Squamatephylogeny,taxonsampling,anddatacongruence · Organisms,Diversity&Evolution5(2005)25–45...

ARTICLE IN PRESS

1439-6092/$ - se

doi:10.1016/j.od

�CorrespondiMuseum, North

E-mail addre

lee.mike@saugo

Organisms, Diversity & Evolution 5 (2005) 25–45www.elsevier.de/ode

Squamate phylogeny, taxon sampling, and data congruence

Michael S.Y. Leea,b,�

aEarth Sciences Section, South Australian Museum, North Terrace, Adelaide SA 5000, AustraliabSchool of Earth and Environmental Sciences, University of Adelaide, Adelaide SA 5005, Australia

Received 16 December 2002; accepted 25 May 2004

Abstract

To investigate the affinities of snakes, amphisbaenians and dibamids, the phylogenetic relationships among themajor lineages (families) of extinct and extant squamates are assessed through a combined analysis of 248 osteological,133 soft anatomical, and 18 ecological traits. The osteological data set represents a revision of previous data, takinginto account recent criticism; the ecological data set is new. In addition, potentially critical fossil taxa(polyglyphanodontids and macrocephalosaurs) are included for the first time. The osteological and soft anatomicaldata sets each place snakes within anguimorphs, with dibamids and amphisbaenians near gekkotans. The putativeprimitive fossil amphisbaenian Sineoamphisbaena groups with macrocephalosaurs and polyglyphanodontids, togetherthe sister group to scleroglossans. All three data sets are congruent, and these results are reinforced by combinedanalyses. In these, as in the osteological analyses, snakes are nested within marine lizards. However, exclusion of fossiltaxa from the osteological data set results in a ‘limbless clade’ consisting of snakes, amphisbaenians and dibamids, andintroduces significant conflict between osteology and soft anatomy. Also, deletion tests and character weighting revealthat the signal in the reduced osteological data set is internally contradictory. These results increase confidence in thearrangement supported by the all-taxon osteological, the soft anatomical, and the combined data, and suggest thatexclusion of fossils confounds the signal in the osteological data set. Finally, the morphological data support thenesting of snakes within marine lizards, and thus a marine origin of snakes. This result still holds when relationshipsbetween living forms are constrained to the topology suggested by molecular sequences: if marine lizards are allowed to‘float’ within this molecular framework, they form the stem group to snakes, and do not group with varanids aspreviously suggested.r 2005 Elsevier GmbH. All rights reserved.

Keywords: Squamata; Snakes; Amphisbaenians; Dibamids; Partitioned branch support; Character congruence

See also Electronic Supplement at: http://www.senckenberg.de/odes/05-04.htm.

e front matter r 2005 Elsevier GmbH. All rights reserved.

e.2004.05.003

ng author. Earth Sciences Section, South Australian

Terrace, Adelaide SA 5000, Australia.

sses: [email protected],

v.sa.gov.au (M.S.Y. Lee).

Introduction

It has been eight decades since Charles Camp (1923)produced his landmark monograph, ‘Classification ofthe lizards’. Yet, many long-standing uncertaintiesremain regarding higher-level squamate relationships,despite detailed studies of a plethora of informative

http://www.senckenberg.de/odes/05-04.htm

www.elsevier.de/ode

ARTICLE IN PRESSM.S.Y. Lee / Organisms, Diversity & Evolution 5 (2005) 25–4526

traits from a variety of systems: osteology (Estes et al.1988; Wu et al. 1996; Lee 1998), musculature (Russell1988), visceral anatomy (Gabe and Saint-Girons 1972),oral glands (Kochva 1978), the brain (Northcutt 1978),olfactory organs (Parsons 1970), the ear (Miller 1966;Wever 1978; Manley 2003), reproductive structures(Gabe and Saint-Girons 1965; Arnold 1984), tongueand hyoid apparatus (Schwenk 1988, 1993), spermato-zoa (Jamieson 1995), and behaviour (Stamps 1977;Cooper 1996). Recent molecular studies of squamatesare promising (see Forstner et al. 1995; Saint et al. 1998;Harris et al. 2001; Vicario et al. 2002; Whiting et al.2003; Townsend et al. 2004; Vidal and Hedges 2004),but relationships between many major lineages remainpoorly supported. While additional sequence data willundoubtedly clarify some remaining problems, it cannotresolve all outstanding uncertainties (Lee 2005). Manyradiations, such as basal metazoan divergences, haveproven notoriously intractable to molecular (as well asmorphological) resolution despite initial optimism.The major areas of consensus and contention in

higher-level squamate phylogeny are illustrated inFig. 1. The following major clades of limbed squamatesare widely accepted, based primarily on morphologicaldata (e.g. Estes et al. 1988; Schwenk 1988; Wu et al.1996; Evans and Barbadillo 1997; Lee 1998; Lee andCaldwell 2000): Iguania, Acrodonta, Scleroglossa, Gek-kota, Lacertoidea, Anguimorpha, and Varanoidea. Allexcept Scleroglossa are also corroborated (or at leastcurrently uncontradicted) by molecular data (e.g.Donnellan et al. 1999; Vicario et al. 2002; Whiting etal. 2003; Vidal and Hedges 2004). However, muchuncertainty remains regarding the affinities of many

Igua

nida

e

Leio

lepi

dina

e

Aga

min

ae

Cha

mae

leon

idae

Xan

tusi

idae

Eub

leph

arin

ae

Gek

koni

nes

s.l.

Dip

loda

ctyl

inae

Pyg

opod

idae

Lace

rtid

ae

Tei

idae

Iguania

AcrodontaGekkota

Lac

S

A,SA,

A,D

A

A

Fig. 1. Major areas of agreement or disagreement between mor

proposed positions of the three highly modified limb-reduced group

major squamate groups, even when only recent cladisticmorphological studies are considered. Major areas ofuncertainty include the position of gekkotans, relation-ships (and monophyly) of the scincomorph taxa, andbasal relationships within Scleroglossa and Anguimor-pha. Gekkotans have been associated with scinco-morphs (Presch 1988; Schwenk 1988), or positionedmore basally, as the sister group of scincomorphs plusanguimorphs (Camp 1923; Estes et al. 1988). Of thescincomorph groups, xantusiids are particularly proble-matic: though traditionally allied with lacertoids (e.g.Estes et al. 1988; Wu et al. 1996; Evans and Barbadillo1997), they also have been aligned with gekkotans (e.g.McDowell and Bogert 1954; Northcutt 1978; Greer1985a); molecular data suggest scincoid affinities(Vicario et al. 2002; Whiting et al. 2003; Townsend etal. 2004). Inferred relationships within the remainingscincomorph groups also differ between morphologicalstudies (e.g. Estes et al. 1988; Presch 1988; Evans andBarbadillo 1997), and this diverse assemblage might noteven be monophyletic (Lee 1998; Townsend et al. 2004).Within anguimorphs, the relationships between anguids,xenosaurids and varanoids are uncertain, with allthree possibilities having some support in certainmorphological analyses (Rieppel 1980; Presch 1988;Wu et al. 1996).In addition to the above problems involving limbed

squamates (lizards), there remains uncertainty regardingthe phylogenetic position of three highly modified limb-reduced taxa (Fig. 1). Estes et al. (1988) conceded thatrelationships of snakes, dibamids and amphisbaenianscould not be resolved beyond the likelihood that eachbelonged ‘somewhere within Scleroglossa’. Snakes have

Gym

noph

thal

mid

ae

Sci

ncid

ae

Cor

dylid

ae

Ang

uida

e

Shi

nisa

urus

Xen

osau

rus

Hel

oder

ma

Lant

hano

tus

Var

anus

ertoidea

Scincoidea

cleroglossa

Anguimorpha

Varanoidea

D,S

A,D

S

phological analyses of squamate phylogeny. Arrows denote

s: amphisbaenians (A), dibamids (D), and snakes (S).

ARTICLE IN PRESSM.S.Y. Lee / Organisms, Diversity & Evolution 5 (2005) 25–45 27

been allied with either anguimorphs, burrowing scinco-morphs, amphisbaenians and dibamids, or placed out-side lizards altogether (see Rieppel 1988 for review).Dibamids have been placed with scincomorphs (Rieppel1984), gekkotans (Wu et al. 1996), or amphisbaenians(Greer 1985b; Hallermann 1998), or outside othersquamates altogether (Greer 1985b; see also Vidal andHedges 2004). Amphisbaenians have been associatedwith lacertoids (Wu et al. 1996), teioids (Schwenk 1988),gekkotans (Presch 1988) or dibamids (Greer 1985b;Hallermann 1998).The above disagreements between morphological

studies appear substantial, and might suggest that fewhigher-level groupings of squamates are strongly sup-ported by this data set. However, much of this apparentincongruence might be a methodological artefact. Moststudies of squamate phylogeny have focused onparticular sources of data to the exclusion of othersources: for instance, Schwenk (1988, 1993) focused ontongue characters, Jamieson (1995) on sperm ultra-structure, and even Estes et al.’s (1988) comprehensiveanalysis relied mainly on osteological data, whichcomprised 130 out of their 148 characters. It is notsurprising, therefore, that studies that have considereddifferent and often small subsets of characters haveoften yielded different phylogenies. However, it is quitepossible that a very similar common signal underlies allthese data sets, but in some (or even all) of them mightbe swamped by homoplasy and thus ‘hidden’ (Gatesy etal. 1999). Thus, the ‘taxonomic congruence’ approach(Miyamoto and Fitch 1995) of analysing data setsseparately and looking for groupings in common mightnot merely underestimate the amount of agreementbetween data sets, but completely overlook such‘hidden’ groupings (Kluge 1989; Barrett et al. 1991;Nixon and Carpenter 1996; Gatesy et al. 1999). Asimultaneous analysis of all data needs to be performedto see if any common signal underlies these apparentlyconflicting data sets, and if the overall data stronglysupport a particular phylogenetic arrangement. Theseinsights cannot be gained from comparing the results ofseparate analyses of small, isolated data sets. However,analyses of subsets of characters or taxa can also bevaluable to probe the influence of particular portions ofthe data. Thus, a combination of the simultaneous (e.g.Kluge 1989; Nixon and Carpenter 1996) and partitioned(Miyamoto and Fitch 1995) approaches can yield moreinsights than adherence to a single method.An analysis of all the (non-molecular) data bearing on

squamate phylogeny is attempted here. Informativeosteological, soft anatomical, and ecological charactershave been compiled, scored across all the majorsquamate lineages, and analysed in a simultaneousparsimony analysis. The amount of congruence betweenthese three data sets is also investigated, using standardtests that assess overall levels of agreement between data

sets, and partitioned branch support (Baker and deSalle1997) which assesses congruence with respect toparticular groupings. All methods indicate that, con-trary to expectation, all three data sets are consistentwith the same phylogenetic arrangement. The effect offossils, both on tree topology and congruence, is alsoinvestigated. If fossils are ignored, the osteological dataimply a tree which is intrinsically problematic, as well ashighly incongruent with the tree from the soft anatomyand ecology. When fossils are added, the osteologicaldata yield a more biologically reasonable tree, and onewhich is almost identical to the tree implied by softanatomy. Other tests involving successive deletion oflong convergent branches (see Scanlon 1996; Siddall andWhiting 1999) also suggest that the fossils are not justchanging the osteological tree, but are changing it inarguably the ‘right’ direction, by amplifying hiddensignals that are congruent with the soft anatomy.If the resultant tree from the simultaneous analysis is

taken as the most highly corroborated phylogeny, therelative ‘quality’ of each of the data sets can also beevaluated. Ecological (including behavioural) traits areno different from morphological traits in termsof phylogenetic informativeness (e.g. Proctor 1996;Wimberger and de Queiroz 1996). In both cases, initialhypotheses of homology are formulated based onseveral criteria (internal similarity, location or context,existence of intermediate forms), and tested via con-gruence with other characters in a phylogenetic analysis.In both behaviour and morphology, there can beproblems with delimiting character states where acontinuum exists. Both behavioural and morphologicalcharacters show ontogenetic variability, can be geneti-cally fixed at birth or phenotypically plastic, can evolvevery rapidly or slowly, and exhibit similar ranges ofheritability. Empirical studies show that behaviour andmorphology are equally informative or, conversely,equally likely to be homoplastic (e.g. Wenzel 1992; deQueiroz and Wimberger 1993), and that neither class ofcharacters is consistently more useful in phylogeneticreconstruction (e.g. Sanderson and Donoghue 1989).Here, the reliability of osteology, soft anatomy andbehaviour is assessed by comparing their consistencyand retention indices in the context of a combinedanalysis.

