The Auk 115(3):621-634, 1998

MOLECULAR PHYLOGENETICS OF THE GENUS PIRANGA: IMPLICATIONS FOR BIOGEOGRAPHY AND THE EVOLUTION OF MORPHOLOGY AND BEHAVIOR

KEVIN J. BURNS• Museumof VertebrateZoology and Department of IntegrativeBiology, University of California, Berkeley,California 94720, USA

ABSTRACT.--Speciesin the genusPiranga vary in degreeof sexualdimorphism, plumage coloration,morphology, song, migratory patterns, and geographicdistribution. To study theseaspects of Pirangabiology in an evolutionarycontext, I constructeda phylogeny for thisgenus using cytochrome-b sequence data. Parsimony and maximum-likelihood analyses of DNA datareveal three possible phylogenies for speciesin thisgenus. All threetrees iden- tify a cladecontaining P. rubriceps, P.leucoptera, and P. erythrocephala and a cladecontaining P.rubra, P. ludoviciana, P.olivacea, P. bidentata,and P.fiava. The trees differ in theplacement of P.roseogularis. Morphology, song, and plumage data did not agreewith these phylogenies. Levelsof sequencedivergence and the phylogeny of haplotypesare consistent with theidea thatP. fiava as currently described contains more than one evolutionary unit. Mapping the evolutionof seasonalmigration onto the DNA treesindicates that migrationevolved once within Piranga.Received 15 November1996, accepted 20 November1997.

STUDYING PATTERNS OF CHARACTER VARIA- plumage.One exceptionto this marked sexual TIONin conjunctionwith a phylogenyprovides dichromatismis P. rubriceps,in which males a powerful methodfor understandinghow and femalesdiffer only in the extentof red col- charactersevolve and for uncovering evolution- orationon the upper breast. ary associationsamong traits (Brooks and Brush(1967) and Hudon (1990)characterized McLennan1991, Harvey and Pagel1991). The the chemicalbasis underlying some of the dra- genusPiranga (Passeriformes: Thrau- maticplumage color in four speciesof Piranga. pidae) providesan excellentmodel for study- In threeof thesespecies (rubra, fiava, and oIiva- ing the evolutionof differenttypes of charac- cea),several common carotenoids were respon- ters.Species in thisgenus vary in the degreeof siblefor red plumage(Hudon 1990). Piranga Iu- sexualdimorphism, plumage coloration, mor- dovicianawas unique among the four speciesin phology,song, migratory patterns, and geo- that the only red pigmentfound was rhodox- graphicdistribution. None of theseaspects of anthin,a dietary-basedcarotenoid. The relative Pirangabiology has been examined in a histor- significanceof thesefindings and patternsof ical, evolutionarycontext because of the lackof plumage-colorevolution need to be addressed a rigorousphylogeny for the group. in a phylogeneticcontext. Although plumage The nine speciesof Pirangahave been placed colorand patternoften are usedto definegen- in this genusand arrangedin a linearclassifi- era and to determinespecies status, their rele- cation(Storer 1970) based on plumagecharac- vanceas phylogeneticcharacters has not been teristics,a stoutlyshaped bill, geographicdis- thoroughlyinvestigated. tribution, and the presenceof sexual dimor- Body size and shape vary considerably phism.One of themore striking features of this among speciesof Piranga.Two species,P. Ieu- genusis theconspicuous plumage coloration of copteraand P.erythrocephaIa, average 33% small- males.Males typically are more colorful than er than the otherseven species. These smaller- females,and males of all specieshave at least sized Pirangawere consideredby Howell and somered plumage.Females are morecryptic, Webb(1995) to belongto a separategenus, Sper- with mainly brown, yellow,or olive-colored magra(Swainson 1827). In addition, somespe- ciesof Piranga(Ieucoptera, erythrocephaIa, and ro- • Presentaddress: Department of Biology,San Di- seoguIaris)have shorter and less-pointed wings ego State University, San Diego, California 92182, than the otherspecies in the genus(Ridgway USA. E-mail: kburns@sunstroke. sdsu.edu 1902).Whether there is a phylogeneticcompo- 621 622 KEVINJ. BURNS [Auk, Vol. 115 nent to this variation cannot be addressed group.Although no singlecharacter defines the without a phylogenyfor this group. genus,two molecularstudies confirm that Pi- Shy (1983,1984a, b, c) detailedintraspecific rangais monophyleticwith respectto othertan- and interspecificsong variation in P.olivacea, P. agers. A study of allozyme variation (Mc- ludoviciana,P.rubra, and P.fiava. In a comparison Donald 1988) showedthat three morphologi- of songcharacteristics of the four species,Shy cally similar species(rubra, olivacea, and ludovici- (1984c)provided data on nine songvariables: ana)form a monophyleticgroup. Furthermore, maximum frequency,minimum frequency,av- a comprehensivecytochrome-b phylogeny of erage maximum frequency,average minimum Thraupidae(Burns 1997) showed that two mor- frequency,average interval between figures, phologicallydistinct species (olivacea and eryth- average duration of figures, frequencyrange, rocephala)form a clade with respectto the 48 averagefrequency range of figures, and num- othertanager genera sampled. Thus, an addi- ber of figuretypes per . In conjunctionwith tionalgoal of this studyis to furtherclarify the a phylogeny for Piranga, Shy's information limits of this genus. could addressthe usefulnessof songcharac- teristicsas systematicdata. Vocal information METHODS has provenimportant to systematicquestions at and below the specieslevel, yet few studies Taxonsampling and outgroupchoice.--I used indi- have used songdata to establishrelationships viduals representingall nine speciescurrently as- signedto Piranga(Storer 1970). For all speciesexcept among species(see Payne 1986). P.rubriceps, more than oneindividual was used to re- The geographicdistributions of Pirangaare duce problemsassociated with using only a single centered in Mexico and Central America. Three exemplar per taxon (Table 1). Tanager outgroups species(erythrocephala, bidentata, and roseogular- used included representativesof the generaHabia, is) are endemicto this region,and all but one ,Calochaetes, and Ramphocelus(Gen- occur there during at least sometime of the Bank Accession numbers AF006213, AF006219, year Pirangafiava and P. leucopteraare wide- AF006233,and U15717 to U15724). Habia and Chlo- spreadin Mexico, Central America, and South rothraupiswere chosenbased on a closerelationship America,and the distributionof P.fiava also ex- to Pirangaidentified by a molecularphylogenetic tendsnorthward to North Americaduring the study of tanagergenera (Burns 1997). I alsoincluded Calochaetesand Ramphocelusbecause the traditional borealsummer Three other species (rubra, lu- linear arrangement of (e.g. Storer 1970) doviciana,and olivacea)are seasonalmigrants placesthem near Piranga. Multiple sequences of Ram- that spendthe breeding season in NorthAmer- phocelus(from Hackett 1996) were included in order ica and the nonbreedingseason in parts of to increase tree balance and to reduce the branch Mexico, Central America, and South America. lengthleading to this genus(Smith 1994). As an out- The onespecies not foundin Mexicoor Central groupto thesetanager sequences, I used sequence of America during at least part of the year is P. a non-emberizidpasserine, Pomatostomus temporalis rubriceps,which occurssolely in the Andesof (from Edwards et al. 1991;GenBank Accession num- Colombia, Ecuador,and Peru. Many authors ber X60936). havehypothesized that closelyrelated species DNA isolationand sequencing.--DNAextracts were prepared from liver or muscletissue preserved in of North Americanbirds (Rand 1948,Mengel 95% ethanol or frozen at -80øC. Extractions were 1964, Hubbard 1973) and closelyrelated spe- performedusing either a 5% Chelexsolution (Walsh ciesof SouthAmerican (Haffer 1985) di- et al. 1991) or by NaC1 extraction (Miller et al. 1988). vergedfrom eachother as a result of isolation Usingstandard protocols (Burns 1997), specific frag- in refugia during Pleistoceneglaciation. The mentsof the cytochrome-bgene were then amplified biogeographichistory of Pirangaremains unex- usingthe polymerasechain reaction (PCR) and seven plored due to lack of knowledgeabout evolu- different primers: L14987, L15236, H15304, H15706 tionaryrelationships within this group. from Cicero and Johnson(1995); H15916 from Ed- Here, I use datafrom the cytochrome-bgene wards et al. (1991); H16065 from Smith and Patton (1993); and L15661 (ACCTCCTAGGAGA[C/ of themitochondria to reconstructthe first phy- T]CCAGA[C/A/T]AA[C/T]T), which is new to this logenetichypothesis for the genusPiranga. I study. Fragmentswere sequencedusing manual and then use the phylogenyto assesshow aspects ABI dye-terminator methods following standard of Piranga biology mentioned above have procedures(Burns 1997). SequenceNavigator Ver- changedduring the evolutionaryhistory of the sion 1.0.1 (Applied Biosystems,Perkin Elmer) was July1998] SystematicsofPiranga 623

