Data Matrix

A number of studies have investigated the relationships of varanids and the interrelationships of Varanoidea over the last five decades (McDowell and Bogert, 1954; Rieppel, 1980a; Branch, 1982; Borsuk-Bialynicka, 1983, 1984; Pregill et al., 1986; Estes et al., 1988; Norell et al., 1992; Pianka, 1995; Caldwell, 1996; Dal Sasso and Pinna, 1997; Lee, 1997; Caldwell, 2000; Lee and Caldwell, 2000; Nydam, 2000; Rieppel and Zaher, 2000a; Pepin, 2001; Rieppel et al., 2006; Conrad, 2008; Norell et al., 2008). Most recently, Conrad (2008) presented an extensive morphological analysis of squamate relationships. Norell et al. (2008) augmented this analysis by adding 29 taxa (including the basal varanine Ovoo gurvel) and nine new characters. To this matrix of 251 taxa and 373 characters we have now added four more Varanus species (Varanus albogularis, Varanus gilleni, Varanus rudicollis, and Varanus salvator) and coded Saniwa ensidens based on the complete skeleton (FMNH PR 2378) and a cast of a partial skull (FMNH PR 2380) recently described by Rieppel and Grande (2007).

Analysis 1

We analyzed these data using the computer program T.N.T. (Goloboff et al., 2003) using the New Technology Search (1000 replicates) and three subsequent ratchet runs (each of 1000 replicates). Our analysis recovered 2373 equally short trees, each with a length of 3846 steps and a retention index of 0.7138.

We report both the strict and Adams consensus trees for taxa of interest (fig. 2), the latter as reconstructed by the computer program PAUP* (Swofford, 2001). Adams consensus trees are reported rather than majority-rule trees. Adams consensus collapses problematic/volatile taxa to their most basal supported position and retains branches that are supported in all of the principle trees, regardless of the position of volatile taxa.

Figure 2

Strict (left) and Adams rule (right) consensus cladograms representing areas of interest from the global analysis of squamate relationships described in the text (note, non-varaniform taxa omitted). Extinct taxa are denoted by daggers (†). Groups A and B (from the text) are identified here and referred to again in figure 4.

Analysis 2

To further test the position of Saniwa ensidens and other fossil varanids, such as Megalania prisca and Varanus rusingensis, we created a subset of the data containing only “Saniwa” feisti, Necrosaurus cayluxi, Saniwides mongoliensis, Telmasaurus grangeri, Aiolosaurus oriens, Lanthanotus borneensis, Cherminotus longifrons, Ovoo gurvel, Saniwa ensidens, Megalania prisca, and all of the Varanus species present in the global analysis described above (reduced matrix A). These data were then analyzed in the same manner as the global data (see above). Because the results of this analysis were very similar to those of the global analysis (see below), the RMA was used for further tests that would have been prohibitively time-consuming or impossible given the global data matrix.

Analysis 3

We ran a third analysis in which only extant taxa were included for comparison with a recent molecular phylogeny of varanids (Ast, 2001, 2002a). Ast's (2001) study was chosen for comparison because it is one of the most taxonomically inclusive and one of the most recent molecular-based studies of extant varanoid interrelationships. Fuller et al. (1998) and King et al. (1999) have also recently analyzed varanid interrelationships using molecular data and produced a phylogenetic hypothesis that is generally similar to that of Ast (2001), but the latter was slightly more taxonomically inclusive. Pianka (1995) recovered a slightly different result, using different techniques. Fitch et al. (2006) analyzed relationships among a subset of Varanus, and their results generally agree with those of Ast (2001).

Analysis 4

Finally, we constructed a constraint tree for reduced matrix A (RMA) in which the extant taxa common both to it and the analysis of Ast (2001) were forced into the topology of the latter study. This constraint tree was treated as a “backbone” constraint in PAUP* so that nonoverlapping taxa were could be placed within the constraint tree.