Character analysis

Character selection

A previous study of squamate phylogeny (Lee andCaldwell 2000) compiled a data set of 258 cladisticallyinformative osteological characters. Most of thesecharacters had been employed in previous studies (e.g.


Pregill et al. 1986; Estes et al. 1988; Presch 1988; Wuet al. 1996; Evans and Barbadillo 1997; Lee 1998;Reynoso 1998). Criticisms of certain character codingsin this data set (Rieppel and Zaher 2000) are hereevaluated: suggested recodings of taxa and redefinitionsof character states are assessed and, if valid, incorpo-rated into the current matrix. Ten characters (corre-sponding to characters 12, 41, 50, 58, 71, 89, 124, 134,136, and 256 in Lee and Caldwell 2000) have beendeleted, for reasons discussed in a section concluding thecharacter list (see the accompanying Organisms Diver-sity and Evolution Electronic Supplement: http://www.senckenberg.de/odes/05-04.htm, Part 1). This re-vised osteological data set has been supplemented withsoft anatomical and ecological characters. Many ofthese characters were included in a recent phylogeneticanalysis of squamate soft anatomy (Lee 2000). However,some terminal taxa of dubious monophyly in thatanalysis (agamids, gekkonids, xenosaurids; see below)have been subdivided in the present analysis, and as aresult additional characters have been added that helpresolve the relationships among these smaller terminaltaxa. Ecological (including behavioural) characters havealso been added, although this data set is much smallerthan the previous two. These include chemosensorycharacters (Cooper 1995, 1996, 1997), territoriality(Stamps 1977; Carpenter and Ferguson 1977), andreproductive mode (Lee and Shine 1998).The character list and data matrix are presented in

Electronic Supplement 05-04, Parts 1 and 2, respectively.Nearly all the osteological characters have beendiscussed more fully, and the relevant sources cited, inprevious papers (Lee 1998; Lee and Caldwell 2000) andthus are listed only briefly here. However, charactersthat have been debated recently (Rieppel and Zaher2000) are discussed more fully to address thesecriticisms. Most of the soft anatomical characters havebeen discussed elsewhere (Lee 2000), and this informa-tion is not repeated here. New soft anatomicalcharacters included in this study (see above) areidentified with an asterisk, and discussed and referencedmore fully. Ecological characters were not included inthe above analyses of squamate phylogeny, and thus arealso discussed and referenced more fully.The phenotypic character set compiled in this

analysis, though one of the largest of this kind, has byno means exhausted the possibilities. Many otherinformative characters no doubt exist and should beadded to the data set in the future. For instance,relatively few visceral characters have been included,largely because of the previous lack of interest inphylogenetic information contained in this system withrespect to higher-level squamate phylogeny. However,studies on relationships within snakes have shown thatmany visceral characters are overly variable (e.g.Wallach 1985, 1998; Cundall et al. 1993; Keogh 1996).

Similarly, only a few cranial, trunk, and limb muscleswere included. Without doubt, a comprehensive surveywill reveal many more informative myological charac-ters in this region; similar surveys of the tongue andgenital musculature have revealed many such characters.However, some studies suggest that both visceral andgeneral myological features might not be very informa-tive. Lee (1997) reported that in elapid snakes there wasextensive character conflict between visceral characters,and no clear phylogenetic signal. In Russell’s (1988)extensive survey of limb muscles across squamates, fewappeared as phylogenetically informative, there beingextensive variability for these muscles in many taxa.

Terminal taxa, polarity, and character-state coding

The terminal taxa used in this analysis (see Electr.Suppl. 05-04, Part 2) are very similar to those used inprevious analyses of higher-level squamate phylogeny(Estes et al. 1988; Presch 1988; Wu et al. 1996; Evansand Barbadillo 1997). They include all extant lineagestraditionally termed families. However, the terminaltaxa have been refined and expanded in two importantways.Firstly, many extant squamate ‘families’ used in

previous analyses were of uncertain monophyly. Here,the evidence for monophyly of each of these groups isreassessed based on the most recent informationavailable. If the evidence is found to be still weak, thegroup is subdivided into smaller (and demonstrablymonophyletic) lineages. Use of only such clades asterminals addresses the criticism (Zaher and Rieppel1999) that some terminal taxa employed in previousanalyses (e.g. Estes et al. 1988; Lee 1998) were ofdubious monophyly.

Iguanidae: Until recently, the monophyly of iguanidswith respect to acrodontans was uncertain (Estes et al.1988; Frost and Etheridge 1989). However, molecularanalyses now have strongly supported the monophyly ofiguanids (Harris et al. 2001; Vidal and Hedges 2004):Zaher and Rieppel’s (1999) statement that the mono-phyly of iguanids remains poorly supported appears tobe based on the older studies. Iguanids are treated hereas a single terminal taxon.

Agamidae: The monophyly of ‘agamids’ remainsuncertain, as chameleons might be nested within them.The most comprehensive analysis of the problem interms of taxon sampling (Frost and Etheridge 1989)identified a trichotomy between chamaeleons and twosubgroups of agamids: leiolepidines (Leiolepis andUromastyx) and agamines sensu lato (all other aga-mids). There is both morphological (Moody 1980; Frostand Etheridge 1989) and molecular (Macey et al. 2000)evidence for monophyly of agamines. Leiolepidinemonophyly is strongly corroborated by morphological




data (Moody 1980; Frost and Etheridge 1989) and hasvariable support from molecular data (Macey et al.2000; Townsend et al. 2004). Thus, the metataxonAgamidae is divided here into these two putativelymonophyletic subgroups in order to attempt to resolvethis trichotomy.

Gekkonidae: ‘Gekkonids’ (limbed gekkotans) mightbe monophyletic (Estes et al. 1988) or, more likely,paraphyletic (Kluge 1987; Donnellan et al. 1999; Harriset al. 2001; Townsend et al. 2004) with respect topygopodids. For this reason, limbed gekkotans havebeen divided into three sub-groups whose monophyly iswidely accepted and well-corroborated (Kluge 1987;Donnellan et al. 1999): eublepharines, diplodactylines,and gekkonines (sensu lato, including sphaerodactyinesand Teratoscincus).

Scincidae: It has been suggested that dibamids mightbe nested within scincids, thus making the latterparaphyletic (Rieppel 1984; Greer 1985b). However, arecent cladistic analysis (Hallermann 1998), which codeddibamids and their possible long-bodied, limb-reducedskink relatives as separate terminal taxa, retrieved amonophyletic Scincidae. Molecular data also supportmonophyly, although taxon sampling remains sparse(Harris et al. 2001; Townsend et al. 2004). Accordingly,Scincidae is treated here as a single monophyleticterminal taxon.

Xenosauridae: Though widely assumed in the past(e.g. McDowell and Bogert 1954; Estes et al. 1988),monophyly of the Xenosauridae has been questioned,with molecular studies suggesting that xenosauridsmight not be monophyletic (Townsend et al. 2004).Thus, Xenosauridae is divided here into two terminaltaxa of uncontested monophyly, Xenosaurus and Shini-

saurus.Serpentes: The monophyly of modern snakes (here

called Serpentes), to the exclusion of the Cretaceousmarine snake-like squamates, has been suggested in thepast (e.g. Haas 1980; Lee and Caldwell 1998) butquestioned recently (Tchernov et al. 2000), with thesuggestion that Pachyrhachis and Haasiophis are deeplynested within modern snakes as the sister group ofadvanced snakes (macrostomatans). However, as dis-cussed more fully elsewhere (Lee and Scanlon 2002), thelatter study is problematic. The characters proposed aslinking Pachyrhachis and Haasiophis with macrostoma-tans are based on unlikely interpretations of poorlypreserved skull elements. Furthermore, even if theseinterpretations are accepted, the proposed characters arenot unique to Pachyrhachis, Haasiophis, and macro-stomatans, but are present in other snakes and in‘lizards’, and thus might be primitive for snakes. Arecent re-analysis of snake phylogeny (Lee and Scanlon2002), using a larger suite of characters, confirms thatpachyophiids are basal to all other snakes (Serpentes).This result holds even if the data of Tchernov et al.

(2000) are added. Serpentes, excluding the pachyophiids,thus is treated here as a single, monophyletic terminaltaxon (fide Haas 1980; Rage and Escuillie 2000; Lee andScanlon 2002). The primitive condition in modernsnakes is inferred based on taxa identified as basal inrecent analyses (e.g. Lee and Scanlon 2002; Slowinskiand Lawson 2002; Vidal and Hedges 2002, 2004;Scanlon 2005): scolecophidians, ‘anilioids’, trophido-phiines, Dinilysia, and madtsoiids.The second change in the present study involves the

inclusion of additional fossil taxa of potential relevanceto the affinities of snakes, amphisbaenians and diba-mids. These additional taxa obviously could be codedfor very few of the soft anatomical and behaviouralcharacters. Sineoamphisbaena has been interpreted pre-viously as a basal amphisbaenian (Wu et al. 1996), thesister group to the amphisbaenian–dibamid clade (Lee1998), or as only distantly related to amphisbaeniansand dibamids (Kearney 2003). Affinities between snakesand Cretaceous marine lizards (aigialosaurids, mosa-saurids, dolichosaurs, and Adriosaurus) have also beenproposed (Cope 1869; Nopcsa 1923; McDowell andBogert 1954; Lee and Caldwell 2000) and disputed(Zaher and Rieppel 1999; Rieppel and Zaher 2000).These taxa, along with well-known Cretaceous marinesnakes, Pachyrhachis and Haasiophis, are all included.Character data from Dinilysia and madtsoiids areincorporated as well (see Serpentes above). Mostprevious analyses of squamate phylogeny have notincluded all these taxa (e.g. Estes et al., 1998; Wu et al.1996; Evans and Barbadillo 1997), and the only studywhich did (Lee and Caldwell 2000) only consideredosteological characters. In addition, affinities betweenpolyglyphanodontids, macrocephalosaurs, Sineoamphis-

baena and amphisbaenians have been proposed (Wuet al. 1996). Polyglyphanodontids are represented hereby Polyglyphanodon, and macrocephalosaurs by Macro-

cephalosaurus, the eponymous and most completelyknown forms. These taxa were scored based onreexamination of very complete and undistorted materi-al in Washington and Warsaw, respectively (see Esteset al. 1988). There are, of course, additional fossil taxawhich could have been included (e.g. Evans andBarbadillo 1997; Evans and Chure 1998; Reynoso1998). However, none of these fossil taxa have everbeen proposed to be closely related to snakes, amphis-baenians or dibamids. Thus they are unlikely to have animpact on which taxa emerge as the immediate relativesof the limb-reduced taxa. A preliminary analysis of someof the better-known excluded fossil lineages confirmedthis: Estesia and paramacellodids were added in theanalysis based on descriptions in the literature, but hadvery little effect on tree topology and support, and didnot influence the positions of the limb-reduced taxa. Forthis reason, their exclusion in this study is not likely tohave an important effect on the main question


addressed, i.e. the affinities of snakes, amphisbaeniansand dibamids.The monophyly of the ingroup (Squamata) is

corroborated by a large suite of derived characters froma variety of trait classes (Estes et al. 1988; Gauthier et al.1988a; Pianka and Vitt 2003). The successively moredistant outgroups used to polarise characters wererhynchocephalians, Marmoretta, and kuehneosaurs(see Gauthier et al. 1988a; Evans 1991). The tree wasrooted with an ancestral taxon possessing the inferredprimitive states for all characters; where more than onestate could be primitive (e.g. because it was highlyvariable, or not applicable, in the outgroups); theancestral taxon was coded with all possible primitivestates.In some previous analyses (e.g. Wu et al. 1996; Zaher