ZZZ Z ZZ Z ZZZ OOZZZ 624 KEVINJ. BURNS [Auk, Vol. 115 used to reversecomplement opposing directions, to TABLE2. Plumagecharacters and character states of align differentfragments from the sameindividual, speciesof Piranga.• and to translate complete sequencesinto amino ac- ids. Character number Accuracyof DNA sequencingwas verified in four 111111 ways:(1) sequencingboth heavy and light strandsof Species 123456789012345 most PCR fragments, (2) using overlapping frag- Pirangaroseogularis 022320003331213 mentsof cytochromeb (approximately30% of theto- Pirangabidentata 112241111111111 tal sequenceis overlappedby two fragments),(3) se- Pirangafiava 112220002111112 quencingsome individuals more than once,and (4) Pirangarubra 122220002221122 comparinglevels of sequencedivergence separately Pirangaolivacea 112210002221112 for the threefragments sequenced in eachindividual Pirangaludoviciana 112111214221112 (seeHackett et al. 1995). Pirangaleucoptera 012211112241132 The resultingsequences include 1,045 base pairs of Pirangaerythrocephala 012130005241122 012110205242132 the cytochrome-bgene and part of the threoninet- Pirangarubriceps RNA with the intergenicspacer region. These se- • Key for charactersand characterstates: (1) juvenalplumage, 1 - quencesbegin at base14,991 and end at base16,064 streaked,0 - unstreaked;(2) bill color,1 - black,2 = yellow;(3) male crownand throat, 1 - olive green,2 - red; (4) malebreast, 1 - yellow, relative to the publishedsequence of Gallusgallus 2 = red,3 = buffygray; (5) malewing and tail, 1 - black,2 - reddish, (Desjardinsand Morais 1990).Sequences obtained in 3 - green/yellow,4 = brownish;(6) greatercovert tips, 0 = same,1 this study have been depositedinto GenBank(Ac- - white, 2 = yellow;(7) mediancovert tips, 0 - same,1 - white, 2 cession numbers AF011759 to AF011781, AF006247, = yellow;(8) tertialtips, 0 = same,1 = white;(9) maleback color, 1 - streaked black or red-brown, 2 - red, 3 - olive brown, 4 - black, AF006248).Percent sequence divergence was calcu- 5 = yellowish/greenish;(10) male ear coverts, 1 - duskyred, 2 - red, lated as the number of nucleotide differences be- 3 - gray;(11) male lores and ocular, 1 - duskyred, 2 = red,3 = gray, tweentwo sequencesdivided by the totalnumber of 4 - black;(12) female crown and throat color, 1 = olivegreen/yellow, nucleotidescompared (p-distance of Nei [1987]).Dis- 2 = red; (13) femalebreast, 1 - olive gray,2 - gray;(14) femalewing and tail, 1 - dark olive green,2 - light olive green,3 - black/gray; tancebetween two sequenceswas also calculated us- (15) female back, 1 = streaked,2 = olive green/yellow, 3 = olive ing Kimura's (1980) two-parameter model of se- brown. quenceevolution. This measureof distanceis a max- imum-likelihood estimate of sequencedifference that correctsfor multiple substitutionsat sitesand additional analysisto correctfor noiseat third-po- the higher rate of transitionrelative to transversion sition sites by giving third-position transitionsa changes. weight one-sixthof that given to other characters. Phylogeneticanalysis of DNA data.--Phylogenetic This was accomplishedusing the stepmatrixoption analyseswere carriedout usingparsimony and max- of PAUPto assignthird-position transitions a weight imum-likelihoodapproaches as implementedin test of one,and third-positiontransversions a weight of version4.0.0d55 of PAUP*provided by D. L. Swof- six. All first- and second-positionchanges were as- ford. Forthe parsimonyanalysis, sequences were an- signeda weight of six. The ratio of 6:1 was obtained alyzedusing the heuristicoption with 100random- empirically by examining multiple sequencesfrom addition replicatesfor eachanalysis. Data were an- eachof two generaused in this study:Ramphocelus alyzedwith all charactersgiven equal weight and us- (Hackett 1996) and Piranga. ing a weightingscheme designed to correctfor mul- I performedmaximum-likelihood analyses using tiple substitutionsat a given site. Relative support the heuristicsearch option and 10 randomaddition for different nodeswas assessedusing 100bootstrap replicates.Empirical base frequencies were used and replicates(Felsenstein 1985) and by calculatingBre- the transition-to-transversion ratio was estimated ruer indices(i.e. decayindices; Bremer 1988). from a neighbor-joiningtree of Kimura distances. To explorethe possibilitythat sometypes of base Log-likelihood values of the maximum-likelihood substitutionshave become saturated, I plottedp-dis- tree were comparedwith likelihoodvalues of other tance versus Kimura's two-parameter distancesfor topologiesfound in this study.I usedKishino-Has- first-, second-,and third-positionsites. Saturation egawa(1989) tests to determineif the likelihoodtree wasjudged to havetaken place if the resultingscat- was significantlymore likely than the parsimony ter plots did not indicate a linear relationshipbe- trees found. A topology was judged to be signifi- tween the two typesof distancesand if the data did cantlydifferent from the maximum-likelihood tree if not fall on the line y = x (seeBerbee et al. 1995).A the rangeof valueswithin _+1.96 SD of thetree being nonlinear relationshipwould indicate that the Ki- testeddid not includethe log-likelihoodvalue of the mura estimatehas adjustedthe originaldistance es- maximum-likelihood tree. timate based on the differences in substitution rates Parsimonyanalysis of plumagedata.--In addition to between transitions and transversions. DNA sequencedata, plumage characters(Table 2) After determiningthat third-positionsites were were used to provide information about evolution saturatedfor transitions(see below), I performedan within Piranga.Plumage characters were analyzed July1998] SystematicsofPiranga 625

