poe 2004 phylogeny of anoles[1]

7/27/2019 Poe 2004 Phylogeny of Anoles[1]

1/53

Herpetological Monographs, 18, 2004, 3789 2004 by The Herpetologists League, Inc.

PHYLOGENY OF ANOLES

STEVEN POE1

Department of Biology, and Museum of Southwestern Biology, Castetter Hall, University of New Mexico, Albuquerque,NM 87131, USA

ABSTRACT: I present a phylogenetic analysis of the lizard genus Anolis using new morphological data incombination with diverse data from the literature. Ninety-one characters of osteology and external anatomywere completely or partially scored for 174 Anolis species and seven outgroups. These data were combinedwith data from chromosomes, DNA sequences, allozymes, and immunology and analyzed with parsimony toproduce an estimate ofAnolis relationships. The genus Anolis was supported as monophyletic. Anolis occultusis sister to the rest of the genus. A South American and Southern Lesser Antillean clade is sister to the GreaterAntillean and Northern Lesser Antillean Anolis and a clade of mainland species. Successive clades ofCaribbean Alpha Anolis are sister to the Beta Anolis (Norops) clade. Within the Betas, the Jamaican Betas aresister to the remaining Betas, and the Cuban Betas are sister to the monophyletic mainland forms. Otherhigher-level groupings of previous authors were not supported, but members of some previously-recognized

lower-level groups formed clades: roquet series, Phenacosaurus, cybotes series, cristatellus species group,bimaculatus series/species group, hendersoni species group, chlorocyanus series, equestris series, grahamiseries, sagrei series, crassulus species group, gadovii species group, laeviventris species group, nebulosusspecies group.

Key words: anoles; Anolis; morphology; phylogeny.

THE EXTRAORDINARILY SPECIES-RICH, biogeo-graphically complex, ecologically diverse neo-tropical group of lizards called anoles (genus

Anolis) has long captivated the interest of

vertebrate biologists. Hundreds of studies havedocumented the behavioral, ecological, andmorphological diversity in these lizards, whichare probably among the best-studied verte-brates.Anolis lizards have been used to addressfundamental biological issues such as commu-nity ecology (e.g., Williams, 1983), biogeogra-phy (e.g., Lazell, 1972), sexual dimorphism(e.g., Fitch, 1975), competition (e.g., Pacalaand Roughgarden, 1982), energetics (e.g.,Naganuma and Roughgarden, 1990), func-

tional morphology (e.g., Losos, 1990), adaptiveradiation (e.g., Williams, 1972), ecomorphol-ogy (e.g., Collette, 1961), character displace-ment (e.g., Schoener, 1970), social behavior(Stamps, 1973), perception (e.g., Fleishman etal., 1993), reproduction (e.g., Sexton et al.,1971), and communication (e.g., Rand and

Williams, 1970).Many aspects of comparative biology have

been recorded for numerous anole species(e.g., Fitch, 1975), and it is clear that properinterpretation of comparative data requires

knowledge of phylogeny (e.g., Felsenstein,1985a). Recent papers on the comparativebiology of Anolis have incorporated informa-tion on phylogeny (e.g., Losos et al., 1998).

However, comparative studies of the entiregenus have not been possible due to the lack ofa comprehensive phylogeny.

The phylogeny of Anolis is a notoriouslydifficult problem (Williams, 1989). Anolis isthe largest amniote genus, containing approx-imately 369 species (personal observation).The great size of the genus and the perceivedmorphological conservativeness have hinderedthe reconstruction of Anolis phylogeny. Hillis(1996) referred to the genus as a huge group

where all the species look virtually the same.Richard Etheridges (1959) analysis of 150species for his Ph.D. dissertation remains themost comprehensive treatment of anole evo-lution. Considering the wealth of comparativedata currently available for Anolis, the avail-ability of much new osteological material(Williams, 1989), and the effect that speciessampling can have on hypotheses of relation-ship (e.g., Gauthier et al., 1988) and compar-ative biology (e.g., Ackerly, 2000), an update ofEtheridges (1959) landmark work is longoverdue.

The goal of this paper is to estimate thephylogeny ofAnolis. I collected morphological1 CORRESPONDENCE: e-mail, [email protected]

37


2/53

data and analyzed these in combination withdiverse types of data from the literature.

Previous Hypotheses of Anole Relationships

Etheridge (1959) divided Anolis into twomajor sections based on a single osteologicalcharacter (absence of caudal transverse pro-cesses in Alpha Anolis, presence in Beta

Anolis), and identified several informal groupscalled series within these sections. Williams(1976a,b) further subdivided the genus intosubsections, subseries, species groups, andsuperspecies. Although based on availabledata, this work was not a formal phylogeneticanalysis. Still, the informal groups proposed by

Williams often have been treated as clades(Guyer and Savage, 1992). The contents ofmany of the Williams groups were rearrangedby Gorman and collaborators (e.g., Gormanand Kim, 1976; Gorman et al., 1983), Lieb(1981), and Shochat and Dessauer (1981).Shochat and Dessauer (1981) used immuno-logical data to present the first challenge to theAlpha-Beta dichotomy, suggesting that Alphaand Beta species from the Caribbean are moreclosely related to each other than to other

Anolis.

Implicitly or explicitly, each of the abovestudies accepted most of the Etheridge and

Williams framework. Guyer and Savage (1986)broke from this perspective and attempted toapply a rigorous and modern phylogeneticapproach to Etheridges (1959) data. However,the study and resulting taxonomic rearrange-ment suggested by Guyer and Savage (1986;see also Savage and Guyer, 1989) wereseverely criticized by Williams (1989) andCannatella and de Queiroz (1989) for its

problematic data, methods, and classification(e.g., one of their proposed genera is notmonophyletic in their preferred tree). Guyerand Savage (1992) demonstrated that many oftheir results were robust to these criticisms,and their taxonomy has been followed bysome researchers, especially those working onCentral American forms (e.g., Kohler andMcCranie, 2001). Still, the Guyer and Savagetreatment, like more recent papers addressinglarge-scale Anolis phylogeny (e.g., Burnell and

Hedges, 1990; Poe, 1998; Jackman et al.,1999), suffers from its paucity of charactersand the number of taxa included. Guyer andSavage (1992) suggested generic rearrange-

ment for over 300 species based on an analysisof 27 species, and the most comprehensivemodern treatment, by Jackman et al. [1999],analyzed only about 15% of the species of

Anolis. Four of the Guyer and Savage generahave been shown to be paraphyletic (Poe,1998; Jackman et al., 1999), and recognition oftheir remaining genus Norops would necessi-tate an entirely new Anolis taxonomy. Thus,their generic-level taxonomy is not followedhere. However, unless otherwise noted, theirinfrageneric taxonomy (Savage and Guyer,1989) is employed because it is the mostcomprehensive treatment and its groups aresimilar or identical to those recognized byother authors (e.g., Burnell and Hedges,1990).

Like the majority of pre-Guyer and Savage(1989) studies, most modern work has focusedon phylogenies of smaller groups of Anolis(e.g., Hedges and Burnell, 1990; Poe, 1998;Giannisi et al., 2000; Schneider et al., 2001;Creer et al., 2001; Nicholson, 2002). Twoexceptions are Burnell and Hedges (1990),

who studied 50 Caribbean species usingallozyme data, and Jackman et al. (1999),

who studied 44 Caribbean and 9 mainland

species using DNA sequences. The Burnelland Hedges (1990) study included only 12informative loci, and so was unable to resolvemany relationships (a strict consensus of most-parsimonious trees from their data is un-resolved but for two clades). The Jackman etal. (1999) study found well-supported speciesgroups that were in accord with part of theEtheridge-Williams paradigm and a monophy-letic Beta section. But in spite of a relativelylarge number of data (861 parsimony-infor-

mative sites), deep relationships were onlyweakly resolved.

MATERIALS AND METHODS

Taxa

At least two species were included fromeach of the infrageneric groups of Savage andGuyer (1989) except for the laevis series andthe onca series, for which only one species wasincluded. More than two species were in-

cluded for all of the better-studied groups forwhich several species were available. Repre-sentatives of the formerly recognized anolinegenera Chamaeleolis, Chamaelinorops (synon-

38 HERPETOLOGICAL MONOGRAPHS [No. 18


3/53

ymized with Anolis by Hass et al., 1993), andPhenacosaurus (synonymized with Anolis byPoe, 1998; see also Etheridge, 1959) also wereincluded. This selection of species spans the

geographic and phenotypic diversity in Anolisand allows testing of the monophyly of almostall named genera and smaller groups ofAnolis.A total of 174 Anolis species was analyzed.

Species from the genera Enyalius, Urostro-phus, Anisolepis, Polychrus, and Leiocephaluswere included as outgroups. Frost and Ether-idge (1989) and Frost et al. (2001) found thesegenera to be close relatives of Anolis. Leioce-

phalus was included because it has beenscored for the DNA sequence data of Hasset al. (1993) and Jackman et al. (1999), re-spectively. Schulte et al. (2003) found addi-tional taxa to be close relatives of Anolis (e.g.,Liolaemus of their Trodidurinae*). Howeverthe interrelationships among putative Anolisrelatives were resolved with only weak sup-port (e.g., decay index of 1 supporting someTrodidurines as sister taxa to Anolis Leioce-

phalus). Under these conditions of weak sup-port, choice of outgroup becomes somewhatarbitrary. Although the outgroup sample ofthis paper is not comprehensive, all seven out-

group species included in this paper have beenfound to be close relatives of Anolis in otherphylogenetic analyses.