Data Acquisition

Morphological data were acquired through direct observations of specimens (appendix 1) and published data. Published data were used to supplement the codings for some extant Varanus (Mertens, 1942a, c, b; Bellairs, 1949; McDowell and Bogert, 1954; Rieppel, 1980a; Jenkins and Goslow, 1983; Landsmeer, 1983; Rieppel, 1993; Pianka, 1995; Zaher and Rieppel, 1999a; Rieppel and Zaher, 2000b; Maisano, 2001b; Pepin, 2001; Rieppel and Grande, 2007; Uetz, 2007), Lanthanotus (McDowell and Bogert, 1954; Rieppel, 1980a, b, 1983; Zaher and Rieppel, 1999b, a; Maisano, 2001a; Rieppel and Zaher, 2000b; Maisano et al., 2002), the included Gobi taxa (Gilmore, 1943; Borsuk-Bialynicka, 1983; Estes, 1983; Borsuk-Bialynicka, 1984; Gao and Hou, 1996; Gao and Norell, 1998, 2000), Saniwa ensidens (Gilmore, 1928; Estes, 1983; Caldwell, 2003; Rieppel and Grande, 2007), Megalania prisca (Hecht, 1975; Molnar, 1990, 2004; Lee, 1995; Erickson et al., 2003), and Varanus rusingensis (Clos, 1995).

Monospecific Taxa

Note that Aiolosaurus oriens, Cherminotus longifrons, Lanthanotus borneensis, Saniwides mongoliensis, Telmasaurus grangeri are each monospecific and may, hereafter, be referred to by their generic names only.

Analysis 1

Our global analysis recovered 2373 equally short trees, each with a length of 3846 steps and a retention index of 0.7138. Both strict and Abrams consensuses are illustrated (fig. 2).

Analysis 2

Our analysis of reduced matrix A (RMA) recovered 144 equally short trees, each with a length of 288 steps and a retention index of 0.5621. Only the Adams consensus of this search is illustrated (fig. 3).

Figure 3

Analysis from reduced matrix A (RMA) analysis. See the text for details. Groups A and B are identical in their topology to those denoted in figure 2.

Analysis 3

The analysis limited to extant taxa recovered three equally short trees, each with a length of 204 steps and a retention index of 0.4729. The strict and Adams consensuses for this analysis are identical (see fig. 4).

Figure 4

Phylogenetic hypothesis generated by analysis of the RMA when all fossil taxa are deleted. Note that this topology shows similarities with the morphological analyses in figures 2 and 3, and also with that of the molecular analysis of Ast (2001) (fig. 5).

Analysis 4

The analysis run with a backbone constraint for extant varanids recovered 70 equally short trees, each with a length of 308 steps and a retention index of 0.5056. The results of this analysis are described more fully below.

Character list

The complete list of characters and character states are found in Conrad (2008) and Norell et al. (2008). Both of these papers are available for free in pdf format from  http://digitallibrary.amnh.org/dspace. The data are presented in a format that allows them to be copied and pasted to a new document in nexus format, T.N.T. (Goloboff et al., 2003) format, etc.

Clade Diagnoses

According to the current study, “Saniwa” feisti, Necrosaurus cayluxi, Saniwides, Telmasaurus, and crown-group Varanidae form a clade based on five unambiguous synapomorphies. These are 58(0) straight (linear) interorbital frontal margins, 61(1) subolfactory processes of the frontals partly surrounding the olfactory tracts of the brain, 65(1) presence of an external maxillofrontal contact, 115(2) presence of a single line of palatine teeth, and 233(1) presence of at least weak precondylar vertebral constriction. “Saniwa” feisti and Necrosaurus cayluxi form a polytomy at the base of this clade. This and other polytomies within the results of this analysis may be present, in part, because of the inclusion of incomplete specimens. Saniwides, Telmasaurus, and crown-group varanids form a clade based on 55(0) presence of separate (unfused) frontals in adults, 70(0) presence of a U-shaped frontoparietal suture, 124(1) presence of an ectopterygoid-palatine contact anterior to the suborbital fenestra, and 183(1) dorsal dentary contribution to the anterior inferior alveolar foramen. Within this clade, Telmasaurus and Saniwides form an exclusive clade diagnosed by 115(0) presence of palatine tooth patches, and 118(0) pterygoid teeth arranged in multiple rows/patches.