0.08'

'llll14- Among genera of Thraupidae

Among speciesof Piranga

• -I• Withill Piranga (except P. fl•va) ++

Percent Sequence Divergence

FIG. 1. Uncorrectedpercent sequencedivergence

(p-distance;Nei 1987) among individuals includedin 0.02 - thisstudy. Each cross represents a pairwise compar- ison between two individuals.

0.00 0.00 0.62 cladisticallyby coding different body regions as charactersand assigningdifferent colorsof these Kimura distance regionsas unorderedcharacter states (see Cracraft and Prum 1988,Hackett and Rosenberg1990). The resultingdata matrix was subjectedto parsimony 0.03 analysis as implemented in PAUP 3.1.1 (Swofford 1993). Songand morphologicalsimilarity.--To study mor- phologicalevolution, I measureda seriesof museum skinsof malesrepresenting each species of Piranga 0.02 (seeAppendix). All characters(length of exposed culmen,height of bill at base,width of bill at base, lengthof closed wing, length of tail,and length of tarsometatarsus)weremeasured following Baldwin '•, et al. (1931).In addition,mean body mass of each 0.01. specieswas obtainedfrom Islet and Islet (1987).Pri- or to the analysis,the meanbody mass of eachspe- cieswas linearizedby dividing by its cuberoot. To scalethe charactersequally, each character was di- vided by the varianceof that particularcharacter amongspecies ofPiranga. Means of each character 0'%.00 0.5• 0.62 0.63 measured were used to calculate a matrix of taxo- Kimura distance nomic distances(Sheath and Sokal 1973).These dis- tanceswere used to constructan UPGMA pheno- gram(using NTSYS-pc; Rohlf 1988) summarizing 0.4, overall similarity in morphology(see Hackett and Rosenberg1990 and Johnsonand Marten 1992). Thesedistances were alsoused to build a neighbor- joining tree using PHYLIP version3.5c (Felsenstein 0.3- 1993),which, unlike UPGMA, does not assume equal lengthsofbranches. Because phenetic analyses such as thesedo not distinguishbetween shared-derived andshared-primitive characters, treesresulting 0.2- from neighbor-joiningand UPGMA analysesare in- terpretedas diagramsdepicting overall similarity andnot asphylogenetic hypotheses of the speciesin 0.1

0.0 FIG.2. Levelsof saturationamong sequences for 0.0 o.i o.i 03 first-(top), second- (middle), and third-position sites Kimura distance (bottom). 626 KEVINJ. BURNS [Auk, Vol. 115 question.Data on songcharacteristics of four species roup) I of Pirangawere obtainedfrom Shy (1984c:table 2). tlavahepatica (hepatica group) Thesequantitative characters were comparedwith roup) tlavatextacea (luteagroup) thephylogeny of Pirangato determinewhich (if any) of the songtraits tracked the phylogeny.