Characters

Non-morphological characters.Non-mor-phological data is included from chromo-somes, allozymes, DNA, and immunology.Three chromosome characters could be scoredfor 97 Anolis species and two outgroup species.Chromosome data were taken from Webster

(1974), Webster et al. (1972), and Gorman(1973). Allozyme data are incorporated fromGorman and Kim (1976) for the bimaculatusseries (nine species; six parsimony-informativeloci), Gorman et al. (1983) for the cristatellusseries (10 species; 16 loci), Lieb (1981) for the

gadovii and nebulosus series (7 species; 13loci), Yang et al. (1974) for the roquet series(7 species, 9 loci), and Burnell and Hedges(1990) for 44 Caribbean species (12 loci).These data were coded with the locus as the

character and the modal allele as the characterstate. Each study was coded separately. Al-though many of these studies analyze the sameloci, there is very little overlap in taxonomic

coverage. Data from Hedges and Burnell(1990) on the Jamaican grahami series andfrom Case and Williams (1987) for seven

Anolis species is omitted because of extensive

overlap in taxa and loci with the other studies.The cytochrome b mitochondrial DNAsequence data of Giannisi et al. (2000) wereincluded for the roquet series (7 species;97 parsimony-informative characters). Cyto-chrome b data also were included fromSchneider et al. (2001) for the bimaculatusseries (9 species, 193 parsimony-informativecharacters), although this could not be aligned

with the roquet series data because of the greatdivergence between these groups (see Poe[2000] for an example of long branch attractionbetween these groups). The 16S mitochondrialDNA sequence data from Hass et al. (1993) for25 Anolis species and two outgroups wereincluded (67 parsimony-informative sites).Jackman et al.s (1999) mitochondrial DNAsequence data from the NADH dehydroge-nase subunit 2 gene and five transfer RNAs

was combined with recent mitochondrial datafrom Glor et al. (2001), Jackman et al. (2002),Losos et al. (2003), and Harmon et al. (2003)for a total of 859 parsimony-informative char-

acters for 98 Anolis species and one outgroup.Nicholsons (2002) nuclear ITS data wereincluded for 42 Anolis species (317 parsimony-informative characters). A single characterreflecting immunological distance was codedfor eight species (Hass et al., 1993). Onlyparsimony-informative characters are analyzedand discussed in the character list (see below).

Morphological characters.Morphologicaldata were obtained by examination of museumspecimens and from the literature. Skull

characters were taken from Poe (1998) andnew characters were discovered during exam-ination of specimens. States for skull charac-ters were recorded from dry skulls. Postcranialosteological characters are from Etheridge(1959), and states for these were taken fromoriginal data sheets provided by RichardEtheridge, or from dry skeletons or dissectionof preserved specimens. External characters

were taken from literature sources and werediscovered during examination of specimens.

External character states were scored fromalcohol-preserved specimens, with some ex-ceptions discussed in the character descrip-tions.

2004] HERPETOLOGICAL MONOGRAPHS 39


4/53

Morphological character selection criteriafollowed the recommendations of Poe and

Wiens (2000). That is, I included charactersthat varied intraspecifically and continuously

(as well as discrete, intraspecifically invariantcharacters) if they appeared to vary indepen-dently. Intraspecifically varying characters

were coded with frequency coding (Wiens,1995) if two states were recognized, or with themodal state if more than two states wererecognized. The frequency approach has beenshown to perform well in simulation studies(Wiens and Servedio, 1997). Characters wereconsidered ordered if a morphocline ofstates was more believable than independentevolution of states. In the character descrip-tions (below), binary characters are describedin terms of alternative states a and z. InAppendix 2, matrix entry a corresponds to100% of specimens scored with the conditionlisted as state a in the character descriptions,entry z corresponds to 100% of specimensscored with the state listed as state z, andintermediate letters correspond to intermedi-ate frequencies of these states (see Wiens,1995).

Some quantitatively measured characters

were corrected for size effects using linearregression (see below). In these cases, severaltransformations of independent and depen-dent variables were attempted and the trans-formation(s) (or nontransformation) that bestmet regression assumptions was used. Mensu-ral characters were coded with Thieles (1993)gap-weighting method, with extreme condi-tions coded as ordered states a and z, re-spectively, and interrmediate measurementscoded as intermediate letters. Outgroup scores

were excluded from initial assignment of gap-weighted states, so that extreme conditions inoutgroups would not affect weighting of statechanges between ingroup species (e.g., Enya-lius iheringi has a relatively shorter head thanany Anolis species). Outgroup species wereassigned states that corresponded to theconditions in Anolis that were closest to theirconditions (e.g., Enyalius iheringi was as-signed the same state as the Anolis speciesthat possesses the relatively shortest head).

The considerable size of this morphologicaldata matrix (91 characters3 181 taxa) entreatsseveral difficulties with characterization of

variation and accurate scoring of characters.

To minimize bias in scoring, I attempted todefine states that could be scored unambigu-ously relative to observable landmarksi.e.,contact vs. noncontactrather than rely on

my own subjective impression of differencesi.e., large vs. small. After initial scoring, Irechecked almost every morphological char-acter state in the matrix relative to my own datasheets, published observations, and additionalspecimens, to confirm codings for particularcharacters. Thus the number of specimensexamined is much larger than the number inAppendix 1. In many cases I re-scored the samespecimens to ensure that my initial observa-tions were repeatable. However, it of course

was impossible to compare all species andspecimens relative to all others simultaneously.

The morphological character matrix for thispaper is an updated version of the matrix inPoe (2000). Three characters were omitted(Numbers 40, 77, 78) from Poe (2000) becauseof characterization difficulties that were dis-covered upon examination of additionalspecimens. Also, several matrix entries werecorrected due to information from additionalspecimens examined.

One-hundred seventy-three of 174 Anolis

species and all 7 outgroup species were scoredfor most of the 50 external characters (Anolis[Phenacosaurus] nicefori, included for itsDNA data, was scored for only a few charac-ters), 162 Anolis species and 7 outgroupspecies were scored for most of the 7 post-cranial osteological characters, and 162 Anolisspecies and 6 outgroup species were scored formost of the 38 skull characters. Over 2000specimens were examined for this study.Specimens examined for initial scoring are

listed in Appendix 1.Characters listed below are described if theyhave not appeared before in a phylogeneticanalysis ofAnolis. If they have been describedpreviously, the reference is given. Manycharacters, such as keeling of scales, are oftraditional importance for Anolis (indeed, forlizard) systematics, and it is difficult to find theabsolute first reference where this characterhas been used. Most of such public domaincharacters are listed without reference, with

no claim implied for proprietary originality.Character state assignments for each speciesare listed in Appendix 2 for the morphologicalcharacters (Nos. 191). States for other



5/53

characters have been published previously;references for these are given.

Several characters in addition to those usedin this paper were scored but discarded due to

characterization difficulties. These may beuseful for analyses of smaller groups ofAnolis,or of all of Anolis under some different char-acterization than I attempted. Some of thesecharacters include the relative lengths of thefourth and fifth toes, color of dorsum and

venter, color of dewlap, keeling of limbs,sleeping posture (limbs flexed in some species,straight in others), voice (some anoles squeak

when handled), scalation posterior to supra-ciliaries (enlarged regular rows in somespecies, jumbled scales in others), structureof ilium (Lazell, 1969), middorsal crest, struc-ture of teeth, lateral body scalation (homoge-neous in some species, mixture of large andsmall scales in others), blue pigment at edge ofmouth (Myers, 1971), number of toe lamellae,display behavior, elongate nasal appendage(Williams, 1979), and prehensality of the tail.

Character Descriptions

1. Maximum male snout-to-vent length 38mm (a); 188 mm (z). Gap weighted. Poe

(1998). Data are from the Anolis handlist(Williams et al., 1995).

2. Ratio of maximum female snout-to-ventlength to maximum male snout to ventlength 1.06 (a); 0.57 (z). Gap weighted. Poe(1998). Data are from the Anolis Handlist(Williams et al., 1995).

3. Length of thigh short (a); long (z). Gapweighted. The leg was measured from theventral midline of the specimen lateral tothe knee, with the limb aligned perpendic-

ular to the body and bent at the knee.Bivariate plots of snout-to-vent length vs.femoral length demonstrated that thischaracter is correlated with size. In orderto correct for size, thigh length wasregressed on (independent variable) SVLusing natural-log-transformed mean valuesfor each species as data points. Residual

values of these regressions for each specieswere input as raw data for gap weighting.Although this character has been suggested

to be an ecomorph character in someCaribbean forms (e.g., Williams, 1983;Losos et al., 1998) and thus subject toconsiderable convergence, it remains a use-

ful and widely-used systematic character inAnolis from other areas (e.g., CentralAmerica; Savage and Talbot, 1978).

4. Length of head short (a); long (z). Gap

weighted. Some Anolis species (e.g., A.longiceps) have long snouts whereas others(e.g., A. capito) have short snouts. Headlength was measured from the anterior ofthe ear opening to the tip of the snout.Head length was found to be correlated

with SVL. Head length was regressed onSVL using natural-log-transformed mean

values for each species as data points.Residuals were gap-weighted. This charac-ter also is a likely ecomorph character(Williams, 1983; Losos et al., 1998). How-ever, many species (e.g., A. porcatus and A.allisoni) apparently share similar headshapes owing not to convergence but ratherto common ancestry.

5. Width of head narrow (a); broad (z). Gapweighted. Anolis species may have broad(e.g., A. cybotes) or narrow (e.g., A.

sheplani) heads. Head width was measuredbetween the posteroventral corners of theorbits (generally the widest part of theskull). Head width was regressed on SVL

using mean raw values for each species asdata points. Raw values were found tobetter meet regression assumptions thantransformed values. Residuals for eachspecies were gap weighted. Surprisingly,this character does not appear to bestrongly correlated with character 4 (headlength) after correction for size.

6. Height of ear small (a); large (z). Gapweighted. Ear size varies strikingly inAnolis. Some species (e.g., A. vermiculatus)

have a prominent oval tympanum, whereasin others (e.g., A. darlingtoni) the earopening is barely visible. This character

was found to be correlated with size. Inorder to correct for size, ear height wasregressed on head length using natural-log-transformed mean values for each speciesas data points. Residuals for each species

were gap weighted. Differences in earshape also occur, but these were difficultto characterize.

7. Interparietal scale large (a); about equal tosurrounding scales (z). Gap weighted. Insome Anolis species (e.g., A. argenteolus),the interparietal is several times the size of



6/53

the surrounding scales. In others (e.g., A.fraseri), the interparietal is approximatelythe same size as the surrounding scales. Oneach specimen examined, I measured the

length of the interparietal and the length ofthe second-largest scale in lateral contactwith the interparietal, excluding the supra-orbital semicircles. Mean values of thesemeasurements were compiled for eachspecies, natural-log transformed, and ana-lyzed with linear regression using interpar-ietal length as the independent variable.Residuals were gap weighted.