and Rieppel 1999), where multiple character statesoccurred within a terminal taxon, that taxon was oftencoded as polymorphic (‘all polymorphisms’ coding).Here, an attempt was made to determine, on the basis ofwell-corroborated phylogenetic relationships within theterminal taxon, which state was primitive (‘inferred-ancestor’ coding). For diverse terminal taxa (e.g.iguanids, agamines, scincids, gekkonines, snakes), theprimitive condition was that found in basal lineages asidentified in recent cladistic analyses. Only if theprimitive state could not be determined readily (due tovariability of the character in basal representatives, orpoor knowledge of relationships and/or characterdistribution within the terminal taxon) was the terminaltaxon coded as polymorphic. The inferred-ancestormethod is widely used (e.g. Estes et al. 1988; Frost andEtheridge 1989; Lee 1998) and in simulations has beenshown to give better results than the all-polymorphismsmethod, which allows characters occurring only in asingle highly nested species to be potentially primitivefor a large terminal higher taxon (Bininda-Emonds et al.1998; Wiens 1998a). These studies contradict claims(Zaher and Rieppel 1999, p. 833) that the all-poly-morphisms coding method is superior, and that theinferred-ancestor method, as employed in Lee (1998),‘fails to recognise variability (polymorphism) of char-acters within families’, giving results that ‘continue toobscure the debate’. The latter method, in fact, betteraccounts for variability by recognising highly derivedstates within terminal taxa as unrepresentative of thebasal condition.Many of the characters in the present analysis are

multistate. Two approaches to analysing these wereemployed. In one analysis, all multistate characters weretreated as unordered (non-additive). In another, multi-state characters were ordered, where possible, intomorphoclines. The extremes in morphoclines were codedas being derivable from each other only via intermediatestages (the transformation from 0 to 2 entails two steps:0 to 1, and 1 to 2), i.e. they were coded as ordered

(additive). Only characters that could not be orderedinto clear morphoclines were left unordered. Because thelatter approach discriminates against large changeswithin a character (e.g. those between extremes in amorphocline), it has been argued to result in cladogramsthat entail less overall evolutionary change thancladograms constructed by treating all multistatecharacters as unordered (Lipscomb 1992; Wilkinson1992; Slowinski 1993). The phylogenetic results from theunordered and unordered analyses were compared toascertain any possible effects of the alternative treat-ments of multistate characters.The following multistate characters formed clear

morphoclines and thus were ordered 0–1–2 (or0–1–2–3–4–5, etc., where applicable): 23, 31, 37, 41,42, 57, 73, 85, 109, 118, 119, 122, 123, 125, 126, 141, 144,147, 164, 165, 171, 172, 181, 185, 186, 203, 216, 218, 222,225, 228, 240, 246, 267, 269, 272, 277, 280, 290, 297, 303,308–311, 317, 321, 327, 329, 334, 336, 337, 351, 353, 355,363, 369, 371, 376–378, 390.The following multistate characters formed clear

morphoclines and thus were ordered 1–0–2 (or1–0–2–3, 1–0–2–3–4, 1–0–2–3–4–5, as applicable): 56,110, 145, 180, 223.The following multistate characters did not form clear

morphoclines and were left unordered, even in the‘ordered’ analyses: 55, 63, 67–70, 83, 94, 124, 134, 152,153, 157, 257, 285, 299, 340, 351, 354, 356, 361, 380, 394,395.The complete character list and data matrix are

available from the two-part Electronic Supplement, thematrix (Part 2) in Nexus (PAUP and MacClade) fileformat. The character list and a similar matrix (withadditional molecular data) are also available fromTreeBASE (http://www.treebase.org/treebase/; acces-sion number SN1715).

Phylogenetic analyses and results

In each analysis below, two runs were performed, onewith multistate characters ordered, the other, unor-dered. Data were entered using MacClade (Maddisonand Maddison 2000), and tree searches were performedusing PAUP* (Swofford 2000); the heuristic algorithmemploying 100 random addition sequences was used.Tree(s) were rooted with a taxon coded with the inferredprimitive state for each character, as discussed above.For tree length calculations, variability within terminaltaxa was interpreted as uncertainty over the ancestralcondition, rather than as polymorphism in the ancestrallineage. The robustness of each grouping was ascer-tained using branch support (Bremer 1988) and, incombined analyses, partitioned branch support (Bakerand deSalle 1997), calculated in PAUP using commands

http://www.treebase.org/treebase/


generated by TreeRot Version 2 (Sorenson 1999). Thesecommands were modified so that each heuristic reversesearch employed 100 rather than 20 random additionsequences. Bootstrapping (Felsenstein 1985) was used aswell, employing 1000 replicates each with 100 randomaddition sequences.

Partitioned analyses

The trees obtained from the separate analyses ofosteology, soft anatomy, and behaviour are shown inFigs. 2–5. Two analyses of the osteological partition wereperformed: with all taxa (including fossils), and withextant taxa only. Only extant taxa were included inanalyses of the soft anatomical and behavioural partitions,since fossils could not usually be scored for these traits.The trees obtained from osteology (all taxa included)

and soft anatomy agree remarkably. The strict con-sensus trees are shown in Figs. 2 and 3. The monophylyof many traditionally accepted squamate groupings isupheld: Iguania, Acrodonta, Scleroglossa, Gekkota,Anguimorpha, Xenosauridae, and Varanoidea. In bothanalyses, amphisbaenians and dibamids are sistergroups, together related to gekkotans, xantusiids aremore closely related to gekkotans than to lacertiforms,scincids and cordylids occupy a heterodox positionnearer to anguimorphs than to lacertiforms, and snakesare related to varanids. The marine varanoids form aparaphyletic assemblage leading from varanids to snakesin the osteological analysis (they could not be included inthe soft anatomical analysis). Sineoamphisbaena groupsnot with amphisbaenians but with Polyglyphanodon andMacrocephalosaurus (fide Kearney 2003), as a basal cladeof scleroglossans. These taxa are thus not closely relatedto teioids, as often assumed (e.g. Wu et al. 1996), andtheir dental similarities are convergent.The two trees only differ in three areas. In the

osteological tree, xantusiids are the sister group togekkotans plus the amphisbaenian–dibamid clade,whereas in the soft anatomical tree, xantusiids are thesister group to gekkotans alone. Also, amphisbaeniansand dibamids are the sister group of pygopodids alonein the osteological tree, but of all gekkotans in the softanatomical tree. Finally, in the osteological tree, scincidsand cordylids are successive outgroups to anguimorphs,whereas in the soft anatomical tree a scincid–cordylidclade forms the sister group to anguimorphs. However,in no case are incompatible clades strongly supported.The exact positions of xantusiids and the amphisbae-nian–dibamid clade are poorly supported in both datasets, suggesting that the differences might be the resultof sampling error and a weak overall signal, rather thanof strongly incongruent data sets. Similarly, while thescincid–cordylid clade is strongly supported in the softanatomical data set, the conflicting scincid–anguimorph

clade in the osteological data set is very weak. Again, itwould be reasonable to view the result in the latteranalysis as an unresolved trichotomy involving scincids,cordylids, and anguimorphs, an arrangement consistentwith the strong scincid–cordylid clade obtained fromsoft anatomy. Thus, where there is disagreement, eitheror both of the conflicting clades are poorly resolved. Ifpoorly resolved nodes are collapsed into polytomies,there is no conflict between data sets.Obviously, there are too few behavioural and

ecological characters in the present analysis (18) toobtain robust phylogenetic results. An heuristic searchyielded over 20,000 equally parsimonious trees beforethe computer ran out of memory. The consensus treeobtained (Fig. 4) is largely unresolved, although all theresolved clades except one are congruent with thoseobtained by both osteology and soft anatomy. Theremaining clade (the gekkotan-xantusiid clade) wasobtained by soft anatomy but not osteology. As notedbefore, however, this incongruence is not significant,since the conflicting clades in the behavioural (and softanatomical) and the osteological trees are not well-supported. The limited behavioural data, therefore,appear to be consistent with the other two (much larger)data sets. Future studies might identify new behaviouralcharacters allowing more detailed comparisons betweenbehaviour and the other data sets. One might predictthat, as more and more behavioural characters areadded, the unresolved polytomies in the behavioural treewill gradually resolve to a topology similar to those inthe other analyses.When fossil taxa were deleted, the results were as

follows (Fig. 5). In the ‘ordered’ analysis, six MPTsresulted. In all of these, amphisbaenians and dibamidsgrouped robustly with snakes, and this ‘limbless clade’grouped with varanids. Pygopodids, in contrast,grouped with other gekkotans. However, trees a singlestep longer unite the amphisbaenian–dibamid-snakeclade with pygopodids, forming a single limblessradiation. The movement of a limbless clade acrossfrom varanids to pygopodids with only a single extrastep means that all intervening branches are very weaklysupported (Bremer 1, bootstrap o50%). In the‘unordered’ analysis, the results were very similar. TwoMPTs resulted, and the relationships were largely asbefore. Again, a limbless clade (amphisbaenians, diba-mids and snakes) is retrieved and clusters either withvaranoids or, with a single extra step, pygopodids,leading to weak branch supports within Scleroglossa.

Osteological convergence in limbless forms – further

tests

It was suspected that, when fossils are ignored, thelimbless taxa, namely amphisbaenians, dibamids, snakes

ARTICLE IN PRESS

59,1

100,10

81, 2

85,4

93,14

10010

91,10

956

91,12

964

69,179,4

82,5

52,2

68,451,2

10019

35,156,2

64,4

70,2

97,3

49,1

98,10

71,1

93,5

852

99,8

74,1

84,5

92,9

10010

854

82,688,9

52,2

985

57,2

63,2

661

29,1

46,130,1

100,1333,1

18,1

67,2

96,3

43,143,1

98,8

35 ,1

27,1

69,1

86,3

85,2

Igua

nida

e

Leio

lepi

dina

eA

gam

inae

Cha

mae

leon

idae

Xan

tusi

idae

Eub

leph

arin

ae

Gek

koni

nes

s.l.

Dip

loda

ctyl

inae

Pyg

opod

idae

Lace

rtid

ae

Tei

idae

Gym

noph

thal

mid

ae

Sci

ncid

ae

Cor

dylid

ae

Ang

uida

eS

hini

saur

us

Xen

osau

rus

Hel

oder

ma

Lant

hano

tus

Var

anus

Ser

pent

es

Pac

hyrh

achi

sH

aasi

ophi

s

Adr

iosa

urus

Dol

icho

saur

idae

Mos

asau

ridae

Aig

ialo

saur

idae

Dib

amid

ae

Pol

ygly

phan

odon

Sin

eoam

phis

baen

a

Mac

roce

phal

osau

rus

Am

phis

baen

ia

Igua

nida

e

Leio

lepi

dina

eA

gam

inae

Cha

mae

leon

idae

Xan

tusi

idae

Eub

leph

arin

ae

Gek

koni

nes

s.l.

Dip

loda

ctyl

inae

Pyg

opod

idae

Lace

rtid

ae

Tei

idae

Gym

noph

thal

mid

ae

Sci

ncid

ae

Cor

dylid

ae

Ang

uida

eS

hini

saur

us

Xen

osau

rus

Hel

oder

ma

Lant

hano

tus

Var

anus

Ser

pent

es

Pac

hyrh

achi

sH

aasi

ophi

s

Adr

iosa

urus

Dol

icho

saur

idae

Mos

asau

ridae

Aig

ialo

saur

idae

Dib

amid

ae

Pol

ygly

phan

odon

Sin

eoam

phis

baen

a

Mac

roce

phal

osau

rus

Am

phis

baen

i a

A

B

Fig. 2. Cladograms resulting from analysis of the osteology data set with both living and fossil taxa included. At each node, first

number refers to bootstrap frequency, second number to branch support. (A) Multistate characters ordered; tree is strict consensus

of nine most-parsimonious trees, each with length 669, consistency index 0.47, retention index 0.71. (B) Multistate characters

unordered; tree is strict consensus of two most-parsimonious trees, each with length 634, consistency index 0.49, retention index

0.71.

