulla: evidence from morphogenesis. Acta Theriol. 45, 253–270. Ivanoff, D.V., 2006. Unlocking the ring: occurrence and development of the uninterrupted intrabullar septum in Canidae. Mamm. Biol. 72, 145–162. Jenner, R.A., 2004. Accepting partnership by submission? Morphological phylogenetics in a molecular millennium. Syst. Biol. 53, 333–342. Jernvall, J., 1995. Mammalian molar cusp patterns: developmental mechanism of diversity. Acta Zool. Fenn. 198, 1–61. Källersjö, M., Albert, V.A., Farris, J.S., 1999. Homoplasy increases phylogenetic structure. Cladistics 15, 91–93. Kangas, A.T., Evans, A.R., Thesleff, I., Jernvall, J., 2004. Nonindependence of mammalian dental characters. Nature 432, 211–214. Kearney, M., 2002. Fragmentary taxa, missing data, and ambiguity: mistaken assumptions and conclusions. Syst. Biol. 51, 369–381. Kearney, M., Clark, J.M., 2003. Problems due to missing data in phylogenetic analyses including fossils: a critical review. J. Vert. Paleontol. 23, 263–274. Kluge, A.G., 1989. A concern for evidence and a phylogenetic hypothesis for relationships among Epicrates (Boidae, Serpentes). Syst. Zool. 38, 1–25. Koepfli, K.-P., Gompper, M.E., Eizirik, E., Ho, C-C., Linden, L., Maldonado, J.E., Wayne, R.K., 2007. Phylogeny of the Procyonidae (Mammalia: Carnivora): molecules, morphology and the Great American Interchange. Mol. Phylogenet. Evol. 43, 1076– 1095. Kornet, D.J., Turner, H., 1999. Coding polymorphism for phylogeny reconstruction. Syst. Biol. 48, 365–379. Kraglievich, L., 1928. Contribución al conocimiento de los grandes cánidos extinguidos de Sud América. Anales de la Sociedad Cientı́fica Argentina 106, 25–66. Kraglievich, L., 1930. Craneometrı́a y clasificación de los cánidos sudamericanos, especialmente los argentinos actuales y fósiles. Physis 10, 35–73. Langguth, A., 1980. El origen del género Speothos y la evolución hacia Speothos venaticus. I Reunión Iberoamer. Zool. Vert. 1, 587– 600. Lecointre, G., Deleporte, P., 2004. Total evidence requires exclusion of phylogenetically misleading data. Zool. Scr. 34, 101–117. Lee, D.G., Bryant, H.N., 1999. A reconsideration of the coding of inapplicable characters: assumptions and problems. Cladistics 15, 373–378. Lindblad-Toh, K., Wade, C.M., Mikkelsen, T.S., Karlsson, E.K., Jaffe, D.B., Kamal, M., Clamp, M., Chang, J.L., Kulbokas, E.J. III, Zody, M.C., Mauceli, E., Xie, X., Breen, M., Wayne, R.K., Ostrander, E.A., Ponting, C.P., Galibert, F., Smith, D.R., de Jong, P.J., Kirkness, E., Alvarez, P., Biagi, T., Brockman, W., Butler, J., Chin, C.-W., Cook, A., Cuff, J., Daly, M.J., DeCaprio, D., Gnerre, S., Grabherr, M., Kellis, M.M., Kleber, M., Bardeleben, C., Goodstadt, L., Heger, A., Hitte, C., Kim, L., Koepfli, K.-P., Parker, H.G., Pollinger, J.P., Searle, S.M.J., Sutter, N.B., Thomas, R., Webber, C., Broad Institute Genome Sequencing Platform, Lander, E.S., 2005. Genome sequence, comparative analysis and haplotype structure of the domestic dog. Nature 438, 803–819. Luckow, M., Bruneau, A., 1997. Circularity and independence in phylogenetic test of ecological hypotheses. Cladistics 13, 145–151. Lyras, G.A., Van der Geer, A.A.E., 2003. External brain anatomy in relation to the phylogeny of the Caninae (Carnivora: Canidae). Zool. J. Linn. Soc. 138, 505–522. Maddison, W.P., 1993. Missing data versus missing characters in phylogenetic analysis. Syst. Biol. 42, 576–581. Marshall, L.G., Butler, R.F., Drake, R.E., Curtis, G.H., Tedford, R.H., 1979. Calibration of the Great American Interchange. Science 204, 272–279. Matthew, W.D., 1930. The phylogeny of dogs. J. Mammal. 11, 117– 138. Moore, W.J., 1981. The Mammalian Skull. Cambridge University Press, Cambridge, UK. Morgenstern, B., 2004. DIALIGN: multiple DNA and protein sequence alignment at BiBiServ. Nucleic Acids Res. 32, 33–36. Munthe, K., 1989. The skeleton of the Borophaginae (Carnivore, Canidae), morphology and function. Univ. Calif. Publ. Geol. Sci. 133, 1–115. Naylor, G.J.P., Adams, D.C., 2001. Are the fossil data really at odds with the molecular data? Morphological evidence for Cetartiodactyla phylogeny reexamined. Syst. Biol. 50, 444–453. Nixon, K.C., Carpenter, J.M., 1996. On simultaneous analysis. Cladistics 12, 221–242. Nowak, R.M., 1979. North American Quaternary Canis. Monogr. Univ. Kansas Mus. Nat. Hist. 6, 1–154. Nowak, R.M., 2002. The original status of wolves in eastern North America. Southeast. Nat. 1, 95–130. Palmqvist, P., Arribas, A., Martinez-Navarro, B., 1999. Ecomorphological study of large canids from the lower Pleistocene of southern Spain. Lethaia 32, 75–88. Patterson, C., 1982. Morphological characters and homology. In: Joysey, K.A., Friday, A.E. (Eds.), Problems of Phylogenetic Reconstruction. Academic Press, London, pp. 21–74. F.J. Prevosti / Cladistics 26 (2010) 456–481 Peigné, S., Salesa, M.J., Antón, M., Morales, J., 2005. Ailurid carnivoran mammal Simocyon from the late Miocene of Spain and the systematics of the genus. Acta Palaeontol. Pol. 50, 219– 238. de Pinna, M.C.C., 1991. Concepts and tests of homology in cladistic paradigm. Cladistics 7, 367–394. Pol, D., Gasparini, Z., 2009. Skull anatomy of Dakosaurus andiniensis (Thalattosuchia: Crocodylomorpha) and the phylogenetic position of Thalattosuchia. J. Syst. Palaeontol. 7, 163–197. Polly, P.D., 1998. Variability in mammalian dentitions: size-related bias in the coefficient of variation. Biol. J. Linn. Soc. 64, 83–99. Prates, L., Berón, M., Prevosti, F.J., 2009. Los perros prehispánicos del cono sur: tendencias y nuevos registros. In: Berón, M., Luna, L., Bonomo, M., Montalvo, C., Aranda, C., Carrera Aizpitarte, M. (Eds), Mamül Mapu: pasado y presente desde la arqueologı́a pampeana. Editorial Libros del Espinillo, Ayacucho, pp. 215–228. Prevosti, F.J. 2001. The Fossil Record of Canis (Carnivora: Canidae) in South America. Comments on the Systematic Status of Canis gezi. Abstracts Canid Biology and Conservation Conference. Oxford University, Oxford, pp. 94. Prevosti, F.J., 2006. Grandes Cánidos (Carnivora, Canidae) del Cuaternario de la Republica Argentina: Sistemática, Filogenia, Bioestratigrafı́ay Paleoecologı́a. PhD Thesis, Universidad Nacional de La Plata, La Plata, Argentina. Prevosti, F.J., Chemisquy, M.A., 2009. The impact of missing data on real morphological phylogenies: influence of the number and distribution of missing entries. Cladistics, in press; doi: 10.1111/ j.1096-0031.2009.00289.x. Prevosti, F.J., Palmqvist, P., 2001. Análisis ecomorfológico del cánido hipercarnı́voro Theriodictis platensis (Mammalia, Carnivora) basado en un nuevo ejemplar del Pleistoceno de Sudamérica. Ameghiniana 38, 375–384. Prevosti, F.J., Rincón, A.D., 2007. A new fossil canid assemblage from the late Pleistocene of northern South America: the canids of the Inciarte asphalt pit (Zulia, Venezuela), fossil record and biogeography. J. Paleontol. 81, 1053–1065. Prevosti, F.J., Zurita, A.E., Carlini, A.A., 2005. Biostratigraphy, systematics and palaeoecology of the species of Protocyon Giebel, 1855 (Carnivora, Canidae) in South America. J. South Am. Earth Sci. 20, 5–12. Prevosti, F.J., Tonni, E.P., Bidegain, J.C., 2009. The stratigraphic range of the large canids (Carnivora, Canidae) in South America, and its relevance to the Quternary Biostratigraphy. Quat. Int. 210, 76–81. Ramı́rez, M.J., 2003. The spider subfamily Amaurobioidinae (Araneae, Anyphaenidae): a phylogenetic revision at generic level. Bull. Am. Mus. Nat. Hist. 277, 1–262. Rieppel, O., Kearney, M., 2002. Similarity. Biol. J. Linn. Soc. 75, 59– 82. Rossie, J.B., 2006. Ontogeny and homology of the paranasal sinuses in Platyrrhini (Mammalia: Primates). J. Morphol. 