RESULTS a bidentata I

Sequencevariation.--The portion of cyto- chromeb sequencedfor this studyaligned eas- -- Piranga olivacea 1 Pirangarubra rubra I ily without gaps and insertions.However, Pirangarubra rubra 2 within the intergenicspacer region between cy- piranga leucopteraamens 1 P•r•ga leucopteraamens 2 tochrome b and the tRNA, all seven individuals -- Pirangaleucoptera leucoptera • Pirangarubriceps of Pirangatiara had a deletionrelative to the Pirangaerythrocephala canida 1 other spedessequenced and to the chickense- Pir•ga erythrocephalacanida 2 Pir•ga roseogularisI quence(Desjardins and Morais 1990). These Pirangaroseogularis 2 sevensequences were aligned by handby add- Habia rubica 9,.[--.Chlorothraupis carmioli ing a gap at baseposition 16,038 relative to the Ramphoceluspasserinii chickensequence. Rampboceluspasserinii -- Ramphoceluspasserinii Of the 1,073 sites, 484 (45%) are variable. • Ramphocelusicteronotus Ramphoceluscarbo Levelsof uncorrectedsequence divergence are Ramphocelusbresilius structuredhierarchically (Fig. 1). Spedfically, -- Ramphocelusnigrogularis Ramphocelussanguinolentus divergencesbetween Pomatostomustemporalis Calochaetes coccineus and the tanagersequences range from 15.6 to Pomatostomustemporalis 19.8%, divergences among tanager genera FIG. 3. One of two most-parsimonioustrees re- rangefrom 9.6 to 13.3%,and divergenceamong suitingfrom a parsimonyanalysis in whichall char- speciesof Pirangarange from 4.2 to 10.5%.For acterswere givenequal weight. The othertree re- most species,levels of intraspecificsequence suitingfrom this analysisis identicalexcept that P. divergenceare relativelylow. For speciesother olivacea 1 and P. olivacea 2 are sister taxa. Tree is root- than Pirangatiara, levels of divergencewithin ed at Pomatostomastemporalis. Numbers on nodes re- speciesrange from 0.1 to 2.0%.Within P.tiara, fer to relativesupport of eachclade. The firstnumber thereis a greaterrange of sequencedivergence. indicatesthe numberof bootstrapreplicates (out of The lowestlevel (0.7%) was observedbetween 100)in whichthat clade is found.The second number an individual from Mexico and one from the is theBremer (or decay)index indicating the number of additionalsteps needed before a particularclade UnitedStates, whereas the highestlevel (6.1%) breaks down. was observed between an individual from Mexico and one from Bolivia. Base composition(adenine 27.4%, cytosine were phylogeneticallyinformative for parsi- 34.3%, guanine 13.1%, thymine 25.2%) was mony analyses.Giving all charactersequal similarto that observedin otherspedes in the weight resulted in two most-parsimonious Thraupidae(Burns 1997) and that reportedin trees(one shown in Fig. 3) of 1,098steps (con- otherbirds (Edwardset al. 1991,Kornegay et sistencyindex, excludinguninformative char- al. 1993,Hackett 1996). Changes at third-posi- acters[CI] = 0.44, retention index [RI] = 0.71). tion siteswere more commonthan changesat The two treesdiffered only in the arrangement second-and first-positionsites (Fig. 2). Among of individual sequencesof P. olivacea.All spe- thePiranga sequences, transitions were approx- des were retainedas monophyleticunits, and imately six times more commonthan transver- the genusPiranga was monophyleticas well. sions. Plots of Kimura distances versus uncor- Within Piranga,there was a cladeof taxa that rected sequencedivergence are linear and occurs at high elevations:P. erythrocephala roughlyfall alongthe line y = x for first- and (highlandsof Mexico), P. rubriceps(northern second-positionsites (Fig. 2). For third-posi- and central Andes), and P. leucoptera(wide- tion sites, transitions are saturated relative to spread in highlandsof Middle America,the transversions(Fig. 2). Andes, and Tepuisof SouthAmerica). Habia to- Phylogenetics.---Ofthe 484 variablesites, 294 getherwith Chlorothraupisformed a monophy- July1998] SystematicsofPiranga 627

that largelyagrees with the equallyweighted trees (one shown in Fig. 3). Two differences (luteagroup) distinguish the transversion-weightedtree

, Pirangafiava rosacea(tiara group) from the equallyweighted tree. In the equally

a ludovici•a 2 weightedtree, P. roseogularisis the mostbasal a ludoviciana 3 Piranga;however, in the transversion-weighted

a bidentata 2 tree, P. roseogularisfalls within Pirangaas the sistertaxon to the clade containingP. rubra,P. 55/1 Pir•ga olivacea2 olivacea,P. fiava, P. bidentata,and P. ludoviciana. The two treesalso differ in thatthe positionsof P. rubra and P. olivacea are reversed. 00l>10[• pirangaPiranga roseogularis 21 Levelsof supportof parsimonytrees as mea- suredby Bremerindices and bootstrapvalues variedamong nodes (Figs. 3, 4). The monophy- ly of eachspecies (except P. fiava) had a rela- 95/>10 Habia rubica [• Chlorothraupiscarmioli tively high degreeof support.In addition,the •/• Ramphoceluspasserinii •> 1o Ramphaceluspasserinii monophylyof Pirangawas stronglysupported, as was the highlandclade containing P. eryth- 821>10 Ramphocelusicteronotus 69/9 Ramphoceluscarbo rocephala,P. leucoptera,and P. rubriceps.Lower sat.10 Ramphocelusbresilius bootstrap values and smaller Bremer indices />1o•i•Ramphocelus passerlniinigrogtdaris [ •.Ramphocel....gmnol ..... were found on branchesindicating relation- Calochaetes coccineus Pomatostomustemporalis shipsamong P. rubra, P. olivacea, P.fiava, and the cladecontaining P. bidentata and P.ludoviciana. FIG. 4. Most-parsimonioustree resulting from In all 10 random-additionreplicates of the the transversion-weightedanalysis described in the text. Tree is rooted at Pomatostomastemporalis. Num- maximum-likelihoodanalysis, the samemost- berson nodesrefer to relativesupport of eachclade. likely tree was found (-In likelihood = The first numberindicates the numberof bootstrap 7,241.46).This tree was similar to the transver- replicates(out of 100) in which that cladeis found. sion-weightedparsimony tree (Fig. 4) and to The secondnumber is the Bremer(or decay)index resultsof the equallyweighted analyses (Fig. indicatingthe number of additionalsteps needed be- 3). The sametwo nodesin contentionbetween fore a particularclade breaks down. the two differentparsimony-weighting schemes also differed between the maximum-likelihood analysisand the parsimonyanalyses. Like the letic sister group to Piranga (see also Burns equally weighted trees, the maximum-likeli- 1997). Speciesof Ramphocelusformed a mono- hoodtree placed P. roseogularis basal to all other phyleticsister group to the cladecontaining the Piranga.However, both the n•aximum-likeli- Piranga, Habia, and Chlorothraupissequences. hood and transversion-weightedtrees agreed Relationshipsamong the Ramphocelussequenc- in the positionof P.rubra and P.olivacea. Other es found in all analysesof the currentstudy thanthe placementof thesethree taxa, the trees were identicalto thosefound by Hackett(1996). As foundby Burns(1997), the sequenceof Cal- obtained from all analyses of this study ochaetesdid not clusterwith either Ramphocelusshowed complete agreement in their depiction or Piranga,suggesting that plumagesimilari- of relationshipsamong species. ties of this speciesand membersof the genus Log-likelihoodvalues of the equallyweight- ed trees are similar to that of the maximum- Ramphocelusare the result of convergence,not sharedevolutionary history. likelihoodtree (Table3). Although the trans- Becauseof evidenceof saturationat third-po- version-weightedtree differsmore substantial- sition sites(Fig. 2), ! performedanother parsi- ly in log-likelihoodvalue, none of the parsi- mony analysisin which third-positiontransi- mony treeshas likelihoodvalues that are sig- tionswere given a weightone-sixth that of other nificantlydifferent from the likelihoodvalue of types of changes.This transversion-weightedthe n•aximum-likelihood tree. The n•aximum- analysisresulted in a singlemost-parsimonious likelihoodtree and equallyweighted trees are tree (3,219 steps,CI = 0.46, RI = 0.75; Fig. 4) also more similar to each other in terms of the 628 KEVINJ. BURNS [Auk, Vol. 115