M.S.Y. Lee / Organisms, Diversity & Evolution 5 (2005) 25–4532

and (sometimes) pygopodids, might cluster togetherspuriously based on characters correlated with bodyelongation and limb reduction. To test this possibility,two further analyses were performed on the ‘extant-only’ osteological data set.If taxa are ‘attracting’ one another due to convergence

or long-branch attraction, and thus their relationshipsare obscured, this confounding signal can be eliminatedby including only one problematic taxon at a time in ananalysis (Scanlon 1996; Siddall and Whiting 1999). Taxacannot attract one another unless they are simulta-

neously present in an analysis. Therefore, an analysiswas performed on the extant osteological data, butincluding only snakes and excluding the other limblesstaxa (pygopodids, dibamids, amphisbaenians). Similaranalyses were performed sequentially, including onlypygopodids, only dibamids, and only amphisbaenians(Fig. 6). Snakes emerged within anguimorphs, as thesister taxon to varanids. Pygopodids emerged withingekkotans, in a polytomy with diplodactylines andeublepharines. Dibamids emerged as the sister taxonto Gekkota. There was strong support for the positions

ARTICLE IN PRESS

Igua

nida

e

Leio

lepi

dina

e

Aga

min

ae

Cha

mae

leon

idae

Xan

tusi

idae

Eub

leph

arin

ae

Gek

koni

nes

s.l.

Dip

loda

ctyl

inae

Pyg

opod

idae

Lace

rtid

ae

Tei

idae

Gym

noph

thal

mid

ae

Sci

ncid

ae

Cor

dylid

ae

Ang

uida

e

Shi

nisa

urus

Xen

osau

rus

Hel

oder

ma

Lant

hano

tus

Var

anus

Ser

pent

es

Dib

amid

ae

Am

phis

baen

ia

98,7

93 ,4

71,3

95 ,4100,10

72,4

32 ,1

89,5

37,1

10 0,1194,9

67,1

80,3

65,1

96,7.2

92,4

29,1.251,1.2

68,3.2

90,5

32,2

41,2

31,1

100,11

49,1

51,195,8

29,1

94,5100, 7

86,4.1

Igua

nida

e

Leio

lepi

dina

e

Aga

min

ae

Cha

mae

leon

idae

Xan

tusi

idae

Eub

leph

arin

ae

Gek

koni

nes

s.l.

Dip

loda

ctyl

inae

Pyg

opod

idae

Lace

rtid

ae

Tei

idae

Gym

noph

thal

mid

ae

Sci

ncid

ae

Cor

dylid

ae

Ang

uida

e

Shi

nisa

urus

Xen

osau

rus

Hel

oder

ma

Lant

hano

tus

Var

anus

Ser

pent

es

Dib

amid

ae

Am

phis

baen

ia

A

B

Fig. 3. Cladograms resulting from analysis of the soft anatomy data set. At each node, first number refers to bootstrap frequency,

second number to branch support. (A) Multistate characters ordered; tree is strict consensus of 84 most-parsimonious trees, each

with length 350, consistency index 0.50, retention index 0.67. (B) Multistate characters unordered; tree is strict consensus of eight

most-parsimonious trees, each with length 329, consistency index 0.53, retention index 0.66.

Igua

nida

e

Leio

lepi

dina

e

Aga

min

ae

Cha

mae

leon

idae

Xan

tusi

idae

Eub

leph

arin

ae

Gek

koni

nes

s.l.

Dip

loda

ctyl

inae

Pyg

opod

idae

Lace

rtid

ae

Tei

idae

Gym

noph

thal

mid

ae

Sci

ncid

ae

Cor

dylid

ae

Ang

uida

e

Shi

nisa

urus

Xen

osau

rus

Hel

oder

ma

Lant

hano

tus

Var

anus

Ser

pent

es

Dib

amid

ae

Am

phis

baen

ia

50 ,1

48,1

63,1

66,1

58,147,1

Fig. 4. Cladogram resulting from analysis of the ecology data set. Both ordered and unordered analyses produced 420,000 most-parsimonious trees, all with length 26, consistency index 0.77, retention index 0.90. upper pair of numbers refers to ordered analysis,

lower pair to unordered analysis; within each pair, first number refers to bootstrap frequency, second number to branch support.

M.S.Y. Lee / Organisms, Diversity & Evolution 5 (2005) 25–45 33

ARTICLE IN PRESS

100,8

100,18

34,1

100,9

50,1

96,5

33,1

40,1

42,1

45,1

69,1

45,1

92,8

96,4

72, 2

85,2

84,6

71,2

99,10

91,2

8,1

98,5

99,15

27,1

98,6

34,1

70,1

33,1

33,1

42, 1

76,1

50,1

83,6

93, 3

62,1

84,2

76, 4

69,2

Igua

nida

e

Leio

lepi

dina

e

Aga

min

ae

Cha

mae

leon

idae

Xan

tusi

idae

Eub

leph

arin

ae

Gek

koni

nes

s.l.

Dip

loda

ctyl

inae

Pyg

opod

idae

Lace

rtid

ae

Tei

idae

Gym

noph

thal

mid

ae

Sci

ncid

ae

Cor

dylid

ae

Ang

uida

e

Shi

nisa

urus

Xen

osau

rus

Hel

oder

ma

Lant

hano

tus

Var

anus

Ser

pent

es

Dib

amid

ae

Am

phis

baen

ia

Igua

nida

e

Leio

lepi

dina

e

Aga

min

ae

Cha

mae

leon

idae

Xan

tusi

idae

Eub

leph

arin

ae

Gek

koni

nes

s.l.

Dip

loda

ctyl

inae

Pyg

opod

idae

Lace

rtid

ae

Tei

idae

Gym

noph

thal

mid

ae

Sci

ncid

ae

Cor

dylid

ae

Ang

uida

e

Shi

nisa

urus

Xen

osau

rus

Hel

oder

ma

Lant

hano

tus

Var

anus

Ser

pent

es

Dib

amid

ae

Am

phis

baen

ia

A

B

Fig. 5. Cladograms resulting from analysis of the osteology data set, with fossils deleted. (A) Multistate characters ordered; six

most-parsimonious trees, each with length 513, consistency index 0.53, retention index 0.69. (B) Multistate characters unordered;

two most-parsimonious trees, each with length 482, consistency index 0.55, retention index 0.69.


of each of these taxa. When only amphisbaeniansamong the limbless taxa were included in an analysis,they clustered weakly with varanoids. Thus, even whenonly extant taxa are considered, there is an osteologicalsignal placing snakes with varanids, dibamids withgekkotans, and pygopodids with diplodactylines. Therelationships of amphisbaenians are more obscure, asthe smallest clade that they fall robustly within isScleroglossa (Bremer 15/12, bootstrap 98/97 in ordered/unordered analyses). However, when all the limblesstaxa are simultaneously present in the analysis, thesesecondary signals (which place them in different regionsof the squamate tree) are overwhelmed and hidden: the

limbless taxa attract each other. It is of interest that thehidden phylogenetic signal for the limbless taxa in the‘extant-only’ osteological data set revealed using thisdeletion procedure matches closely the phylogeneticsignal that emerges when fossils are added, i.e. the signalpresent in the full osteological data set. It also matchesthe signal present in the soft anatomical data set. Theecological data set has very little signal due to the lownumber of characters, thus useful comparisons are notpossible. The same analyses were performed on the fullosteological, and soft anatomical data sets, where only asingle limbless taxon was included at a time: theinterrelationships between the remaining taxa were

ARTICLE IN PRESS

86/3,75/2

95/8,91/4

97/13 96/11

76/5,38/1

84/3,97/5

100/15100/12

72/1,69/1

100/12 100/12

80/5,87/4

99/6,98/4

73/1,73/1

Eub

leph

arin

ae

Gek

koni

nes

s.l.

Dip

loda

ctyl

inae

Pyg

op

od

idae

Eub

leph

arin

ae

Gek

koni

nes

s.l.

Dip

loda

ctyl

inae

Pyg

op

od

idae

Hel

oder

ma

Lant

hano

tus

Var

anus

Ser

pen

tes

Hel

oder

ma

Lant

hano

tus

Var

anus

Am

ph

isb

aen

ia

Fig. 6. Subtrees showing positions of limb-reduced taxa in the

‘extant-only’ data set, when only one taxon is included in each

analysis. Respective first pair of numbers refer to ordered

analysis, second pair to unordered analysis; within each pair,

first number refers to bootstrap frequency, second number to

branch support. Well-corroborated nodes containing the

limbless taxon are indicated.


largely unchanged in each of these analyses (results notshown).The other way to account for functional convergence

is to identify, and downweight, the characters suspectedto be functionally correlated. This approach, however, isinvariably subjective and imprecise. Characters will becorrelated with a particular function to varying degrees,and dividing the data into characters which are totally‘independent’ or totally ‘correlated’ with that functionresults in an artificial dichotomy. Furthermore, ofcourse, there will be other functions with associatedsuites of correlated characters, and these suites canoverlap to a greater or lesser extent. To arbitrarily singleout a single function, and then arbitrarily dividecharacters ‘correlated’ with that function and ‘indepen-dent’ ones, will often represent an oversimplification.However, some characters will tend to be more stronglycorrelated than others. Downweighting them can havegreat heuristic value (e.g. McCracken et al. 1999) andthus was attempted here. The method is not circular,since the weighting scheme is not derived from existingphylogenetic assumptions, but from an independentsource of evidence (e.g. functional morphology, devel-opmental genetics).The following osteological characters appear (almost

by definition) to be tightly correlated functionally withbody elongation and limb reduction, respectively.

Correlated with elongation of respective body region:increased number of presacral vertebrae (character 180),increased number of cervical vertebrae (181).Correlated with the more uniform trunk region

and more complex axial musculature in long-bodied animals: presence of cervical rib on thirdpresacral vertebra (197), anteroventral pseudotubercu-lum of rib (198), posterodorsal pseudotuberculum of rib(199).Correlated with loss of sacral contact in taxa with

reduced pelvis: free lymphapophyses (200).Correlated with reduction of appendicular skeleton:

scapulocoracoid reduction or loss (203), clavicle loss(207), interclavicle loss (210), sternum reduction or loss(213), reduction of number of rib attachment points tosternum (216), reduction or loss of forelimbs (218),reduction or loss of pelvis (222), reduction or loss ofhindlimbs (228).Limb reduction and body elongation is functionally

relevant in small, burrowing squamates. Thus, thefollowing characters linked to miniaturisation andfossoriality can be considered to be correlated with limbreduction and body elongation as well. Not all limb-reduced squamates possess these traits, since limbreduction and body elongation also occur in largeaquatic squamates (for anguilliform swimming) and inlarge surface-active terrestrial squamates (for slidingthrough dense vegetation). Furthermore, these traits canoccur in tetrapodal squamates. Thus, the correlationbetween these traits and limb reduction/body elongationis less precise than for the previous set of traits.Nevertheless, they are at least partly correlated withlimb reduction and body elongation, and all these traitsoccur in two or more of the limb-reduced groupsconsidered here, but not in the majority of squamategroups.Correlated with miniaturisation of the skull, resulting

in fewer centres of ossification and thus reduced numberof discrete bones: lacrimal loss (character 11), jugal loss(15), postfrontal loss (31), postorbital loss (35), supra-temporal loss (51), splenial loss (125), angular loss (137),articular loss (144).Correlated with reduction in skull size: maxillary teeth

reduced in number (164). dentary teeth reduced innumber (165).Correlated with the relatively larger size of the brain

in small animals, pushing the jaw musculature onto theskull roof and the region of the temporal arch: parietaljaw adductors cover skull table (42), loss of temporalarch (45).Correlated with cranial consolidation needed for

head-first burrowing: frontoparietal suture complex(30).Correlated with the cylindrically symmetrical trunk

region in burrowing animals: neural spines reduced(176).