Varanidae is typically considered to include extant Varanus, Lanthanotus, and all descendants of their last common ancestor (Pregill et al., 1986; Estes et al., 1988; Pianka, 1995; Lee, 1998; Bernstein, 1999; Conrad, 2004, 2008). The current analysis recovers five unambiguous synapomorphies for Varanidae including Aiolosaurus. Aiolosaurus forms a polytomy with lanthanotines and varanines in this analysis (for a review on the phylogenetic placement of Aiolosaurus, see Norell et al., 2008). The five character states supporting Varanidae are 3(1) presence of a blunt and rounded snout, 65(0) absence of an external maxillo-frontal contact, 149(1) presence of a broad crista tuberalis that incorporates the paroccipital processes, 151(1) an anteriorly placed spheno-occipital tubercle, and 178(0) a ventrally arched dentary. The Cherminotus-Lanthanotus clade is supported by 32(1) presence of a medial palatine flange on the maxilla, 83(1) presence of a nuchal fossa at the posterior margin of the parietal, and 107(0) presence of a rod-shaped vomer.

Varaninae (sensu Pregill et al., 1986 as modified by Conrad, 2008; all taxa sharing a more recent common ancestor with Varanus varius than Lanthanotus borneensis) is here diagnosed by 27(1) presence of a midline contact of the maxillae posterior to the premaxillary nasal process, 51(1) absence of a jugal-postorbital contact, and 62(1) midline contact of the subolfactory processes on the frontals. Saniwa ensidens is found to be the sister taxon to crown-group Varanus in all the principle trees in this analysis. The six unambiguous synapomorphies uniting this clade are 3(0) presence of a tapering snout (fig. 1) (reversed in some Varanus), 10(1) presence of dermal sculpturing on the frontal and parietal, 42(1) presence of a posterolateral lacrimal flange, 43(1) a double lacrimal foramen, 44(1) an enlarged primary lacrimal foramen, and 111(0) palatine longer than wide. The species of Varanus are united to the exclusion of Saniwa ensidens based on the following synapomorphies: 52(2) Jugal lying mostly posterior to the maxilla, with little or no overlap, 69(1) presence of parietal tabs on the frontal, 74(1) frontal tabs present on the parietal, 115(1) absence of palatine teeth, and 118(2) absence of pterygoid teeth.

Varanus rusingensis and Megalania prisca (hereafter Varanus prisca) are both nested within the extant Varanus radiation according to this analysis. Varanus prisca is the sister taxon to Varanus salvadorii, sharing with it 69(0) absence of parietal tabs on the frontal. Varanus komodoensis forms a clade with Varanus prisca and Varanus salvadorii based on 32(1) presence of a medial palatine flange on the maxilla and 111(0) palatine longer than wide. Varanus varius is the next most proximal sister taxon and shares with the other species listed 3(1) a blunt, rounded snout and 372(0) distal placement of the fleshy external nares. Varanus salvator shares with these other taxa three unambiguous synapomorphies, including 72(1) broad and squared lateral flange of the parietal at the frontoparietal suture, 138(1) reduced crista prootica without a distinct ventral extension, and 189(1) splenial extending for less than one-half the dentary length. Varanus gouldii and Varanus bengalensis form a polytomy with the salvator-varius-komodoensis-salvadorii-prisca group in a clade hereafter referred to as “group A”. Group A is diagnosed by 1(3) an extremely elongate antorbital snout (constituting more than half the craniobasal length) and 233(2) vertebrae with strong precondylar constriction.

Group A forms a polytomy with Varanus flavescens and a clade containing the mostly African taxa Varanus albigularis, Varanus exanthematixus, Varanus griseus, Varanus niloticus, and Varanus rusingensis; hereafter referred to as “group B”. Group B is united by a single unambiguous synapomorphy, which is 212(5) presence of globidont posterior teeth. All of group B besides the Miocene Varanus rusingensis form an unresolved clade united by 292(0) presence of a separate astragalus and calcaneum.

Group A, group B, and Varanus flavescens form a clade with Varanus indicus; united by 141(1) presence of a double facial foramen and 312(0) absence of keeled scales on the body. This clade, in turn, forms a polytomy with Varanus rudicollis and a clade composed of Varanus dumerilii and Varanus olivaceus. This clade is diagnosed by 20(1) fused nasals and 111(1) palatine subequal in length and width, and forms a polytomy with other Varanus (fig. 2).

