What’s in an Outgroup? The Impact of Outgroup Choice on the Phylogenetic Position ofThalattosuchia (Crocodylomorpha) and the Origin of Crocodyliformes
ERIC W. WILBERG1,2,∗1Department of Geoscience, University of Iowa, Iowa City, IA, USA and 2Current Address – Department of Geology and Geography, Georgia Southern
University, Statesboro, GA, USA∗Correspondence to be sent to: Department of Geology and Geography, Georgia Southern University, P.O. Box 8149, Statesboro, GA 30460, USA
Received 16 August 2014; reviews returned 24 October 2014; accepted 30 March 2015Associate Editor: Norm MacLeod
Abstract.—Outgroup sampling is a central issue in phylogenetic analysis. However, good justification is rarely given foroutgroup selection in published analyses. Recent advances in our understanding of archosaur phylogeny suggest thatmany previous studies of crocodylomorph and crocodyliform relationships have rooted trees on outgroup taxa that areonly very distantly related to the ingroup (e.g., Gracilisuchus stipanicicorum), or might actually belong within the ingroup.Thalattosuchia, a group of Mesozoic marine crocodylomorphs, has a controversial phylogenetic position—they are recoveredas either the sister group to Crocodyliformes, in a basal position within Crocodyliformes, or nested high in the crocodyliformtree. Thalattosuchians lack several crocodyliform apomorphies, but share several character states with derived long-snoutedforms with a similar ecological habit, suggesting their derived position may be the result of convergent evolution. Several ofthese “shared” characters may result from ambiguously worded character state definitions—structures that are superficiallysimilar but anatomically different in detail are identically coded. A new analysis of crocodylomorphs with increasedoutgroup sampling recovers Thalattosuchia as the sister group to Crocodyliformes, distantly related to long-snoutedcrocodyliforms. I also demonstrate that expanding the outgroup sampling of previously published matrices results inthe recovery of thalattosuchians as sister to Crocodyliformes. The exclusion of thalattosuchians from Crocodyliformeshas numerous implications for large-scale evolutionary trends within the group, including extensive convergence inthe evolution of the secondary palate characteristic of the group. These results demonstrate the importance of carefuloutgroup sampling and character construction, and their profound effect on the position of labile clades. [convergence;Crocodyliformes; Crocodylomorpha; outgroup sampling; Thalattosuchia.]
“As parsimony analysis attributes characterstates to the hypothetical stem species of thetree, fixing the position of the root determinesthe direction of character transformations,and so the relative plesiomorphy of features.”
– J.S. Farris (1982) (p. 329)
Since tetrapods first transitioned onto land, manygroups have returned to the water. Secondarilyaquatic lineages have arisen in nearly every majortetrapod group. Crocodylians and their relatives(Crocodylomorpha) are no exception to this pattern.Numerous crocodylomorph lineages have producedmarine or estuarine predators possessing elongatetubular snouts; a putative adaptation for piscivory(Langston 1973). This repeated evolution of similarskull shapes is not limited to slender snouted marineforms. Several snout shapes have recurred throughoutcrocodylomorph evolution (Brochu 2001), and studieshave shown that within crown-group crocodylianssnout shape is more highly correlated with functionalparameters than with phylogeny (e.g., Pierce et al.2008; Sadleir and Makovicky 2008). The patternwithin modern and fossil crocodylians suggests thatMesozoic crocodylomorphs possessing similar rostralmorphology may not be closely related. The evolutionof long slender snouts in marine crocodylomorphsmay have occurred a single time or multiple timesconvergently.
Historically, longirostry (possession of a longsnout) was thought to have arisen multiple times
within crocodylomorphs (Kälin 1955; Langston 1973;Buffetaut 1982). Morphological phylogenetic analysesof the crown-group consistently support at least threeindependent derivations of the long-snouted condition(e.g., Norell 1989; Brochu 1997a) wheras molecularanalyses suggest only two (e.g., Gatesy et al. 2003;McAliley et al. 2006; Meredith et al. 2011; see Brochu2001 for a detailed discussion of this issue).
Three groups of noncrown crocodylomorphswere traditionally thought to have independentlyevolved longirostry: thalattosuchians, dyrosaurids, andpholidosaurs (Kälin 1955; Langston 1973; Buffetaut1982). The thalattosuchians are a unique clade ofcrocodylomorphs demonstrating the most extremearchosaurian adaptation to the marine environment(Hua and Buffetaut 1997; Young and Andrade 2009).The group is known from marine deposits of the LowerJurassic through the Lower Cretaceous, and includes twogroups: Metriorhynchidae and Teleosauridae. Thoughlongirostry is prevalent among thalattosuchians,not all possess an elongate snout. Some taxa (e.g.,Dakosaurus) possess relatively short, robust snouts.This has been used as evidence against functionallycorrelated characters uniting thalattosuchians withpholidosaurs/dyrosaurids (Pol and Gasparini 2009).
Dyrosauridae and Pholidosauridae make up theother two clades of noncrown group longirostrinecrocodylomorphs. The dyrosaursids are found primarilyin coastal marine environments from the Cretaceousthrough the Eocene, whereas Pholidosauridae (latestJurassic through the Cretaceous) contains both marine
FIGURE 1. Generalized phylogeny of Crocodylomorphashowing the three potential positions of Thalattosuchia. a) Sistergroup to Crocodyliformes; b) basal mesoeucrocodyilans; c) Derivedneosuchians, allied with pholidosaurs/dyrosaurids to form a“longirostrine clade.”
and freshwater forms (Hua and Buffetaut 1997).Traditionally the pholidosaurs and dyrosaurids were notthought to be closely related (Buffetaut 1982), however,all phylogenetic analyses including both groups haverecovered them as a clade (e.g., Benton and Clark 1988;Clark 1994; Buckley and Brochu 1999; Wu et al. 1997;Ortega et al. 2000; Sereno et al. 2001, 2003; Brochu et al.2002; Pol and Norell 2004a, 2004b; Jouve et al. 2006;Larsson and Sues 2007; Jouve 2009; Young and Andrade2009; Turner and Sertich 2010; Bronzati et al. 2012).
Historically, thalattosuchians were interpreted asbasal “mesosuchians” distantly related to crown-groupCrocodylia (Kälin 1955; Buffetaut 1982). Parsimonyanalyses support thalattosuchian monophyly but havenot clarified their phylogenetic position. This issuehas been termed the “longirostrine problem” (Clark1994). Three mutually exclusive topologies can be foundin published analyses (Fig. 1). Some analyses recoverthalattosuchians as sister to Crocodyiformes (Fig. 1—position a; Jouve 2009; Pol and Gasparini 2009; Wilberg2015). Some place them in a basal position amongmesoeucrocodylians (Fig. 1—position b; e.g., Serenoet al. 2001, 2003; Larsson and Sues 2007; Sereno andLarsson 2009; Young and Andrade 2009; Young et al.2012), and others as derived mesoeucrocodylians moreclosely related to the crown (Fig. 1—position c; e.g.,Clark 1994; Wu et al. 1997, 2001; Pol and Norell 2004a,2004b; Jouve et al. 2006; Jouve 2009; Pol and Gasparini2009; Turner and Sertich 2010; Andrade et al. 2011;Pritchard et al. 2013). Analyses drawing thalattosuchianscrownward link them with other long-snouted clades,suggesting that convergence related to snout shapemay be confounding efforts to recover a phylogeneticsignal.
A point directly related to the phylogenetic positionof thalattosuchians that has yet to be investigatedinvolves outgroup sampling in published analyses.
Outgroup sampling is of primary importance inphylogenetic analyses, affecting ingroup relationshipsand, in placing the root, polarizing characters (Lyons-Weiler et al. 1998). However, justification is rarelygiven for outgroup selection. There are essentiallythree different outgroup sampling schemes in publishedcrocodyliform phylogenetic analyses. Data sets baseddirectly on the original analyses of Clark (Clark 1994)modified primarily by the addition of ingroup taxaand characters (e.g., Pol and Norell 2004a, 2004b;Pol et al. 2004; Gasparini et al. 2006; Turner andBuckley 2008; Pol and Gasparini 2009; Pol et al. 2009;Turner and Sertich 2010; Pritchard et al. 2013) generallyroot their topologies on Gracilisuchus stipanicicorumand include two “sphenosuchian” crocodylomorphs:Terrestrisuchus gracilis and Dibothrosuchus elaphros (itshould be noted that these analyses greatly reducedthe number of sampled outgroups originally presentin Clark 1994). Analyses based on the data setof Jouve (e.g., Jouve et al. 2006; Jouve 2009) rootthe tree on the rauisuchid (sensu Nesbitt 2011)Postosuchus kirkpatricki, and include the “sphenosuchian”crocodylomorphs Sphenosuchus acutus, Dibothrosuchuselaphros, and Kayentasuchus walkeri. Other analyses roottheir trees on a protosuchian (e.g., Buckley and Brochu1999; Sereno et al. 2001; Brochu et al. 2002; Sereno et al.2003; Larsson and Sues 2007; Sereno and Larsson 2009).Though some published analyses have utilized differentoutgroup sampling schemes (e.g., Young and Andrade2009; Andrade et al. 2011; Young et al. 2012), most includefewer than four taxa.