267, 1–40. Roy, M.S., Smith, D., Wayne, R.K., 1996. Molecular genetics of pre1940 red wolves. Conserv. Biol. 10, 1413–1424. Salesa, M.J., Antón, M., Peigné, S., Morales, J., 2006. Evidence of a false thumb in a fossil carnivore clarifies the evolution of pandas. Proc. Natl Acad. Sci. USA 103, 379–382. Schwartz, M., 1997. A History of Dogs in the Early Americas. Yale University Press, New Haven ⁄ London. Sillero Zubiri, C., Hoffmann, M., Macdonald, D.W., 2004. Canids: Foxes, Wolves, Jackals and Dogs. Status Survey and Conservation Action Plan. IUCN Species Programme, Gland, Switzerland. Simpson, G.G., 1945. The principles of classification and a classification of the mammals. Bull. Am. Mus. Nat. Hist. 8, 1–350. Smith, A.B., 1998. What does palaeontology contribute to a molecular world? Mol. Phylogenet. Evol. 9, 437–447. Smith, T.D., Rossie, J.B., 2006. Primate olfaction: anatomy and evolution. In: Brewer, W.J., Castle, D., Pantelis, C. (Eds), 475 Olfaction and the Brain. Cambridge University Press, New York, pp. 135–166. Smith, N.D., Turner, A.H., 2005. MorphologyÕs role in phylogeny reconstruction: perspectives from paleontology. Syst. Biol. 54, 166– 173. Smith, T.D., Rossie, J.B., Bhatnagar, K.P., 2007. Evolution of the nose and nasal skeleton in primates. Evol. Anthropol. 16, 132– 146. Soibelzon, L., Prevosti, F.J., 2007. Los carnı́voros (Carnivora, Mammalia) terrestres del cuaternario de América del Sur. In: Pons, G.X., Vicens, D. (Eds.), Geomorfologia Litoral i Quaternari. Homenatge a D. Joan Cuerda Barceló. Monogr. Soc. Hist. Nat. Balears 12, 49–68. Springer, M.S., Meredith, R.W., Eizirik, E., Teeling, E., Murphy, W.J., 2007. Morphology and placental mammal phylogeny. Syst. Biol. 56, 673–84. Springer, M.S., Burk-Herrick, A., Meredith, R., Eizirik, E., Teeling, E., OÕBrien, S.J., Murphy, W.J., 2008. The adequacy of morphology for reconstructing the early history of placental mammals. Syst. Biol. 57, 499–503. Stain, H.J., 1975. Calcanea of members of the Canidae. Bull. South. Calif. Acad. Sci. 74, 43–155. Strong, E.E., Lipscomb, D., 1999. Character coding and inapplicable data. Cladistics 11, 363–371. Szuma, E., 2002. Dental polymorphism in a population of the red fox (Vulpes vulpes) from Poland. J. Zool. Lond. 256, 243–253. Szuma, E., 2003. Microevolutionary trends in the dentition of the Red fox (Vulpes vulpes). J. Zoolog. Syst. Evol. Res. 41, 47–56. Szuma, E., 2004. Evolutionary implications of morphological variation in the lower carnassial of red fox Vulpes vulpes. Acta Theriol. 49, 433–447. Szuma, E., 2008a. Evolutionary and climatic factors affecting tooth size in the red fox Vulpes vulpes in the Holarctic. Acta Theriol. 53, 289–332. Szuma, E., 2008b. Geographic variation of tooth and skull sizes in the arctic fox Vulpes (Alopex) lagopus. Ann. Zool. Fenn. 45, 185–199. Szuma, E., 2008c. Geography of sexual dimorphism in the tooth size of the red fox Vulpes vulpes (Mammalia, Carnivora). J. Zoolog. Syst. Evol. Res. 46, 73–81. Tedford, R.H., 1976. Relationship of pinnipeds to other carnivores (Mammalia). Syst. Zool. 25, 363–374. Tedford, R., Taylor, B.E., Wang, X., 1995. Phylogeny of the Caninae (Carnivora: Canidae): the living taxa. Am. Mus. Nov. 3146, 1–37. Tedford, R.H., Wang, X., Taylor, B.E. 2009. Phylogenetic systematics of the North American fossil Caninae. Bull. Am. Mus. Nat. Hist 325, 1–218. Tedrow, A.R., Baskin, J.A., Robison, S.F., 1999. An additional occurrence of Simocyon (Mammalia, Carnivora, Procyonidae) in North America. Utah Geol. Surv. Misc. Pub. 99, 487–493. Torres Roldán, V.E., 1980. El Significado Paleontológico-Estratigráfico de la Mastofaunula Local Algodones, Plioceno tardı́o de Baja California Sur, México. PhD Thesis, Universidad Nacional Autónoma de México, Facultad de Ciencias, México DF. Torres Roldán, V., Ferusquı́a Villafranca, I., 1981. Cerdocyon sp. nov. A (Mammalia, Carnivora) en México y su significación evolutiva y zoogeográfica en relación a los cánidos sudamericanos. Anais do II Congreso Latino-Americano de Paleontologı́a (Porto Alegre, 1981) 2, 709–719. Van Valkenburgh, B., 1991. Iterative evolution of hypercarnivory in canids (Mammalia: Carnivore): evolutionary interactions among sympatric predators. Paleobiology 17, 340–362. Varón, A., Vinh, L.S., Bomash, I., Wheeler, W.C., 2009. POY 4.1.1. American Museum of Natural History. http://research.amnh.org/ scicomp/projects/poy.php. Wang, X., 1993. Transformation from plantigrady to digitigrady: functional morphology of locomotion in Hesperocyon (Canidae: Carnivora). Am. Mus. Nov. 3069, 1–23. 476 F.J. Prevosti / Cladistics 26 (2010) 456–481 Wang, X., 1994. Phylogenetic systematics of the Hesperocyonidae (Carnivora: Canidae). Bull. Am. Mus. Nat. Hist. 221, 1–207. Wang, X., 1997. New cranial material of Simocyon from China, and its implications for phylogenetic relationship to the red panda (Ailurus). J. Vert. Paleontol. 17, 184–198. Wang, X., Tedford, R.H., 1994. Basicranial anatomy and Phylogeny of primitive canids and closely related miacids (Carnivora: Mammalia). Am. Mus. Nov. 3092, 1–34. Wang, X, Tedford, R.H., Taylor, B.E., 1999. Phylogenetic systematics of the Borophaginae (Carnivora: Canidae). Bull. Am. Mus. Nat. Hist. 243, 1–391. Wang, X., Tedford, R.H., Van Valkenburgh, B., Wayne, R.K., 2004a. Ancestry. Evolutionary history, molecular systematics, and evolutionary ecology of Canidae. In: MacDonald, D.W., Sillero Zubiri, C. (Eds.), Biology and Conservation of Wild Canids. Oxford University Press, Oxford, pp. 31–54. Wang, X., Tedford, R. H, Van Valkenburgh, B., Wayne, R.K., 2004b. Phylogeny, classification, and evolutionary ecology of the canidae. In: Sillero Zubiri, C., Hoffmann, M., Macdonald, D.W. (Eds.), Foxes, Wolves, Jackals and Dogs. Status Survey and Conservation Action Plan. IUCN Species Programme, Gland, Switzerland, pp. 8–20. Wang, X., Tedford, R.H., Antón, M., 2008. Dogs: Their Fossil Relatives and Evolutionary History. Columbia University Press, New York. Wayne, R.K., OÕBrien, S.J., 1987. Allozyme divergence within the Canidae. Syst. Zool. 36, 339–355. Wayne, R.K., Geffen, E., Girman, D.J., Koepfli, K.P., Lau, L.M., Marshall, C.R., 1997. Molecular systematics of the Canidae. Syst. Biol. 46, 622–653. Wheeler, W., 1992. Extinction, sampling and molecular phylogenetics. In: Novacek, M., Wheeler, Q. (Eds.), Extinction and Phylogeny. Columbia University Press, New York, pp. 205–215. Wheeler, W., 1996. Optimization alignment: the end of multiple sequence alignment in phylogenetics? Cladistics 12, 1–9. Wheeler, W., 2003. Implied alignment: a synapomorphy-based multiple-sequence alignment method and its use in cladogram search. Cladistics 19, 261–268. Wiens, J.J., 2000. Coding morphological variation within species and higher taxa for phylogenetic analysis. In: Wiens, J.J. (Ed.), Phylogenetic Analysis of Morphological Data. Smithsonian Institution Press, Washington, pp. 115–145. Wiens, J.J., 2003a. Missing data, incomplete taxa, and phylogenetic accuracy. Syst. Biol. 52, 528–538. Wiens, J.J., 2003b. Incomplete taxa, incomplete characters, and phylogenetic accuracy: is there a missing data problem? J. Vert. Paleontol. 