8. Length of tail. Gap weighted. Data weretaken from Williams et al. (1995), who

recognized five states: Tail length aboutequal to SVL (a); about 1.5 times SVL (h);about two times SVL (m); about 2.5 timesSVL (s); more than 2.8 times SVL (z). Gap-

weighted. Species reported to vary for thischaracter were assigned the median score.As with characters 3 and 4, this characterhas been suggested to be an ecomorphcharacter for some species (e.g., Williams,1983; Losos et al., 1998). However, thereare obvious cases where similarities in tail

length are due to phylogeny rather thanadaptive response (e.g., similarly-sized tailsin A. clivicola and A. alutaceus), so I includethis character in phylogenetic analysis.

9. Toepads overlapping first phalanx (0); notdistinct from first phalanx (1); absent (2).Unordered. Williams et al. (1995). State 2 isadded to Poes (1998) character 2.

10. Enlarged postanal scales present in males(a); absent in males (z). Frequency coded.Data and character are from the Anolis

Handlist (Williams et al., 1995).11. Row of large spinose middorsal caudalscales separated by smaller smooth scalesabsent (a); present (z). Frequency coded.This character is a modification of Poes(1998) character 12 (see also Hedges et al.,1989) to note the condition on the tail(derived condition seen in A. insolitus, A.

sheplani, and A. placidus) rather than onthe body (derived condition seen only in

A. sheplani and A. placidus).

12. Tail crest absent (a); present in largestadult males (z). Frequency coded. Pres-ence and size of tail crest appears to belargely size and sex dependent. Thus, the

character is evaluated only in large adultmales.

13. Number of rows of enlarged middorsalscales 04 (a); 5 or more (z). Frequency

coded. Most species of Anolis have eithertwo rows of middorsal scales slightly largerthan the surrounding scales (e.g., A.cybotes) or a middorsal band of 7 or moreabruptly enlarged scales (e.g., A. semi-lineatus).

14. Each ventral scale is bordered posteriorlyby two scales (a); by three scales (z).Frequency coded. State a includesspecies with rectangular ventrals in trans-

verse rows (e.g., A. carolinensis). State zincludes species with the posterior borderof the ventral rounded or pointed suchthat scales appear to be in diagonaloverlapping rows (e.g., A. cybotes).

15. Base of tail round (a); laterally compressed(z). Frequency coded. On each specimen,cross-sectional height and width of the tail

were examined at the point where theknee would reach the tail if the leg werefolded back. If the height of the tail isgreater than the greatest width, state z isassigned.

16. Male dewlap extends posterior past arms(0); to arms or shorter (1); absent (2).Ordered. This character and data weretaken directly from the Anolis Handlist(Williams et al., 1995; supplemented byexamination of specimens) however cod-ing is slightly different. Williams et al.(1995) characterized Anolis dewlaps aslarge if the dewlap extends posteriorlypast the arms, intermediate if it reachesthe arms, small if it does not reach the

arms, and absent if nonexistent. Becausethe arms are a convenient landmark forstate delimitation and intermediate andsmall are morphologically very similarconditions in Anolisthere is more vari-ation within the state large than betweenthe other statesI lump the middle twoconditions into state (1). Species with onlylarge scores in Williams et al. (1995) areassigned state (0); other species areassigned state (1), except for two species

(A. bartschi, A. vermiculatus) for whichdewlap is absent in males, which areassigned state (2).

17. Female dewlap extends posterior past



7/53

arms (0); to arms or shorter (1); absent (2).Ordered. Many female Anolis speciescompletely lack a dewlap. Coding is asfor male dewlap.

18. Tail uniformly patterned (a); base of tailpurple, posterior part green or brown(z). Frequency coded. Underwood and

Williams (1959).19. Mean number of dorsal scales in 5%of SVL

2.5 (a); 17 (z). Gap weighted. Number ofscalesin5%ofSVLiscountedjustlateraltothe dorsal midline. Many authors (e.g.,Lazell, 1972) have used some measure ofscale size as a systematic character.

20. Mean number of ventral scales in 5% of

SVL 2.75 (a); 14.3 (z). Gap weighted.Number of scales in 5% of SVL is countedlongitudinally on the posterior belly area.

21. Scales on dewlap in rows of single scales(a); with at least one double row (z).Frequency coded. Species with scattereddewlap scales or dewlaps mostly lackingscales are coded as ?.

22. Middorsal caudal scale rows single (a);double (z). Frequency coded. Williams etal. (1970).

23. Axillary pocket absent (a); deep, tubelike(z). Frequency coded.24. Scales of midnuchal area similar to mid-

dorsal scales (a); in continuous row ofbulbous scales distinct from dorsal scales(z). Frequency coded. Anolis darlingtoniand A. insolitus possess state z.

25. Transparent scales in lower eyelid absent(a); present (z). Frequency coded. Wil-liams and Hecht (1955). Anolis argenteo-lus has two transparent scales in the lower

eyelid, A. lucius has three, and otherAnolis have none.26. Mental scale partially divided (a); com-

pletely divided (z). Frequency coded. Insome anole species (e.g., A. griseus) themental is longitudinally split posteriorlybut not anteriorly (state a); in others (e.g.,

A. carolinensis) this division is complete(state z).

27. Mental scale broader than rostral scale (a);rostral broader than mental (z). Frequency

coded. The comparison is made alongthe rim of the mouth. In most species (e.g.,A. sagrei), the mental extends fartherposteriorly along the mouth than the

rostral (state a) whereas in others (e.g.,A. baleatus) the rostral is broader (state z).

28. Subocular scales and supralabial scales incontact (a); separated by one or more rows

of scales (z). Frequency coded.29. Mean number of scales across the snout2.5 (a); 19 (z). Gap weighted. This isa minimum count, made between thesecond canthals. Poe (1998) used median

values from the Anolis Handlist (Williamset al., 1995) for this character. Specimens

were scored for the data for this paper.30. Mean number of postmental scales 3.25

(a); 9.75 (z). Gap weighted. Poe (1998)used median values from the Anolis

Handlist (Williams et al., 1995) for thiscount. Specimens were scored for the datafor this paper.

31. Posterior border of mental scale concave(a); straight or convex (z). Frequencycoded.

32. Supraorbital semicircles separated by oneor more rows of scales (a); in contact (z).Frequency coded.

33. Preoccipital scale absent (a); present (z).Frequency coded. The preoccipital is an

enlarged scale directly anterior to theinterparietal.34. Middorsal scales of the snout not in

regular pattern (a); arranged in twoparallel rows that extend from the levelof the second canthals to the nares (z).Frequency coded.

35. Posterodorsal edge of rostral smooth (a);cleft (z). Frequency coded.

36. Anteromost aspect of rostral scale is evenwith lower jaw (a); overlaps lower jaw (z).Frequency coded. Although extensive in-terspecific variation occurs in this charac-ter and scoring for most specimens isstraightforward, continuous variation pre-cludes erection of a simple character stateboundary. Operationally, I compared bor-derline specimens to USNM (NationalMuseum of Natural History) 347286 (A.oxylophus). If overlap was equal to orgreater than that seen on this specimen,state z was assigned.

37. Color of iris dark brown (0); yellow (1);blue or grey (2); green (3) red. Unordered.Data for this character were taken fromthe literature, mainly from Schwartz and



8/53

Henderson (1991) and the Anolis Handlist(Williams et al., 1995).

38. Modal number of supraciliary scales zero(0); one (1); two (2); three (3). Ordered.

Supraciliaries are elongate scales along thedorsal rim of the orbit.39. Modal nasal scale type: anterior nasal in

contact with rostral (0); divided anteriornasal in contact with rostral (1); circum-nasal separated from rostral by one scale,not in contact with supralabial (2); externalnaris separated from rostral by two scales,not in contact with supralabial (3); externalnaris separated from rostral by three ormore scales, not in contact with supra-labial (4); circumnasal in contact withrostral (5); circumnasal in contact withsupralabial, separated from rostral by onescale (6); circumnasal in contact withsupralabial, separated from rostral by twoor more scales (7). Most of these con-ditions are described and figured in

Williams et al. (1995). States 3 and 4generally involve circumnasal scales butoccasionally incorporate an anterior nasal.

Each species is coded with its modal state.Transformations are weighted according

to the following step matrix. Change costswere multiplied by 1000 in the analyses tomaintain comparability with other charac-ters (see below).

40. Keeling of dorsals, ventrals, supradigitals,and head scales. Unordered. Even weakkeeling was scored as keeled. I originallytook data on the keeling of each of theseregions with the expectation that each

would be a separate character. However itbecame apparent in the course of this

study that there are strong constraints onwhich combinations of scalation are pos-sible. For example, 89% of Anolis speciesin this study display one of only four of the

sixteen possible combinations of keeling(states 0, 1, 2, and c; see below). Thisdistribution is very different from thatexpected if there is random covariation

(results not shown). For the purposes ofthis study, each unique combination ofscalation is treated as a separate state. Thefollowing states are listed in order of

whether keels (k) or smooth scales (s) areobserved on head scales, ventrals, dorsals,and supradigitals, respectively. For exam-ple, sssk means that all surfaces exceptsupradigitals are smooth. (0) kkkk; (1) ssss;(2) kskk; (3) kksk; (4) kkks; (5) kkss; (6)ksks; (7) kssk; (8) ksss; (9) skkk; (a) skks; (b)sksk; (c) sskk; (d) sssk; (e) ssks; (f) skss.

41. Scales in supraocular disc vary continu-ously in size and are bordered medially byan unbroken row of small scales (0); varycontinuously in size and are borderedmedially by an incomplete row of smallscales (1); with one to three abruptlyenlarged scales and bordered medially byan unbroken row of small scales (2); withone to three abruptly enlarged scales andbordered medially by an incomplete rowof small scales (3); about equal in size (4).

Species are assigned their modal condi-tion. Abruptly enlarged is operationallydefined as a difference in size by a factor oftwo. Thus, species with states 2 and 3display two discrete size classes of scales inthe supraocular disca group of one ormore scales that are all at least twice thesize of the next largest scale, and a series ofsmaller scales that vary in size continu-ously down to the smallest scales present.

Character changes were weighted accord-ing to the following step matrix. Changecosts were multiplied by 1000 in theanalyses to maintain comparability withother characters (see below).

42. Lining of throat light (a); black (z).Frequency coded. Character and dataare from Underwood and Williams (1959).

43. Fold of skin extending over dorsal rim of

0 1 2 3 4 5 6 7

[0] 0 1 1 1 2 2 2 2[1] 1 . 1 1 2 2 2 2[2] 1 1 . 1 2 1 1 2[3] 1 1 1 . 1 2 2 1[4] 2 2 2 1 . 3 2 1[5] 2 2 1 2 3 . 1 2[6] 2 2 1 2 2 1 . 1[7] 2 2 2 1 1 2 1 .