Some analyses recovering thalattosuchians basallywithin Mesoeucrocodylia (e.g., Sereno et al. 2001,2003; Larsson and Sues 2007; Sereno and Larsson2009) utilize a protosuchian outgroup. This outgroupsampling scheme a priori excludes the possibility thatthalattosuchians are the sister group to Crocodyliformes(Fig. 1—position a). These analyses assumed ingroupmonophyly, presumably based on the results of earlycladistic analyses (Benton and Clark 1988; Clark 1994).When sampling only a single outgroup, it is not possibleto test the monophyly of the ingroup (Nixon andCarpenter 1993; Barriel and Tassy 1998; Sanderson andShaffer 2002; Puslednik and Serb 2008). Addition of moredistant outgroups (without constraining their topologyas in Maddison et al. 1984) allows for a more severe test ofphylogenetic relationships by removing this assumption(Meacham 1986; Nixon and Carpenter 1993) and shouldrender a more stable ingroup topology (Puslednik andSerb 2008).
A separate problem involves the selection ofindividual outgroup taxa. Smith (1994) suggested thatsampling the two closest successive sister groups isthe ideal scheme for choosing outgroups. However,sampling beyond the two closest sister groups allowsa further test that the presumed sister group is,in fact, the sister group (Graham et al. 2002). Asimilar suggestion is made by Brusatte et al. (2010)for determining the minimum number of speciesnecessary to represent a supraspecific taxon in an
2015 WILBERG—OUTGROUP CHOICE AND THE ORIGIN OF CROCODYLIFORMES 623
analysis of higher level relationships. Yet cautionshould be exercised to avoid sampling a very distantlyrelated taxon (without sampling other outgroup taxato break up the resulting long phylogenetic branch).Very distantly related taxa are expected to share fewerrelevant character states with the ingroup and havehad greater time to accumulate homoplasy. Publishedanalyses rooting on Gracilisuchus and sampling fewother outgroup taxa recover thalattosuchians as sisterto the other longirostrine crocodyliforms (Fig. 1—position c). The phylogenetic position of Gracilisuchusis contentious. While some analyses of archosaurrelationships have recovered Gracilisuchus as a closerelative of Crocodylomorpha (e.g., Brusatte et al. 2010),many have recovered it as somewhat more distantlyrelated (e.g., Benton and Clark 1988; Parrish 1993).Recent work on archosaur phylogenetics suggests thatGracilisuchus is actually a basal suchian, very distantlyrelated to Crocodylomorpha (Nesbitt 2011). Thus,Gracilisuchus might not represent an ideal outgroupchoice in the absence of additional, more proximaloutgroups.
Previous work has demonstrated that outgroupsampling can have a great effect on ingrouprelationships, particularly for labile clades (e.g.,Spaulding et al. 2009). Spaulding et al. (2009) showedthat Mesonychia, an archaic group of carnivorousungulates sometimes linked with the origin of whales(e.g., Luo and Gingrich 1999), is highly labile withinthe ungulate tree. When few outgroups were sampled,Mesonychia was recovered in a derived position (closeto Cetacea), but when the full suite of outgroupswas sampled, Mesonychia was recovered in a basalposition (Spaulding et al. 2009). This conflict, at leastsuperficially, mirrors that seen with Thalattosuchia andthe longirostrine crocodyliforms—an instance of strong,but conflicting, character support for multiple highlydivergent topologies.
The three contentious longirostrine groups areall found in marine environments suggesting theyshare similar ecological/functional pressures. Earlyphylogenetic analyses recovered all three long-snoutedgroups as a clade (e.g., Benton and Clark 1988; Clark1994; Wu et al. 1997; Pol and Norell 2004a, 2004b; Polet al. 2004). This grouping was treated skeptically fromthe beginning and was originally dismissed as the resultof convergence (Clark 1986; Benton and Clark 1988).Later works no longer dismissed the longirostrine clade,but it has remained suspect for three primary reasons(Pol and Gasparini 2009). First, thalattosuchians possessthe plesiomorphic condition for many characters (whichmust be optimized as reversals when sister to otherlongirostrine groups; Benton and Clark 1988). Second,because crocodyliforms have demonstrably evolvedsimilar skull shapes numerous times (Langston 1973;Busbey 1995; Brochu 2001), it was assumed that snoutshape is not a reliable phylogenetic character. Finally,the sister group relationship between thalattosuchiansand pholidosaurs/dyrosaurids requires a long ghostlineage because thalattosuchians first appear in the
Early Jurassic, whereas pholidosaurs/dyrosaurids firstappear in the latest Jurassic. Additionally, analyseshave demonstrated that when characters suspected ofcorrelation with a slender snout shape (as suggested byClark; 1994) are removed, thalattosuchians are recoveredas either basal mesoeucrocodylians (e.g., Buckley andBrochu 1999; Jouve et al. 2006) or as the sister group toCrocodyliformes (e.g., Pol and Gasparini 2009).
Brochu (2001) proposed that the inclusion ofdyrosaurids and pholidosaurs in an analysis draws thethalattosuchians up the tree. When dyrosauridsand pholidosaurs are excluded from analyses,thalattosuchians move back toward the base of thetree (e.g., Buckley et al. 2000; Tykoski et al. 2002; Turnerand Calvo 2005—though removal of dyrosaurids whileretaining pholidosaurs does not change the positionof Thalattosuchia; Jouve et al. 2006). There are threepotential nonmutually exclusive hypotheses to explainthese results.
(1) The three groups of marine crocodylomorphsare indeed monophyletic and thalattosuchians arecharacterized by multiple character state reversalsthat, in the absence of evidence from dyrosauridsor pholidosaurs, are mistaken for plesiomorphies.
(2) They are not monophyletic, but characters relatedto convergence on the long-snouted morphotypeare pulling them together.
(3) Character information supporting thenoncrocodyliform affinities of thalattosuchians ispresent, but the lack of proper outgroup samplingforces the reconstruction of plesiomorphiccharacter states as synapomorphies forThalattosuchia.
This is a difficult question to address by analyzing thecurrent literature. Early attempts at reconstructing thecrocodyliform tree sampled thalattosuchians mostly ascomposite taxa (Metriorhynchidae, Teleosauridae; e.g.,Clark 1994). More recent analyses based on species-level data generally include only the most derivedforms (e.g., Cricosaurus and Dakosaurus), and sampleonly one or two teleosaurids (e.g., Pol and Gasparini2009; Turner and Sertich 2010), or include severalteleosaurids, but weakly sample metriorhynchids (e.g.,Jouve 2009). These sampling schemes may obscureimportant character information that could be obtainedby more comprehensive sampling of thalattosuchiantaxa. Additionally, as most teleosaurid species have yet tobe included in a phylogenetic analysis, much data fromthe early history of the clade has not been sampled. Someof these taxa may retain plesiomorphic character statesthat will impact the relationship of thalattosuchiansamong other crocodylomorphs. The only publishedanalyses sampling a wide range of thalattosuchians arebased on the matrix of Young (e.g., Young and Andrade2009; Young et al. 2012; Parrilla-Bel et al. 2013). However,this data set samples only a single protosuchian taxonand relatively few outgroup taxa.
Thus, a large amount of character and taxoninformation has not yet been included in previous effortsto address the longirostrine problem. Its addition willbe an important test of the competing scenarios. Thisstudy will address these issues by expanding characterinformation and ingroup and outgroup sampling. Iwill also test the effects of outgroup sampling onthe phylogenetic relationships of crocodylomorphs ingeneral, and the position of Thalattosuchia in particular,through modification to the outgroup sampling schemeof this and previously published data sets.
Institutional AbbreviationsCNRST-SUNY, Centre National de Recherche
Scientifique et Technologique du Mali–Stony BrookUniversity, Stony Brook, New York, USA; HMN,Museum für Naturkunde Humboldt-Universtät, Berlin,Germany; NHMUK, Natural History Museum, London,UK; SMC, Sedgwick Museum, Cambridge, UK;UOMNCH, University of Oregon Museum of Naturaland Cultural History, Eugene, Oregon, USA.
MATERIALS AND METHODS
Taxon and Character SamplingTo test the relationships of the thalattosuchian
crocodylomorphs, I performed a phylogeneticanalysis of 394 morphological characters scoredfor eight outgroup and 78 ingroup taxa, including24 thalattosuchian species (online Appendix 1available as Supplementary Material on Dryad athttp://dx.doi.org/10.5061/dryad.00ss6). This new dataset is a modified version of that presented in Wilberg(2015) with the addition of 10 new characters andthe modification of many others (online Appendix2 available as Supplementary Material on Dryad athttp://dx.doi.org/10.5061/dryad.00ss6). To minimizeerrors in character coding, I focused ingroup samplingon specimens I could observe firsthand or those withdetailed published descriptions. I made an effort tosample widely from all major Crocodylomorph groups.Taxon sampling within Thalattosuchia focused oncapturing the broad range of morphologies presentin the group across their entire temporal duration.Outgroup sampling was increased from previousanalyses with the intent of better characterizing thedistribution of character states in noncrocodyliforms.The basal suchian Gracilisuchus was used to root thetree based on its position in the broad scale analysis ofArchosauria by Nesbitt (2011). The rauisuchid (sensuNesbitt 2011) Postosuchus kirkpatricki was includedfor two primary reasons. First, Rauisuchidae hasfrequently been recovered as the sister group toCrocodylomorpha, just outside the phylogeneticallyunstable “Sphenosuchia” (e.g., Benton and Clark 1988;Parrish 1993; Juul 1994; Nesbitt 2011). Second, Postosuchuskirkpatricki is well known from multiple specimens
representing nearly the complete skeleton allowing thescoring of most characters. Six “sphenosuchian” taxawere also sampled. Three of these have been recovered asthe sister taxon to Crocodyliformes in previous analyses(Junggarsuchus sloani, Clark et al. 2004; Kayentasuchuswalkeri, Nesbitt 2011; Almadasuchus figarii, Pol et al. 2013).The inclusion of these taxa will provide a more stringenttest of the potential placement of Thalattosuchia asthe sister group to Crocodyliformes. To assess thesensitivity of the topology to outgroup sampling, theanalysis was also run in three permutations: Excludingthe basal suchian Gracilisuchus (rooting on Postosuchus);excluding the noncrocodylomorph taxa Gracilisuchusand Postosuchus (rooting on Hesperosuchus agilis);and excluding all noncrocodyliforms and rooting onthe protosuchian Orthosuchus stormbergi as in somepublished analyses (e.g., Sereno and Larsson 2009).