TABLE3. Log-likelihood(-In L) valuesand number of stepsrequired for eachof the four topologiesob- tainedusing maximum-likelihood and parsimonyanalyses. Difference in log likelihood(A -In L _+SD) reportedrelative to the maximum-likelihoodtree.

Parsimony(length) Maximum likelihood Equally Transversion- Topology -In L A -ln L weighted weighted Maximum-likelihood tree 7,241.46 -- 1,099 3,229 Equalweight tree 1 7,241.69 0.23 +--3.73 1,098 3,233 Equalweight tree 2 7,241.47 0.01 +--5.91 1,098 3,233 Transversion-weightedtree 7,247.19 5.73 --- 9.67 1,104 3,219 numberof stepsrequired by the tree'spartic- doesnot shareany nodesin commonwith the ular topology. analysesof the cytochrome-bdata. Plumage.--Incontrast to the DNA data, Song.--Noneof the songcharacteristics de- plumagedata were muchless successful in re- scribedby Shy (1984c:table 2) showssimilari- solving relationshipsin Piranga.Analysis of tiesto the phylogenyof Piranga(Fig. 6). Species plumage charactersproduced 12 most-parsi- with more similar valuesof songtraits are not monious trees (35 steps, CI = 0.71, R! = 0.57), eachother's closest relative. For example, of the the strict consensus of which showed little res- speciesstudied by Shy(1984c), P. fiava and P. lu- olution.Only two nodeswere recoveredby all dovicianaare the most closelyrelated (Figs. 3, plumagetrees (the monophyly of erythrocephala4); however,these two speciesdo nothave sim- and rubriceps,and of rubra and roseogularis).ilar songvalues relative to the other Piranga Neitherof thesenodes was found in any of the studied. analysesof theDNA data.However, half of the 12 plumage trees show the highland clade DISCUSSION (erythrocephala,rubriceps, and leucoptera)that was recoveredin the DNA phylogenies. Previousphylogenetic hypotheses.--Few previ- Morphology.--Boththe UPGMA phenogram oushypotheses have been made concerning re- and the neighbor-joiningtree of the external lationshipswithin Piranga.Based on hybrids morphologicaldata (Table4) resultedin the between P. ludovicianaand P. olivacea(Tordoff same topology (Fig. 5). Speciesof Pirangaare 1950,Mengel 1963), Mengel (1963) argued that divided into three main groups.The small- these two specieswere closely related and bodiedP. erythrocephala and P.leucoptera form a shared a recent common ancestor. However, group;the moremedium-sized P. roseogularis, these two speciesare not closelyrelated in any P. ludoviciana,and P. olivaceaform another clus- of the DNA phylogenies.Thus, the ability for ter; and the largerP. bidentata, P.fiava, P. rubri- very limited hybridizationto occur between ceps,and P. rubraform yet a third group.The thesespecies represents the retention of a prim- phenogramproduced from morphologicaldata itive characteristic and should not be used as a

TABLE4. Mean externalmeasurements (mm) and body mass(g). Valuesfor body massobtained from Isler and Isler (1987).

Bill Bill Bill Body Species length height width Wing Tail Tarsus mass P. bidentata(n = 9) 17.22 9.44 10.29 95.37 83.87 21.45 34.0 P.erythrocephala (n = 10) 13.30 7.89 8.64 72.15 72.42 20.07 22.0 P.fiava (n = 21) 17.73 9.64 10.37 95.30 82.81 21.81 35.3 P.leucoptera (n = 8) 12.96 7.43 8.33 68.67 64.79 18.72 16.0 P.ludoviciana (n = 10) 15.14 7.98 9.08 94.70 75.14 20.59 30.0 P. olivacea(n = 11) 14.98 8.45 9.56 96.71 75.64 19.72 25.0 P.roseogularis (n = 7) 15.31 8.91 9.25 79.02 74.50 21.85 24.0 P.rubra (n = 20) 19.10 9.62 10.35 96.77 82.61 19.74 26.0 P.rubriceps (n = 15) 15.33 9.04 10.60 96.51 85.33 22.79 35.0 July1998] SystematicsofPiranga 629