0 1 2 3[0] . 1 1 2[1] 1 . 2 1[2] 1 2 . 1[3] 2 1 1 .



9/53

ear opening absent (a); present (z).Frequency coded.

44. Modal number of enlarged sublabialscales zero (0); one (1); two or more (2).

Ordered. Sublabials are scales in the gularregion, medial and running parallel to theinfralabials (e.g., Williams et al., 1995).These are counted back from the mental,and are considered to be enlarged if atleast twice the size of any medial scale.

45. Frontal depression present (a); absent, topof snout is flat (z). Frequency coded.Although there are clear differencesbetween several species in degree offrontal depression (e.g., A. garmani hasa very flat head, A. reconditus has a pro-nounced depression), I found it impossi-ble to erect a satisfactory objectivelyrecognizable character state boundary forthis character. Thus, any borderline spe-cies are assigned ?.

46. Interparietal scale separated from supra-orbital semicircles by one or more rows ofscales (a); in contact with supraorbitalsemicircles (z). Frequency coded.

47. Modal postxiphisternal inscriptional ribformula 4: 3 (0); 5: 1 (1); 4: 2 (2); 5: 0 (3);

4: 1 (4); 3: 2 (5); 4: 0 (6); 3: 1 (7); 2: 2 (8); 1:3 (9); 2: 1 (a); 5: 2 (b). Etheridge (1959).Data is from original data sheets providedby Richard Etheridge. Character changes

are weighted according to the followingstepmatrix, adapted from Jackman et al.

(1999) Change costs were multiplied by1000 in the analyses to maintain compa-rability with other characters (see below).

48. Modal number of sternal ribs two (0);

three (1); four (2). Ordered. Etheridgeand de Queiroz (1988). Data is fromEtheridge and de Queiroz (1988), Frostand Etheridge (1989), and Williams

(1989).49. Caudal vertebrae are Alpha type (0); Betatype (1); Chamaelinorops type (2); Basi-liscus type (3); Sceloporus type (4). Ether-idge (1959). Unordered. Data is from datasheets provided by Richard Etheridge for

Anolis and from Etheridge and de Queiroz(1988) for outgroups.

50. Interclavicle arrow-shaped (a); T-shaped(z). Frequency coded. Etheridge (1959).Data is from data sheets provided by

Richard Etheridge and from Williams(1989).51. Modal number of presacral vertebrae 24

(0); 23 (1); 22 (2). Ordered. Etheridge(1959). Data is from original data sheetsprovided by Richard Etheridge.

52. Modal number of lumbar vertebrae three(0); four (1); five (2); six (3). Ordered.Etheridge(1959).Dataisfromoriginaldatasheets provided by Richard Etheridge.

53. Modal number of caudal vertebrae ante-

rior to first autotomic vertebrae eleven (0),ten (1), nine (2), eight (3), seven (4), six(5), five (6). Ordered. Etheridge (1959).Data is from original data sheets providedby Richard Etheridge. Species that lackautotomy were coded ?.

54. Caudal autotomy septa present (a); absent(z). Frequency coded. Etheridge (1959).Data is from Etheridge (1959) and Wil-liams (1989).

55. Supraoccipital cresting continuous acrosssupraoccipital (0); lateral processes dis-tinct from supraoccipital crest (1); singlenarrow central process (2). Unordered.Species are assigned their modal condi-tion. State 2 is added to character 105 ofPoe (1998).

56. Dorsal surface of skull smooth (a); rugosewith bony tubercles (z). Frequency coded.In most Anolis the surface of the skull issmooth. But in some species (e.g., A.equestris) the rugosity is developed into

hard pustulate tubercles. Individual speci-mens were assigned whatever conditionpredominates on the surface of the skull.Species with pronounced wrinkling but

0 1 2 3 4 5 6 7 8 9 a b[0] . 2 1 3 2 2 3 3 3 3 4 1[1] 2 . 1 1 1 2 2 2 3 4 3 1[2] 1 1 . 2 1 1 2 2 2 3 3 1[3] 3 1 2 . 1 2 1 2 3 4 3 2[4] 2 1 1 1 . 1 1 1 2 3 2 2

[5] 2 2 1 2 1 . 2 1 1 2 2 2[6] 3 2 2 1 1 2 . 1 2 3 2 3[7] 3 2 2 2 1 1 1 . 1 2 1 3[8] 3 3 2 3 2 1 2 1 . 1 1 3[9] 3 4 3 4 3 2 3 2 1 . 2 4[a] 4 3 3 3 2 2 2 1 1 2 . 4[b] 1 1 1 2 2 2 3 3 3 4 4 .



10/53

not pustules (e.g., A. darlingtoni) werecoded as intermediate (state n).

57. Parietal crests form a trapezoid (0); V (1);Y (2); Y with parietal spur (3). Ordered.

Etheridge (1959; see also Poe [1998]).Each species is coded with its modal state.58. Anterolateral corners of parietal crests

reach posterolateral corners of frontal(a); reach medial to posterolateral cornersof frontal (z). Frequency coded.

59. Parietal casque absent (a); present (z).Frequency coded. In some species (e.g.,

A. equestris) the roof of the parietalextends posterolaterally, shelf-like, overthe supratemporal processes of the parie-tal to form a casque (state z).

60. Pineal foramen at parietal/frontal suture(a); in parietal (z). Frequency coded.Etheridge (1959).

61. Supratemporal processes leave supraocci-pital exposed above (a); extend oversupraoccipital (z). Frequency coded.Etheridge (1959; see also Poe [1998]).

62. Postfrontal present (a); absent (z). Fre-quency coded Poe (1998).

63. Prefrontal contacts nasal (a); is separatedfrom nasal by frontal and maxilla (z).

Frequency coded Poe (1998).64. Frontal sutures only with nasals anteriorly

(0); is separated from nasals by opengap (1); contacts premaxilla and nasalsanteriorly (2). Unordered. Poe (1998).State 2 is added to Poes (1998) charac-ter 94. Species are assigned their modalstate.

65. Parallel crests extending longitudinallydown nasals from frontal to nares absent(a); present (z). Frequency coded.

66. Anterior edge of nasal forms posteriorborder of naris (a); does not reach naris(z). Frequency coded. Species for whichthe nasal forms a partial medial border tothe naris are coded with the intermediatestate (n).

67. Dorsal process of jugal terminates onposterior or medial aspect of postorbital(a); on lateral aspect of poorbital (z).Frequency coded. Poe (1998).

68. Contact between jugal and squamosal

absent (a); present (z). Frequency coded.Poe (1998).69. Posteroventral corner of jugal is anterior

to posterior edge of jugal (a); posterior to

posterior edge of jugal (z). Frequencycoded. In species with state z (e.g., A.cybotes), the posterior edge of the jugalis concave; species with state a (e.g.,

A. limifrons) have a straight or convexposterior jugal.70. Epipterygoid contacts parietal (a); does

not contact parietal (z). Frequency coded.Poe (1998).

71. Pterygoid teeth present (a); absent (z).Frequency coded. Etheridge (1959; seealso Poe [1998]).

72. Lateral edges of vomer smooth (a); withposteriorly directed lateral processes (z).Frequency coded. This character is similarto Poes (1998) character 102 (palatine-

vomer suture). However in this paper Icombine states 0 (transverse suture, noprocesses) and 1 (posterolateral suture, noprocesses) into state (a) because the widersample of taxa examined for this paperdemonstrated continuous variation thatmade objective distinction between states0 and 1 of Poe (1998) impossible.

73. Maxilla extends posteriorly to ectoptery-goid (a); beyond ectopterygoid (z). Fre-quency coded. Poe (1998).

74. Basipterygoid crest absent (a); present (z).Frequency coded. Poe (1998). Jackman etal. (1999) called this structure a parabasi-sphenoid.

75. Quadrate lateral shelf absent (a); present(z). Frequency coded. Poe (1998).

76. Black pigment on skull absent (a); presentover most bones on the dorsal surface ofthe skull (z). Frequency coded. Specimens

were scored as state z if a majority ofdorsal skull bones had black pigment;

other conditions were scored as state a.77. Premaxilla overlaps nasals laterally or isflush with them (a); nasal overlapslateral edge of premaxilla (z). Frequencycoded.

78. Posterior of skull slopes superiorly or is flat(a); slopes inferiorly (z). Frequency coded.If the posterior edge of the parietal isinferior to its anterior edge, a score of (z) isassigned.

79. Crenulation along lateral edges of parietal

absent (a); present (z). Frequency coded.Poe (1998).80. Parietal roof flat (a); convex (z). Frequency

coded. Poe (1998).



11/53

81. Posteriormost tooth is at least partiallyposterior to anterior mylohyoid foramen(a); posteriormost tooth is at least partiallyanterior to anterior mylohyoid foramen

(m); posteriormost tooth is completelyanterior to anterior mylohyoid foramen(z). Coding follows the frequency analysisof Poe (1998).

82. Angular process of articular present, large(a); reduced or absent (z). Frequencycoded. Williams (1989; see also Poe[1998]).

83. Posterior suture of dentary pronged (a);blunt (z). Frequency coded. Poe (1998).

84. Anteriormost aspect of posterior border ofdentary is anterior to mandibular fossa (a);

within mandibular fossa (z). Frequencycoded. Poe (1998).

85. Splenial present, large (0); absent (1);present as anteromedial sliver (2). Un-ordered. Etheridge (1959; see also Poe[1998]). Each species is coded with itsmodal state. Although some Anolis speciesin the bimaculatus series demonstratea fourth statesplenial present as ventralfragmentnone of these species has thisstate as its modal condition.

86. Anteromedial process of coronoid extendsanteriorly (a); ventral aspect of anterome-dial process projects posteriorly (z). Fre-quency coded. Poe (1998).

87. Surangular foramen completely in suran-gular (a); bordered laterally by dentary (z).Frequency coded. Poe (1998).

88. Coronoid labial process absent (a); present(z). Frequency coded. Poe (1998; see alsoEtheridge and de Queiroz [1988]).

89. Posterolateral aspect of coronoid termi-

nates anterior to supra-angular foramen(a); extends into or beyond supra-angularforamen (z). Frequency coded.