As with any paleontological phylogenetic analysis,the study data set contains relatively high amounts ofmissing data (40.75% missing or inapplicable). Muchof the missing data is concentrated in the postcranialcharacters as numerous crocodylomorph taxa are knownprimarily from cranial material. Three taxa (Zaraasuchusshepardi, Eoneustes gaudryi, and Steneosaurus brevidens)are highly incomplete (80–82%), whereas medianincompleteness per taxon is ∼36%. However, whilemissing data has been shown to reduce phylogeneticaccuracy (e.g., Wiens 2003; Prevosti and Chemisquy2010 and references therein), the quantity of missingdata does not directly correlate with the informationcontent of a taxon. A highly incomplete taxon maystill increase resolution if it contains informativesynapomorphic information (Kearney and Clark 2003;Wiens 2003).
Parsimony AnalysisThe phylogenetic data set was analyzed in TNT v1.1
(Goloboff et al. 2008) using equally weighted parsimony.Minimal length trees were found using a heuristicsearch with 1000 replicates of Wagner trees usingrandom addition sequences followed by tree bisectionand reconnection (TBR) branch swapping. The shortesttrees obtained from these replicates were subjected toa final round of TBR branch swapping to ensure allminimum length trees were discovered. Zero lengthbranches were collapsed if they lacked support underany of the minimal length trees (Rule 1 of Coddingtonand Scharff 1994). Two separate analyses were run. Inthe first, to test the effect of potentially nested sets ofhomologies present in some multistate characters, 36characters were treated as ordered (online Appendix2 available as Supplementary Material on Dryad athttp://dx.doi.org/10.5061/dryad.00ss6). In the second,multistate characters were treated as unordered toavoid making a priori assumptions about the process ofevolution (though whether treating such characters asunordered involves better justified assumptions has beenquestioned; e.g., Lipscomb 1992; Slowinski 1993).
2015 WILBERG—OUTGROUP CHOICE AND THE ORIGIN OF CROCODYLIFORMES 625
Nodal SupportNodal support was assessed using jackknife
resampling as applied to character data (Farris et al.1996). Jackknife support was calculated in TNT using1000 replicates with the probability of independentcharacter removal set at 0.37 (∼e−1; as recommendedin Farris et al. 1996). A heuristic search was employedwith each replicate consisting of 10 random additionsequences, saving 10 trees per replicate. The resultingtopologies were summarized using GC frequencies(difference between the frequency of recovering a givengroup and the most frequent contradictory group;Goloboff et al. 2003). GC frequencies are preferred overabsolute frequencies (the standard method of countingfrequencies in bootstrap and jackknife analyses) becausethey account for the evidence in support of a clade aswell as the amount of evidence falsifying that clade.
Comparative MatricesTo assess the effect of outgroup sampling on tree
topology, two previously published crocodylomorphtaxon-character matrices (Turner and Buckley 2008;Sereno and Larsson 2009) were investigated. Theanalysis of Turner and Buckley (2008) consists of 75taxa and 290 characters and includes Gracilisuchusstipanicicorum, Terrestrisuchus gracilis, and Dibothrosuchuselaphros as outgroup taxa (rooted on Gracilisuchus).The analysis of Sereno and Larsson (2009) includes 43taxa and 252 characters (rooted on the protosuchianOrthosuchus stormbergi). Both matrices were unalteredwith the exception of the addition of new outgrouptaxa. In the case of Turner and Buckley (2008),the single terminal taxon Postosuchus kirkpatricki wasadded. For comparative purposes, both Postosuchusand Gracilisuchus were added to the data set ofSereno and Larsson (2009). These data sets wereanalyzed using unweighted parsimony in TNT v. 1.1and the same search parameters described above. Bothanalyses incorporated additive characters, and thesewere retained as such. Gracilisuchus was set as theroot for both matrices. All phylogenetic data setsare available as Supplementary Material on Dryad athttp://dx.doi.org/10.5061/dryad.00ss6.
RESULTS
Parsimony analysis of the ordered study data setresulted in 42 most parsimonious trees (MPTs) of length1691 (Fig. 2; consistency index [CI] = 0.306, retentionindex [RI] = 0.710). The unordered analysis yielded566 MPTs (length = 1648; CI = 0.312; RI = 0.703). Thestrict consensus of the unordered analysis is highlycongruent with that of the ordered analysis, but muchresolution is lost within Notosuchia and Neosuchia(Fig. S1 available as Supplementary Material on Dryad athttp://dx.doi.org/10.5061/dryad.00ss6). Details of theresults discussed below pertain to the ordered analysisunless otherwise noted.
In all MPTs, thalattosuchians are sister toCrocodyliformes, rather than within. This resultoccurs in spite of the inclusion of numerousnonthalattosuchian long-snouted taxa (dyrosauridsand pholidosaurs). Twenty unambiguous andnine ambiguous synapomorphies support theexclusion of Thalattosuchia from Crocodyliformes(Table 1). An additional six steps are required forThalattosuchia to rejoin Crocodyliformes, where theyare recovered as basal mesoeucrocodylianas (Fig. S2available as Supplementary Material on Dryad athttp://dx.doi.org/10.5061/dryad.00ss6), whereasinclusion of Thalattosuchia in the longrostrineclade demands 12 additional steps (Fig. S3available as Supplementary Material on Dryad athttp://dx.doi.org/10.5061/dryad.00ss6). Monophylyof Thalattosuchia is robustly supported by a GCjackknife index of 99. Within Thalattosuchia, the genusSteneosaurus is paraphyletic, and in need of taxonomicrevision (as suggested also by the analysis of Mueller-Töwe, 2006). Pelagosaurus typus, a thalattosuchian ofcontroversial affinity, is here found as the basal-mostmetriorhynchoid, a relationship recovered in somerecent analyses (e.g., Pol et al. 2009; Young et al.2012, 2013; Adams 2013; Parrilla-Bel et al. 2013) and isfairly well supported (GC jackknife: 69). Relationshipswithin Metriorhynchoidea differ slightly from those ofYoung and Andrade (2009) and subsequent analysesbased on the same data set (e.g., Young et al. 2012).However, the few incongruent nodes are not particularlystrongly supported here, and the analysis by Youngand Andrade (2009) includes a more comprehensivesampling of metriorhynchids.
Kayentasuchus is here recovered as the sister taxonto Crocodyliformes + Thalattosuchia (consistent withthe topology of Nesbitt 2011). This differs from theanalyses of Clark et al. (2004) and Pol et al. (2013) whichrecovered Junggarsuchus and Almadasuchus, respectively,as the sister taxon to Crocodyliformes. However, theposition of Kayentasuchus is not strongly supportedand only two additional steps are required to makethe clade of Junggarsuchus + Almadasuchus sisterto Crocodyliformes + Thalattosuchia. Many of thefeatures of the braincase linking Junggarsuchus andAlmadasuchus with crocodyliforms are not preserved inKayentasuchus, thus additional fossil material may berequired to resolve this portion of the tree with anyconfidence.
This analysis recovers a monophyletic Protosuchia,an uncommon result found in some previous analyses(e.g., Wu et al. 1994, 1997, 2001; Jouve et al. 2006).The analysis also recovers a monophyletic Notosuchia(including a monophyletic Peirosauridae in some MPTs),Atoposauridae, Goniopholididae, Dyrosauridae, andPholidosauridae—clades commonly recovered in otherpublished analyses. Elosuchus, a longirostrine taxononce proposed to be closely related to Stolokrosuchus(Lapparent de Broin 2002), is recovered as sister to a cladeformed by Dyrosauridae and Pholidosauridae, distantfrom Stolokrosuchus.
FIGURE 2. Strict consensus of 42 MPTs of length 1691 (CI = 0.306, RI = 0.710). Values above nodes represent GC jackknife support scores.Thalattosuchia is highlighted in gray.
2015 WILBERG—OUTGROUP CHOICE AND THE ORIGIN OF CROCODYLIFORMES 627
TABLE 1. Synapomorphies supporting the exclusion of Thalattosuchia from Crocodyliformes. Characters indicated in bold have a CI= 1.0. Character state transformations indicated with an asterisk support this topology only under ACCTRAN (accelerated transformation)optimization; transformations indicated with a “^” support this topology only under DELTRAN (delayed transformation) optimization. Allother synapomorphies are unambiguous
Analysis of the ordered data set excludingGracilisuchus and rooting on Postosuchus resultedin 42 MPTs (length 1676). Analysis of the ordered dataset excluding the noncrocodylomorph outgroups(Gracilisuchus and Postosuchus) and rooting onHesperosuchus resulted in the same 42 MPTs (length1632). The strict consensus for each of these reducedanalyses is identical to that of the full data set (Fig. S4available as Supplementary Material on Dryad athttp://dx.doi.org/10.5061/dryad.00ss6). Permutationsof the unordered analysis with restricted outgroupsampling yielded results congruent with the orderedanalyses and will not be discussed here.