ARTICLE IN PRESS

Table 1. Results (P values; highly significant results in italics)

for the ILD tests comparing various data partitions; note that

the osteological comparison is congruent with the soft

anatomy when all taxa are considered, but highly incongruent

when only extant taxa are considered—this trend influences

the three-way tests, which are insignificant when all taxa are

included, but highly significant when only extant taxa are

considered

ILD comparison Ordered Unordered

Three-way (all taxa) 0.91 0.89

Three-way (no fossils) 0.023 0.030

Three-way (no fossils,

weighted)

0.79 0.57

Osteology (all taxa) vs. soft

anatomy

0.62 0.78

Osteology (no fossils) vs. soft

anatomy

0.0020 0.063

Osteology (no fossils,

weighted) vs. soft anatomy

0.92 0.47

Osteology (all taxa) vs. ecology 0.83 0.75

Osteology (no fossils) vs.

ecology

0.88 0.59

Osteology (no fossils,

weighted) vs. ecology

0.86 0.43

Soft anatomy vs. ecology 0.94 0.99


Correlated with the proportionally large braincase insmall forms pushing the opisthotic towards the quad-rate, and the supraoccipital upwards and backwardsagainst the skull roof: suspensorial process of parietalshort (44), quadrate articulates directly with opisthotic(55), supraoccipital situated behind parietal (86), post-temporal fenestra closed (89).Correlated with adaptation in fossorial animals to

detect ground-borne rather than airborne sounds:tympanic crest lost (57), stapes robust (78).Correlated with eye reduction in subterranean ani-

mals: scleral ossicles lost (241).When the ‘extant’ osteological data are re-analysed

with the above characters deleted, the topology changes.Again, snakes become separated from the amphisbae-nian–dibamid clade. In both the ordered and unorderedanalyses, snakes cluster with varanids, and pygopodidswith the three groups of limbed gekkotans, but amphis-baenians and dibamids remain sister groups. Thus, thesignal in the weighted ‘extant’ osteological data set seemsto produce a tree congruent with the tree produced by theunweighted (and weighted) complete osteological data set,and with the tree produced by the soft anatomy.

Congruence between data sets

The incongruence length difference (ILD) test (Farriset al. 1995; for caveats see Lee 2001; Barker and Lutzoni2002; Darlu and Lecointre 2002), implemented inPAUP* as the partition homogeneity test, was used toassess congruence between the three data sets, in a three-way test and in pairwise tests. As before, two separatetreatments of the osteological data (with and withoutfossils) were used. Uninformative characters weredeleted before each ILD test (Lee 2001).As might be expected from examination of the tree

topologies obtained in the partitioned analyses, theosteological data are congruent with the soft anatomicaldata when fossils are included, but incongruent whenfossils are excluded (Table 1). Again it appears that thefossils are not only changing the nature of the signal inthe osteological data, but changing it so that it matchesthe signal in the soft anatomy. The ‘extant-only’osteological data set, however, becomes congruent withthe soft anatomy if characters associated with burrow-ing are deleted; thus it appears that the weightingscheme has increased congruence. The tests comparingecological traits with the other two data partitions didnot detect significant incongruence, which might havebeen expected due to the small number of characters andweak phylogenetic signal in the ecological data set.

Fossils and phylogeny

The signal in the osteological data, whether measuredby tree topology or by congruence with other data sets,

therefore is affected drastically by the inclusion of fossiltaxa. While all the limbed taxa stay in approximately thesame position, the affinities of the limb-reduced taxachange significantly. If both living and fossil taxa areconsidered, amphisbaenians and dibamids cluster withSineoamphisbaena and gekkotans, and snakes areembedded within varanoids. If only extant taxa areconsidered, amphisbaenians, dibamids, snakes (andsometimes pygopodids) cluster together.It is reasonable to interpret the results in the following

fashion (although other interpretations are possible).Living amphisbaenians, dibamids and snakes share ahost of similarities correlated with limb reduction andfossoriality. Fossils exhibit combinations of characterstates that reveal the convergent acquisition of thesetraits in amphisbaenians and dibamids, and in snakes.In particular, the aquatic varanoids and Cretaceousmarine snakes share many synapomorphies with snakes,but also retain well-developed limbs and lack almost allfossorial specialisations. When these fossils are included,therefore, the amphisbaenian–dibamid clade, andsnakes, do not cluster together. In the absence of thesecritical fossils, however, the evidence for convergencedisappears and all the limbless taxa form a spuriousgrouping based on traits correlated with body elonga-tion and limb reduction. This reasonable hypothesis,that fossils are not just affecting the phylogenetic signalin the osteological data, but improving it, is supportedby several observations.


Firstly, the phylogenetic signal in the osteology withonly extant taxa considered is highly incongruent withthe signal in the soft anatomy (whether measured bycomparing tree topologies in partitioned analyses, or viathe ILD test). However, when fossils are considered, theosteological signal becomes almost identical to thesignal in the soft anatomy. This strong coincidence ofsignals suggests they are reflecting common evolutionaryhistory, although some other common (confounding)factor cannot be ruled out. Secondly, when only extanttaxa are considered, analyses including problematic taxaone at a time reveal that there is a strong but hiddensignal grouping snakes with varanoids, and amphisbae-nians and dibamids with gekkotans. This signal isoverwhelmed by an even stronger signal uniting alllimb-reduced groups. The hidden signal, however,coincides exactly with the signal found in the full (fossilplus extant) osteological data set, and in the softanatomical data set. Again, it increases the confidencein the tree from the full osteological data set, since thesame signal is not only present in other data sets, butalso present (albeit latent) in the extant-only osteologicaldata. Thirdly, if characters functionally correlated withlimb reduction and fossoriality are deleted, the topologyof the extant-only osteological analysis again changes tomore closely match the topology found in the full (fossilplus extant) osteological data set, and in the softanatomical data set—at least in the ordered analysis.In other words, the hidden signal is again revealed, andsupports a topology found in the full osteological andthe soft anatomical data set.These results together strongly suggest that when only

extant taxa are considered, the osteological data set ismisleading (at least regarding the affinities of snakes,amphisbaenians and dibamids) due to extensive con-vergence, and fails to produce relationships consistentwith the most highly corroborated tree for all taxa.When fossils are included, a very different tree, with adifferent arrangement of extant taxa, is retrieved. Thesoft anatomy, however, produces the latter tree based onextant taxa alone. There is thus the question of why,when only extant taxa are considered, convergenceapparently confounds the osteological data, but not thesoft anatomical data. In this example, there is aplausible reason. Limb reduction and fossoriality isone of the most pervasive adaptive trends in squamates,and involves an extremely large suite of characters. Alarge proportion of the osteological characters used inthis analysis are reductions or losses of bones correlatedwith limb reduction and fossoriality: for instance, manydeal with the reduction or loss of cranial, girdle, andlimb elements. In contrast, a smaller proportion of softanatomical traits appears to be correlated with limbreduction and fossoriality: this problem only appears toseriously influence a few of the cranial cartilage andvisceral characters. The tongue, hemipenial, histological

and spermatozoal characters which dominate the softanatomical data set are not as obviously linked to limbreduction and fossoriality (though they are undoubtedlycorrelated with other functional complexes). Thus, forthis particular data set there is some reason to expectthat convergence would affect the osteological data setmore severely.

Combined analyses

In cases with significant conflict between data subsets,the validity of the combined-analysis approach has beendebated (e.g. Kluge 1989; Bull et al. 1993; Miyamotoand Fitch 1995; Nixon and Carpenter 1996; Wiens1998b). In the present study, the (complete) osteological,soft anatomical and behavioural data partitions are notin significant conflict, and indeed the two informativedata sets give almost identical phylogenetic signals.Therefore, combined analyses were performed, with allcharacters included, and both with and without thefossil taxa. It has been demonstrated that fossil taxa canhave major effects even in analyses where nearly all softanatomical, ecological, and molecular traits must bescored as unknown (e.g. Gauthier et al. 1988b; O’Leary1999).The support for a particular clade from each data

partition in the context of a combined analysis is thepartitioned branch support, or PBS (Baker and deSalle1997). A positive PBS means that the data partitionsupports that node in the context of a combinedanalysis, a negative PBS means that the data partitioncontradicts that node in the context of a combinedanalysis. Note that partitioned branch supports are notidentical to branch supports in partitioned analyses (fora fuller discussion, see Gatesy et al. 1999). For eachclade, PBS was calculated for the osteology, the softanatomy, and the ecological partitions. TreeRot version2 (Sorenson 1999) was used in all these calculations.In the combined analyses that included all taxa, three

most-parsimonious trees were obtained when multistatecharacters were ordered, and 12 when multistatecharacters were unordered. The consensus trees forboth analyses (Fig. 7) are almost identical, the onlydifference concerning a lack of resolution of relation-ships within gekkotans in the unordered analysis. Asexpected, the topology of the combined tree is almostidentical to those found in the (all-taxon) osteologicaland soft anatomical analyses. It is also consistent withthe (few) resolved nodes in the ecological tree. Regard-ing the two areas where there were disagreementsbetween the three data sets, xantusiids emerge as thesister group of the gekkotan–dibamid–amphisbaenianclade (fide osteology, contra soft anatomy and beha-viour), whereas scincids and cordylids emerge as sistertaxa (fide soft anatomy, contra osteology).

ARTICLE IN PRESS

58,1/0/0

100,10/0/0

84,2/0/0

88,4/0/0

98,17/0/1

100

10/0/0936/0/0

97,11/10/0

97,14/0/0

975/0/1

78,2/1/0

93,7.5/2.5/0

74,5/-2/0

975/0/1

79,6/0/0

953.5/5/0.5

100

3.5/11/0.5

55,2/0/0

582/0/0

57,2/0/095,-2/11/1

61,0/0/1

520.5/0/0.5

10017/0/0

411/0/0

371/0/0

99,11/0/0

973.3/4.3/1.3

976/2/1

842/0/0

Igua

nida

e

Leio

lepi

dina

eA

gam

inae

Cha

mae

leon

idae

Xan

tusi

idae

Eub

leph

arin

aeG

ekko

nine

ss.

l.D

iplo

dact

ylin

ae

Pyg

opod

idae

Lace

rtid

ae

Tei

idae

Gym

noph

thal

mid

aeS

cinc

idae

Cor

dylid

aeA

ngui

dae

Shi

nisa

urus

Xen

osau

rus

Hel

oder

ma

Lant

hano

tus

Var

anus

Ser

pent

es

Pac

hyrh

achi

sH

aasi

ophi

s

Adr

iosa

urus

Dol

icho

saur

idae

Mos

asau

ridae

Aig

ialo

saur

idae

Dib

amid

ae

Pol

ygly

phan

odon

Sin

eoam

phis

baen

aM

acro

ceph

alos

auru

s

Am

phis

baen

ia

99,8/0/0

78, 1/0/0

92,5/0/0

97,15/0.5/0.5

97,16/0/0

97,9/0/0

64,2/0/086,4.7/1.8/-0.5

66, 5/-2/0

970/6/0

986/0/1

914/1/0

100

10/0/0

65,4.2/0.3/0.5

923/3.5/0.5

1002/7.5/0.5

9910.9/1/0.1100

11/0/0

99, 9/0/0

961.8/5.2/1

955/1/1

852/0/0

Mac

roce

phal

osau

rus

Igua

nida

e

Leio

lepi

dina

eA

gam

inae

Cha

mae

leon

idae

Xan

tusi

idae

Eub

leph

arin

aeG

ekko

nine

ss.

l.D

iplo

dact

ylin

ae

Pyg

opod

idae

Lace

rtid

ae

Tei

idae

Gym

noph

thal

mid

aeS

cinc

idae

Cor

dylid

aeA

ngui

dae

Shi

nisa

urus

Xen

osau

rus

Hel

oder

ma

Lant

hano

tus

Var

anus

Ser

pent

es

Pac

hyrh

achi

sH

aasi

ophi

s

Adr

iosa

urus

Dol

icho

saur

idae

Mos

asau

ridae

Aig

ialo

saur

idae

Dib

amid

ae

Pol

ygly

phan

odon

Sin

eoam

phis

baen

a

Am

phis

baen

ia

A

B

Fig. 7. Cladograms resulting from analysis of the combined data set, with both extant and fossil taxa included. Respective first

number refers to bootstrap frequency, next three numbers to partitioned branch support for osteology, soft anatomy, and ecology,

respectively; overall branch support is the sum of these three numbers. (A) Multistate characters ordered; strict consensus of three

most-parsimonious trees, each with length 1057, consistency index 0.48, retention index 0.70. (B) Multistate characters unordered;

strict consensus of 12 most-parsimonious trees, each with length 1002, consistency index 0.50, retention index 0.69.