23, 297–310. Wiens, J.J., 2004. The role of morphological data in phylogeny reconstruction. Syst. Biol. 53, 653–661. Wiens, J.J., 2006. Missing data and the design of phylogenetic analyses. J. Biomed. Inform. 39, 34–42. Wiens, J.J., Servedio, M.R., 1997. Accuracy of phylogenetic analysis including and excluding polymorphic characters. Syst. Biol. 46, 332–345. Wilkinson, M., 1995a. A comparison of two methods of character construction. Cladistics 11, 297–308. Wilkinson, M., 1995b. Coping with abundant missing entries in phylogenetic inference using parsimony. Syst. Biol. 44, 501–514. Wilkinson, M., 2003. Missing entries and multiple trees: instability, relationships, and support in parsimony analysis. J. Vert. Paleontol. 23, 311–323. Wilson, D.E., Reeder, D.M., 2005. Mammal Species of the World. A Taxonomic and Geographic Reference, 3rd edn. Johns Hopkins University Press, Baltimore. Wilson, P.J., Grewal, S.K., Mallory, F.F., White, B.N., 2009. Genetic characterization of hybrid wolves across Ontario. J. Hered. 100, S80–S89. Winge, H., 1895. Jordfunde og nulevende Rovdyr (Carnivora) fra Lagoa Santa, Minas Geraes, Brasilien. E Museo Lundii 2, 1–103. Wozencraft, W.C., 2005. Order Carnivora. In: Wilson, D.E., Reeder, D.M. (Eds.), Mammal Species of the World: A Taxonomic and Geographic Reference. Johns Hopkins University Press, Baltimore, pp. 279–348. Zar, J.H., 1984. Biostatistical Analysis, 2nd edn. Prentice-Hall, Englewood Cliffs, New Jersey. Zhao, Z., Weiss, K.M., Stock, D.W., 2000. Development and evolution of dentition patterns and their genetic basis. In: Teaford, M.F., Smith, M.M., Ferguson, M.W.J. (Eds.), Development, Function and Evolution of Teeth. Cambridge University Press, Cambridge, UK, pp. 152–172. Zrzavý, J., Řičánková, V., 2004. Phylogeny of recent Canidae (Mammalia, Carnivora): relative reliability and the utility of morphological and molecular datasets. Zool. Scr. 33, 311–333. Supporting Information Additional Supporting Information may be found in the online version of this article: Appendix S1. Specimens studied for the character analysis of osteological features. Appendix S2. Figure of some characters used in the phylogenetic analysis. Appendix S3. Combined matrix in TNT format. Appendix S4. Genbank accesion numbers. Please note: Wiley-Blackwell is not responsible for the content or functionality of any supplementary materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article. Appendix 1 List of the osteological characters used in the phylogenetic analysis Continuous 0. Length of the rostrum (modified from character no. 1 of Wang, 1994 and no. 1 of Wang et al., 1999): condylobasal length ⁄ length from the occipital condyles to the anterior margin of the orbits (hereafter LOO). 1. Palatal width (modified from character no. 6 of Berta, 1989, no. 41 of Tedford et al., 1995; and no. 14 of Wang et al., 1999): LOO ⁄ width of the palate at the P4–M1 junction. 2. Width of the frontals (modified from character no. 3 of Wang et al., 1999): width of the braincase ⁄ width of the frontals at the postorbital process. 3. Size of the tympanic bulla: LOO ⁄ anteroposterior length of the bulla. 4. Width of the scar left by the superficial masseteric muscle on the zygomatic bone (character no. 2 of Berta, 1987; 1988, no. 36 of Tedford et al., 1995; and no. 6 of Wang et al., 1999): zygomatic height ⁄ width of the anterior portion of the superficial masseteric muscle scar. Previous authors had codified this character as 0: scar uniformly wide; 1: very widened at its anterior portion (e.g. Berta, F.J. Prevosti / Cladistics 26 (2010) 456–481 1988; Tedford et al., 1995). Our observations showed that this character presents a continuous variation, so it is not possible to separate it unambiguously into discrete states. 5. Robustness of the body of the mandible (modified from character no. 5 of Tedford et al., 1995 and no. 23 of Wang et al., 1999): (height · width of the mandible at the distal border of the m1) ⁄ LOO. 6. Shape of the coronoid process (character no. 11 of Berta, 1988, and no. 37 of Tedford et al., 1995): height of the coronoid process ⁄ length of the coronoid process. 7. Relative size of the I3 (modified from character no. 12 of Berta, 1988, no. 26 of Wang, 1994; no. 53 of Tedford et al., 1995; and no. 32 of Wang et al., 1999): Length of the I3 ⁄ length of the I2. 8. Height of the canine (modified from character no. 27 of Tedford et al., 1995): height of the crown of the C1 ⁄ length of the C1. 9. Relative size of the p4 (modified from character no. 20 of Tedford et al., 1995; and no. 63 of Wang et al., 1999): length of the p4 ⁄ length of the p3. 10. Relative size of the lower carnassial (m1) (modified from character no. 15 of Berta, 1988, no. 34 of Wang, 1994; no. 39 of Tedford et al., 1995; and no. 43 of Wang et al., 1999): LOO ⁄ length of the m1. 11. Relative length of the trigonid of the m1 (modified from characters no. 66–67 of Wang et al., 1999): length of the m1 ⁄ length of the trigonid of the m1. 12. Length of anterior limb relative to the posterior limb (Tedford et al., 2009): functional length of the tibia, from proximal to distal articulation surface excluding the lateral malleolus ⁄ functional length of the radius, from proximal to distal articulation surface excluding the styloid process. 13. Functional length of the tibia, from proximal to distal articulation surface excluding the lateral malleolus ⁄ length of the femur, from the head to the distal articulation (Tedford et al., 2009). Discrete–Cranial 14. Orientation of the paraoccipital process (modified from characters no. 10 and 31 of Tedford et al., 1995; no. 14, 15 and 17 of Wang, 1994; and no. 20 of Wang et al., 1999): 0: directed posteriorly; 1: expanded ventrocaudally; 2: mostly ventral, but with a gentle posterior inclination that forms a small angle (ca. 15) with the distal border of the tympanic bulla; 3: directed ventrally and parallel to the distal border of the bulla in lateral view. 15. Level of fusion of the paraoccipital process with the tympanic bulla (modified from characters no. 10 and 31 of Tedford et al., 1995; nos. 14, 15 and 17 of Wang, 1994): 0: the paraoccipital process is basally fused to the bulla; 1: fused along all its extension (e.g. Vulpes, Cerdocyon); 2: as in state 1 but with a small free distal process. 16. Shape of the paraoccipital process (character no. 16 of Wang, 1994 and no. 18 of Tedford et al., 1995): 0: narrow with a straight medial border; 1: wide with a convex medial border. 17. Shape of the supraoccipital shield in caudal view (modified from character no. 4 of Berta, 1988, no. 55 of Tedford et al., 1995; and no. 3 of Wang et al., 1999): 0: wide and rectangular ⁄ convex (e.g. Vulpes, L. griseus, L. gymnocercus (Fischer, 1814), Cerdocyon); 1: narrow and triangular, with the occipital crests forming an acute angle where they converge (e.g. C. latrans, C. aureus, L. pictus, D. avus, L. culpaeus); 2: narrow but convex (e.g. Theriodictis, Protocyon, C. lupus, C. dirus, C. nehringi); 3: very narrowed (C. brachyurus) (Appendix S2). 