Analysis of the ordered data set excluding theeight “outgroup” taxa and rooted instead on theprotosuchian Orthosuchus resulted in 32 MPTs (Fig. 3;length 1493). Importantly, thalattosuchians join a cladewith other longirostrine crocodyliforms (dyrosaurids,pholidosaurs, and Elosuchus). Protosuchia is no longermonophyletic (which it cannot be if the tree isrooted on a single protosuchian taxon), and muchresolution is lost among notosuchians. Most other cladesare retained. Internal relationships of Thalattosuchiaare drastically different from the original analysis.Teleosaurids form a paraphyletic grade leading toMetriorhynchoidea. Steneosaurus remains paraphyletic,but Teleosaurus becomes the basal-most memberof Thalattosuchia. Metriorhynchoid relationships areunchanged. Two additional steps are required to recoverthe thalattosuchian topology present in the full analysis(with teleosaurid monophyly and S. brevior + S. brevidensas the basal-most teleosaurids).
The recovery of Thalattosuchia as part of thelongirostrine clade reverses the polarity of twocharacters (41 and 304). Character 41 describes theorientation of the orbit. In all outgroup taxa (aswell as basal crocodyliforms), the orbits are laterallyoriented. The same is true of metriorhynchoids andof the basal-most teleosaurids (e.g., S. brevior). Morederived teleosaurids (e.g., Teleosaurus), however, possessmore dorsally directed orbits. When Thalattosuchiais a part of the longirostrine clade, dorsally directed
orbits are optimized as the primitive state, with laterallydirected orbits arising secondarily in metriorhynchoidsand S. brevior. Character 304 describes the number ofpremaxillary teeth. When thalattosuchians are sister toCrocodyliformes, four premaxillary teeth are optimizedas primitive for the clade, with an increase to fivein Teleosaurus, and a decrease to three in someteleosaurids and in metriorhynchoids more derivedthan Pelagosaurus. However, when thalattosuchians jointhe longirostrine clade, five premaxillary teeth areoptimized as primitive, drawing Teleosaurus to the baseof the Thalattosuchia.
Comparative MatricesReanalysis of the data matrix of Turner and Buckley
(2008) resulted in 120 MPTs of length 1122 (Fig. 4a; CI =0.320, RI = 0.706). Addition of Postosuchus to the matrixdrastically alters the resulting topology (Fig. 4b). Thisanalysis resulted in 16 trees of length 1157 (CI = 0.310,RI = 0.692). Thalattosuchians are recovered as the sistergroup to Crocodyliformes (an additional two steps arerequired to place them within Crocodyliformes). Theexclusion of Thalattosuchia from Crocodyliformes issupported by 13 unambiguous and five ambiguoussynapomorphies (Table 1). Protosuchia becomesmonophyletic including Fruitachampsa and Hsisosuchus.Notosuchian relationships remain largely unchanged,but the peirosaurids (including Mahajangasuchus) joinNotosuchia. This result is very similar to the topologyrecovered by Turner and Sertich (2010). Relationships inthe rest of the tree are largely identical.
Analysis of the data set of Sereno and Larsson (2009)resulted in two MPTs of length 984 (Fig. 5a; CI =0.34, RI = 0.66). This result differs slightly from thatreported by Sereno and Larsson (2009). The differencein the number of MPTs is merely a result of the strictercollapsing rule employed here. However, the tree lengthis two steps shorter than in Sereno and Larsson (2009),and the relationships between Sphagesaurus, Notosuchus,and Malawisuchus are fully resolved. The cause of this
FIGURE 3. Strict consensus of 32 MPTs of length 1493 when outgroup taxa are excluded and the tree is rooted on Orthosuchus stormbergi(ordered analysis). Thalattosuchians are highlighted in gray.
discrepancy is unknown; however, the difference isminor and should not affect interpretations drawn fromthis study.
Addition of Gracilisuchus and Postosuchus causesmajor topological changes (Fig. 5b). This analysis
resulted in 12 MPTs (length = 1029; CI = 0.323,RI = 0.641). Thalattosuchia is again sister toCrocodyliformes, rather than within. One ambiguousand 19 unambiguous synapomorphies support theexclusion of Thalattosuchia from Crocodyliformes
FIGURE 4. Trees resulting from the reanalysis of the matrix of Turner and Buckley (2008). a) Original analysis: strict consensus of 120 MPTsof length 1222 (CI = 0.320, RI = 0.706); b) Modified analysis including Postosuchus: Strict consensus of 16 MPTs of length 1157 (CI = 0.310, RI =0.692). Thalattosuchians are highlighted in gray.
(Table 1). An additional 10 steps are required to placeThalattosuchia within Crocodyliformes. Another majorchange involves taxa normally allied with Notosuchia.Araripesuchus becomes polyphyletic and, with the cladeformed by Uruguaysuchus, Simosuchus, and Anatosuchus,forms a paraphyletic grade leading to neosuchians.
DISCUSSION
The analysis presented here (Fig. 2) recoversthalattosuchians as the sister group to Crocodyliformes.This result occurs in spite of relatively dense samplingof other longirostrine taxa (four pholidosaurs, threedyrosaurs, Elosuchus, Stolokrosuchus, and Gavialis). Theeffect of excluding other longirostrine taxa on thephylogenetic position of Thalattosuchia has beendiscussed by several authors (e.g., Buckley and Brochu1999; Jouve et al. 2006; Jouve 2009; Pol and Gasparini
2009). When pholidosaurs and dyrosaurs were excludedfrom these analyses, thalattosuchians were recoveredeither as basal mesoeucrocodylians (when rooting ona protosuchian; Buckley and Brochu 1999), or sisterto Crocodyliformes (when rooting on Gracilisuchus;Pol and Gasparini 2009). Thus, the results presentedhere are novel in recovering a basal position forthalattosuchians in spite of extensive longirostrine taxonsampling. The analysis of Young and Andrade (2009)also sampled numerous longirostrine taxa, includingpholidosaurs, dyrosaurs, and thalattosuchians, butrecovered thalattosuchians as basal mesoeucrocodyliansrather than outside of Crocodyliformes. However, thisanalysis sampled only a single protosuchian, potentiallyobscuring character state optimizations at the base of thetree.
The addition of Postosuchus to the outgroup acts tochange the polarity of several character transformations
FIGURE 5. Trees resulting from the reanalysis of the matrix of Sereno and Larsson (2009). a) Original analysis: strict consensus of two MPTsof length 984 (CI = 0.340, RI = 0.660); b) Modified analysis including Gracilisuchus and Postosuchus: strict consensus of 12 MPTs of length 1029(CI = 0.323, RI = 0.641). Thalattosuchian taxa are highlighted in gray.
in the phylogenetic matrices. Many of these characterstate changes are synapomorphies for Crocodyliformesthat were previously optimized as reversals inthalattosuchians. Inclusion of Postosuchus also increasedthe number of character states that were scored for theroot. For example, character 206 of Turner and Buckley(2008; cranial table width with respect to ventral portionof skull: as wide as ventral portion—0); or narrower thanventral portion of skull—1) was scored as unknownfor all noncrocodyliform outgroups, leading to state 1being reconstructed at the root. Addition of Postosuchus(scored for state 0), reconstructs state 0 as plesiomorphic.This state is shared with thalattosuchians, and makes anarrow skull table a synapomorphy of crocodyliforms(though the state is reversed in some protosuchians andBaurusuchus).
Clark (in Benton and Clark 1988) noted thatthalattosuchians lack several crocodyliformsynapomorphies, and suggested that thalattosuchiansmight be the sister group to crocodyliforms (though thistopology was not among the most parsimonious).The plesiomorphic character states retained bythalattosuchians were reconstructed as secondarylosses in topologies in which Thalattosuchia fell withinCrocodyliformes. Resolving thalattosuchians as thesister group to crocodyliforms instead reconstructs thelack of these features as plesiomorphic, making severalcharacters perfectly congruent (CI = 1). I shall discussseveral of these characters that exclude thalattosuchiansfrom Crocodyliformes.
All crocodyliforms, exclusive of thalattosuchians,possess a relatively flat skull table (char. 24 of Turnerand Buckley 2008), whereas no noncrocodyliformcrocodylomorphs possess this morphology (thoughAlmadasuchus comes close). Perhaps more importantly,characters relating to the relationship of the quadratewith the braincase and the presence of a prearticularbone are more congruently optimized whenthalattosuchians are not members of Crocodyliformes.One major morphological feature distinguishingcrocodyliforms from other crocodile-line archosaursis the extensive reinforcement of the skull, with thejaw joint tightly articulated with the braincase (Clark1994). This allows for expanded jaw musculature andthe ability to resist torsion caused by struggling prey(Busbey 1995). This transformation appears to haveoccurred in mosaic fashion, with the quadrate graduallyincreasing in articulation with the braincase, beginningin derived “sphenosuchian” crocodylomorphs (Polet al. 2013). In all crocodyliforms, the quadrate is tightlysutured to the braincase, contacting the laterosphenoid(covering most of the prootic externally; char. 47 ofTurner and Buckley 2008; char. 161 of Sereno andLarsson 2009). Such is not the case in thalattosuchians,where the quadrate is expanded onto the braincase,but is not directly sutured to the laterosphenoid,basisphenoid, or pterygoid (Holliday and Witmer 2009;Jouve 2009; Fernández et al. 2011). This leaves a broadlyexposed prootic that forms much of the lateral surfaceof the braincase. When thalattosuchians form a clade
2015 WILBERG—OUTGROUP CHOICE AND THE ORIGIN OF CROCODYLIFORMES 631
with pholidosaurs and dyrosaurs, they are optimized assecondarily losing the tight articulation of the quadratewith the braincase; a scenario which seems unlikelyfrom a functional standpoint given that numerousmembers have been interpreted as large macrophagouspredators (Young et al. 2012).