Combining data sets greatly increased the support foralmost all groupings, whether measured by branchsupport or bootstrap frequency. This was expected,given that usually the same groupings were found in thetwo separate analyses. Combining data sets thusreinforces these common signals, whereas conflictingsignals (presumably random noise) are unlikely toreinforce each other. Many groupings which werepresent in both analyses, but only weakly supported(and thus might be interpreted as insignificant), emerge

as robust in the combined analysis. In particular, theposition of dibamids, amphisbaenians, and snakes isreinforced. It is notable that, for nearly every clade inboth the ordered and unordered analyses, partitionedBremer support is positive or zero for all data sets. Thisindicates that nearly all clades are supported by (or atleast compatible with) all three data sets. There are onlytwo clades in the ordered analysis, and three clades inthe unordered analysis, with negative partitioned branchsupports. Furthermore, these negative values have very

ARTICLE IN PRESS

100,14/4/0

79,22/-12/1

57,2/1/0

969.5/2.5/0

78,22/-12/1

63,2/2/0

985/0/1

76,13.5/-3/0.5

61,5/-2/0

63,0/1/5

980/8/0

1004/12/0

973/6/0

100,16.5/6.5/1

621/0/0

78,1/1/1

98,18/-1/2

71,4.5/0/2.5

82 ,2/0/0

98,6/3/1

982.5/6.5/1

18/-8.5/1.5

95 99,10/5/0

79,16/-9.5/1.52/1/0

64,14/-9.5/1.5

995/641/1

73,16/-9.5/1.5

52, 2 /0/098 ,0/7/0

53,5/-2/0

1003/9/0

893.5/3.5/0

48, - 2/0/5

100,13.5/2.5/1

76, 2 /1/2

100,15/2/2

66,1/0/1

531/0/0

96,2/5/1

95, 6 /2/1

872/0/0

Igua

nida

e

Leio

lepi

dina

e

Aga

min

ae

Cha

mae

leon

idae

Xan

tusi

idae

Eub

leph

arin

ae

Gek

koni

nes

s.l.

Dip

loda

ctyl

inae

Pyg

opod

idae

Lace

rtid

ae

Tei

idae

Gym

noph

thal

mid

ae

Sci

ncid

ae

Cor

dylid

ae

Ang

uida

e

Shi

nisa

urus

Xen

osau

rus

Hel

oder

ma

Lant

hano

tus

Var

anus

Ser

pent

es

Dib

amid

ae

Am

phis

baen

ia

Igua

nida

e

Leio

lepi

dina

e

Aga

min

ae

Cha

mae

leon

idae

Xan

tusi

idae

Eub

leph

arin

ae

Gek

koni

nes

s.l.

Dip

loda

ctyl

inae

Pyg

opod

idae

Lace

rtid

ae

Tei

idae

Gym

noph

thal

mid

ae

Sci

ncid

ae

Cor

dylid

ae

Ang

uida

e

Shi

nisa

urus

Xen

osau

rus

Hel

oder

ma

Lant

hano

tus

Var

anus

Ser

pent

es

Dib

amid

ae

Am

phis

baen

ia

A

B

Fig. 8. Cladograms resulting from analysis of the combined data set, with only extant taxa included. Respective first number refers

to bootstrap frequency, next three numbers to partitioned branch support for osteology, soft anatomy, and ecology, respectively;

overall branch support is the sum of these three numbers. (A) Multistate characters ordered; single most-parsimonious tree with

length 908, consistency index 0.52, retention index 0.68. (B) Multistate characters unordered; single most-parsimonious tree with

length 857, consistency index 0.54, retention index 0.67.


small (absolute) values, the largest being only �2.Agreement between the data sets therefore is very good,with nearly all the phylogenetic signal in the three datasets being retained as branch support in the combinedtree.When the combined analyses are performed with

fossils deleted, however, a very different result is

obtained (Fig. 8). Not only is the tree very poorlysupported, but there is great disagreement between thedata sets. The number of nodes with at least onenegative PBS is five in the ordered analysis, four in theunordered analysis. Despite there being fewer nodes inthe extant-only data sets (due to fewer taxa), thereare more nodes with negative PBS. Moreover, these

ARTICLE IN PRESS

Table 2. The relative informativeness of each data partition,

as measured by their consistency and retention indices (CU,

RI); Farris, 1989 in the context of the overall combined tree;

note that all means fall within one standard deviation of each

other, and thus are not significantly different

Data

partition

Ordered analysis Unordered analysis

CI RI CI RI

Osteology 0.6170.30 0.7570.48 0.6270.30 0.7170.30Soft anatomy 0.6470.30 0.7270.49 0.6670.30 0.6870.32Ecology 0.7970.25 0.8070.31 0.7670.25 0.7770.32


negative values are often rather large: in the orderedanalysis, two nodes have a PBS of �12; in the unorderedanalysis, three nodes have a PBS of �9.5, and anotherhas a PBS of �8.5. Thus, a smaller proportion of thephylogenetic signal in the three data sets is retained assupport for the combined tree. The amount of disagree-ment between the data sets appears to be substantial; itis also rather localised. In particular, all the highnegative values (implying disagreement between datasets over particular clades) occur within the cladeconsisting of anguimorphs and three limbless taxa(snakes, amphisbaenians and dibamids), indicating thatthere is severe disagreement between data sets concern-ing this clade. This is explicable because the (extant-only) osteology strongly supports this arrangement,while the soft anatomy strongly suggests that snakesnest within anguimorphs, but that amphisbaenians anddibamids are more basal; the combined data setpreserves the osteological signal over the conflicting softanatomical signal.These two combined analyses, with and without

fossils, reinforce the results obtained in the separateanalyses and the ILD tests. When fossils are included inthe analyses, the data sets are congruent with each other,the phylogenetic signal in the separate data sets isretained in the combined analysis, and the resultant treeis well-supported. When fossils are ignored, there is agreat deal of incongruence between data sets, much ofthe phylogenetic signal in the separate data sets is lost inthe combined analysis, and the resultant tree is weaklysupported. Again, the strong and arguably positiveeffect that fossils have on data congruence and treetopology is upheld.

Relative informativeness of osteology, soft anatomy

and behaviour

The relative phylogenetic informativeness of osteol-ogy, soft anatomy and behaviour was investigated in thecontext of the combined analyses. As there is nosignificant incongruence between the three data sets,the tree from the combined analysis can be accepted asthe most highly corroborated estimate of squamatephylogeny. The congruence of osteology, soft anatomy,and behaviour on this ‘best-estimate’ tree can beevaluated by comparing the average consistency andretention indices for each set of characters whenoptimised on this combined tree. All three classes ofcharacters have very similar values that do not differsignificantly (Table 2). While osteology has slightlylower CIs and RIs than soft anatomy and ecology, thismight be due to osteological characters being observableon more taxa (varying numbers of fossils), and thushaving more opportunity for homoplasy.

Conclusions, prospects, and snake origins

When different data sets strongly support radicallydifferent phylogenies, at least one of them is wrong (i.e.does not reflect evolutionary history). There areinstances where both data sets can be right, involvinglineage sorting, hybridisation, and other processes (e.g.Wiens 1998b). However, when dealing with higher taxa,these processes generally can be assumed to be ratherminor. It has been shown that increased taxon samplingis one of the factors most important to retrieving thecorrect tree topology (e.g. Hillis 1996), and that thecritical additional sampled taxa are often fossils (e.g.Gauthier et al. 1988b). In an analysis of higher taxa, iftwo data sets disagree, at least one of them is wrong.Increasing the taxon sampling in both data sets shouldcause both to converge on the true tree, and thus oneach other (cf. Cunningham 1997). This prediction isupheld by the above analyses. If only extant taxa areconsidered, the osteological data set contains a phylo-genetic signal that is driven by a suite of correlatedcharacters, and is incongruent with the soft anatomicaldata set. When additional (fossil) taxa are sampled,however, the osteological data set becomes congruentwith the soft anatomical data. This suggests that thefossil taxa are not merely changing the signal in theosteological data, but changing by retreiving secondarysignals that match those in other data sets. Moreempirical studies are needed, however, because anotherrecent study (Mitchell et al. 2000) found that addingtaxa decreased congruence.Assessing phylogenetic accuracy is usually proble-

matic because (with a few special exceptions) one cannever know the true phylogeny. Increased congruencebetween data sets, however, is a valuable indicator thatthe data sets are each converging on the real phylogeny.Previous studies have used this assumption to test thevalidity of tree construction methods (Cunningham1997) and coding and weighting criteria (Allard andCarpenter 1996; Wiens 1998c). If a method produces thesame tree when applied to different sources of data, this


would suggest that the method is converging on thereal phylogeny. Conversely, if a method producesradically different trees when applied to different datasets, this would suggest that the method fails (at least insome applications). There is, however, the addedcomplication (not considered by Cunningham 1997)that different methods might be applicable to differentdata sets (e.g. likelihood to molecules, parsimony tomorphology), and this might be suggested if applyingdifferent methods to different data sets increases treecongruence, compared to using a single method on alldata sets.Tree congruence can be used to evaluate other

problems. There currently is a debate over whetheradding characters or adding taxa is more important torecovering the correct tree topology. Most argumentsfor either position have been based on rather simplifiedsimulations that might be of limited relevance to thebiological world. However, if congruence between datasets is an indicator of phylogenetic accuracy, there mightbe a way to evaluate the problem using real data afterall. If adding taxa is the better way to improve chancesat retrieving the real phylogeny, then one would expectthat doubling the number of sampled taxa (e.g. speciessequenced) will be better for reducing incongruence thandoubling the number of characters (e.g. getting twice asmuch sequence data for each taxon). Conversely, ifadding characters is the better way, then one wouldexpect that doubling the number of characters will bebetter for reducing incongruence than doubling thenumber of sampled taxa. The current study shows thatadding taxa can have a great effect on reducingincongruence and presumably improving phylogenetic

Fig. 9. Strict consensus of the two trees (753 steps) that result if oste

(for which gene sequences are known) constrained to the molecula

(indicated by asterisks) are allowed to ‘float’ within this molecular ba

with snakes, implying a marine phase in snake origins. See Lee (200

accuracy. However, it remains to be seen if addingcharacters can have just as profound an effect. A morecomprehensive study using multiple data sets andvarying both taxon sampling and character sampling isrequired.In particular, recent molecular work on the higher-

level phylogeny of squamates (e.g. Saint et al. 1998;Harris et al. 2001; Vicario et al. 2002; Vidal and Hedges2004; Townsend et al. 2004) should soon offer a largeindependent data set to test these ideas. These molecularstudies tend to place dibamids, then gekkotans, as thebasal squamates (an arrangement with some morpholo-gical support: Greer 1985b; Underwood and Lee 2001),challenging the idea of a basal iguanian–scleroglossandichotomy. The poor resolution at the base of squa-mates found in the present analysis is mirrored in recentmolecular studies, which have also found short branchesand/or poor support in this region. The failure of bothmorphology and molecules to resolve many of thesebranches is consistent with the hypothesis that basaldivergences within Scleroglossa occurred rapidly, gen-erating short internal branches that are difficult toreconstruct.The new molecular data also tend to group snakes

with iguanians (Vidal and Hedges 2004; Townsend et al.2004). This is highly heterodox, given that the two taxarepresent opposite morphological and behaviouralextremes within squamates (e.g. Schwenk 2000; Piankaand Vitt 2003). However, the snake–iguanian clade isweakly supported: snakes and iguanians essentially forma polytomy with anguimorphs, and a snake–angui-morph clade (strongly supported by morphology) thusremains consistent with the molecular data (Townsend

ology data are analysed with relationships between extant taxa

r topology suggested by Vidal and Hedges (2004); other taxa

ckbone. Note that marine lizards and pachyophiids both group

5) for further discussion.