18. Caudal extension of the inion (modified from character no. 55 of Tedford et al., 1995): 0: not extended beyond the occipital condyles; 1: prolonged beyond the occipital condyles, with a triangular outline in lateral view. 19. Postparietal foramen (character no. 11 of Tedford et al., 1995): 0: present; 1: absent. 20. Orbital fissure and optic canal (character no. 7 of Berta, 1988): 0: are independent foramina; 1: open in a common pit. 21. Shape of the mastoid process (modified from character no. 18 of Wang, 1994; no. 33 of Tedford et al., 1995, and no. 19 of Wang, 1999): 0: a small bulge transversally compressed (like a crest) and separated 477 from the lambdoid crest by a shallow depression; 1: a strong tubercle separated from the lambdoid crest by a deep depression (Appendix S2). 22. Shape of the mastoid process: 0: long and directed anteroventrally; 1: short and oriented ventrally. 23. Fontal sinuses (modified from Huxley, 1880; character no. 1 of Berta, 1988, no. 4 of Wang, 1994; and no. 32 of Tedford et al., 1995): 0: absent, with a clear depression on the dorsal surface of the postorbital process; 1: present. For scoring the skulls that were not dissected I followed the criterion of Berta (1988) and Tedford et al. (1995). Based on ontogenetic studies of living primates, recent authors (Rossie, 2006; Smith and Rossie, 2006; Smith et al., 2007), separated recesses from sinuses due to the existence of a secondary pneumatization. Albeit that it is not possible to classify these structures in adult specimens, several authors (Moore, 1981; Rossie, 2006; Smith and Rossie, 2006) considered them as homologues to the frontal sinuses of other mammals. 24. Development of the frontal sinuses (modified from character no. 1 of Berta, 1988, no. 4 of Wang, 1994; no. 32 of Tedford et al., 1995; and no. 4 of Wang et al., 1999): 0: not invading the postorbital processes; 1: invading the postorbital processes. 25. Caudal extension of the frontal sinuses (modified from character no. 1 of Berta, 1988, no. 4 of Wang, 1994; no. 32 of Tedford et al., 1995; and no. 4 of Wang et al., 1999): 0: not expanded; 1: expanded caudally, reaching the frontoparietal suture and ⁄ or going beyond it. 26. Orbital region of the zygomatic bone (character no. 35 of Tedford et al., 1995 and no. 7 of Wang et al., 1999): 0: presence of a lateral flare and eversion of the dorsal border; 1: dorsal border thickened. 27. Length of the orbital process of the zygomatic bone (character no. 52 of Tedford et al., 1995): 0: extended and contacting the lacrimal bone; 1: not extended anteriorly and not contacting the lacrimal bone. 28. Shape of the zygomatic arch in lateral view (modified from character no. 3 of Berta, 1988 and no. 4 of Tedford et al., 1995): 0: flat to moderately arched; 1: strongly arched dorsoventrally (Appendix S2). 29. Position of the scar of the superficial massseteric on the zygomatic bone: 0: oriented ventrolaterally; 1: oriented ventrally. 30. Caudal extension of the palate (modified from character no. 5 of Berta, 1988 and no. 26 of Tedford et al., 1995): 0: up to the distal border of the last upper molar or not even reaching it; 1: beyond the last upper molar. 31. Shape of the tip of the frontal process of the maxillary (modified from character no. 2 of Wang, 1994): 0: convex; 1: with its medial border very straight. 32. Shape of the infraorbitary foramen (character no. 5 of Wang, 1994): 0: rounded; 1: elliptic, laterally narrowed. 33. Premaxillary–frontal contact (modified from character no. 2 of Wang et al., 1999): 0: absent; 1: present. 34. Postorbital constriction: 0: abrupt and narrow; 1: gradual and wide (Appendix S2). 35. Sagittal crest (modified from characters no. 9 and 10 of Wang et al., 1999): 0: absent or limited to the occipital part of the braincase, the temporal ‘‘crests’’ forming a lyriform area; 1: similar to state 0, but the temporal ‘‘crests’’ very sharp and strong; 2: well developed and extended, at least from the frontoparital suture to the occipital suture. The polymorphism of this character can be partially explained by the presence of sexual dimorphism. Females usually have sagittal crests less developed than males. 36. Position of the anterior border of the orbits related to the cheekteeth: 0: over the mesial part of the P4; 1: over the distal half of the P4; 2: over the M1. To score this character the anterior margin of the orbit was established as a plane tangent to this margin and perpendicular to the palate (Appendix S2). 37. Development of the lateral wall of the alisphenoid canal: 0: long and hiding the foramen rotundum; 1: short and not covering the foramen rotundum (Appendix S2). 478 F.J. Prevosti / Cladistics 26 (2010) 456–481 38. Nasal extension related to the frontomaxillar suture (modified from character no. 19 of Tedford et al., 1995): 0: short and before the suture; 1: at the level of the suture; 2: extended beyond the suture. 39. Nasal shape in dorsal view: 0: with a widened middle section; 1: gradually narrowed in caudal direction and with a concave lateral border (Appendix S2). 40. Caudal extension of the vomer: 0: with a little tip beyond the palate; 1: clearly expanded beyond the palate, but not covering all the width of the mesopterygoid fossa; 2: very expanded and covering all the width of the mesopterygoid fossa (Appendix S2). 41. Shape of the presphenoid: 0: distal half short and wide; 1: distal half long and narrow. This bone has a rectangular outline in ventral view, with two small processes on the middle its sides. The anterior half is usually long and very narrowed (Appendix S2). 42. Shape of the nasal spine: 0: reduced, nearly absent; 1: well developed, but simple; 2: bifid. 43. Position of the canal of the facial nerve related to the canal of the acoustic nerve in the dorsal face of the petrosal: 0: anterodorsal; 1: anteriorly placed (Appendix S2). 44. Suprameatal fossa (character no. 7 of Wang, 1994 and no. 8 Wang and Tedford, 1994). 0: present; 1: absent. 45. Laterocaudal angle of the tympanic bulla: 0: rounded; 1: with a sharp angle (Appendix S2). 46. Shape of the tympanic bulla: 0: inflated; 1: depressed (Appendix S2). 47. Ventral intrabular septum (modified from Ivanoff, 2000, 2006): 0: incomplete, interrupted at the middle of the lateral wall of the bulla; 1: continuous all over the lateral wall of the bulla (Appendix S2). Speothos venaticus was coded following Ivanoff (2000, 2006). In some specimens of Cerdocyon thous and Canis mesomelas, and in M. coryphaeus, this septum is continuous but very low, thus was coded as ‘‘0’’. 48. External auditive meatum (character no. 8 of Berta, 1988 and no. 18 of Wang et al., 1999): 0: well developed; 1: scarcely developed (Appendix S2). 49. Depression of the orbits at the interorbital constriction: 0: absent, the border of the orbit is smoothly continuous; 1: present, with a sharp medial indention. 50. Subangular lobule of the mandible (modified from character no. 9 of Berta, 1988, no. 24 of Tedford et al., 1995; and no. 27 of Wang et al., 1999): 0: smooth or absent 1: sharp and wide; 2: well developed but laterally compressed (Appendix S2). 