90. Jaw sculpturing in large adult males absent(0); Chamaeleolis type (1); krugi type (2);cristatellus type (3); cybotes type (4);

wrinkled (5). Unordered. See Etheridge(1959) for descriptions of these conditions.Data for this character is from personalobservation, supplemented by an unpub-lished manuscript by Ernest Williams and

Susan Case provided by Ernest Williams.Each species is coded with its modal state.91. Angular bone present (a); absent (z).

Frequency coded. Etheridge (1959).

92104. Allozyme loci. Lieb (1981).105116. Allozyme loci. Burnell and Hedges

(1990).117183. Ribosomal RNA sequence sites

from the 16S region. Hass et al.(1993). Sequences were down-loaded form Genbank and alignedusing Clustal W (Thompson et al.,1994) and by eye.

184189. Allozyme loci. Gorman and Kim(1976).

190205. Allozyme loci. Gorman et al. (1983).206. Number of macrochromosomes 12 (0); 14

(1); 16 (2); 17 (3); 18 (4); 19 (5); 20 (6); 22(7); 24 (8). Unordered. Webster (1974);

Webster et al. (1972); Gorman (1973).207. Number of microchromosomes 8 (0); 10

(1); 12 (2); 14 (3); 16 (4); 18 (5); 20 (6); 22(7); 24 (8). Unordered. Webster (1974);

Webster et al. (1972); Gorman (1973).208. Sex chromosomes heterogeneity absent

(0); XY (1); XXY (2). Unordered. Webster(1974); Webster et al. (1972); Gorman(1973).

209217. Allozyme loci. Yang et al. (1974).

218. Immunological distance. Hass et al.s(1993) table 2 of immunological distan-

ces was entered as a step matrix (scaledsuch that maximum change was atequivalent parsimony cost to a changeat a DNA site):

1: A carolinensis, 2: A. cuvieri, 3: A.cybotes, 4: A. valencienni, 5: A. crista-

tellus, 6: A. evermanni, 7: A. bimaculatus,8: A.roquet (this species was assigned thestate for A. extremus, a species notincluded here that is sister to A. roquetbased on chromosome [Gorman and

Atkins, 1967] and DNA [Giannisi et al.,2000] data).

219411. Mitochondrial DNA sequence sites

1 2 3 4 5 6 7 8

[1] . 712 1000 806 849 813 777 784[2] 712 . 619 525 504 619 633 705[3] 1000 619 . 820 655 597 734 763[4] 806 525 820 . 424 388 604 755[5] 849 504 655 424 . 223 583 683[6] 813 619 597 388 223 . 489 727[7] 777 633 734 604 583 489 . 791[8] 784 705 763 755 683 727 791 .



12/53

from the cytochrome b region.Schneider et al. (2001). Sequences

were downloaded from Genbankand aligned using Clustal W

(Thompson et al., 1994) and by eye.412508. Mitochondrial DNA sequence sitesfrom the cytochrome b region.Giannasi et al. (2000). Sequences

were downloaded form Genbankand aligned using Clustal W(Thompson et al., 1994) and byeye. Their Anolis oculatus outgroupsequence was excluded because thissequence has been shown to sufferfrom long branch attraction relativeto the roquet series (Poe, 2000).

509820. Nuclear DNA sequence sites fromthe internal transcribed spacer re-gion (Nicholson, 2002). Alignmentused was provided by Nicholson(personal communication).

8211680. Mitochondrial DNA sequencesites from NADH dehydrogenasesubunit 2 and five transfer-RNAgenes. Jackman et al. (1999). Gloret al. (2001). Losos et al. (2003).Harmon et al. (2003). Sequences

were provided by Rich Glor. I usedthe alignment and site-exclusionscheme recommended by Glor(personal communication).

Analyses

I performed parsimony analyses of themorphological data in combination with thechromosome, DNA, immunological, and allo-zyme data using PAUP* 4 (Swofford, 2000).

Parsimony was selected as an optimalitycriterion because it performed well in simula-tion studies of various types of characters (e.g.,DNA, Huelsenbeck and Hillis, 1993; intra-specifically varying characters, Wiens andServedio, 1997) and because the evolutionaryprocess assumptions of this method seemreasonable for data matrices such as this onethat incorporate several diverse types of data.Furthermore, most other optimality criteria(e.g., likelihood-based methods) are not yet

suitable for matrices like this one that includeweighted, ordered, and step-matrix characters.More than 2000 searches for optimal trees

were performed. For each search, taxa were

added in random order to construct an initialtree, and the tree-bisection-reconnection(TBR) algorithm was used to find optimaltrees for that replicate. Several additional

searches were performed with an optimal tree(rather than a random-addition tree) as thestarting tree and retention of trees slightlylonger than the optimal tree for swapping; thisapproach allowed discovery of shorter treesthan those found during initial rounds ofswapping.

With some exceptions (noted in the charac-ter descriptions), characters were weightedsuch that a change between extreme statesoccurred at a parsimony cost of 1000. Binarymorphological characters scored with frequen-cy coding were coded with 26 ordered states(az), with change between adjacent states atacostof40(5 1000/25). Thus, change from 0 to100% fixation costs 1000 (see Wiens, 1995).Gap-weighted characters were scaled such thatchanges between extreme values (a , z) cost2000. This scaling was adopted rather thana scaling of 1000 because the latter coding

would effectively downweight gap-weightedcharacters relative to all other characters. Thisdownweighting would occur because taxa are

scored on a continuous (mensural) scale andonly the extreme values are at maximumparsimony cost (see Thiele, 1993). This meansthat changes between any two taxa exceptbetween the two (or few) taxa with the highestand lowest measurements costs less than themaximum. If the maximum is set at 1000, then

virtually all changes for these characters willcost less than change between non-gap codedcharacters, and these characters will contribute

very little to tree construction.

Support for individual clades was evaluatedusing the bootstrap (Felsenstein, 1985b). Forbootstrap analysis, 100 replicate matrices ofthe same size as the original were constructedby sampling characters with replacement.Replicate matrices were analyzed with parsi-mony using one round of random taxonaddition and TBR branch swapping, andresulting trees were summarized by mappingbootstrap values greater than 50% onto theoptimal parsimony tree.

For comparison, I also performed a parsi-mony analysis of the morphological data aloneusing the same tree-searching strategies de-scribed for the combined matrix.



13/53

RESULTS AND DISCUSSION

Optimal Tree

Analysis of the entire matrix producedone optimal tree of length 16834634 withconsistency index (CI; Kluge and Farris, 1969)of 0.20 and retention index (RI; Farris, 1989)of 0.46 (Figs. 14). Very few clades are well-supported. This result is unsurprising, as manytaxa were scored for only the 91 morphologicalcharacters, or even fewer for those species for

which a dry skeleton was unavailable (e.g.,A. roosevelti). Although these morphological

characters appear to be adequate phylogeneticmarkers (RI for morphological characters onthe combined tree is 0.54), there simply wasnot enough of them to resolve the relationships

of 174 taxa with great confidence. An analysisof a more restricted sample of taxa (i.e., those

that are completely scored) undoubtedlywould produce higher bootstrap levels; how-ever this tactic seems unlikely to result ina better estimate of relationships (e.g., Gauth-ier et al., 1988).

Many clades in the optimal tree areconcordant with previously-held ideas ofanole relationships (see below). For purposesof mapping character changes, this tree wasrooted on the branch leading to Leiocepha-lus. Rather than review character support for

each of the 179 clades in this tree, I only dis-cuss support for clades that are likely to be ofspecial interest, such as previously named or

well-studied groups and clades that are con-

FIG. 1.Overview ofAnolis phylogeny recovered in this paper. NLA 5 Northern Lesser Antilles. PR 5 Puerto Rico.Hisp 5 Hispaniola.



14/53

gruent with geographic breaks. I list all mor-phological support for discussed clades. SeeAppendix 2 for a complete list of morphologi-cal support for all clades under acceleratedtransformation optimization (ACCTRAN;Swofford, 2000). A full list of all changes

(i.e., including DNA, allozyme, and chromo-some data, under delayed or acceleratedtransformation optimization) is available onrequest.

The monophyly ofAnolis (node 354 of Fig.2; 70% bootstrap) is supported by 18 un-ambiguous synapomorphies (11 morphologi-cal), including greater sexual size dimorphism(character 2: a fi e), longer snout (4: e fi i),smaller ear (6: v fi r), larger interparietal

(7: vfi

p), presence of toepads (9: 2fi

0),reduced size of dorsal scales (19: nfi s), morepostmentals (30: a fi f), Alpha type caudal

vertebrae (49: 3 fi 0), prefrontal bone

FIG. 2.Phylogenetic estimate for basal Anolis. Species names are followed by series groupings according to Savageand Guyer (1989). Numbers above clades are bootstrap values. Numbers to the right of splits indicate nodes for purposesof listing apomorphic support in Appendix 3.



15/53

separated from nasal (63: a fi i), lengtheneddentary (84: ffi z), and loss of angular (91:a fi z).

Anolis occultus, the most basal species, has

been difficult to place due to its confusingmosaic of plesiomorphic and derived charac-ters. Its placement here requires extensivehomoplasy; 16 unambiguous morphological

autapomorphies and 131 unambiguous DNAautapomorphies occur on the A. occultusbranch.

The basal clade ofAnolis (node 352) that is

sister to the Caribbean Anolis radiationincludes the South American giant anoles(latifrons series) and punctatus- and tigrinus-group species, the well-studied Southern

FIG. 3.Phylogenetic estimate for Caribbean Alpha Anolis. Species names are followed by series groupings accordingto Savage and Guyer (1989). Numbers above clades are bootstrap values. Numbers to the right of splits indicate nodes forpurposes of listing apomorphic support in Appendix 3.



16/53

FIG. 4.Phylogenetic estimate for Beta Anolis. Species names are followed by series groupings according to Savageand Guyer (1989). Numbers above clades are bootstrap values. Numbers to the right of splits indicate nodes for purposes

of listing apomorphic support in Appendix 3.