et al. 2004). The nesting of snakes within marine lizardsfound here strongly supports an aquatic origin of snakes(Nopcsa 1923; Lee and Caldwell 2000). The recentmolecular evidence has been interpreted as refuting thisscenario (Vidal and Hedges 2004), suggesting thatsnakes are only distantly related to living varanidlizards. As marine lizards such as mosasaurs anddolichosaurs usually are considered to be varanoids(e.g. Lee and Caldwell 2000), this would seem topreclude them being close relatives of snakes. However,these marine lizards share similarities with both snakesand living varanids. In fact, the phylogeny found here –(varanids (marine lizards (snakes))) – indicates thatderived traits shared between marine lizards andvaranids will also tend to be found in snakes, and thatmarine lizards must share additional synapomorphieswith snakes alone. Thus, if the molecular phylogeny isaccepted and varanids and snakes are assumed to bewidely separated, marine lizards must ‘choose’ whatother group they align with; they could be expected tocluster more strongly with snakes than with varanids.Accordingly, an analysis of the osteological datasuggests that, if relationships between living taxa areconstrained to the proposed molecular topology, andmarine lizards are allowed to ‘float’, they group mostparsimoniously with snakes, rather than with varanids(Lee 2005; see present Fig. 9). This tree is significantlybetter than the one resulting when relationships betweenliving taxa are constrained to the molecular topology,and when marine lizards are forced to cluster withvaranids (the arrangement suggested by Vidal andHedges 2004). Very similar results occur if the fullmorphological data set (not just the osteology) is used;this is to be expected given that the majority of ‘floating’taxa (including all fossils) are scored for osteology only.Thus, contrary to recent suggestions, the molecular datado not refute the idea that marine lizards are related tosnakes. Instead, it contradicts the idea that marinelizards are related to varanids.

Acknowledgements

The author thanks the late Garth Underwood, andMichael Caldwell, John Scanlon, Jean-Claude Rage,Tod Reeder, Ken Kardong, Susan Evans and NickArnold for discussion, as well as anonymous reviewersfor comments on two versions of this manuscript. Theauthor is especially grateful to Kurt Schwenk for hishelp in clarifying some tongue characters, and toMagdalena Borsuk-Bialynicka (Polish Academy ofSciences, Warsaw) and Bill Purdy (US NationalMuseum of Natural History, Washington) for accessto macrocephalosaur and polyglyphanodontid material.

References

Allard, M.W., Carpenter, J.M., 1996. On weighting and

congruence. Cladistics 12, 183–198.

Arnold, E.N., 1984. Variation in the cloacal and hemipenial

musculature of lizards and its bearing on their relation-

ships. Symp. Zool. Soc. Lond. 52, 47–85.

Baker, R., deSalle, R., 1997. Multiple sources of character

information and the phylogeny of Hawaiian drosophilids.

Syst. Biol. 46, 654–673.

Barker, F.K., Lutzoni, F.M., 2002. The utility of the

incongruence length difference test. Syst. Biol. 51, 625–657.

Barrett, M., Donoghue, M.J., Sober, E., 1991. Against

consensus. Syst. Zool. 40, 486–493.

Bininda-Emonds, O.R., Bryant, H.N., Russell, A.P., 1998.

Supraspecific taxa as terminals in cladistic analysis: implicit

assumptions of monophyly and a comparison of methods.

Biol. J. Linn. Soc. 64, 101–133.

Bremer, K., 1988. The limits of amino acid sequence data in

angiosperm phylogenetic reconstruction. Evolution 42,

795–803.

Bull, J.J., Huelsenbeck, J.P., Cunningham, C.W., Swofford,

D.L., Waddell, P.J., 1993. Partitioning and combining data

in phylogenetic analysis. Syst. Biol. 42, 384–397.

Camp, C.L., 1923. Classification of the lizards. Bull. Am. Mus.

Nat. Hist. 48, 289–481.

Carpenter, C.C., Ferguson, G.W., 1977. Variation and

evolution of stereotyped behaviour in reptiles. In: Gans,

C., Tinkle, D.W. (Eds.), Biology of the Reptilia, vol. 7.

Academic Press, London, pp. 335–554.

Cooper Jr, W.E., 1995. Foraging mode, prey chemical

discrimination, and phylogeny in lizards. Anim. Behav.

50, 973–985.

Cooper Jr, W.E., 1996. Preliminary reconstruction of nasal

chemosensory evolution in Squamata. Amphibia–Reptilia

17, 395–415.

Cooper Jr, W.E., 1997. Independent evolution of squamate

olfaction and vomerolfaction and correlated evolution of

vomerolfaction and lingual structure. Amphibia–Reptilia

18, 85–105.

Cope, E.D., 1869. On the reptilian orders Pythonomorpha

and Streptosauria. Proc. Boston Soc. Nat. Hist. 12,

250–266.

Cundall, D., Wallach, V., Rossman, D.S., 1993. The systema-

tic relationships of the snake genus Anomochilus. Zool.

J. Linn. Soc. 109, 275–299.

Cunningham, C.W., 1997. Is congruence between data

partitions a reliable predictor of phylogenetic accuracy?

Empirically testing an iterative procedure for choosing

among phylogenetic methods. Syst. Biol. 46, 464–478.

Darlu, P., Lecointre, G., 2002. When does the incongruence

length difference test fail? Mol. Biol. Evol. 19, 432–437.

Donnellan, S.C., Hutchinson, M.N., Saint, K.M., 1999.

Molecular evidence for the phylogeny of Australian

gekkonoid lizards. Biol. J. Linn. Soc. 67, 97–118.

Estes, R., de Queiroz, K., Gauthier, J., 1988. Phylogenetic

relationships within Squamata. In: Estes, R., Pregill, G.K.

(Eds.), Phylogenetic Relationships of the Lizard Families:

Essays Commemorating Charles L. Camp. Stanford Uni-

versity Press, Stanford, pp. 119–281.


Evans, S.E., 1991. A new lizard-like reptile (Diapsida:

Lepidosauromorpha) from the Middle Jurassic of England.

Zool. J. Linn. Soc. 103, 391–412.

Evans, S.E., Barbadillo, J., 1997. Early cretaceous lizards from

Las Hoyas, Spain. Zool. J. Linn. Soc. 119, 23–49.

Evans, S.E., Chure, D.C., 1998. Paramacellodid lizard

skulls from the Jurassic Morrison formation at

Dinosaur National Monument, Utah. J. Vert. Paleont.

18, 99–114.

Farris, J.S., 1989. The retention index and the rescaled

consistency index. Cladistics 5, 417–419.

Farris, J.S., Kallersjo, M., Kluge, A.G., Bult, C., 1995.

(‘‘1994’’). Testing significance of incongruence. Cladistics

10, 315–319.

Felsenstein, J., 1985. Confidence limits on phylogenies: an

approach using the bootstrap. Evolution 39, 783–791.

Forstner, M.R.J., Davis, S.K., Arevalo, E., 1995. Support for

the hypothesis of anguimorph ancestry for the suborder

Serpentes from phylogenetic analysis of mitochondrial

DNA sequences. Mol. Phyl. Evol. 4, 93–102.

Frost, D.R., Etheridge, R., 1989. A phylogenetic analysis and

taxonomy of iguanian lizards (Reptilia: Squamata). Misc.

Publ. Mus. Nat. Hist. Univ. Kansas 81, 1–65.

Gabe, M., Saint-Girons, H., 1965. Contribution a la morpho-

logie comparee du cloaque et des glandes epidermoıdes de

la region cloacale chez les lepidosauriens. Mem. Mus. Nat.

Hist., Ser. A, Zool. 33, 151–292.

Gabe, M., Saint-Girons, H., 1972. Contribution a l’histologie

de l’estomac des lepidosauriens (reptiles). Zool. Jahrb.

Anat. 89, 579–599.

Gatesy, J., O’Grady, P., Baker, R.H., 1999. Corroboration

among data sets in simultaneous analysis: hidden support

for phylogenetic relationships among higher-level artiodac-

tyl taxa. Cladistics 15, 271–313.

Gauthier, J., Estes, R., de Queiroz, K., 1988a. A phylogenetic

analysis of the Lepidosauromorpha. In: Estes, R., Pregill,

G.K. (Eds.), Phylogenetic Relationships of the Lizard

Families: Essays Commemorating Charles L. Camp.

Stanford University Press, Stanford, pp. 15–98.

Gauthier, J., Kluge, A.G., Rowe, T., 1988b. Amniote

phylogeny and the importance of fossils. Cladistics 4,

105–209.

Greer, A.E., 1985a. Facial tongue-wiping in xantusiid lizards:

its systematic implications. J. Herp. 19, 174–175.

Greer, A.E., 1985b. The relationships of the lizard genera

Anelytropsis and Dibamus. J. Herp. 19, 116–156.

Haas, G., 1980. Pachyrhachis problematicus Haas, snakelike

reptile from the lower Cenomanian: ventral view of skull.

Bull. Mus. Natn. Hist. Nat. Paris, 4me Ser, Sect. C 2,

87–104.

Hallermann, J., 1998. The ethmoidal region of Dibamus taylori

(Squamata: Dibamidae), with a phylogenetic hypothesis on

dibamid relationships within Squamata. Zool. J. Linn. Soc.

122, 385–426.

Harris, D., Marshall, J.C., Crandall, K.A., 2001. Squamate

relationships based on C-mos nuclear DNA sequences:

increased taxon sampling improves bootstrap support.

Amphibia–Reptilia 22, 235–242.

Hillis, D.M., 1996. Inferring complex phylogenies. Nature 383,

130–131.

Jamieson, B.G.M., 1995. The ultrastructure of the spermato-

zoa of the Squamata (Reptilia) with phylogenetic con-

siderations. Mem. Mus. Natn. Hist. Nat. 166, 359–383.

Kearney, M., 2003. The phylogenetic position of Sineoam-

phisbaena hextabularis reexamined. J. Vert. Paleont. 23,

394–403.

Keogh, J.S., 1996. Evolution of the colubrid snake tribe

Lampropeltini: a morphological perspective. Herpetologica

52, 406–416.

Kluge, A.G., 1987. Cladistic relationships in the Gekkonoidea

(Squamata, Sauria), Misc. Publ. Mus. Zool. Univ. Michi-

gan 173, 1–54.

Kluge, A.G., 1989. A concern for evidence and a phylogenetic

hypothesis of relationships among Epicrates (Boidae,

Serpentes). Syst. Zool. 38, 7–25.

Kochva, E., 1978. Phylogeny of the oral glands in reptiles as

related to the origin and evolution of snakes. In: Rosen-

berg, P. (Ed.), Proceedings of the Fifth International

Symposium on Animal, Plant, and Microbial Toxins.

Pergamon Press, Oxford, pp. 29–37.

Lee, M.S.Y., 1997. Phylogenetic relationships among Austra-

lian elapid snakes: the soft anatomical data reconsidered.

Herp. J. 7, 93–102.

Lee, M.S.Y., 1998. Convergent evolution and character

correlation in burrowing squamates: a phylogenetic analy-

sis. Biol. J. Linn. Soc. 65, 369–453.

Lee, M.S.Y., 2000. Soft anatomy, diffuse homoplasy, and the

relationships of lizards and snakes. Zool. Scr. 29, 101–130.

Lee, M.S.Y., 2001. Uninformative characters and apparent

conflict between molecules and morphology. Mol. Biol.

Evol. 18, 676–680.

Lee, M.S.Y., 2005. Molecular evidence and snake origins. Biol.

Lett. (in press).

Lee, M.S.Y., Caldwell, M.W., 1998. Anatomy and relation-

ships of Pachyrhachis, a primitive snake with hindlimbs.

Philos. Trans. R. Soc. Lond., Biol. Sci. 353, 1521–1552.

Lee, M.S.Y., Caldwell, M.W., 2000. Adriosaurus and

the affinities of mosasaurs, dolichosaurs, and snakes.