Pirangaroseogularis MXFRQ MNFRQ FRQR INTR DURA Piranga olivacea

I Pirangaludoviciana Pirattga ludoviciatta 4420 1993 2427 256 222 • Pirangabidentam Piranga rubra 4378 1945 2433 232 148 • Pirangaflava Piranga olivacea 5466 2161 3305 212 230 Piranga rubriceps • Pirangaflava474418922852295199 Pirangarubra F•G.6. Songcharacters and phylogenyof Piranga. Piranga leucoptera Phylogenyshown is thetransversion- weighted tree and maximum-likelihoodtree. The songcharacters I Pirangaerythrocephala shown(MXFRQ = maximumfrequency, MNFRQ = F•a. 5. Morphologicalsimilarity amongspecies minimum frequency,FRQR = frequency range, of Pirangaas determinedby UPGMA and neighbor- INTR = averageinterval, DURA = averageduration) joininganalyses. Branch lengths are not proportion- representfive of the nine charactersof Shy (1983c). al. Shy'sother four charactersalso do not showagree- mentwith the phylogeniesof this study.Frequency values are in Hz, and duration and interval are in ms. guidelinefor relationships(Zink andMcKitrick 1995).Piranga olivacea has also hybridized with P. rubra (McCormick 1893), and P. ludoviciana Distance-Wagnertree, rubra and olivaceawere has hybridized with P. bidentata(Morse and sister taxa with ludoviciana outside of these two Monson1985). The hybridizationof P. olivacea species.This set of speciesrelationships was and P. rubradoes not agreewith phylogeny; notidentified in anyof the treesresulting from however,the hybridization of P.ludoviciana and my study.It is difficultto determinewhether P.bidentata is in agreementwith the closephy- thesedifferences are the resultof differentpat- logeneticrelationship of these two species. terns of evolution of different characters, vari- Nevertheless,hybridization is notnecessarily a ationin the abilityof differentmethods to re- good indicatorof phylogeneticrelationships. coverrelationships among taxa, or artifactsof Evenspecies with extensivehybrid zones often taxonsampling. Many studies have shown that are not eachother's closest relatives (e.g. Fox incompletetaxon sampling can have profound Sparrows[Passerella iliaca], Zink 1994;orioles, effectson final topologicalstructure (Gauthier Freeman and Zink 1995; and flickers, Moore et et al. 1988,Weller et al. 1992).I sampledall spe- al. 1991; see Zink and McKitrick 1995). ciesof Pirangaand usedcladistic-based char- In contrast to current systematictreatments acter-statedata, which resulted in a greaterca- of taxawithin Piranga(Storer 1970, AOU 1983, pabilityto recoverphylogenetic relationships. Sibley and Monroe 1990), Howell and Webb Specieslimits of P. flava.--TheHepatic Tana- (1995) placedP. leucopteraand P. erythrocephalager (P.fiava), with 15described subspecies (Sto- in the genusSpermagra apart from otherPiranga rer 1970),exhibits more geographicvariation that occurin Mexico.Given the DNA phylog- than any other Piranga.Some authors(AOU eniesresulting from the currentstudy, the use 1983,Isler and Isler 1987)have divided these15 of the genusSpermagra solely for thesetwo spe- subspeciesinto threegroups based on plumage ciesis unwarranted.For this genus to bemono- differences,disjunct distributions, and ecolog- phyletic,P. rubriceps would need to be included ical differences.Although an early reviewof in Spermagraas well. In addition, if P. roseogu- variation in P.tiara (Zimmer 1929) concluded larisis basalto all otherPiranga (as suggested that only a singlespecies was involved,differ- by the equallyweighted and maximum-likeli- encesamong these subspecies groups are pro- hoodtrees), use of the genusSpermagra would found enoughto suggestthat P.tiara as pres- makePiranga paraphyletic. Because taxonomic ently definedconsists of threespecies (Ridgely namesshould reflect monophyletic groups, the and Tudor 1989). Distributionsof these sub- genusSpermagra should not be used. speciesgroups are largely nonoverlapping. McDonald (1988) surveyedallozyme varia- Subspeciesin the hepaticagroup occur from the tion in severalwarblers and tanagers(includ- southwestern United States to northern Nica- ing P.rubra, P. olivacea, and P. ludoviciana)to de- ragua;the luteagroup occurs mostly at higher terminethe closestrelatives of thegenus Phaen- elevations from Costa Rica to Bolivia, as well as icophilus.In bothan UPGMA phenogramand a in partsof Venezuelaand northernBrazil; and 630 KEVINJ. BURNS [Auk, Vol. 115 thetiara groupoccurs mainly in southernBra- ysesof plumagedata failed to recoverphylo- zil, Paraguay,southeastern Bolivia, and north- geneticrelationships among the speciesof Pi- ern Argentina. Plumage colorationof males ranga.Two similar studies (Christidis et al. variesacross their distributionfrom deep red 1988,Hackett and Rosenberg1990) that inves- to dusky scarlet.Plumage differences between tigatedrelationships among closely related avi- the fiava and luteagroups often are greatest an speciesalso found discrepanciesbetween where individuals of these forms co-occur plumagedata and geneticallybased phyloge- (Ridgely and Tudor 1989).Habitat preferences nies. The fact that plumagecharacters show of P.fiava change across its rangewith thethree morehomoplasy than DNA datamay indicate subspeciesgroups. Representatives of the he- that plumagepatterns and colorsare not as paticgroup are found mostly in pine or pine- usefulin systematicstudies as are othertypes oak forest,the luteagroup occursin broad- of characters. leavedwoodlands, and individualsof thefiava Oneproblem with usingplumage characters group are commonlyfound in savannaand se- in phylogeneticreconstruction is thatsexual se- miopen situations(Isler and Isler 1987). lectionmay causerapid plumagedivergence Levelsof sequencedivergence and phyloge- amongtaxa and obscure phylogenetic relation- neticrelationships observed among P. fiava cor- ships.A more fundamentalissue is that the respondwell with the threesubspecies groups complexmechanisms and multiple sourcesre- of P.fiava. Sequence-divergence values between sponsiblefor plumagecolor make the identifi- someindividuals of P.fiava are muchgreater cationand codingof homologouscharacters thanthose found within otherspecies of Piran- difficult.Little is knownabout how the plum- ga(Fig. 1). Levelsof sequencedivergence with- age colorsand patternsthat we see are pro- in P.fiava at 2.5% and below representcompar- duced. The color itself may be structurally isonsbetween representatives of the samesub- basedor may comefrom a variety of pigments, speciesgroup. Levels of sequencedivergence at most notably melanins and carotenoids.Birds 3.8% and greater representcomparisons be- cannotproduce carotenoid pigments on their tween representativesof different subspecies own and must obtain them from their diets. groups.Thus, levelsof sequencedivergence Birdscan either deposit these pigments direct- amongsubspecies groups of P.fiava are asgreat ly,or thepigment may be modified into another as levelsof divergenceobserved among other pigmentprior to beingdeposited in the feath- speciesof Piranga(Fig. 1). For example,an in- ers. dividual of P. fiava representingthe hepatica The carotenoidpigments underlying plum- group differs by 6.1% from an individual rep- age color(i.e. red, orange,and yellow) in four resentingthe fiava group. Individuals of P. lu- speciesof Piranga(fiava, rubra, olivacea, and lu- dovicianaand P. olivacea show, on average,a sim- doviciana)have been studied in detail (Brush ilar levelof sequencedivergence. 1967, Hudon 1990). One conclusionof these In all phylogeneticanalyses in this study, studiesis that colorsthat appearidentical may membersof the samesubspecies group cluster comefrom completelydifferent sources.For ex- together into monophyleticgroups. These ample,P. ludovicianaand P. olivaceahave a red monophyleticgroups show strong support headthat is indistinguishableto thehuman eye both in termsof bootstrapvalues and Bremer under natural, artificial, and near-ultraviolet il- indices.Thus, both levels of sequencediver- luminationand alsois indistinguishableusing genceand the phylogeny of sequencesare con- reflectancespectrophotometry (Hudon 1990). sistentwith theidea that P. fiava as currently de- However,isolation of the carotenoidpigments scribedcontains more than one evolutionary responsiblefor the red colorationin thesetwo unit. The different subspeciesgroups are speciesrevealed that differentcarotenoid pig- monophyleticand diagnosablydistinct based mentsare responsiblefor producingthe iden- on plumagecharacters. The DNA data of this tical head color(Hudon 1990).In olivacea,sev- studyadd to themorphological, distributional, eral commoncarotenoid pigments produce the and ecologicalevidence that suggestthat the red coloration. In contrast, the carotenoid re- threesubspecies groups of P.fiava represent dif- sponsiblefor the red colorationin P.ludoviciana ferentphylogenetic, if not biological,species. is rhodoxanthin,which is depositeddirectly Evolutionof plumagecolors.--Parsimony anal- into the feathersin an unmodifiedform (Hudon July1998] SystematicsofPiranga 631