Presence of the prearticular bone in the mandiblepredates Tetrapoda (present in Osteichthyes; Friedmanand Brazeau 2010). Absence of a prearticular isan apomorphic feature of mesoeucrocodylians,but a prearticular is present in noncrocodyliformcrocodylomorphs, protosuchians, and thalattosuchians.In topologies in which thalattosuchians are alliedwith longirostrine crocodyliforms, thalattosuchiansare inferred to have re-evolved this feature. Loss ofthe feature is a synapomorphy for Mesoeucrocodyliain topologies where thalattosuchians are sister tocrocodyliforms, or Metasuchia if thalattosuchians arebasal mesoeucrocodyilans.
Another potentially important feature involvesthe groove on the lateral surface of the squamosalfor external ear valve musculature attachment(char. 56 of Sereno and Larsson). This feature isabsent in noncrocodyliform crocodylomorphs (andthalattosuchians) with the possible exception ofKayentasuchus, but is present in all crocodyliforms.The ear valve acts to close the otic aperture whilesubmerged and the absence of this feature in highlyaquatic thalattosuchians seems counterintuitive (unlessthey evolved a different solution, or closed off theaperture completely as in cetaceans). However, otherhighly aquatic crocodyliforms such a dyrosauridsretain the insertion for this musculature. The presenceof a squamosal groove in Kayentasuchus (the sistertaxon to Thalattosuchia + Crocodyliformes), if trulyhomologous with that of crocodyliforms, rendersancestral state reconstruction ambiguous in this part ofthe tree.
In spite of the numerous plesiomorphic featuresretained in thalattosuchians, they possess some featuresuniting them with other long-snouted crocodyliforms.Many of these characters have been hypothesizedto be correlated with longirostry (Benton andClark 1988; Clark 1994; Jouve et al. 2006; Pol andGasparini 2009). This issue was extensively discussedby Pol and Gasparini (2009), and they presentedevidence demonstrating that many of these putative“longirostrine characters” are incompatible withbiological dependence (eight of the 12 investigated).Only the extensive participation of the splenial inthe mandibular symphysis was optimized as anunambiguous synapomorphy for the longirostrineclade. However, exclusion of the four potentiallynonindependent characters in their data set didnot change their topology and thalattosuchians andpholidosaurs/dyrosaurids remained a clade.
Some of the other characters uniting thalattosuchiansand pholidosaurs/dyrosaurids in the analysis of Pol andGasparini (2009) may be misleading due to vagueness ofcharacter state descriptions failing to accurately reflect
a) b)
na
sq
FIGURE 6. Dorsal surface of two crocodylomorph skullsillustrating anatomical differences of similarly shaped and identicallycoded character states. a) A thalattosuchian - Peipehsuchus teleorhinus;b) A dyrosaurid - Dyrosaurus maghribensis (redrawn from Jouveet al. 2006). Nasal bones demonstrating their different contributionto the “tubular” snout; Squamosals showing their differentialcontribution to the elongation of the supratemporal fenestra.Anatomical abbreviations: na, nasal; sq, squamosal.
anatomy. The first step in character construction ofteninvolves the search for similarity in structures betweentaxa. However, in looking for this similarity, one shouldnot resort to gross similarity in shape or size. Whenpossible the systematist should refer to topologicalsimilarity, connectivity, and structural correspondence(Rieppel and Kearney 2002). As noted by Nesbitt(2011), characters describing the shape of complexstructures formed by multiple bones can be problematic.Modifications to different bones could produce the samegeneral shape. These character states would be scoredthe same, even though the architecture of the structureis anatomically different. For example, dyrosaurids andthalattosuchians share greatly enlarged supratemporalfenestrae (optimized as an ambiguous synapomorphyof the longirostrine clade in Pol and Gasparini 2009).But, in thalattosuchians the elongation of these fenestraeis caused primarily by posterior elongation of thepostorbital bone, whereas in dyrosaurids, it is causedby anterior elongation of the squamosal (Fig. 6).
An elongate “tubular” rostrum is another suchfeature (char. 3 of Turner and Buckley 2008; similar tochar. 4 of Sereno and Larsson 2009). The snout is acomplex structure composed of several bones. Whilemost thalattosuchians share a tube-shaped snout withpholidosaurs/dyrosaurids, the bones forming the snoutare quite different. In thalattosuchians, the snout iscomposed of the maxillae and premaxillae, whereas inpholidosaurs and dyrosaurids, the nasals extend thelength of the snout, forming the dorsal portion (Fig. 6).This major anatomical difference suggests these taxashould not be scored the same for this character, andthat characters such as this might best be elaborated toinclude references to the individual bones forming thestructure.
Interpretation of the rostral form character is furtherclouded because this character is often treated asadditive (e.g., Turner and Buckley 2008). However, it isunclear to this author how these character states (narroworeinirostral [0], broad oreinirostral [1], nearly tubular[2], or platyrostral [3]) form a logical transformationseries. Is “platyrostral” really more similar to “tubular”than to “broad oreinirostral”? Reanalysis of the Turnerand Buckley matrix with character 3 treated asunordered greatly reduces resolution in the strictconsensus, but has no effect on the longirostrine clade.
Another synapomorphy of the longirostrine clade isthe transversely flattened postorbital bar (char. 26 ofTurner and Buckley 2008). However, constructionof the postorbital bar differs radically betweenthalattosuchians and pholidosaurs/dyrosaurids.Thalattosuchians are unique among crocodylomorphsin that the postorbital is positioned lateral to the jugalalong the postorbital bar. Ancestrally, the postorbitallies anterior to the jugal, but in all crocodyliforms, thepostorbital lies medial to the jugal (with the exceptionof the phylogenetically enigmatic Hsisosuchus, wherethe ancestral state is retained). Thus, thalattosuchiansand pholidosaurs/dyrosaurids should not be scoredidentically for this character. This character should onlybe scored for taxa in which the postorbital lies medialto the jugal. Thalattosuchians and taxa possessing apostorbital that lies anterior to the jugal should be codedas inapplicable.
A further synapomorphy uniting longirostrinetaxa could result from insufficient fossil material.Thalattosuchians and pholidosaurs/dyrosaurids wereunited in lacking a mastoid antrum that extends throughthe supraoccipital (char. 63 of Turner and Buckley 2008;char. 170 of Sereno and Larsson 2009). This characterstate has no obvious functional relation to the possessionof a long snout and thus seems good evidence of sharedancestry. However, observation of this character in fossiltaxa is often difficult, due to infilling by sediment orpoor preservation. Its visibility relies on the availabilityof CT data, hemisected specimens, or specimens thatare serendipitously broken to expose this passage.
The lack of a continuous mastoid antrum is visiblein exquisitely preserved acid prepared specimensof Pelagosaurus typus (pers. obs; NHMUK PV 32600;
dvs
fm
so
oc
pf
icc
mec mascc
ma
otc
b)
c)
d)
a)
FIGURE 7. Specimen photographs demonstrating continuityor discontinuity of the mastoid antra. a) A hemisected teleosauridbraincase (“Teleosaurus macrocephalus”; SMC J35177a) with a solidsupraoccipital. b, c) natural endocast of the ear canal region fromPholidosaurus meyeri (HMN R. 2066) in anterior (b) and dorsal (c)views showing a large and fully connected mastoid antrum. d) partialbraincase of Rhabdognathus keinensis (CNRST-SUNY 277) showing thegreatly enlarged otic capsules and a narrow, but continuous tubethrough the supraoccipital here interpreted as the transverse passageof the mastoid antra. Anatomical abbreviations: dvs, dorsal venoussinus; fm, foramen magnum; icc, internal carotid canal; ma, mastoidantrum; mec, middle ear canal; oc, occipital condyle; otc, otic capsules;pf, pituitary fossa; scc, semicircular canals; so, supraoccipital.
Clark 1986), as well as hemisected teleosauroidbraincases (Fig. 7a; SMC J35177a; SMC J35177b; Seeley1880). Recent CT scans of Metriorhynchus cf. westermani(Fernández et al. 2011), and the fractured braincase of the
2015 WILBERG—OUTGROUP CHOICE AND THE ORIGIN OF CROCODYLIFORMES 633
metriorhynchoid Zoneait nargorum (UOMNCH F39539;Wilberg 2015) also demonstrate a solid supraoccipital.Noncrocodyliform crocodylomorphs for which thischaracter can be scored all possess a solid supraoccipital(Sphenosuchus acutus—Walker 1990; Kayentasuchuswalkeri—Clark and Sues 2002; Junggarsuchus sloani—Clark et al. 2004). Wu and Chatterjee (1993) state thatDibothrosuchus elaphros possesses a mastoid antrumextending through the supraoccipital, but whether thepneumatic recesses actually connect is unobservable inthe holotype (J. Clark, pers. com.). Thus, based on thephylogenetic hypothesis presented here, the presenceof a continuous mastoid antrum is a synapomorphy ofCrocodyliformes. If it is present in Dibothrosuchus this isthe result of convergence.