Comparisons with molecular studies: The analysis above differs significantly from the phylogenetic hypothesis of the interrelationships of Varanus proposed by Ast (2001), although there are several points of congruence (compare figs. 2, 3, and 5). Notably, both analyses recover a clade containing the large-bodied taxa Varanus komodoensis, Varanus salvadorii, and Varanus varius and a clade containing the African taxa Varanus exanthematicus, Varanus griseus, and Varanus niloticus. In contrast to Ast (2001), we find no support for major Indo-Asian clades. Instead these clades are found to be paraphyletic or polyphyletic with respect to African and Indo-Australian clades. The Indo-Australian group (as identified by Ast, 2001) is not monophyletic. Varanus bengalensis and Varanus salvator are found to belong to the otherwise mostly Indo-Australian clade termed group A above. Ast (2001) found Varanus bengalensis and Varanus salvator to be nested members of the Indo-Asian A clade in her study. Most of the other members of Ast's (2001) Indo-Asian clades included in our global analysis form a paraphyletic assemblage representing proximal sister taxa to the clade including groups A and B.

Taxonomic inclusiveness is a key consideration when comparing and contrasting these analyses. Ast's (2001) analysis is more taxonomically inclusive amongst extant taxa, and she included multiple “subspecies” and/or specimens from multiple localities for many species. By contrast, our morphological analyses (above and below) benefit from the inclusion of fossil taxa (e.g., Saniwa ensidens) that may, in some cases, offer a clearer understanding of the primitive morphology for some nodes and certainly help to polarize characters throughout the tree. We will further investigate these considerations below.

Analysis 2 (rma Analysis)

Analysis of RMA reveals a very similar Adams consensus tree to that of the global analysis (compare figs. 2 and 3). Clades A and B were identical between the two analyses. Adams consensus of the RMA analysis was unable to resolve the relative position of Varanus flavescens with respect to clades A and B, but resolved the position of Varanus acanthurus as a sister-taxon to Varanus rudicollis. All of the more basal taxa on the varanid lineage remained in the same position as with the global analysis. The RMA is used in the following analyses.

Analysis 3 (extant-Only Analysis) (Fig. 4)

We re-analyzed RMA with all fossil taxa deleted, in part, to test the importance of fossils in our analysis and to isolate possible similarities in the signal between the molecular data and our morphological data.

When all fossil taxa are deleted from the RMA, we recover an Indo-Asian monitor clade consisting of the taxa Varanus olivaceus, Varanus bengalensis, and Varanus dumerilii. The interrelationships of these three taxa is that revealed by Ast's (2001) analysis, but with several other taxa removed; that is, some of the Indo-Asian taxa common to both analyses (see the right side of fig. 5) fall outside of the Indo-Asian clade recovered in our extant-only analysis. Ast (2001) finds Varanus flavescens and a clade composed of Varanus rudicollis and Varanus salvator to be successively more remote sister taxa to her Varanus bengalensis–Varanus dumerilii clade. She also finds Varanus prasinus and Varanus indicus to be nested members of a clade including Varanus olivaceus (Ast, 2001). In the current analysis of extant taxa, Varanus flavescens, Varanus indicus, and Varanus rudicollis are suggested to be basal Varanus species. Varanus salvator is recovered as the sister taxon to a mostly Indo-Austalian clade including taxa typically assigned to the clade Odatria (e.g., Varanus acanthurus) and some large-bodied varanids (e.g., Varanus gouldii and Varanus komodoensis); Varanus prasinus is nested within our Indo-Australian clade. Our extant-only analysis recovered Varanus gouldii as a basal member of the clade containing Varanus varius, Varanus komodoensis, and Varanus salvadorii, whereas Ast (2001) recovered it as more closely related to the Indo-Asian clades (but see Ast, 2002b).

Figure 5

Phylogenetic hypothesis for the extant Varanus species as presented by Ast (2001, 2002a). The left side shows all of the species included in her analysis. The right side is a cladogram of the same topology, but with only the species included in both Ast (2001, 2002a) and the current morphological analysis.