17/53

Lesser Antillean roquet group (node 351), andthe anoles formerly placed in the genusPhenacosaurus. All these species have beenconsidered primitive or basal since Etheridges

(1959) work. The roquet group clade (node351; 74% bootstrap) is supported by sixunambiguous morphological synapomorphies:greater sexual size dimorphism (2: j fi m),smaller interparietal (7: n fi k), more post-mentals (30: m fi o), posterior border ofmental straight (31: w fi z), supraorbitalsemicircles in contact (32: afi z), interparietalin contact with supraorbital semicircles (46:a fi z). The unusual Ecuadorian anole A.

proboscis is sister to the phenacosaur clade.Monophyly of the phenacosaurs (node 326;bootstrap 94%) is supported by six unambig-uous morphological changes: reduced inter-parietal (7: n fi y), fewer scales across snout(29: h fi d), interparietal in contact withsupraorbital semicircles (46: a fi t), 5: 1 ribformula (47: 6fi 1), two sternal ribs (48: 1 fi0), 23 presacral vertebrae (51: 0 fi 1). Node352 corresponds to Guyer and Savages (1986)Dactyloa but for the inclusion of Phenaco-

saurus (heterodermus, nicefori).The grouping (Fig. 3: node 353) of the

Caribbean Alpha Anolis with the Beta Anolisclade is supported by five morphologicalchanges: longer head (4: o fi r), broader head(5: jfi l), rib formula 3:1 (47: 6 fi 7), parietalcrests Y-shaped (57: 1 fi 2), supratemporalprocesses extend over supraoccipital (61: ffiz). Within this clade are clades that include allor most of the members of the relatively well-studied Northern Antillean Alpha series. Theclade of Cuban giants of the equestris series(node 318; 98% bootstrap) is supported by 92

unambiguous synapomorphies, including 24morphological changes: larger size (1: mfi z),broader head (5: l fi q), smaller ear (6: l fi i),smaller interparietal (7: s fi t), round tail (15:m fi a), large male dewlap (16: 1 fi 0), largefemale dewlap (17: 1fi 0), larger dorsal scales(19: rfi k), completely divided mental (26: ifiu), rostral broader than mental (27: a fi w),fewer scales across snout (29: i fi d), fewerpostmentals (30: lfij), zero supraciliaries (38:1 fi 0), nasal type 3 (39: 0 fi 3), supraocular

scales about equal in size (41: 0fi

4), threelumbar vertebrae (52: 1 fi 0), nine anterioraseptate caudal vertebrae (53: 4 fi 2),pustulate dorsal surface of skull (56: d fi z),

parietal casque (59: a fi z), nasal does notreach naris (66: afi z), vomers with posteriorlydirected processes (72: afi z), maxilla extendsbeyond ectopterygoid (73: q fi u), reduced

posterior extent of dentary (84: zfi

a),presence of splenial (85: 1 fi 0). TheHispaniolan chlorocyanus series (node 315;97% bootstrap) is supported by 46 unambig-uous synapomorphies, including eight mor-phological changes: shorter thigh (3: o fi n),compressed tail (15: m fi z), posterior borderof mental concave (31: t fi d), greater keeling(40: 1 fi 2), parietal crests V-shaped (57: 2 fi1), supraoccipital exposed (61: z fi a), maxillaextends to ectopterygoid (73: q fi a), blackpigment on skull (76: afi z). Monophyly of theHispaniolan hendersoni series (node 320;100%) is supported by 48 unambiguoussynapomorphies, 11 of which are morpholog-ical changes: smaller size (1: g fi f), longerhead (4: rfiy), smaller ear (6: rfi q), smaller

ventral scales (20: o fi n), mental scalecompletely divided (26: i fi t), more post-mentals (30: m fi o), rostral overlap (36: a fig), supraoccipital exposed (61: r fi n), jugalsquamosal contact (68: efi a), maxilla extendsto ectopterygoid (73: q fi g), splenial present

(85: 1fi 0). The cuvieri series (Xiphosurus)plus A. christophei (node 186; 54% bootstrap)is supported by 22 unambiguous synapomor-phies, including presence of a tail crest (12:a fi z). Within this group the three Hispanio-lan species (A. ricordi, A. baleatus, A. bar-ahonae) form a clade (node 184; 63%bootstrap).

The bimaculatus, cristatellus, and disti-chus groups form a clade (node 216; 79%bootstrap) supported by 25 unambiguous

synapomorphies, including five morphologicalchanges: more dorsal scales (19: t fi u), fewerscales across snout (29: ffi e), supraorbitalsemicircle contact (32: a fi t), interclavicleT-shaped (50: z fi a), lack of jugal-squamosalcontact (68: g fi a). The Northern LesserAntillean bimaculatus series (node 214; 96%bootstrap) is supported by 27 unambiguoussynapomorphies, including two morphologicalchanges: dewlap scales in double rows (21: lfim), blunt posterior suture of dentary (83: gfi

z). The Hispaniolan distichus series (node 207;bootstrap 97%) is supported by 10 unambig-uous synapomorphies, all of which are mor-phological: shorter thigh (4: s fi m), broader



18/53

head (5: ofi p), shorter tail (8: pfi m), fewerscales across snout (29: e fi d), preoccipitalscale (33: afim), parallel rows of snout scales(34: a fi z), cleft rostral (35: a fi t), all scales

smooth (40: 0fi

1), interparietal contactssupraorbital semicircles (46: afi g), prefrontalcontacts nasal (63: z fi i). The Puerto Ricancristatellus series (node 205; 93% bootstrap) issupported by 19 unambiguous synapomor-phies, including presence of pterygoid teeth(71: z fi f).

The Cuban carolinensis series (node 300;81%) is supported by 28 unambiguous syna-pomorphies, including four morphologicalchanges: greater SVL (1: efi f), fewer dorsals(19: s fi o), supraocular scales grade contin-uously in size (41: 2 fi 0), more anteriormandibular toothline (81: t fi z).

This series of Caribbean clades also includesthe unusual terrestrial Hispaniolan species A.(Chamaelinorops ) barbouri, which groups

with some Caribbean grass Anolis (node 194).This species appears tenuously placed, re-quiring 24 unambiguous autapomorphic mor-phological changes along its branch (142 total).

The cybotes series is sister group to the BetaAnolis. This grouping (node 293) is supported

by 13 unambiguous synapomorphies, includ-ing diagnol ventral scalation pattern (14: a fin). Although Etheridge (1959) considered thecybotes anoles to be Alpha Anolis (and they

were coded as such in this study), he noted thatthe condition of their caudal vertebrae appearssomewhat intermediate between true Alphaand Beta forms. The Hispaniolan cybotes clade(node 292; 99% bootstrap) is supported by 58unambiguous synapomorphies, including 11morphological changes: longer thigh (3: sfiv),

broader head (5: o fi t), subocular scalesseparated from supralabials (28: a fi p),supraorbital semicircles in contact (32: a fir), interclavicle T-shaped (50: z fi a), antero-lateral corners of parietal are medial to edgesof frontal (58: a fi n), postfrontal absent (62:a fi z), concave posterior aspect of jugal (69:a fi z), quadrate lateral shelf (75: n fi z),cybotes-type jaw sculpturing (90: 0 fi 4).

Monophyly of the Betas (node 286, Fig. 4) issupported by 27 unambiguous synapomor-

phies, including five morphological changes:shorter head (4: t fi o), two supraciliaries(38: 1 fi 2), transverse processes on caudal

vertebrae (Beta condition; 49: 0 fi 1), pineal

foramen in parietal (60: afi z), nasal overlapspremaxilla (77: a fi n). The two CaribbeanBeta cladesthe Cuban sagrei series (node225; 76% bootstrap) and the Jamaican grahami

series (node 284; 74% bootstrap)are succes-sive monophyletic outgroups to a mainlandBeta clade (node 278). The sagrei series cladeis supported by 27 unambiguous synapo-morphies, including seven morphologicalchanges:greater sexual size dimorphism (2:l fi p), larger interparietal (7: n fi j), shortertail (8: p fi m), tail crest (12: a fi z), roundtail (15: gfi a), pineal foramen in parietal (60:rfi z), basipterygoid crest (74: a fi z). Mono-phyly of the grahami series is supported by19 unambiguous synapomorphies, includingsix morphological changes: greater SVL (1:i fi l), transverse ventral scales (14: n fi a),presence of female dewlap (17: 2fi 1), smaller

ventral scales (20: m fi r), partially dividedmental scale (26: g fi a), wrinkled jawsculpturing (90: 0 fi 5).

Monophyly of the mainland Betas is sup-ported by seven unambiguous synapomor-phies: shorter head (4: o fi n), narrowerhead (5: o fi n), larger ventrals (19: t fi r),concave mental (31: z fi i), 23 presacral

vertebrae (51: 0 fi 1), parietal crests V-shaped(57: 2 fi 1), supratemporal processes leavesupraoccipital exposed (61: zfi a). Within themainland Beta clade there are few well-supported clades and few clades that corre-spond to recognized informal groupings. Fourpreviously recognized species groups (Lieb,1981; Savage and Guyer, 1989) are upheld asmonophyletic: crassulus (node 261; 87% boot-strap), supported by eleven unambiguous mor-phological changes: shorter thigh (3: t fi q),

narrower head (5: n fi o), smaller ear (6: qfio), smaller interparietal (7: p fi r), longer tail(8: p fi s), larger dorsal scales (19: o fi l),fewer scales across snout (29: h fi f), dorsalskull rugose (56: a fi n), parietal creststrapezoidal (57: 1 fi 0), jugal-squamosalcontact (68: a fi z), splenial present (85: 1 fi0); gadovii (node 271), supported by fourunambiguous morphological changes: loss ofenlarged middorsal scales (13: n fi a), roundtail (15: m fi i), presence of female dewlap

(17: 2fi

1), keeling pattern (40: 0fi

c);nebulosus (node 272), supported by fourunambiguous morphological changes: smallersize (1: ffi d), shorter thigh (3: sfi o), smaller



19/53

FIG. 5.Phylogenetic estimate ofAnolis based on morphological data analyzed alone.



20/53

ear (6: p fi n), compressed tail (15: m fi z);intermedius (node 256; 70% bootstrap), sup-ported by four unambiguous morphologicalchanges: shorter thigh (3: o fi n), broader

head (5: lfi

n), loss of female dewlap (17: 1fi

2), 23 presacral vertebrae (51: 0 fi 1). Themonophyly of the intermedius series is un-surprising considering that the two represen-tatives of this series analyzed here haverecently been suggested to be conspecific(Kohler, 2004).