The state of this character inpholidosaurs/dyrosaurids has been somewhatuncertain. In the first cladistic matrices of Clark(Benton and Clark 1988; Clark 1994), Dyrosaurus andPholidosaurus are coded as lacking a mastoid antrumthat extends through the supraoccipital. These codingshave persisted in most matrices based directly on theoriginal work of Clark (e.g., Turner and Buckley 2008; Polet al. 2009), and Sarcosuchus imperator and Terminonarisrobusta have also been coded as lacking this feature.However, natural endocasts of the middle ear region andassociated structures of Pholidosaurus meyeri (HMN R.2066; Koken 1887) clearly demonstrate a fully connectedand well-expanded mastoid antrum above the braincavity (Fig. 7b, c). Unfortunately, these endocastswere attributed to Pholidosaurus on the sole basis thatPholidosaurus was the only known crocodylomorph fromthat stratigraphic level of the Hastings sandstone (D.Schwarz-Wings pers. comm.). Evidence for a continuousmastoid antrum in pholidosaurs would be greatlystrengthened by CT scanning of pholidosaur braincasematerial, but this is beyond the scope of this study. Thestate in dyrosaurids is slightly more difficult to interpret.
Dyrosaurids possess greatly expanded otic capsules,often so large that they actually meet at the midline,separating a dorsal and ventral portion of the braincavity (Fig. 7d). This inflation forces the roof ofthe brain cavity to expand dorsally, narrowing thesupraoccipital. Thus, it could be interpreted that the ex-panded otic capsules of dyrosaurids caused theclosure of the transverse canal connecting the mastoidantra. However, a specimen of Rhabdognathus keinensis(CNRST-SUNY 277) shows a narrow tube extendingthrough the supraoccipital in the dorsal roof of thebraincase (Fig. 7d). This is interpreted as a narrowed,but continuous mastoid antrum connecting the middleear regions. This interpretation is supported by thereport of a similar diverticulum present in CT scansof Rhabdognathus aslerensis (CNRST-SUNY 190; Dufeau2011). While I have no evidence to contradict previouscodings for Sarcosuchus, Dyrosaurus, or Terminonaris,given the presence of a fully connected mastoid antrumin both Pholidosaurus and Rhabdognathus, two relativelydistantly related taxa, it seems likely that pholidosaursand dyrosaurs do possess a continuous mastoid antrum
as in all other crocodyliform taxa for which this charactercan be scored.
The character modifications suggested abovemay have a profound effect on the phylogeneticplacement of Thalattosuchia, even in the absence ofexpanded outgroup sampling. To test this proposition,I recoded three characters in the matrix of Turner andBuckley (2008): Characters 3 (rostrum proportions),26 (postorbital bar shape), and 63 (mastoid antrumconnectivity). The rostrum proportions character wasrephrased based on the shape of the maxilla andtopological relationships between the maxilla andadjacent bones as follows: 0: maxilla much tallerthan wide, lateral surface nearly vertical (“narroworeinirostral”); 1: maxilla much taller than broad, dorsalsurface inclined toward midline (“broad oreinirostral”);2: maxilla width and height approximately equal, morethan three times as long as tall, bordered medially bynasals (“nearly tubular” with maxillae not meetingat midline); 3: maxillae wider than tall, dorsomedialsurface near horizontal (“platyrostral”); 4: maxilla widthand height approximately equal, more than three timesas long as tall, maxillae meeting at midline (“nearlytubular” w/ abbreviated nasal contribution). In essence,this modification splits the original state 3 “nearlytubular” to take into account the contribution of thenasals to the rostrum. All thalattosuchians (except thetwo species of Dakosaurus) were recoded as state 4, aswas Gavialis. This character is here treated as unordered.
Modification of character 26 (postorbital barshape) involved the recoding of thalattosuchiantaxa, Gracilisuchus, Terrestrisuchus, Dibothrosuchus,and Hsisosuchus as inapplicable (though, of course,current computer algorithms treat inapplicable datain the same fashion as missing data). Character 63(continuity of mastoid antrum) was recoded as presentin Rhabdognathus and Pholidosaurus based on evidencepresented above. This feature was originally coded asabsent for Sarcosuchus, Terminonaris, and Dyrosaurusand these codings were retained in the absence ofcontradictory evidence.
The data set of Turner and Buckley (2008)was reanalyzed using the original matrix andsearch parameters, but with modification to thescoring of the three characters mentioned above(the modified taxon-characer matrix is availableavailable as Supplementary Material on Dryadat http://dx.doi.org/10.5061/dryad.00ss6). Thissearch resulted in 2038 MPTs of length 1121. Thestrict consensus is largely unresolved (Fig. S5available as Supplementary Material on Dryad athttp://dx.doi.org/10.5061/dryad.00ss6), but in someof the MPTs Thalattosuchia is recovered as the sistergroup to Crocodyliformes. These results demonstratethe importance of character formulation and scoring.
Secondary Palate EvolutionUnderstanding of the evolutionary history
vomers and maxillaemaxillae onlypalatines onlypterygoids only
FIGURE 8. Strict consensus tree showing ACCTRAN optimization of the secondary palate character. Taxa highlighted in gray possess asecondary palate constructed differently from other taxa possessing the same bones forming the anterior border of the choana. It should benoted that while both “sphenosuchians” and protosuchians possess an internal choana bordered anteriorly by the maxillae, the secondary palateis more extensive among protosuchians.
development of the bony secondary palate through threeevolutionary grades: “Protosuchia”, “Mesosuchia”, andEusuchia (e.g., Langston 1973). Protosuchians have amodest secondary palate, with the internal choanabordered anteriorly by the maxillae. “Mesosuchians”have a more extensive palate, with the choana borderedanteriorly by the palatines, and eusuchians have anelaborate secondary palate with the internal choanaenclosed entirely within the pterygoids. However, ifthalattosuchians are the sister group to Crocodyliformes,this scenario is more complicated than it once seemed.
Thalattosuchians possess a well-developed secondarypalate of the “mesosuchian” grade. Optimizationof the palatal development character (char. 163)suggests that the “mesosuchian” palate is primitive forCrocodyliformes, and an extensive secondary palate hasbeen lost at least twice (Fig. 8). Other nonthalattosuchiantaxa further complicate the evolution of this character.Zossuchus davidsoni and Fruitachampsa callisoni(protosuchians) possesses a palate with two choanalopenings; the posterior (and presumably functional)opening is enclosed anteriorly by the palatines, as in
2015 WILBERG—OUTGROUP CHOICE AND THE ORIGIN OF CROCODYLIFORMES 635
“mesosuchian”—grade crocodyliforms, whereas theanterior is bordered anteriorly by the maxillae as inthe “protosuchian condition” (Pol and Norell 2004a;Clark 2011). North American goniopholidids (derivedneosuchians) such as Eutretauranosuchus delfsi andCalsoyasuchus valliceps, share an internal choana that isanteroposteriorly elongate and bordered anteriorly bythe maxillae (the “protosuchian” condition; Pritchardet al. 2013). Evolution of the “eusuchian condition” isalso homoplastic. Mahajangasuchus insignis possessesan internal choana entirely enclosed by the pterygoids,but in a manner unlike that of eusuchians (Turner andBuckley 2008). Thus, the evolution of the secondarypalate is much more complicated than once thought,likely involving multiple gains and/or losses.
CONCLUSIONS
Previous studies have demonstrated the importanceof taxon and character sampling to the issue ofthalattosuchian relationships (e.g., Clark 1994; Buckleyand Brochu 1999; Jouve et al. 2006; Jouve 2009; Pol andGasparini 2009). This study is the first to demonstrate thegreat effect of outgroup sampling on the phylogeneticstructure of Crocodylomorpha. The goal of phylogeneticanalysis may be to recover structure within the ingroup(Maddison et al. 1984), but broad outgroup samplingacts as a test of ingroup monophyly. While there areno guarantees that a particular outgroup taxon ornumber of outgroups will yield the “correct” ingrouptopology (Nixon and Carpenter 1993), it seems prudentto sample numerous outgroup taxa especially wherelarge-scale relationships between clades remain anactive question. Analyses of crocodyliform relationshipsshould include outgroups sampled from closely relatednoncrocodyliform clades. If the relationships betweenthese groups are not well known, the effects of outgroupsampling should be assessed by investigating changesto ingroup topology under different outgroup samplingschemes.
This study also highlights issues related tocharacter construction and state description. Existingcrocodyliform character sets include only brief statedescriptions and are rarely figured. Some of the includedcharacters may treat homology at too superficial of alevel (e.g. general shapes of complex, multipartitestructures). These characters need to be reassessed andtheir states should be described in a higher level of detail.Future work should aim for more clarity in characterdescriptions as interpretations of these characters arecentral to the repeatability of phylogenetic analyses.Legitimate differences in interpretation of morphologyexist, but many differences in character state coding incurrent published literature are likely due to ambiguityin character state descriptions.
Overall, the phylogenetic results presented hereare consistent with numerous previously publishedphylogenetic hypotheses (with the exception of theposition of thalattosuchians). However, while support
for several individual clades is high, the backbone ofthe tree demonstrating the relationships between thesegroups is only weakly supported. It is these nodes thatare of the greatest interest when investigating large-scalepatterns and timing of evolution in crocodylomorphs.Future efforts at resolving these issues should carefullyconsider both outgroup sampling and characterconstruction.
SUPPLEMENTARY MATERIAL
Data available from the Dryad Digital Repository:http://dx.doi.org/10.5061/dryad.00ss6.
FUNDING
Funding for this project was provided by the NationalScience Foundation (Doctoral Dissertation ImprovementGrant DEB-1011097 to C. Brochu) and the University ofIowa Department of Geoscience.