Thus, similarities between the extant-only analysis performed here and the molecular analysis (Ast, 2001) not present in analyses 1 or 2 above include the recovery of an Indo-Asian clade and consolidation of Indo-Australian taxa into a single clade (figs. 2, 3, 5). This result raises the possibility that the absence of historical morphological data provided by fossils may be a factor in the recovery of those clades by molecular analyses. Inclusion of fossil outgroups to Varanus (e.g., Saniwa ensidens) and fossil Varanus (e.g., Varanus rusingensis) apparently repolarize character states such that the attraction of some Indo-Asian taxa and the two main clades of Indo-Australian taxa is reduced. However, the possibility remains that there also may be undetected morphological tendencies associated with becoming large, leading to convergence between Varanus salvator, Varanus gouldii, and the large-bodied Indo-Australian taxa within morphological phylogenetic analyses. Molecular data may be less susceptible to this bias, and the problem of subjectivity present in morphological character formulation.

Analysis 4 (constrained Tree Analysis)

Our constrained tree analysis was designed to test the placement of fossil taxa given the context of the molecular phylogeny presented for Varanus by molecular evidence.

Saniwa ensidens was found to be the sister taxon to Varanus in all of the principle trees recovered in this analysis. Varanus rusingensis was somewhat variable in its placement. It was variably recovered as the sister taxon to all of other Varanus, to Varanus olivaceus, or to Varanus flavescens. Varanus prisca was consistently recovered in a nested position within the extant Varanus radiation. It was variably found to be the sister taxon to Varanus salvadorii, the sister taxon to Varanus varius, the sister taxon to a clade containing Varanus varius and Varanus komodoensis, or the outgroup to a clade containing Varanus komodoensis, Varanus salvadorii, and Varanus varius.

Fossil Varanidae Conclusions

Saniwa ensidens is a pivotal taxon for understanding the evolution of the varanid lineage because of its placement and the completeness of the known remains. Our analyses confirm the placement of Saniwa ensidens as the sister taxon to Varanus within Varaninae (fig. 2). Saniwa ensidens differs from Varanus mainly in possessing a more anteriorly extensive jugal, a simpler frontoparietal suture, and in the plesiomorphic retention of palatal teeth. The overall appearance of the skull roof of Saniwa ensidens is similar to that of extant Varanus, especially given the diversity present in the latter (see fig. 1 and Mertens, 1942a, b, c, 1942c).

Some other taxa sometimes considered close relatives of Varanus are actually outgroups to the crown-group Varanidae. Saniwides mongoliensis and Telmasaurus grangeri form a clade that is one of the proximal outgroups to Varanidae in our analysis. This conclusion contrasts with those of some earlier studies (e.g., Lee, 1997 and Pepin, 1999), which suggested Saniwides formed a clade with Varanus exclusive of Saniwa ensidens. The recently described specimens of Saniwa ensidens have clarified similarities between that taxon and Varanus.

Varanus rusingensis is probably the oldest known member of the extant Varanus radiation. The only doubt about this hypothesis is raised by a small subset of the trees generated from the constrained analysis described above (analysis 4). This taxon demonstrates that the Varanus clades had diversified by and reached Africa no later than the Early Miocene (Clos, 1995). “Megalania” prisca is confirmed to fall within the extant radiation of Varanus.

Inclusion of Incomplete Fossils

As is often the case, inclusion of fossils in this analysis also means the inclusion of taxa that cannot be scored for every character state. Indeed, many of the extant taxa were coded from incomplete specimens or species for which specific knowledge of some character systems (e.g., some soft characters) could not be scored. Inclusion of these additional data will benefit future iterations of this analysis pending fuller description of specific taxa and/or discovery of more complete fossils.

Conclusions about Clades of Varanus

The interrelationships of the Varanus clades have been approached both using molecular and morphological data. Ast's (2001) relatively inclusive analysis of extant Varanus produced a result that differs significantly from our morphology-based global analysis, which included numerous nonvaranids, fossil varanids, and several extant Varanus. Taxon sampling is at least partly responsible for this disparity in results; the morphological analysis includes fewer extant taxa and the molecular study includes no fossils. However, the specific limitations of each approach also must be further investigated. A combined analysis of morphological and molecular data may offer a way of further investigating this problem in the future. Further discoveries of complete specimens of fossil taxa (like the recently described Saniwa ensidens), more complete descriptions of fossil and extant taxa, and increased sampling of molecular data for extant taxa will also further improve future studies of varanid systematics.