1990).When ! codedplumage characters in the Piranga roseogularis parsimonyanalysis, I assignedthe samechar- acter statefor head color in both P.olivacea and __ Piranga rubriceps P.ludoviciana. Because the ultimate sourceof the red colorationof thehead is quitedifferent in __ Piranga leucoptera thesetwo species,my originalcoding is an oversimplifieddepiction of plumagecolor in Piranga erythrocephala thesespecies. Unfortunately, not enough pig- Piranga olivacea mentsand not enoughtaxa have been studied for thesedata to be incorporatedinto a more Piranga rubra thoroughphylogenetic analysis. The results of this studysuggest that plumagecolor (es- • Piranga tiara peciallyin ignoranceofits chemical basis) may beless useful than other characters for phylo- Key __ Pirangabidentata geneticanalysis. • non-migratory Sexualdimorphism in plumage.--The phylo- • migratory Piranga ludoviciana geneticposition of P.rubriceps (Figs. 3, 4) indi- FIG.7. Evolutionof seasonalmigration mapped catesthat the relatively smaller degree of sex- ontothe maximumlikelihood tree of Pirangarela- ualdimorphism in plumageof thisspecies is a tionships.Key: nonmigratory = species does not mi- derivedcondition. In thisspecies, females have grateseasonally, migratory - speciesmigrates sea- evolvedmore brightly coloredplumage, sonally in at leastpart of itsrange. whereasmale plumage has remained brightly colored.Such changes in femaleplumage rath- erthan male plumage are common in tanagers(Van Buskirk 1997).In contrast,none of the in- (Burns1996) and often cause changes in thede- dividualsong characters ofPiranga tracks phy- greeof sexualdimorphism. This suggests that logeny(Fig. 6). In additionto phylogeny, other theevolution ofsexual dimorphism inplumage factors can influence song. Body size constrains is affectedby other factors(such as socialse- songfeatures such as frequency, and physical characteristicsof the habitatcan have selective lectionon females; Irwin 1994)as well as sexual selectionfor brightly coloredmales. Unfortu- influenceson songbecause some types of vo- nately,little is known of the habits and nothing calizationstransmit better than others in a giv- isknown of the breeding biology of P. rubriceps en habitat(Morton 1975, Badyaev and Leaf (Isler and lsler 1987).Thus, the reasonsfor the 1997).These factors have probably had a strong relativelysmall degreeof plumagedimor- influenceon shapingsong variation among phisinin P.rubriceps remain unknown. speciesof Piranga.Habitat openness is corre- Externalmorphology and song variation.--As in latedwith vocal characteristics among popu- lationsof P.rubra (Shy 1983). Additional anal- otherlineages of birds (e.g. Hackett and Rosen- ysesmay showthat this relationshipexists berg1990, Zink and Blackwell 1996), morpho- amongother species of Pirangaas well. metricvariation among Piranga is not correlat- Biogeographyand the evolution of seasonalmi- edwith genetic variation. Overall similarity in gration.--Thephylogeny of Pirangapresented morphology(Fig. 5) showsno similaritiesto in thisstudy provides a frameworkfor address- phylogeniesderived from DNA sequence data. ing distributionalpatterns among species, in- Thus,size and shape have changed frequently cluding the evolutionof seasonalmigration. in Piranga.For example, the clade of P.erythro- Fourof thenine species of Piranga(fiava, ludov- cephala,P. leucoptera,and P.rubriceps contains iciana,olivacea, and rubra) are seasonal migrants thetwo smallest Piranga and one of the largest thatbreed in NorthAmerica. Some P. fiava also (i.e. rubriceps). migrate seasonallyin southernSouth America. Recentstudies have shown a relationshipbe- Mappingthe evolutionof seasonalmigration tweenphylogeny and certainvocal character- ontoany of the DNA trees (e.g. Fig. 7) indicates istics.For example, fundamental frequency cor- thatthe most-parsimonious explanation for the respondswith phylogenyin herons (Mc- evolutionof thistrait is thatmigration evolved Crackenand Sheldon1997), and durationof onceon the branch leading to the monophyletic notesis associatedwith phylogenyin warblers groupof P.rubra, P. olivacea, P.fiava, P. bidentata, 632 KEVINJ. BURNS [Auk, Vol. 115 and P. ludoviciana.Thus, thesefive taxa proba- grant (DEB-9321316),the AmericanMuseum of Nat- bly evolvedfrom a migratoryancestor, and lat- ural History (FrankM. ChapmanFund), the