51. Inferointernal facet of the angular process (modified from character no. 10 of Berta, 1988, nos. 20–21 of Wang, 1994; and nos. 38, 54 of Tedford et al., 1995): 0: not expanded; 1: expanded, with a rectangular or quadrangular outline (Appendix S2). The inferior branch of the medial pterygoid muscle gets inserted in this fossa (Gaspard, 1964; Berta, 1988; Tedford et al., 1995). I did not create another state for the variation observed in the state one (rectangular vs. quadrangular outline) because this shape is correlated with the shape of the angular process (character no. 52, vide infra). 52. Shape of the angular process (modified from character no. 10 of Berta, 1988, nos. 20–21 of Wang, 1994; and nos. 38–54 of Tedford et al., 1995): 0: dorsoventrally narrowed, with a concave dorsal border (forming a dorsal hook); 1: high and without a dorsal hook, but longer than higher (‘‘rectangular’’); 2: quadrangular, as long as high (Appendix S2). Nyctereutes procyonoides has a deep process that looks like the quadrangular state, but it is more triangular than in Cerdocyon thous or Speothos, thus I preferred to score it as ‘‘0’’. 53. Height of the mandibular condyle: 0: low, at the level of the talonid of the m1; 1: high, reaching the tip of the trigonid of the m1. Discrete–dental 54. I2–I3 diastema (modified from character no. 23 of Tedford et al., 1995): 0: I2–I3 in contact (diastema absent); 1: present (Appendix S2). 55. Mesiolingual cingulum of the I3 (modified from character no. 12 of Berta, 1988 and no. 53 of Tedford et al., 1995): 0: reduced; 1: well developed, forming a sharp crest (Appendix S2). 56. Mesial accessory cusp of the I3 (modified from character no. 17 of Tedford et al., 1995): 0: present; 1: absent (Appendix S2). 57. Mesial accessory cusp of the I1 (modified from character no. 17 of Tedford et al., 1995): 0: present; 1: absent (Appendix S2). 58. Mesial accessory cusp of the I2 (modified from character no. 17 of Tedford et al., 1995): 0: present; 1: absent (Appendix S2). 59. Distal accessory cusp of the I2 (modified from character no. 17 of Tedford et al., 1995): 0: present; 1: absent (Appendix S2). 60. Diastemata between the lower premolars (i.e. p2–p3–p4; modified from character no. 7 of Tedford et al., 1995 and no. 42 of Wang et al., 1999): 0: present; 1: absent (Appendix S2). 61. Morphology of the principal cusp of the premolars (i.e. P3, p3– p4; modified from character no. 20 of Tedford et al., 1995 and no. 37 of Wang et al., 1999): 0: slender, tall and acute, with a concave distal border and the mesial border with an inclination of more than 45º related to the crown base in labial view; 1: similar to state 0, but the cusp is more robust and wide; 2: very tall, weak and acute, but compressed mesiolingually and with distal and mesial borders almost vertical (forming a ca. 75º angle with the plane defined by the base of the crown); 3: robust and expanded mesiolingually, with a very inclined mesial border (45º or less); large and robust, but higher and acuter than in state 3, with more vertical distal and mesial borders (Appendix S2). 62. Distal accessory cusp of the P2 (modified from character no. 14 of Berta, 1988, no. 30 of Wang, 1994; no. 7 of Tedford et al., 1995; and no. 36 of Wang et al., 1999): 0: absent; 1: present. 63. Distal accessory cusp of the P3 (modified from character no. 14 of Berta, 1988, no. 30 of Wang, 1994; no. 7 of Tedford et al., 1995; and no. 36 of Wang et al., 1999): 0: absent; 1: present. 64. Development of the protocone of the P4 (modified from character no. 16 of Berta, 1988, no. 32 of Wang, 1994; and no. 44 of Wang et al., 1999): 0: forming a tall, well developed cusp, clearly separated from the rest of the tooth (with a circular outline in the occlusal view); 1: the protocone forms a well developed cusp but little separated from the paracone and its diameter is less than half the width of the paracone; 2: protocone very reduced, forming a lobe in the mesiolingual face of the tooth (Appendix S2). 65. Parastyle of the P4 (modified from character no. 46 of Wang et al., 1999): 0: absent; 1: present. 66. Parastyle of the M1–M2 (modified from character no. 35 of Wang, 1994 and no. 14 of Wang and Tedford, 1994): 0: well developed; 1: reduced. 67. Mesial cingulum of the M1 (modified from character no. 24 of Berta, 1988): 0: well developed, forming a strong crest that could contact the hypocone; 1: reduced, as a weak rounded cingulum (Appendix S2). 68. Paraconule of the M1 (character no. 56 of Wang et al., 1999): 0: absent; 1: present (Appendix S2). 69. Hypocone of the M1 (modified from characters no. 20 and 22 of Berta, 1988, no. 1 of Tedford et al., 1995; and nos. 51 and 54 of Wang et al., 1999): 0: well developed; 1: reduced (Appendix S2). 70. Size of the paracone of the M1 (modified from character no. 18 of Berta, 1988, no. 36 of Wang, 1994; no. 45 of Tedford et al., 1995; and no. 55 of Wang et al., 1999): 0: lower to slightly higher than the metacone in labial view; 1: very high and conspicuously taller than the metacone. 71. Labial cingulum of the M1 (modified from cf. character no. 19 of Berta, 1988 and no. 46 of Tedford et al., 1995): 0: well developed along all the labial border of the crown; 1: reduced, restricted to the distal part of the metacone and the mesial part of the paracone (Appendix S2). 72. Metaconule of the M1 modified from character no. 53 of Wang et al., 1999).: 0: well developed forming a cusp; 1: reduced to an inflection of the postprotocrist (Appendix S2). 73. M2–M3 modified from character no. 32 of Berta, 1988, no. 40 of Wang, 1994; and no. 50 of Tedford et al., 1995): 0: M2 absent; 1: M2 present; 2: M3 present. F.J. Prevosti / Cladistics 26 (2010) 456–481 74. Size of the metacone of the M2 (modified from character no. 60 of Wang et al., 1999): 0: higher than the paracone; 1: well developed but slightly lower than the paracone; 2: reduced to the half of the paracone size; 3: very reduced, like a raised crest (Appendix S2). 75. Mesial extension of the hipocone of the M2 (modified from character no. 24 of Berta, 1988): 0: extended around the mesiolingual part of the protocone, and joining the mesial cingulum; 1: not surrounding the protocone (Appendix S2). 76. Number of roots of the M2 (modified from character no. 50 of Tedford et al., 1995): 0: three; 1: two. 77. Paraconule of the M2 (character no. 56 of Wang et al., 1999): 0: absent; 1: present. 78. Distal accessory cusp of the p2 (modified from character no. 14 of Berta, 1988, no. 30 of Wang, 1994; no. 7 of Tedford et al., 1995; and no. 36 of Wang et al., 1999): 0: present; 1: absent (Appendix S2). 79. Distal accessory cusp of the p3 (modified from character no. 14 of Berta, 1988, no. 30 of Wang, 1994; no. 7 of Tedford et al., 1995; and no. 36 of Wang et al., 1999): 0: present; 1: absent (Appendix S2). 80. Second distal accessory cusp of the p4 (modified from character no. 26 of Berta, 1988 and no. 34 of Tedford et al., 1995): 0: absent; 1: present (Appendix S2). 