J. Paleontol. 74, 915–937.

Lee, M.S.Y., Scanlon, J.D., 2002. Snake phylogeny based on

osteology, soft anatomy and behaviour. Biol. Rev. 77,

333–401.

Lee, M.S.Y., Shine, R., 1998. Reptilian viviparity and Dollo’s

law. Evolution 52, 1441–1450.

Lipscomb, D., 1992. Parsimony, homology, and the analysis of

multistate characters. Cladistics 8, 45–66.

Macey, J.R., Schulte II, J.A., Larson, A., Ananjeva, N.B.,

Wang, Y., Pethiyagoda, R., Rastegar-Pouyani, N., Papen-

fuss, T.J., 2000. Evaluating trans-Tethys migration: an

example using acrodont lizard phylogenetics. Syst. Biol. 49,

233–256.

Maddison, D.R., Maddison, W.P., 2000. MacClade 4:

Analysis of Phylogeny and Character Evolution. Sinauer

Associates, Sunderland, MA.

Manley, G.A., 2003. Evolution of structure and function of the

hearing organ of lizards. J. Neurobiol. 51, 202–211.

McCracken, K.C., Harshman, J., McClellan, D.A., Afton,

A.D., 1999. Data set incongruence and correlated character

evolution: an example of functional convergence in the hind

limbs of stifftail diving ducks. Syst. Biol. 48, 683–714.


McDowell, S.B., Bogert, C.M., 1954. The systematic position

of Lanthanotus and the affinities of anguinomorphan

lizards. Bull. Am. Mus. Nat. Hist. 105, 1–142.

Miller, M.R., 1966. The cochlear duct of lizards and snakes.

Am. Zool. 6, 421–429.

Mitchell, A., Mitter, C., Regier, J.C., 2000. More taxa or more

characters revisited: combining data from nuclear protein-

encoding genes for phylogenetic analyses of Noctuiodea

(Insecta: Lepidoptera). Syst. Biol. 49, 202–224.

Miyamoto, M.M., Fitch, W.M., 1995. Testing species phylo-

genies and phylogenetic methods with congruence. Syst.

Biol. 44, 64–76.

Moody, S., 1980. Phylogenetic and historical biogeographical

relationships of the genera in the Agamidae (Reptilia,

Lacertilia). Ph.D. Dissertation, University of Michigan,

Ann Arbor, MI.

Nixon, K.C., Carpenter, J.M., 1996. On simultaneous analysis.

Cladistics 12, 221–241.

Nopcsa, F., 1923. Eidolosaurus und Pachyophis. Zwei neue

Neocom-Reptilien. Palaeontographica 65, 99–154.

Northcutt, R.G., 1978. Forebrain and midbrain organisation

in lizards and its phylogenetic significance. In: Greenberg,

N., MacLean, P. (Eds.), Behavior and Neurology of

Lizards. NIMH, Rockville, MD, pp. 11–64.

O’Leary, M., 1999. Parsimony analysis of total evidence

from extinct and extant taxa and the cetacean-

artiodactyl question (Mammalia, Ungulata). Cladistics 15,

315–330.

Parsons, T.S., 1970. The nose and Jacobson’s organ. In: Gans,

C., Parsons, T.S. (Eds.), Biology of the Reptilia, vol. 2.

Academic Press, London, pp. 99–191.

Pianka, E.R., Vitt, L.J., 2003. Lizards: Windows to the

Evolution of Diversity. University of California Press,

Berkeley.

Pregill, G., Gauthier, J.A., Greene, H.W., 1986. The evolution

of helodermatid squamates, with description of a new taxon

and an overview of Varanoidea. Trans. San Diego Soc.

Nat. Hist. 21, 167–202.

Presch, W., 1988. Cladistic relationships within the Scinco-

morpha. In: Estes, R., Pregill, G.K. (Eds.), Phylogenetic

Relationships of the Lizard Families: Essays Commemor-

ating Charles L. Camp. Stanford University Press, Stan-

ford, pp. 471–491.

Proctor, H.C., 1996. Behavioural characters and homoplasy:

perception versus practice. In: Sanderson, M.J., Hufford,

L. (Eds.), Homoplasy: the Recurrence of Similarity in

Evolution. Academic Press, San Diego, pp. 131–149.

de Queiroz, A., Wimberger, P.H., 1993. The usefulness of

behaviour for phylogeny estimation: levels of homoplasy in

behavioural and morphological characters. Evolution 47,

46–60.

Rage, J.-C., Escuillie, F., 2000. Un nouveau serpent bipede du

Cenomanien (Cretace). Implications phyletiques. CR Acad.

Sci. Terre Planetes 330, 513–520.

Reynoso, V.H., 1998. Huehuecuetzpalli mixtecus gen. et sp.

nov.: a basal squamate (Reptilia) from the Early Creta-

ceous of Tepexi de Rodriguez, central Mexico. Philos.

Trans. R. Soc. Lond. B 353, 477–500.

Rieppel, O., 1980. The phylogeny of anguimorph lizards.

Denk. Schweiz. Nat. Ges. 94, 1–86.

Rieppel, O., 1984. The cranial morphology of the fossorial

lizard genus Dibamus with a consideration of its phyloge-

netic relationships. J. Zool. 204, 289–327.

Rieppel, O., 1988. A review of the origin of snakes. Evol. Biol.

22, 37–130.

Rieppel, O., Zaher, H., 2000. The intramandibular joint in

squamates, and the phylogenetic relationships of the fossil

snake Pachyrhachis problematicus. Fieldiana Geol 1507,

1–69.

Russell, A.P., 1988. Limb muscles in relation to lizard

systematics: a reappraisal. In: Estes, R., Pregill, G.K.

(Eds.), Phylogenetic Relationships of the Lizard Families:

Essays Commemorating Charles L. Camp. Stanford Uni-

versity Press, Stanford, pp. 493–568.

Saint, K.M., Austin, C.C., Donnellan, S.C., Hutchinson,

M.N., 1998. C-mos, a nuclear marker useful for squamate

phylogenetic analysis. Mol. Phyl. Evol. 10, 259–263.

Sanderson, M.J., Donoghue, M.J., 1989. Patterns of variation

in levels of homoplasy. Evolution 48, 1781–1795.

Scanlon, J.D., 1996. Studies in the Palaeontology and

Systematics of Australian Snakes. Ph.D. Thesis, University

of New South Wales, Sydney.

Scanlon, J.D., 2005. Cranial morphology of the Plio-Pleisto-

cene giant snake Wonambi naracoortensis M.J. Smith 1976

(Squamata, Ophidia, Madtsoiidae). Acta Paleont. Pol.

(in press).

Schwenk, K., 1988. Comparative morphology of the lepido-

saur tongue and its relevance to squamate phylogeny. In:

Estes, R., Pregill, G.K. (Eds.), Phylogenetic Relationships

of the Lizard Families: Essays Commemorating Charles L.

Camp. Stanford University Press, Stanford, pp. 569–598.

Schwenk, K., 1993. The evolution of chemoreception in

squamate reptiles: a phylogenetic approach. Brain Behav.

Evol. 41, 124–137.

Schwenk, K., 2000. Feeding in lepidosaurs. In: Schwenk, K.

(Ed.), Feeding: Form, Function and Evolution in Tetrapod

Vertebrates. Academic Press, San Diego, pp. 175–291.

Siddall, M.E., Whiting, M.F., 1999. Long-branch abstractions.


Slowinski, J.B., 1993. ‘‘Unordered’’ versus ‘‘ordered’’ char-

acters. Syst. Biol. 42, 155–165.

Slowinski, J.B., Lawson, R., 2002. Snake phylogeny: evidence

from nuclear and mitochondrial genes. Mol. Phyl. Evol. 24,

194–202.

Sorenson, M.R., 1999. TreeRot. Version 2. Computer

Program Distributed by the Author. Boston University,

Boston, MA.

Stamps, J.A., 1977. Social behavior and spacing patterns in

lizards. In: Gans, C., Tinkle, D.W. (Eds.), Biology of the

Reptilia, vol. 7. Academic Press, London, pp. 265–334.

Swofford, D.L., 2000. PAUP*—Phylogenetic Analysis Using

Parsimony (*and Other Methods). Beta Version 4. Sinauer

Associates, Sunderland, MA.

Tchernov, E., Rieppel, O., Zaher, H., Polcyn, M., Jacobs,

L.L., 2000. A fossil snake with limbs. Science 287,

2010–2012.

Townsend, T.M., Larson, A., Louis, E., Macey, J.R., 2004.

Molecular phylogenetics of Squamata: the position of

snakes, amphisbaenians, and dibamids, and the root of

the squamate tree. Syst. Biol. 53, 735–757.


Underwood, G., Lee, M.S.Y., 2001. The egg teeth of gekkotan

and dibamid lizards. Amphibia–Reptilia 21, 507–511.

Vicario, S., Caccone, A., Gauthier, J., 2002. Xantusiid ‘‘night’’

lizards: a puzzling phylogenetic problem revisited using

likelihood-based Bayesian methods on mtDNA sequences.

Mol. Phyl. Evol. 26, 243–261.

Vidal, N., Hedges, S.B., 2002. Higher-level relationships of

snakes inferred from four nuclear and mitochondrial genes.

CR Biol. 325, 977–985.

Vidal, N., Hedges, S.B., 2004. Molecular evidence for a

terrestrial origin of snakes. Biol. Lett. 271, 226–229.

Wallach, V., 1985. A cladistic analysis of the terrestrial

Australian Elapidae. In: Grigg, G., Shine, R., Ehmann,

H. (Eds.), Biology of Australasian Frogs and Reptiles.

Royal Zoological Society of New South Wales, Sydney,

pp. 223–253.

Wallach, V., 1998. The lungs of snakes. In: Gans, C., Gaunt,

A.S. (Eds.), Biology of the Reptilia, vol. 19. Society for the

Study of Amphibians and Reptiles, Oxford, OH, pp. 93–295.

Wenzel, J., 1992. Behavioral homology and phylogeny. Annu.

Rev. Ecol. Syst. 23, 361–381.

Wever, E.G., 1978. The Reptile Ear: Its Structure and

Function. Princeton University Press, Princeton, NJ.

Whiting, A.S., Bauer, A.M., Sites Jr, J.W., 2003. Phylogenetic

relationships and limb loss in sub-Saharan African scincine

lizards (Squamata: Scincidae). Mol. Phylogenet. Evol. 29,

582–598.

Wiens, J.J., 1998a. The accuracy of methods for coding and

sampling higher-level taxa for phylogenetic analysis: a

simulation study. Syst. Biol. 47, 397–413.

Wiens, J.J., 1998b. Combining data sets with different

phylogenetic histories. Syst. Biol. 47, 568–581.

Wiens, J.J., 1998c. Testing phylogenetic methods with tree

congruence: phylogenetic analysis of polymorphic morpho-

logical characters in phrynosomatid lizards. Syst. Biol. 47,

427–444.

Wilkinson, M., 1992. Ordered versus unordered characters.


Wimberger, P.H., de Queiroz, A., 1996. Comparing behavioral

and morphological characters as indicators of phylogeny.

In: Martins, E. (Ed.), Phylogenies and the Comparative

Method in Animal Behaviour. Oxford University Press,

New York, pp. 206–233.

Wu, X.-C., Brinkman, D.B., Russell, A.P., 1996. Sineoam-

phisbaena hexatabularis, an amphisbaenian (Diapsida:

Squamata) from the Upper Cretaceous redbeds at

Bayan Mandahu (Inner Mongolia, People’s Republic

of China), and comments on the phylogenetic relation-

ships of the Amphisbaenia. Can. J. Earth Sci. 33,

541–577.

Zaher, H., Rieppel, O., 1999. The phylogenetic relationships of

Pachyrhachis problematicus, and the evolution of limbless-

ness in snakes (Lepidosauria, Squamata). CR Acad. Sci.

Terre Planetes 329, 831–837.

Squamatephylogeny,taxonsampling,anddatacongruence · Organisms,Diversity&Evolution5(2005)25–45...

Documents

Transcript of Squamatephylogeny,taxonsampling,anddatacongruence · Organisms,Diversity&Evolution5(2005)25–45...