Tree from Morphological Data Alone

Figure 5 shows a strict consensus of the four

optimal trees resulting from analysis of themorphological data alone (length 5 1512280;CI 5 0.11; RI 5 0.59). Many groups aremonophyletic or nearly monophyletic in boththe morphological and combined trees, such asthe cybotes group, equestris group, roquetgroup, cristatellus group, distichus group,carolinensis group, and several smaller clades.The mainland Betas are monophyletic exceptfor A. lineatus clustering with a clade of

grahami and sagrei group Anolis and Cuban

A. ophiolepis grouping with A. auratus.Differences between the combined andmorphological trees include the basal para-phyly of the South American and SouthernLesser Antillean anoles that are monophyleticat the base of the combined tree. The giantequestris, chamaeleolis and cuvieri group

Anolis form a clade near the base of themorphological tree. The bimaculatus group

Anolis are paraphyletic relative to distichus,cristatellus, and cybotes group Anolis. Al-though the three geographic clades of Beta

Anolis are each nearly monophyletic, theisland and mainland Beta clades are found indifferent parts of the morphological tree. Thephenacosaurs are sister to the rest of Anolis.Additional data will determine whether any ofthese conflicting morphological clades repre-sent real phylogenetic relationships.

Comparison with Previous Hypotheses ofAnolis Relationship and Taxonomy

Etheridge (1959) was the first to presenta phylogenetic treatment of Anolis, and

virtually all studies that followed are based

implicitly or explicitly on this work. Many ofhis results agree with the optimal tree of thispaper. Notable among this concordance is themonophyly of the Beta section, the basal

position of the South American and SouthernLesser Antillean Alpha Anolis (his latifronsseries), the close relationship of Phenacosau-rus with the latifrons series, the monophyly ofthe mainland Beta species, and the monophylyor near-monophyly of many of his series.Given the small character base and limitedtaxon sample from which he was operating,Etheridges results are remarkably robust.

Many of the series and species groups ofWilliams (1976a,b) are recovered as clades inthe optimal tree (see below). His subsectionsand most other larger groups, however, are notmonophyletic. Similarly, the smaller groupsrecognized by Shochat and Dessauer (1981)based on immunological data are mostlymonophyletic in the optimal tree, but theirlarger Central Caribbean Species Complex isnot. And, the intraisland monophyly of Cubanand Hispaniolan species suggested by Burnelland Hedges (1990) is not evident in theoptimal tree, but many of their series andspecies groups do form clades. This pattern of

easy recovery of smaller groups but non-monophyly of larger groups is a commontheme in Anolis studies:

It is this parallelism that contributes to thenotorious difficulty of the genus Anolis.Narrow groups are rather easy to recognize(though the specific and infraspecific struc-ture within the group may be puzzling in theextreme) but wider relationships (at least

when externals only are considered) are

problematical, becoming obscurer with eachstep more distant from the species group.

Williams (1961)

Jackman et al. (1999) suggested that thispattern may be due to a rapid radiation at thebase of the Anolis tree.

It is encouraging that there is no stronglysupported conflict between the results pre-sented here and those of previous studies of

Anolis phylogeny (as measured by, e.g.,

bootstrap analyses of those data; results notshown; and see Crother [1999]). However withthe exception of Jackman et al. (1999) andNicholson (2002), previous studies of Anolis



21/53

are based on very few characters. Thesestudies, many of which were excellent for theirtimes, are thus unable to produce results thatare in strong conflict with the results of this

paper. Nevertheless, the taxonomies producedin these studies can be compared to thephylogenetic results of this paper. Below, the

Anolis taxonomies of Etheridge (1959),Williams (1976a,b), Savage and Guyer (1989),and Burnell and Hedges (1990) are evaluatedfor their concordance with the results of thispaper. Note that some groups of the samename may have different compositions in eachpaper. Groups discussed below are consideredto be monophyletic or not only relative to thespecies that are common to this study and thelisted study. There is much overlap in contentbetween groups in these studies. The post-

Williams studies generally accept the Williamsgroups with only minor changes. Changes ingroup composition and name are difficult totrack in some cases where authors madechanges without analysis or justification.Among the taxonomic studies discussed below,only Burnell and Hedges (1990) performedphylogenetic analyses that could test themonophyly of Williams groups. However,

many of their named groups reflect conven-tions from Williams rather than their ownresultse.g., six of their named groups are notmonophyletic in any of their presented trees(sagrei series, christophei series, grahamiseries, cristatellus series,pulchellusgroup,stra-

tulus group; compare their Table 1 to their Figs.13) but correspond closely or completely withgroups recognized by Williams (1976a). Thedecision not to accept some of the results oftheir phylogenetic analysis should not neces-

sarily be construed as nonscientific. Rather, itrepresents a recognition by Burnell andHedges (1990) that their limited sampling ofcharacters allows for stochastic errors thatshould be evaluated in light of additionalevidence.

The relatively small number of concordantclades listed below could be interpreted tosuggest great conflict between the aforemen-tioned studies and the tree of this paper. Thisinterpretation would be incorrect. Many of the

groups recognized by these authors are mono-phyletic but for one or a few species orotherwise have some phyletic cohesion (e.g.,sequential paraphyly). And, it should be

remembered that Etheridge and Williamspublished their taxonomies before it wascommon practice to associate names only withclades. Consider for example that both Wil-

liams and Etheridge recognized a genus Tro-pidodactylus (5A. onca) nested within Anolis;and both discussed the possibility of the Betasection arising from within the Alpha section

while simultaneously maintaining the Alpha-Beta dichotomy taxonomically (Williams,1976a: 3; Etheridge, 1959: 194).

Groups recognized by Etheridge (1959) thatare monophyletic in the preferred tree: Betasection, grahami series, and sagrei series.Although monophyletic relative to the othernamed series, the grahami and sagrei serieseach include species which were taxonomicallyunplaced by Etheridge (sagrei: A. ophiolepis;

grahami: A. valencienni) In his phylogenetictree of Betas (Etheridge 1959: fig. 11) A.ophiolepis is monophyletic within the sagreiseries but A. valenciennis is outside of the

grahami group clade.Groups recognized by Etheridge (1959) that

are not monophyletic in the optimal tree:Alpha Section, latifrons series, bimaculatusseries, carolinensis series, coelestinus series,

cristatellus series, angusticeps series, petersiseries, fuscoauratus series, chrysolepis series.

Groups recognized by Williams (1976a) thatare monophyletic in the optimal tree: roquetseries, roquet species group/superspecies,ricordi superspecies, distichus subgroup, bi-

maculatus species group, cybotes subseries/species group, cristatellus subseries, pulchel-lus species group, equestris species group/superspecies, chlorocyanus species group,hendersoni species group, angusticeps sub-

group, argillaceus species group, lucius spe-cies group, vermiculatus species group,sagrei species group, allogus superspecies,carolinensis subgroup/superspecies, grahamisuperspecies.

Groups recognized by Williams (1976a) thatare not monophyletic in the optimal tree:Alpha section, punctatus subsection, luciaespecies group, luciae superspecies, richardisuperspecies, cuvieri series, ricordii speciesgroup, bimaculatus series, stratulus subseries,

stratulus species group, bimaculatus subseries,bimaculatus subgroup, bimaculatus superspe-cies, marmoratus superspecies, cristatellusseries, pulchellus subgroup, carolinensis sub-



22/53

section, occultus series, darlingtoni series,monticola series, monticola species group,carolinensis series, chlorocyanus superspecies,carolinensis species group, lucius series, alu-

taceus series, alutaceus species group, aluta-ceus superspecies, spectrum superspecies,semilineatus species group, grahami series,grahami species group, sagrei series, sagreisuperspecies, homolechis superspecies.

None of the species groups recognized byWilliams (1976b) are monophyletic in theoptimal tree.

Groups recognized by Savage and Guyer(1989; some mainland groupings aretaken fromLieb [1981]) that are monophyletic in this

analysis: roquet series, cybotes series, cristatel-lus species group, bimaculatus series/speciesgroup, hendersoni species group, chlorocyanusseries, equestris series, Norops,grahami series,

sagrei series, crassulus species group, gadoviispecies group, laeviventris species group,

nebulosus species group, pentaprion subseries.Groups recognized by Savage and Guyer

(1989) that are not monophyletic in the com-bined tree: Dactyloa, latifrons series, aequa-

torialis series, punctatus series, tigrinus series,

Semiurus (later recognized as Xiphosurus),Ctenonotus, cristatellus series, Anolis (sensustricto), alutaceus series, alutaceus speciesgroup, semilineatus species group, angusticepsseries, carolinensis series, carolinensis speciesgroup, darlingtoni series, monticola series,lucius series, occultus series, auratus series,

schiedi subseries, schiedi species group, laevi-ventris subseries, subocularis species group,nebuloides species group, auratus subseries,humilis species group, lemurinus species

group, cupreus species group, auratus speciesgroup, fuscoauratus series, fuscoauratus spe-cies group, lionotus species group, petersiseries, petersi subseries.

Groups recognized by Burnell and Hedges(1990) that are monophyletic in the combinedtree: argillaceus series, chlorocyanus series,ricordi group, carolinensis group, angusticepsgroup, equestris series, lucius group, cybotesseries, distichus series, vermiculatus group,

sagrei series, allogus group, hendersoni series,

sheplani series, grahami series, cristatellusseries, pulchellus group, cristatellus group,bimaculatus series, bimaculatus group, gingi-vinus subgroup, marmoratus subgroup, stra-

tulus group, roquet series, roquet group,roquet subgroup.

Groups recognized by Burnell and Hedges(1990) that are not monophyletic in the

optimal tree: alutaceus series, spectrum group,carolinensis series, carolinensis subgroup,lucius series, homolechis group, sagrei group,chlorocyanus group, christophei series, cuvieriseries, semilineatus series, luciae group.

Two larger-scale phylogenetic analyses maybe compared to the results of this paper.Jackman et al. (1999) and Nicholson (2002)collected and analyzed subsets of the dataanalyzed in this paper. The optimal tree of thispaper is similar to the optimal tree of Jackmanet al. (1999:Fig. 10) for shared taxa. This resultis unsurprising, as the mtDNA data from thatpaper accounts for about 51% of scored cells inthe matrix of this paper. Regarding the 52species shared by that study and this one, theonly differences between the optimal com-bined tree in Jackman et al. (1999:Fig. 10) andthe optimal tree of this paper are that papers1) basal position of the equestris group as sisterto South American Alpha Anolis, 2) groupingof the distichoid Anolis group with cristatellusspecies rather than bimaculatus species, 3)

switched phyletic position ofA. lineatopus andA. valencienni, 4) grouping ofA. bahorucoen-sis with A. bartschi and A. vermiculatus ratherthan the Hispaniolan green Anolis, 5) differingplacement of Chamaelinorops. There areadditional differences in deep relationshipsbetween Jackman et al.s (1999) optimal treefrom analysis of their DNA data alone (e.g.,their placement of cybotoid Anolis as sister toCuban Alpha Anolis). These differences arenot well supported in either analysis.