81. Mesial cingular cusp of the p2–p3 (character no. 57 of Tedford et al., 1995): 0: absent; 1: present (Appendix S2). Tedford et al. (1995) interpreted that in Cuon and Lycaon the mesial cusps derive from the mesial cristid and not from the mesial cingulum. In only one specimen of L. pictus (YPMMa 5465) this cusp is separated from the mesial cingulum, and it is not clear whether it comes form this cingulum or from the mesial cristid. In the remaining specimens the cusp is connected with the mesial cingulum, thus they are interpreted as ‘‘cingular cups’’. 82. Mesial cingular cusp of the p4 (modified from character no. 26 of Berta, 1988 and no. 22 of Tedford et al., 1995): 0: absent; 1: present (Appendix S2). See discussion on character 81. 83. Protostylid of the m1 (character no. 28 of Berta, 1988, no. 29 of Tedford et al., 1995; and no. 68 of Wang et al., 1999): 0: absent; 1: present (Appendix S2). 84. Metaconid of the m1 (modified from character no. 27 of Berta, 1988, no. 43 of Wang, 1994; and no. 42 of Tedford et al., 1995): 0: present; 1: absent (Appendix S2). 85. Development of the entoconid of the m1 modified from character no. 27 of Berta, 1988, no. 45 of Wang, 1994; no. 2 of Tedford et al., 1995; and no. 70 of Wang et al., 1999): 0: a sharp crest on the lingual part of the talonid; 1: forming a large, conical cusp, that could contact the base of the hypoconid and closing distally the talonid basin; 2: like state 1 but linked to the hypoconid by transverse crests; 3: reduced to a low lingual cingulum (Appendix S2). 86. Mesoconid of the m1 (character no. 29 of Berta 1988): 0: absent; 1: present (Appendix S2). 87. Entoconulid of the m1: 0: absent; 1: present (Appendix S2). 88. Metastylid of the m1 (modified from Berta, 1988). 0: absent; 1: present. 89. Hypoconulid of the m1 (character no. 13 of Tedford et al., 1995): 0: absent; 1 present (Appendix S2). 90. Development of the hypoconulid of the m1: 0: as a low cingulum that joins the distal margins of the hypoconid and entoconid; 1: elevated as a discrete cusp. 91. Mesiolabial cingulum of the m2 modified from character no. 30 of Berta, 1988 and no. 12 of Tedford et al., 1995): 0: well developed, and expanded to the distal border of the protoconid or the hypoconid; 1: limited to the mesiolingual face of the protoconid; 2: very reduced (Appendix S2). 92. Size of the metaconid of the m2 (modified from character no. 32 of Berta, 1988, no. 15 of Wang and Tedford, 1994; nos. 14 and 43 of Tedford et al., 1995; and no. 76 of Wang et al., 1999): 0: higher than the protoconid; 1: similar or slightly lower and smaller than the protoconid; 2: very reduced, with half of the protoconid size (Appendix S2). 479 93. Development of the entoconid of the m2 (cmodified from character no. 43 of Tedford et al., 1995): 0: reduced to a low cingulum; 1: well developed as a distinct cusp; 2: present, but divided in two cusps (Appendix S2). 94. Entoconulid of the m2: 0: absent; 1: present. 95. m3–m4 (character no. 32 of Berta, 1988 and no. 51 of Tedford et al., 1995): 0: m3 absent; 1: m3 present; 2: m4 present. Discrete–cerebrum 96. Proreal gyrus (Lyras and Van der Geer, 2003): 0: small; 1: well developed (Appendix S2). 97. Outline of the coronal sulci (modified from Lyras and Van der Geer, 2003): 0: not laterally expanded, and aligned with the lateral sulci; 1: ‘‘pentagonal’’, the sulci diverging more posteriorly than anteriorly; 2: ‘‘parenthesis-like’’, with the sulci bowed laterally taking the form of an oval; 3: ‘‘heart-shaped’’, the sulci are bowed laterally but diverge rostrally; 4: ‘‘orthogonal’’, laterally expanded in a more abruptly way, taking a bracketed shape (Appendix S2). Discrete–postcranial 98. Shape of the wings of the atlas: 0: rectangular, with straight lateral borders; 1: rounded or convex and laterally expanded (Appendix S2). 99. Shape of the lateral border of the fourth cervical vertebra in dorsal view: 0: with a small and deep notch immediately after the prezygapophysis; 1: with a depression that occupies the anterior half of the border; 2: straight or slightly concave; 3: concave (Appendix S2). 100. Size of the articulation between the second sacral vertebra and the ilium related to the size of the articulation of the first sacral vertebra with the illium (Hildebrand, 1954): 0: smaller, with less than quarter of the size of the articulation of the first sacral vertebra; 1: with a similar size (Appendix S2). 101. Shape and orientation of the transversal processes of the third sacral vertebra (Hildebrand, 1954): 0: long and caudolaterally expanded; 1: short and directed caudally; 2: similar to state 1 but laterally compressed (Appendix S2). 102. Scar of the attachment of the serratus and rhomboideus muscles in the scapula (Hildebrand, 1954): 0: restricted to the cranial part of the dorsal part of the scapula; 1: like state 0 but slightly expanded in caudal direction; 2: more caudally expanded, reaching the caudal border of the scapula (Appendix S2). 103. Position of the scar of the teres major muscle on the scapula (Hildebrand, 1954): 0: facing laterally (it can be completely seen in lateral view); 1: lateroventral, with an intermediate position between status 0 and 2; 2: limited to the ventral border of the scapula (Appendix S2). 104. Position of the scar of the pronator teres muscle related to the scar of the carpal and digital flexor muscles on the distal portion of the humerus (Hildebrand, 1954): 0: the medial epicondylus is reduced and the scars are at the same level; 1: the facets for the carpal and digital flexor muscles are placed more distally, at the tip of a more developed epicondylar process; 2: similar to state 1, but the process is more distally expanded beyond the distal limit of the trochlea (Appendix S2). 105. Shape of the capitulum of the distal articulation of the humerus: 0: rounded and well differentiated from the rest of the articulation; 1: less differentiated and more depressed (Appendix S2). 106. Shape and height of the head of the humerus related to the greater tubercle (modified from Wang, 1993): 0: similar height, and with a rounded, more or less spherical head; 1: the head is flat and lower than the greater tubercle. 107. Lesser tubercle of the humerus (modified from Wang, 1993): 0: large and laterally expanded; 1: reduced. 108. Development of the lateral epicondylar crest of the humerus (modified from Wang, 1993): 0: well developed and medially extended; 1: reduced. 109. Extension of the deltoid crest of the humerus (modified from Wang, 1993): 0: distally extended beyond the middle point of the shaft; 1: not extended beyond the proximal third of the shaft. 480 F.J. Prevosti / Cladistics 26 (2010) 456–481 110. Olecranon fosa of the humerus (modified from Wang, 1993): 0: shallow; 1: deep. 111. Entoepicondylar foramen (Wang, 1993; character no. 16 of Tedford et al., 1995): 0: present; 1: absent. 112. Diaphysis of the radius (modified from Hildebrand, 1954; Wang, 1993): 0: caudally curved, with a robust distal end; 1: more slender, less curved and more uniform in width; 2: similar to state 1, but more slender and nearly straight (Appendix S2). 113. Medial process of the distal part of the radius: 0: very large, expanded medially and independent from the shaft; 1: reduced as an angle of the shaft. 114. Thickness of the shaft of the ulna: 0: similar to the shaft of the radius (cf. Hildebrand, 1954; Wang, 1993); 1: more slender. 115. Caudal border of the proximal half of the ulna shaft (modified from Wang, 1993; Hildebrand, 1954): 0: straight to convex; 1: concave (Appendix S2). 