ACKNOWLEGMENTS
I thank C. Brochu, M. Spencer, J. Adrain, and theUniversity of Iowa Paleo. Group for discussion andthank J. Clark, M. Benton, A. Busbey, and an anonymousreviewer for helpful comments on an earlier draftof this manuscript. I also thank editors F. Andersonand N. MacLeod for constructive comments. I thankD. Schwarz-Wings for background concerning thepholidosaur endocasts. The phylogenetic analysis wasperformed in TNT v. 1.1, a program provided free ofcharge thanks to a subsidy from the Willi Hennig Society.I thank the following people for access to specimens: M.Carrano and D. Bohaska (USNM), B. Simpson (FMNH),C. Mehling, and M. Norell (AMNH), L. Steel and S.Chapman (NHMUK), A. Henrici (CM), M. O’Leary(SUNY Stony Brook), J. Cundiff and M. Lyons (MCZ),R. Allain (MNHN), P. Jeffery, and D. Siveter (OUM), T.Tokaryk and M. Vovchuk (SMNH), X. Xu and F. Zheng(IVPP), O. Rauhut and M. Moser (BSPG), R. Schoch, R.Boettcher, and M. Rasser (SMNS), P. Havlik (GPIT), R.Hauff (UH), D. Schwarz-Wings (MNHB), A. Folie andA. Dréze (IRSNB).
REFERENCES
Adams T.L. 2013. A new neosuchian crocodyliform from the lowerCretaceous (late Aptian) twin mountains formation of north-centraltexas. J. Vert. Paleontol. 33:85–101.
Andrade M.B., Edmonds R., Benton M.J., Schouten R. 2011. A newBerriasian species of Goniopholis (Mesoeucrocodylia, Neosuchia)from England, and a review of the genus. Zool. J. Linn. Soc.Lond.163:S66–S108.
Barriel V., Tassy P. 1998. Rooting with multiple outgroups: consensusversus parsimony. Cladistics 14:193–200.
Benton M.J., Clark J.M. 1988. Archosaur phylogeny and therelationships of the Crocodylia. In: M. J. Benton, (ed.), Thephylogeny and classification of the tetrapods, Volume 1. Oxford:Clarendon Press. p. 295–338.
Brochu C.A. 1997a. Morphology, fossils, divergence timing, and thephylogenetic relationships of Gavialis. Syst. Biol. 46:479–522.
Brochu C.A. 2001. Crocodylian snouts in space and time: phylogeneticapproaches toward adaptive radiation. Am. Zool. 41:564–585.
Brochu C.A., Bouaré M.L., Sissoko F., Roberts E.M., O’Leary M.A.2002. A dyrosaurid crocodyliform braincase from Mali. J. Paleontol.76:1060–1071.
Bronzati M., Montefeltro F.C., Langer M.C. 2012. A species-levelsupertree of Crocodyliformes. Hist. Biol. 24:598–606.
Brusatte S.L. 2010. Representing supraspecific taxa in higher-levelphylogenetic analyses: guidelines for palaeontologists. Palaeontol.53:1–9.
Brusatte S.L., Benton M.J., Desojo J.B., Langer M.C. 2010. The higher-level phylogeny of Archosauria. J. Syst. Palaeontol. 8:3–47.
Buckley G.A., Brochu C.A. 1999. An enigmatic new crocodile fromthe upper Cretaceous of Madagascar. Spec. Pap. Palaeontol. 60:149–175.
Buckley, G.A., Brochu C.A., Krause D.W., Pol D. 2000. A pug-nosedcrocodyliform from the Late Cretaceous of Madagascar. Nature405:941–944.
Buffetaut E. 1982. Radiation évolutive, paléoécologie et biogéographiedes crocodiliens mésosuchiens. Mém. S.Géo. F. 142:1–88.
Busbey A.B. 1995. The structural consequences of skull flattening incrocodilians. In: Thomason J.J. (ed.), Functional morphology invertebrate paleontology.. New York: Cambridge University Press,p. 173–192.
Clark J.M. 1986. Phylogenetic relationships of the crocodylomorpharchosaurs. Unpublished Ph.D. Dissertation. University of Chicago556 pp.
Clark J.M. 1994. Patterns of evolution in Mesozoic Crocodyliformes.In: Fraser N.C., Sues H.D., editors. In: The shadow of the dinosaurs.Early mesozoic tetrapods. Cambridge (NY): Cambridge UniversityPress p. 84–97.
Clark J.M., Sues H.–D. 2002. Two new basal crocodylomorpharchosaurs from the Lower Jurassic and the monophyly of theSphenosuchia. Zool. J. Linn. Soc.Lond. 136:77–95.
Clark J.M., Xu X., Forster C.A., Wang Y. 2004. A Middle Jurassic‘sphenosuchian’ from China and the origin of the crocodylian skull.Nature 430:1021–1024.
Clark, J.M. 2011. A new shartegosuchid crocodyliform from the UpperJurassic Morrison Formation of western Colorado. Zool. J. Linn.Soc.-Lond. 163:S152–S172.
Coddington J.A., Scharff N. 1994. Problems with zero-length branches.Cladistics 10:415–423.
Dufeau D.L. 2011. The evolution of cranial pneumaticity inArchosauria: patterns of paratympanic sinus development.Unpublished Ph.D. Dissertation. Ohio University 174 pp.
Farris J.S. 1982. Outgroups and parsimony. Syst. Zool. 31:328–334.Farris J.S., Albert V.A., Källersjö M., Lipscomb D., Kluge A.G. 1996.
Fernández M.S., Carabajal A.P., Gasparini Z., Chong Díaz G. 2011.A metriorhynchid crocodyliform braincase from northern Chile.J. Vert. Paleontol. 31:369–377.
Friedman M., Brazeau M.D. 2010. A reappraisal of the origin and basalradiation of the Osteichthyes. J. Vert. Paleontol. 30:36–56.
Gasparini Z., Pol D., Spaletti L.A. 2006. An unusual marinecrocodyliform from the Jurassic-Cretaceous boundary of Patagonia.Science 311:70–73.
Gatesy J., Amato G., Norell M., DeSalle R., Hayashi C. 2003. Combinedsupport for the wholesale taxic atavism in Gavialine crocodylians.Syst. Biol. 52:403–422.
Goloboff P.A., Farris J.S., Källersjö M., Oxelman B., Ramírez M.J.,Szumik C.A. 2003. Improvements to resampling measures of groupsupport. Cladistics 19:324–332.
Goloboff P.A., Farris J.S., Nixon K.C. 2008. TNT, a free program forphylogenetic analysis. Cladistics 24:774–786.
Graham S.W., Olmstead R.G., Barrett S.C.H. 2002. Rooting phylogenetictrees with distant outgroups: a case study from the commelinoidmonocots. Mol. Biol. Evol. 19:1769–1781.
Holliday C.M., Witmer L.M. 2009. The epipterygoid of crocodyliformsand its significance for the evolution of the orbitotemporal regionof eusuchians. J. Vert. Paleontol. 29:715–733.
Hua S., Buffetaut E. 1997. Part V Crocodylia. In: Calloway J.M., NichollsB.L., editors. Ancient marine reptiles.. New York: Academic Press.p. 357–374.
Jouve S. 2009. The skull of Teleosaurus cadomensis (Crocodylomorpha;Thalattosuchia), and phylogenetic analysis of Thalattosuchia. J. Vert.Paleontol. 29:88–102.
Jouve S., Iarochéne M., Bouya B., Amaghzaz M. 2006. A new speciesof Dyrosaurus (Crocodylomorpha, Dyrosauridae) from the earlyEocene of Morocco: phylogenetic implications. Zool. J. Linn. Soc.Lon. 148:603–656.
Juul L. 1994. The phylogeny of basal archosaurs. Palaeontol. Afr. 31:1–38.
Kälin J.A. 1955. Zur Stammesgeschichte der Crocodilia. Rev. SuisseZool. 62:347–356.
Kearney M., Clark J.M. 2003. Problems due to missing data inphylogenetic analyses including fossils: a critical review. J. Vert.Palenotol. 23:263–274.
Koken E. 1887. Die Dinosaurier, Crocodiliden und Sauropterygier desnorddeutschen Wealden. Palaeont. Abh. 3:309–419.
Langston W. 1973. The crocodilian skull in historical perspective In:Gans C., Parsons T., editors. Biology of the Reptilia.Vol. 4. London:Academic Press. p. 263–284
Lapparent de Broin, F. 2002. Elosuchus, a new genus of crocodile fromthe Lower Cretaceous of the North of Africa. C. R. Palevol. 1:275–285.
Larsson H.C.E., Sues H.-D. 2007. Cranial osteology and phylogeneticrelationships of Hamadasuchus rebouli (Crocodyliformes:Mesoeucrocodylia) from the Cretaceous of Morocco. Zoo. J.Linn. Soc. Lon. 149:533–567.
Lipscomb D.L. 1992. Parsimony, homology and the analysis ofmultistate characters. Cladistics 8:45–65.
Luo Z., Gingerich P.D. 1999. Terrestrial Mesonychia to aquatic Cetacea:transformation of the basicranium and evolution of hearing inwhales. Pap. Mus. Paleontol. Univ. Mich. 31:1–98.
Lyons-Weiler J., Hoelzer G.A., Tausch R.J. 1998. Optimal outgroupanalysis. Biol. J. Linn. Soc. 64:493–511.