The optimal tree of this paper agrees withsome of the tip relationships in Nicholson(2002) such as the monophyly of A. interme-dius and A. laeviventris and ofA. crassulus and

A. sminthus, but disagrees with that paper indeeper relationships. In particular, Nicholson(2002) found nonmonophyly for the threegeographic clades of Beta Anolisthe Cuban

sagrei series, the Jamaican grahami series, andthe mainland formssuggested by Etheridge(1959) and Jackman et al. (1999) and corrob-

orated by the larger character and taxonsample of this paper.Although I compared the results of this

paper to earlier papers that revised Anolis



23/53

phylogeny and taxonomy I did not performstatistical tests comparing results, for severalreasons. First, the sample of taxa and charac-ters in this paper is greater than that used in

any previous analysis of Anolis relationships,making comparisons with earlier studiesdifficult to interpret. For example, withoutextensive additional analyses comparing re-sults with and without the additional taxa ofthis paper, and with and without the addi-tional characters in this paper, and underdifferent optimality criteria, it cannot bedetermined whether differences are due toadditional taxa, different character samples,different optimality criteria, different taxo-nomic philosophies, or some combination ofthese factors. Such analyses comparing resultsare valuable, but they would not (in myopinion) change the fact that the current bestestimate of Anolis phylogeny is the tree ofFigs. 24. Second, most named groups in

Anolis were erected without phylogeneticanalysis. There seems little to be gained incontinuing to recognize or discuss thesegroups if they have never been supportedbut have been contradicted in a phylogeneticanalysis. Finally, the data set of this paper

subsumes all the character sets of the earlierpapers. Although there is value in comparingresults between different sources of data toassess degrees of conflict and support, thereseems little point in making a statisticalcomparison between the results from somesample of characters to the results from thatsample combined with another sample. Theresult with the larger sample seems preferableunless of course there is some problem withthe added characters. That is, even if the

result from the smaller sample alone cannotbe rejected by the larger sample, I see noreason to retain such a result when anestimate based on more data exists.

PROSPECTUS

A phylogenetic estimate of 174 Anolisspecies was produced using 91 morphologicalcharacters and 1589 characters from DNA,allozymes, chromosomes, and immunology.

This estimate should facilitate several evolu-tionary studies ofAnolis. First, it may form thebasis for a comprehensive phylogenetic taxon-omy of Anolis. I am currently working with

collaborators to apply the principles of deQueiroz and Gauthier (1992) to classify theanoles. Second, the tree produced here may beused as a framework for comparative analyses.

The copious comparative data collected forAnolis may be analyzed with reference to thistree to enable evolutionary interpretations.Third, the results of this paper may help guidefuture phylogenetic studies of Anolis. Forexample, it is clear from the results of thispaper that much works remains to be done onthe deep relationships of Anolis and on allaspects of the relationships of mainland Anolis.And, it is clear that the monophyly of mostnamed groups ofAnolis should be consideredsuspect until phylogenetic analysis suggestsotherwise (see also Williams, 1992; Nicholson,2002). Finally, the morphological data of thispaper may be combined with new data insubsequent studies ofAnolis phylogeny.

Acknowledgments.My undergraduate advisor, ErnestWilliams, introduced me to Anolis and was an excellentadvisor and friend. The dissertation version of this paperwas dedicted to his memory. Thanks to Tim Rowe, JimBull, David Hillis, David Cannatella, Lee Fitzgerald, Rich

Glor, and two anonymous reviewers for reading variousdrafts of this paper. Thanks to Kevin de Queiroz for hisadvice on many aspects of this paper. Richard Etheridgekindly allowed use of his original anole data sheets. Hisdissertation was an inspiration for this work. Jose Rosado,John Cadle, and the rest of the Museum of ComparativeZoology were very helpful during work done at Harvard.Kevin de Queiroz, Ron Heyer, George Zug, RoyMcDiarmid, Ron Crombie, and the entire staff of theSmithsonians National Museum of Natural History herpdepartment were very helpful during my tenure there.Rich Glor kindly provided the aligned matrix of mtDNAcharacters. Thanks to Jose Ottenwalder in the DominicanRepublic, Miguel Garcia and Jose Luis Chabert Llompart

in Puerto Rico, and Lourdes Rodriguez-Schettino in Cubafor facilitating field work. Thanks to Jonathan Losos forletting me come along to Cuba and for partially financingmy participation with his NSF Grant (DEB 9318642). Thiswork was supported by the NSF (grant to Jonathan Lososand predoctoral fellowship to me), the Smithsonian(predoctoral fellowship), the University of Texas, theDepartment of Zoology at UT (Dorothea Bennet fund,Beth Burnside Fellowship, Hartman Fellowship, UTFellowship, Travel Grants, Research Grants), the NewEngland Herp Society, the Texas Memorial Museum, theMuseum of Comparative Zoology, and the Miller Institute(postdoctoral fellowship). Specimens were generouslyloaned by Harvard (Jose Rosado), the University of Texas

at El Paso (Carl Lieb, Robert Webb, Dominic Lanutti), theUniversity of Texas at Arlington (Johnathan Campbell),and the University of Michigan (Greg Schneider, ArnoldKluge).



24/53

LITERATURE CITEDACKERLY, D. D. 2000. Taxon sampling, correlated evolu-

tion, and independent contrasts. Evolution 54:148092.BURNELL, K. L., AND S. B. HEDGES. 1990. Relationships of

West Indian Anolis (Sauria: Iguanidae): an approach

using slow evolving protein loci. Caribbean Journal ofScience 26:730.

CANNATELLA, D. C., AND K. DE QUEIROZ. 1989. Phyloge-netic systematics of the anoles: is a new taxonomywarranted? Systematic Zoology 38:5769.

CASE, S. M., AND E. E. WILLIAMS. 1987. The cybotoidanoles and Chamaelinorops lizards (Reptilia: Iguani-dae): evidence of mosaic evolution. Zoological Journal ofthe Linnaen Society 91:325341.

COLLETTE, B. B. 1961. Correlations between ecology andmorphology in anoline lizards from Havana. Cuba andsouthern Florida. Bulletin of the Museum of Compar-ative Zoology, 125:137162.

CREER, D. A., K. DE QUEIROZ, T. R. JACKMAN, J. B. LOSOS,AND A. LARSON. 2001. Systematics of the Anolis roquetseries of the southern Lesser Antilles. Journal ofHerpetology 35:428441.

CROTHER, B. I. 1999. Caribbean Amphibians and Reptiles.Academic Press, San Diego. 495 pp.

DE QUEIROZ, K., AND J. GAUTHIER. 1992. Phylogenetictaxonomy. Annual Review of Ecology and Systematics23:449480.

ETHERIDGE, R. E. 1959. The relationships of the anoles(Reptilia: Sauria: Iguanidae): an interpretation based onskeletal morphology. Ph.D. Dissertation, University ofMichigan, Ann Arbor, Michigan, U.S.A.

ETHERIDGE, R. E., AND K. DE QUEIROZ. 1988. A phylogeny

of the Iguanidae. Pp. 283337. In R. Estes and G. Pregill(Eds.), Phylogenetic Relationships of the Lizard Fam-ilies: Essays Commemorating Charles L. Camp. Stan-ford University Press, Stanford, California, U.S.A.

FARRIS, J. S. 1989. The retention index and the rescaledconsistency index. Cladistics 5:417419.

FELSENSTEIN, J. 1985a. Phylogenies and the comparativemethod. American Naturalist 125:115.

. 1985b. Confidence limits on phylogenies: anapproach using the bootstrap. Evolution 39:783791.

FITCH, H. 1975. Sympatry and interrrelationships in CostaRican anoles. Occasional papers of the Museum ofNatural History, University of Kansas. 40:160.

FLEISHMAN, L. J . , E. R. LOEW, AND M. LEAL. 1993.Ultraviolet vision in lizards. Nature 365:397.FROST, D. R., AND R. E. ETHERIDGE. 1989. A phylogenetic

analysis and taxonomy of iguanian lizards (Reptilia:Squamata). Miscellanous publications of the Universityof Kansas Museum of Natural History 81:165.

FROST, D. R., R. E. ETHERIDGE, D. JANIES, AND T. A. TITUS.2001. Total evidence, sequence alignment, evolution ofpolychrotid lizards, and a reclassification of the Iguania(Squamata, Iguania). American Museum Novitates3342:138.

GAUTHIER, J., A. G. KLUGE, AND T. ROWE. 1988. Amniotephylogeny and the importance of fossils. Cladistics4:105209.

GIANNISI, R. , S. THORPE, AND A. MALHOTRA. 2000. Aphylogenetic analysis of body size evolution in the Anolisroquet group (Sauria: Iguanidae): character displace-ment or size assortment? Molecular Ecology 9:193200.

GLOR, R. E., L. J. VITT, AND A. LARSON. 2001. A molecularphylogenetic analysis of diversification in AmazonianAnolis lizards. Molecular Ecology 10:26612668.

GORMAN, G. C. 1973. The chromosomes of the Reptilia,a cytotaxonomic interpretation. Pp. 349424. In A. B.

Chiarelli and E. Capanna (Eds.), Cytotaxonomy andVertebrate Evolution. Academic Press, London.GORMAN, G. C., AND Y. S. KIM. 1976. Anolis lizards of the

eastern Caribbean: a case study in evolution.II. Geneticrelationships and genetic variation of the bimaculatusgroup. Systematic Zoology 20:167185.

GORMAN, G. C., AND L. ATKINS. 1967. The relationships ofAnolis of the roquet group (Sauria: Iguanidae). II.Comparative chromosome cytology. Systematic Zoology16:137143.

GORMAN, G. C., D. B. BUTH, M. SOULE, AND S. YANG. 1983.The relationships of Puerto Rican Anolis:

poe 2004 phylogeny of anoles[1]

Documents