116. Length of the olecranon: 0: long (longer than half of the maximum diameter of the sigmoid cavity); 1: short (less than the half of the maximum diameter of the sigmoid cavity) (Appendix S2). 117. Orientation of the caudal border of the pubis in caudal view (Wang, 1993): 0: forming an angle of ca. 90 between left and right pubis; 1: forming an angle of ca. 180. 118. Caudal ventral illiac spine : 0:absent; 1: present (Appendix S2). 119. Ischiatic tuberosity (cf. Hildebrand, 1954): 0: robust and expanded laterally; 1: scarcelly developed and slightly expanded in lateral direction. 120. Third trochanter of the femur (Wang, 1993): 0: small, but present as a process; 1: reduced to a low crest. 121. Position of the scars of the middle and deep gluteal muscles (Wang, 1993): 0: in contact or near; 1: well separated. 122. Height of the greater trochanter of the femur related to its head: 0: lower; 1: subequal in height. 123. Anterior projection of the tibial tuberosity (modified from Hildebrand, 1954; Wang, 1993): 0: scarcely projected, the tibial crest is gradually lowered in distal direction in lateral view; 1: well expanded, possesses having a truncated shape in lateral view (Appendix S2). 124. Muscular groove of the tibia (cf. Wang, 1993): 0: deep and semicircular, anteriorly limited by a strong knot; 1: very shallow. 125. Scars of the caudal tibial muscle and the middle head of the flexor digitorum longus muscle in the tibia (cf. Wang, 1993): 0: wide and expanded; 1: narrowed and displaced proximomedially. This facet is more narrowed and displaced proximomedially in T. platensis than in the other canines studied (Appendix S2). 126. Shape of the distal articulation of the tibia in anterior view: 0: the groove of the medial ‘‘crest’’ of the astragalum trochlea is very deep and is laterally flanked by a strong ‘‘process’’; 1: this groove is very shallow and the ‘‘process’’ weak (Appendix S2). 127. Shape of the sustentacular facet of the calcaneus: 0: long and triangular, with a long and acute distal end; 1: subelliptic, with the distal end little narrowed; 2: rounded or roughly squared (Appendix S2). 128. Astragalar foramen (Wang, 1993): 0: present; 1: absent (Appendix S2). 129. Plantar tendinal groove of the astragalus (Wang, 1993): 0: present; 1: absent (Appendix S2). 130. First digit of the front foot (character no. 56 of Tedford et al., 1995): 0: present; 1: absent. 131. Development of the first metatarsal (modified from Wang, 1993; and character15 of Tedford et al., 1995): 0: long (larger than half of the length of the third metatarsal); 1: reduced to less than half of the length of the third metatarsal. One specimen of C. latrans (YPMMa 16001) presented a long first metatarsal that approached the state 0, but the development was asymmetric and was larger in the left foot than in the right one. Albeit this aberrant specimen, C. latrans was coded with the state 1 for this character. 132. Shape of the baculum (cf. Wang, 1994): 0: long and curved; 1: short and straight (Appendix S2). Appendix 2 List of osteological synapomorphies and autopomorphies of the tree obtained with the simultaneous analysis under equal weights (Fig. 4). Character numbers and states as in character analysis (Appendix 1), and node numbers as in Fig. 4. Urocyon cinereoargenteus: Char. 2: 1.014–1.018 fi 0.856–0.996; Char. 8: 1.022 fi 0.806–0.975; Char. 9: 1.020 fi 1.066–1.191; Char. 50: 0 fi 2; Char. 53: 0 fi 1; Char. 91: 1 fi 0; Char. 116: 0 fi 1. Atelocynus microtis: Char. 4: 0.855–0.922 fi 0.570–0.779; Char. 6: 0.939–0.943 fi 0.758–0.924; Char. 7: 1.026–1.037 fi 0.850–0.960; Char. 10: 0.930–1.004 fi 1.031–1.124; Char. 13: 0.979–1.000 fi 0.900– 0.971; Char. 24: 1 fi 0; Char. 30: 0 fi 1; Char. 41: 0 fi 1; Char. 42: 1 fi 2; Char. 52: 0 fi 1; Char. 97: 2 fi 4. Canis aureus: Char. 0: 1.043–1.053 fi 0.986–1.010; Char. 4: 1.002– 1.024 fi 0.852–0.905; Char. 5: 0.930–1.091 fi 0.832–0.904; Char. 6: 0.939–0.943 fi 0.846–0.878; Char. 33: 0 fi 1. Canis latrans: Char. 64: 1 fi 0. Canis lupus: Char. 9: 1.018–1.020 fi 0.966–1.006; Char. 71: 0 fi 1; Char. 103: 1 fi 2. Canis mesomelas: Char. 2: 0.969–0.980 fi 0.871–0.954; Char. 9: 1.031–1.043 fi 1.047; Char. 10: 0.899–0.964 fi 0.887; Char. 12: 0.936– 0.938 fi 0.903–0.925; Char. 13: 0.999–1.000 fi 1.016–1.043; Char. 15: 2 fi 1; Char. 79: 1 fi 0; Char. 87: 1 fi 0. Canis simensis: Char. 0: 1.043–1.064 fi 1.082–1.093; Char. 1: 0.946– 0.958 fi 1.084–1.104; Char. 2: 0.969–1.008 fi 1.101–1.129; Char. 3: 1.004–1.029 fi 0.945–1.000; Char. 6: 0.939–0.943 fi 0.874–0.917; Char. 9: 1.018–1.033 fi 0.973; Char. 10: 0.899–0.964 fi 0.981; Char. 36: 1 fi 2; Char. 40: 1 fi 2; Char. 83: 1 fi 0; Char. 87: 1 fi 0; Char. 93: 0 fi 1. Cerdocyon thous: Char. 28: 1 fi 0; Char. 34: 0 fi 1; Char. 50: 0 fi 1; Char. 52: 0 fi 2; Char. 53: 0 fi 1. Chrysocyon brachyurus: Char. 2: 1.014–1.018 fi 0.769–0.879; Char. 6: 0.939–0.943 fi 0.862–0.936; Char. 13: 0.979–1.000 fi 1.016–1.057; Char. 29: 0 fi 1; Char. 31: 0 fi 1; Char. 36: 1 fi 2; Char. 41: 0 fi 1; Char. 58: 0 fi 1; Char. 90: 0 fi 1; Char. 97: 2 fi 3; Char. 101: 0 fi 2; Char. 103: 1 fi 0; Char. 104: 1 fi 0; Char. 112: 1 fi 2; Char. 116: 0 fi 1. Cuon alpinus: Char. 12: 0.936–0.938 fi 1.063; Char. 31: 0 fi 1; Char. 37: 1 fi 2; Char. 40: 1 fi 0. Dusicyon australis: Char. 2: 1.086 fi 1.106; Char. 9: 1.020 fi 1.072; Char. 11: 1.033–1.062 fi 1.267; Char. 50: 0 fi 1; Char. 64: 0 fi 2; Char. 89: 1 fi 0. Lycalopex culpaeus: Char. 5: 0.630–0.711 fi 0.836–0.912; Char. 17: 0 fi 1. Lycalopex gymnocercus: Char. 13: 1.012–1.028 fi 1.030–1.049. Lycalopex sechurae: Char. 1: 1.014–1.027 fi 1.035–1.130; Char. 7: 0.992–1.012 fi 0.875–0.939; Char. 10: 0.979–1.004 fi 1.060–1.227; Char. 79: 0 fi 1. Lycalopex vetulus: Char. 0: 0.982–0.992 fi 0.916–0.970; Char. 3: 0.974–0.998 fi 0.778–0.912; Char. 5: 0.627–0.630 fi 0.470–0.546; Char. 40: 1 fi 0; Char. 77: 0 fi 1. Lycalopex fulvipes: Char. 3: 1.008–1.012 fi 1.034; Char. 4: 1.037– 1.062 fi 1.097; Char. 6: 0.945–0.958 fi 0.904; Char. 30: 1 fi 0. Lycaon pictus: Char. 12: 0.936–0.938 fi 0.868–0.922; Char. 25: 0 fi 1; Char. 61: 3 fi 4; Char. 102: 1 fi 2; Char. 103: 1 fi 2; Char. 130: 0 fi 1. Nyctereutes procyonoides: Char. 48: 0 fi 1; Char. 60: 1 fi 0; Char. 101: 0 fi 1; Char. 102: 0 fi 1; Char. 114: 1 fi 0; Char. 123: 1 fi 0. Otocyon megalotis: Char. 1: 1.037–1.045 fi 1.078–1.164; Char. 3: 0.898–0.900 fi 0.816–0.882; Char. 10: 1.093–1.107 fi 1.727–1.868; Char. 11: 1.014–1.033 fi 1.164–1.337; Char. 13: 1.002–1.004 fi 1.066– 1.094; Char. 35: 02 fi 1; Char. 36: 1 fi 2; Char. 57: 0 fi 1; Char. 58: 0 fi 1; Char. 61: 0 fi 2; Char. 71: 0 fi 1; Char. 73: 1 fi 2; Char. 74: 1 fi 0; Char. 87: 1 fi 0; Char. 95: 1 fi 2; Char. 99: 1 fi 3. F.J. Prevosti / Cladistics 26 (2010) 456–481 Canis adustus: Char. 1: 0.954–0.958 fi 1.014–1.057; Char. 4: 0.862– 0.991 fi 0.624–0.751; Char. 8: 1.022–1.106 fi 1.145–1.202; Char. 31: 0 fi 1; Char. 39: 0 fi 1; Char. 93: 0 fi 2. Speothos venaticus: Char. 0: 1.029–1.048 fi 0.898–0.922; Char. 2: 1.014–1.018 fi 1.079–1.256; Char. 5: 1.169–2.278 fi 0.871–1.092; Char. 10: 0.772–1.004 fi 1.117–1.201; Char. 24: 1 fi 0; Char. 25: 1 fi 0. Char. 27: 0 fi 1; Char. 30: 0 fi 1; Char. 33: 0 fi 1; Char. 36: 1 fi 0; Char. 