McAliley L.R., Willis R.E., Ray D.A., White P.S., Brochu C.A., DensmoreL.D. III. 2006. Are crocodiles really monophyletic?—Evidencefor subdivisions from sequence and morphological data. Mol.Phylogenet. Evol. 39:16–32.
Meacham C.A. 1986. More about directed characters: a reply toDonoghue and Maddison. Taxon 35:538–540.
Meredith R.W., Hekkala E.R., Amato G., Gatesy J. 2011. A phylogenetichypothesis for Crocodylus (Crocodylia) based on mitochondrialDNA: evidence for a trans-Atlantic voyage from Africa to the NewWorld. Mol. Phylogenet. Evol. 60:183–191.
Mueller-Töwe I.J. 2006. Anatomy, phylogeny, and palaeoecology ofthe basal thalattosuchians (Mesoeucrocodylia) from the Liassic ofCentral Europe. Unpublished Ph.D. Dissertation. Universität Mainz422 pp.
Nesbitt S.J. 2011. The early evolution of archosaurs: relationshipsand the origin of major clades. Bull. Am. Mus. Nat. Hist. 352:1–292.
Nixon K.C., Carpenter J.M. 1993. On outgroups. Cladistics 9:413–426.Norell M.A. 1989. The higher level relationships of the extant
Crocodylia. J. Herpetol. 23:325–335.Ortega F., Gasparini Z., Buscalioni A.D., Calvo J.O. 2000. A new species
of Araripesuchus (Crocodylomorpha, Mesoeucrocodylia) from theLower Cretaceous of Patagonia (Argentina). J. Vert. Paleontol. 20:57–76.
Parrilla-Bel J., Young M.T., Moreno-Azanza M., Canudo J.I. 2013.The first metriorhynchoid crocodylomorph from the MiddleJurassic of Spain, with implications for evolution of the subcladeRhacheosaurini. PLoS ONE 8:e54275.
Parrish J.M. 1993. Phylogeny of the Crocodylotarsi, with referenceto archosaurian and crurotarsan monophyly. J. Vert. Paleontol. 13:287–308.
Pierce S.E., Angielczyk K.D., Rayfield E.J. 2008. Patterns ofmorphospace occupation and mechanical performance in extantcrocodilian skulls: a combined geometric morphometric and finiteelement modeling approach. J. Morphol. 269:840–864.
2015 WILBERG—OUTGROUP CHOICE AND THE ORIGIN OF CROCODYLIFORMES 637
Pol D., Gasparini Z. 2009. Skull anatomy of Dakosaurus andiniensis(Thalattosuchia: Crocodylomorpha) and the phylogenetic positionof Thalattosuchia. J. Syst. Palaeontol. 7:163–197.
Pol D., Norell M.A. 2004a. A new crocodyliform from Zos Canyon,Mongolia. Am. Mus. Novit. 3445:1–36.
Pol D., Norell M.A. 2004b. A new gobiosuchid crocodyliform taxonfrom the Cretaceous of Mongolia. Am. Mus. Novit. 3458:1–31.
Pol D., Ji S.-A., Clark J.M., Chiappe L.M. 2004. Basal crocodyliformsfrom the Lower Cretaceous Tugulu Group (Xinjiang, China), and thephylogenetic position of Edentosuchus. Cretaceous Res. 25:603–622.
Pol D., Turner A.H., Norell M.A. 2009. Morphology of the LateCretaceous crocodylomorph Shamosuchus djadochtaensis and adiscussion of neosuchian phylogeny as related to the origin ofEusuchia. Bull. Am. Mus. Nat. Hist. 324:1–103.
Pol D, Rauhut O.W.M., Leucona A., Leardi J.M., Xu X., Clark J.M.2013. A new fossil from the Jurassic of Patagonia reveals the earlybasicranial evolution and the origins of Crocodyliformes. Biol. Rev.Camb. Philos. Soc. 4:862–872.
Prevosti F.J., Chemisquy M.A. 2010. The impact of missing dataon real morphological phylogenies: influence of the number anddistribution of missing entries. Cladistics 26:326–339.
Pritchard A.C., Turner A.H., Allen E.R., Norell M.A. 2013. Osteologyof a North American goniopholidid (Eutretauranosuchus delfsi) andpalate evolution in Neosuchia. Am. Mus. Novit. 3783:1–56.
Puslednik L., Serb J.M. 2008. Molecular phylogenetics of the Pectinidae(Mollusca: Bivalvia) and effect of increased taxon sampling andoutgroup selection on tree topology. Mol. Phylogenet. Evol. 48:1178–1188.
Rieppel O., Kearney M. 2002. Similarity. Biol. J. Linn. Soc. 75:59–82.Sanderson M.J., Shaffer H.B. 2002. Troubleshooting molecular
phylogenetic analyses. Ann. Rev. Ecol. Syst. 33:49–72.Sadleir R.W., Makovicky P.J. 2008. Cranial shape and correlated
characters in crocodilian evolution. J. Evol. Biol. 21:1578–1596.Seeley H. G. 1880. Note on the cranial characters of a large teleosaur
from the Whitby Lias preserved in the Woodwardian museum ofthe University of Cambridge, indicating a new species, Teleosauruseucephalus. Quart. J. Geol. Soc. Lon. 36:627–634.
Sereno P.C., Larsson H.C.E., Sidor C.A., Gado B. 2001. The giantcrocodyliform Sarcosuchus from the Cretaceous of Africa. Science294:1516–1519.
Sereno P.C., Sidor C.A., Larsson H.C.E., Gado B. 2003. A newnotosuchian from the Early Cretaceous of Niger. J. Vert. Paleontol.23:477–482.
Slowinski J.B. 1993. “Unordered” versus “ordered” characters. Syst.Biol. 42:155–165.
Smith A.B. 1994. Rooting molecular trees: problems and strategies. Biol.J. Linn. Soc. 51:279–292.
Spaulding M., O’Leary M.A., Gatesy J. 2009. Relationships of Cetacea(Artiodactyla) among mammals: increased taxon sampling altersinterpretations of key fossils and character evolution. PLoS ONE4(9):e7062.
Turner A.H., Buckley G. A. 2008. Mahajangasuchus insignis(Crocodyliformes: Mesoeucrocodylia) cranial anatomy and newdata on the origin of the eusuchian-style palate. J. Vert. Paleontol.28:382–408.
Turner A.H., Calvo J.O. 2005. A new sebecosuchian crocodyliformfrom the Late Cretaceous of Patagonia. J. Vert. Paleontol. 25:87–98.
Turner A.H. Sertich J.J.W. 2010. Phylogenetic history of Simosuchusclarki (Crocodyliformes: Notosuchia) from the Late Cretaceous ofMadagascar. Soc. Vert. Paleontol. Mem. 10:177–236.
Tykoski R.S., Rowe T.B., Ketcham R.A., Colbert M.W. 2002.Calsoyasuchus valliceps, a new Crocodyliform from the Early JurassicKayenta Formation of Arizona. J. Vert. Paleontol. 22:593–611.
Walker A.D. 1990. A revision of Sphenosuchus acutus Haughton, acrocodylomorph reptile from the Elliot Formation (Late Triassicor Early Jurassic) of South Africa. Phil. T. R. Soc. B 330:1–120.
Wilberg E.W. 2015. A new metriorhynchoid (Crocodylomorpha,Thalattosuchia) from the Middle Jurassic of Oregon and theevolutionary timing of marine adaptations in thalattosuchiancrocodylomorphs. J. Vert. Paleontol. doi:10.1080/02724634.2014.902846
Wu X.-C., Chatterjee S. 1993. Dibothrosuchus elaphros, a crocodylomorphfrom the lower Jurassic of China and the phylogeny of theSphenosuchia. J. Vert. Paleontol. 13:58–89.
Wu X.-C., Brinkman W., Lu J.C. 1994. A new species of Shantungosuchusfrom the Lower Cretaceous of Inner Mongolia (China), withcomments on S. chuhsienensis Young, 1961 and the phylogeneticposition of the genus. J. Vert. Paleontol. 14:210–229.
Wu X.-C., Sues H.-D., Dong Z.M. 1997. Sichuanosuchus shuhanensis,a new? Early Cretaceous protosuchian (Archosauria:Crocodyliformes) from Sichuan (China), and the monophylyof Protosuchia. J. Vert. Paleontol. 17:89–103.
Wu X.-C., Russel A.P., Cumbaa S.L. 2001. Terminonaris (Archosauria:Crocodyliformes): new material from Saskatchewan, Canada, andcomments on its phylogenetic relationships. J. Vert. Paleontol.21:492–514.
Young M.T., Andrade M.B. 2009. What is Geosaurus? Redescriptionof Geosaurus giganteus (Thalattosuchia: Metriorhynchidae) fromthe Upper Jurassic of Bayern, Germany. Zool. J. Linn. Soc. Lon.157:551–585.
Young M T., Brusatte S.L., Andrade M.B., Desojo J.B., BeattyB.L., Steel L., Fernández M.S., Sakamoto M., Ruiz-Omeñaca J.I.,Schoch R.R. 2012. The cranial osteology and feeding ecologyof the metriorhynchid crocodylomorph genera Dakosaurus andPlesiosuchus from the Late Jurassic of Europe. PLoS ONE 7:e44985.
Young M.T., Andrade M.B., Brusatte S.L., Sakamoto M., Liston J. 2013.The oldest known metriorhynchid super-predator: a new genus andspecies from the Middle Jurassic of England, with implications forserration and mandibular evolution in predacious clades. J. Syst.Palaeontol. 11:475–513.