48: 0 fi 1; Char. 53: 0 fi 1; Char. 61: 0 fi 1; Char. 64: 2 fi 1; Char. 65: 0 fi 1; Char. 73: 1 fi 0; Char. 95: 1 fi 0; Char. 97: 2 fi 4; Char. 99: 2 fi 0; Char. 100: 0 fi 1; Char. 104: 1 fi 2; Char. 112: 1 fi 0; Char. 114: 1 fi 0; Char. 115: 1 fi 0; Char. 117: 1 fi 0; Char. 118: 1 fi 0; Char. 123: 1 fi 0; Char. 124: 0 fi 1; Char. 125: 1 fi 0; Char. 127: 2 fi 0. Vulpes lagopus: Char. 4: 1.081–1.308 fi 1.344–1.752; Char. 127: 1 fi 2. Vulpes vulpes: Char. 59: 0 fi 1. Vulpes zerda: Char. 1: 1.037–1.045 fi 1.092–1.115; Char. 2: 1.014– 1.104 fi 1.163–1.245; Char. 3: 0.898–0.900 fi 0.613–0.627; Char. 5: 0.589–0.644 fi 0.260–0.307; Char. 8: 1.022–1.075 fi 1.252–1.291; Char. 11: 1.014–1.033 fi 1.035–1.085; Char. 13: 1.076–1.091 fi 1.143– 1.152; Char. 36: 1 fi 0; Char. 58: 0 fi 1; Char. 81: 0 fi 1. Canis rufus: Char. 11: 0.969–0.971 fi 0.942–0.953. Theriodictis tarijensis: Char. 76: 0 fi 1; Char. 80: 0 fi 1; Char. 93: 0 fi 1. Protocyon scagliorum: Char. 34: 0 fi 1; Char. 75: 0 fi 1. Protocyon troglodytes: Char. 1: 0.833 fi 0.748; Char. 6: 1.287 fi 1.353. Canis dirus: Char. 5: 2.255–2.368 fi 2.390–3.321; Char. 13: 1.000– 1.006 fi 1.012; Char. 39: 1 fi 0. Canis nehringi: Char. 1: 0.852–0.862 fi 0.832; Char. 2: 0.969–1.008 fi 0.837; Char. 3: 1.004–1.081 fi 1.172; Char. 5: 1.382–1.460 fi 2.196; Char. 6: 0.979–1.033 fi 1.057; Char. 9: 1.053–1.063 fi 1.192; Char. 10: 0.844–0.858 fi 0.822; Char. 11: 0.919–0.960 fi 0.889; Char. 14: 3 fi 2; Char. 15: 2 fi 1; Char. 18: 0 fi 1; Char. 20: 0 fi 1; Char. 50: 0 fi 1; Char. 64: 1 fi 2. Canis gezi: Char. 0: 1.029–1.048 fi 1.158; Char. 1: 0.833–0.996 fi 0.772; Char. 5: 1.169–2.278 fi 2.726; Char. 9: 1.020 fi 0.987; Char. 10: 0.772–1.004 fi 0.733; Char. 11: 0.956–0.958 fi 0.867–0.888; Char. 34: 0 fi 1. Dusicyon avus: Char. 0: 0.994–1.006 fi 1.020–1.048; Char. 1: 0.997– 1.014 fi 0.936–0.942; Char. 5: 0.875–0.957 fi 1.121; Char. 6: 0.945– 0.988 fi 1.078; Char. 7: 1.026–1.037 fi 1.125; Char. 10: 0.930– 1.004 > 0.882; Char. 12: 0.982–0.986 fi 0.979. Canis armbrusteri: Char. 86: 0 fi 1. Mesocyon coryphaeus: Char. 2: 1.018–1.116 fi 0.985; Char. 3: 0.900–0.902 fi 0.869; Char. 26: 0 fi 1; Char. 30: 0 fi 1; Char. 41: 0 fi 1; Char. 92: 1 fi 2. Phlaocyon leucosteus: Char. 1: 1.037–1.045 fi 0.967; Char. 3: 0.898– 0.900 fi 0.820; Char. 5: 0.589–0.644 fi 0.850; Char. 6: 0.979–1.026 fi 1.064; Char. 7: 1.026–1.037 fi 0.908; Char. 9: 1.020 fi 1.127; Char. 11: 1.014–1.033 fi 1.090–1.119; Char. 13: 1.002 fi 0.895; Char. 33: 0 fi 1; Char. 61: 0 fi 1; Char. 91: 1 fi 0. Archaeocyon leptodus: Char. 0: 0.880–0.897 fi 0.868; Char. 1: 1.071–1.206 fi 1.228; Char. 3: 0.900–0.902 fi 0.957; Char. 4: 1.081– 1.130 fi 1.237; Char. 5: 0.470–0.543 fi 0.430. Leptocyon vafer: Char. 4: 1.081–1.130 fi 1.080; Char. 6: 0.979– 1.008 fi 0.870; Char. 9: 0.976–1.020 fi 0.939; Char. 45: 0 fi 1. Eucyon davisi: Char. 1: 0.946–0.958 fi 0.936; Char. 6: 0.939–0.943 fi 0.921; Char. 9: 1.020 fi 1.016; Char. 45: 0 fi 1; Char. 85: 2 fi 1; Char. 88: 0 fi 1; Char. 93: 0 fi 2. Node 42: Char. 15: 2 fi 1. Node 43: Char. 0: 0.880–0.897 fi 0.921–0.969; Char. 10: 1.115 fi 1.093–1.107; Char. 12: 1.170 fi 1.070–1.092; Char. 36: 0 fi 1; Char. 106: 0 fi 1; Char. 107: 0 fi 1; Char. 108: 0 fi 1; Char. 109: 0 fi 1; Char. 110: 0 fi 1; Char. 111: 0 fi 1; Char. 112: 0 fi 1; Char. 113: 0 fi 1; Char. 114: 0 fi 1; Char. 121: 0 fi 1; Char. 123: 0 fi 1; Char. 124: 1 fi 0; Char. 125: 0 fi 1; Char. 131: 0 fi 1. Node 44: Char. 1: 1.071–1.206 fi 1.037–1.045. Char. 5: 0.583 fi 0.589–0.644; Char. 51: 0 fi 1; Char. 79: 1 fi 0; Char. 82: 1 fi 0; Char. 83: 0 fi 1; Char. 85: 1 fi 2; Char. 93: 0 fi 1. 481 Node 45: Char. 5: 0.470–0.543 fi 0.583; Char. 22: 0 fi 1; Char. 61: 4 fi 0. Node 46: Char. 19: 0 fi 1; Char. 66: 0 fi 1; Char. 89: 0 fi 1. Node 49: Char. 4: 0.946–0.991 fi 0.861–0.922; Char. 16: 1 fi 0; Char. 68: 1 fi 0; Char. 87: 1 fi 0. Node 50: Char. 0: 0.956–0.988 fi 0.994–1.006; Char. 1: 1.037–1.045 fi 0.997–1.014; Char. 3: 0.898–0.900 fi 0.985–0.998; Char. 5: 0.633– 0.644 fi 0.853–0.957; Char. 10: 1.093–1.107 fi 0.930–1.004; Char. 11: 1.014–1.033 fi 0.990–0.998; Char. 12: 0.994–1.057 fi 0.936–0.986; Char. 13: 1.002–1.004 fi 0.999–1.000; Char. 21: 0 fi 1; Char. 93: 1 fi 0; Char. 122: 0 fi 1. Node 51: Char. 4: 1.081–1.130 fi 0.946–0.991; Char. 6: 0.979–1.026 fi 0.939–0.943; Char. 12: 1.070–1.092 fi 0.994–1.057; Char. 103: 2 fi 1. Node 53: Char. 2: 0.969–1.008 fi 0.920–0.949; Char. 7: 0.987–1.016 fi 0.974–0.977; Char. 25: 0 fi 1; Char. 62: 0 fi 1; Char. 88: 0 fi 1; Char. 122: 1 fi 0. Node 54: Char. 39: 0 fi 1. Node 55: Char. 4: 0.946–0.991 fi 1.002–1.024; Char. 29: 0 fi 1; Char. 63: 0 fi 1; Char. 64: 0 fi 1; Char. 80: 0 fi 1; Char. 127: 1 fi 0. Node 56: Char. 0: 0.994–1.006 fi 1.035–1.042; Char. 3: 0.985–0.998 fi 1.004–1.008; Char. 17: 0 fi 2; Char. 46: 0 fi 1; Char. 49: 0 fi 1. Node 57: Char. 1: 0.997–1.014 fi 0.946–0.958; Char. 2: 1.014–1.018 fi 0.969–1.008; Char. 55: 0 fi 1; Char. 61: 0 fi 3; Char. 79: 0 fi 1. Node 58: Char. 3: 1.004–1.029 fi 1.049–1.083; Char. 5: 0.930–1.091 fi 1.190–1.260; Char. 6: 0.939–0.943 fi 0.952–1.002; Char. 10: 0.899– 0.921 fi 0.875–0.889; Char. 12: 0.936–0.938 fi 0.918–0.926; Char. 13: 0.999 fi 1.000–1.006; Char. 127: 0 fi 1. Node 59: Char. 8: 0.902 fi 0.870–0.895; Char. 18: 0 fi 1; Char. 40: 1 fi 2. Node 60: Char. 5: 1.190–1.260 fi 2.255; Char. 70: 0 fi 1. Node 61: Char. 11: 0.990–0.998 fi 1.052– 1.080. Node 62: Char. 35: 2 fi 0; Char. 122: 1 fi 0. Node 63: Char. 0: 0.994–1.006 fi 1.029–1.048; Char. 3: 0.985–0.998 fi 1.016–1.077; Char. 5: 0.875–0.957 fi 1.169–1.390; Char. 25: 0 fi 1; Char. 71: 0 fi 1; Char. 99: 1 fi 2; Char. 127: 1 fi 2. Node 64: Char. 0: 1.035–1.046 fi 0.975–1.020; Char. 30: 0 fi 1; Char. 46: 1 fi 0; Char. 60: 1 fi 0; Char. 67: 0 fi 1; Char. 71: 0 fi 1; Char. 74: 1 fi 3; Char. 85: 2 fi 3; Char. 87: 1 fi 0; Char. 89: 1 fi 0. Node 65: Char. 1: 0.946–0.958 fi 0.852–0.862; Char. 5: 0.930–1.091 fi 1.382–1.460; Char. 6: 0.939–0.943 fi 0.979–1.033; Char. 8: 1.022– 1.094 fi 0.827–0.846; Char. 9: 1.020–1.043 fi 1.053–1.063; Char. 10: 0.899–0.964 fi 0.844–0.858; Char. 11: 0.969–0.990 fi 0.919–0.960; Char. 69: 0 fi 1; Char. 70: 0 fi 1; Char. 72: 0 fi 1. Node 66: Char. 2: 1.014–1.042 fi 1.086; Char. 4: 0.855–0.922 fi 1.037; Char. 11: 0.994–0.998 fi 1.033–1.062; Char. 30: 0 fi 1. Node 67: Char. 8: 1.022–1.059 fi 1.110. Node 70: Char. 3: 0.974–0.998 fi 1.008–1.012. Node 71: Char. 93: 0 fi 1. Node 72: Char. 7: 1.026 fi 0.992–1.012. Node 73: Char. 8: 1.022 fi 0.928–0.937; Char. 30: 0 fi 1; Char. 50: 0 fi 2; Char. 53: 0 fi 1. Node 74: Char. 69: 0 fi 1; Char. 72: 0 fi 1; Char. 84: 0 fi 1; Char. 89: 1 fi 0. Node 75: Char. 1: 0.997–1.014 fi 0.833–0.996; Char. 11: 0.967– 0.998 fi 0.956–0.958; Char. 14: 3 fi 1; Char. 37: 0 fi 1; Char. 50: 0 fi 1; Char. 64: 0 fi 2; Char. 67: 0 fi 1; Char. 70: 0 fi 1; Char. 91: 1 fi 2; Char. 92: 0 fi 2. Node 76: Char. 0: 0.921–0.969 fi 0.973–0.994; Char. 3: 0.898–0.900 fi 0.929–0.947; Char. 10: 1.093–1.107 fi 0.997–1.004; Char. 12: 1.070– 1.092 fi 1.025–1.048; Char. 40: 0 fi 1. Node 77: Char. 9: 1.020 fi 0.972–0.979; Char. 13: 1.002–1.024 fi 1.076–1.091; Char. 38: 12 fi 0; Char. 45: 0 fi 1; Char. 97: 2 fi 1. Node 78: Char. 43: 0 fi 1; Char. 105: 0 fi 1. Node 79: Char. 6: 1.020–1.252 fi 1.287; Char. 63: 0 fi 1; Char. 74: 2 fi 3.