Amer- er migratorybehavior was lost in P. bidentata icanOrnithologists' Union (Donald L. BleitzAward), and in somepopulations of P.fiava. and the Museum of VertebrateZoology (Wilhelm MartensFund, Annie AlexanderFellowship, and Jo- The phylogenyof Pirangaalso provides a sephMalliard Fellowship).A collectionstudy grant perspectiveon relationshipsamong areasof from the AmericanMuseum of Natural History al- endemism in North America. As mentioned lowed me to visit the AMNH and measurespeci- above,P. ludoviciana(breeds in westernNorth mens.N. K. Johnson,J. L. Patton,J. W. Taylor,S. J. America) and P. olivacea(breeds in eastern Hackett, A. J. Baker, and an anonymousreviewer North America)previously were thoughtto providedhelpful commentson the manuscript. represent an eastern/western sister-species pair (Mengel1963). However, the cytochrome- LITERATURE CITED b data clearlyshow that thesetwo speciesare AMERICAN ORNITHOLOGISTS' UNION. 1983. Check- not closelyrelated. Pirangaludoviciana is more list of North American birds, 6th ed. American closelyallied to P.bidentata (a Mexicanand Cen- Ornithologists'Union, Washington, D.C. tral Americanspecies that has recentlyoc- BADYAEV, A. V., AND E. S. LEAF. 1997. Habitat asso- curred in the southwesternUnited States)and ciationsof song characteristicsin Phylloscopus P.fiava (which occursfrom the southwestern and Hippolaiswarblers. Auk 114:40-46. United Statesto southernSouth America). In BALDWIN, S. P., H. C. OBERHOLSER,AND L. G. WOR- additionto Piranga,many other taxa have close- LEY. 1931. Measurement of birds. Scientific Pub- ly relatedforms in eastern,western, and south- lications of the Cleveland Museum of Natural western North America (Zink and Hackett History Vol. 2. 1988).The phylogenyof Pirangaindicates that BERBEE,M. L., A. YOSKIMURA,J. SUGIYAMA,AND J. W. western North America has a more recent as- TAYLOR.1995. IS Penicilliummonophyletic? An sociation with southwestern North America evaluationof phylogenyin the family Tricho- comaceae from 18S, 5.8S and ITS ribosomal than it has with eastern North America. Two DNA sequencedata. Mycologia87:210-222. othermolecular phylogenies of the 11 taxalist- BREMER,K. 1988.The limits of amino-acidsequence ed by Zink andHackett (1988) also indicate that datain angiospermphylogenetic reconstruction. relationshipsamong these areas follow the pat- Evolution 42:795-803. tern ([west + southwest]+ east).Phylogenies BROOKS,D. R., ANDD. H. MCLENNAN.1991. Phylog- of orioles(Freeman and Zink 1995),and flickers eny,ecology, and behavior:A researchprogram (Mooreet al. 1991)support the samepattern of in comparativebiology. University of Chicago area relationshipsindicated by the phylogeny Press,Chicago. of Piranga.Additional phylogenetic studies of BRUSH,A. H. 1967.Pigmentation in the ScarletTan- ager,Piranga olivacea. Condor 69:549-559. co-distributedtaxa are neededto fully address BURNS,K. J. 1996.Molecular phylogenetics of tana- the strengthof this area cladogram. gersand the evolutionof sexualdimorphism in plumage.Ph.D. dissertation,University of Cali- ACKNOWLEDGMENTS fornia, Berkeley. BURNS,K. J. 1997. Molecular systematicsof tanagers I am grateful to the following for providing tis- (Thraupinae):Evolution and biogeographyof a sues: Museo de Zoologia at Universidad Nacional diverse radiation of Neotropicalbirds. Molecu- Autonoma de Mexico (A. G. Navarro S.); Museum of lar Phylogeneticsand Evolution8:334-348. Natural Scienceat Louisiana State University (F. H. CHRISTIDIS,L., R. SCHODDE,AND P. R. BAVERSTOCK. Sheldonand R. M. Zink); Museode Zoologia,Cole- 1988.Genetic and morphologicaldifferentiation gio de la Frontera Sur (C. Pozo de la Tijera); United and phylogenyin the Australo-Papuanscrub- StatesNational Museum (M. J.Braun); and American wrens (Sericornis,Acanthizidae). Auk 105:616- Museumof Natural History (G. F.Barrowclough and 629. P. R. Sweet).I especiallythank Adolfo Navarroand CICERO,C., AND N. K. JOHNSON.1995. Speciationin personnelat the Museode Zoologiafor obtaininges- sapsuckers(Sphyrapicus): III. Mitochondrial- sentialspecimens and for their hospitalitywhile I DNA sequencedivergence at the cytochrome-b visited their institution. For assistance in the lab, I locus. Auk 112:547-563. thankC. Orrego(who designed several primers used CRACRAFT,J., AND R. O. PRUM. 1988. Patterns and in this study), M. N. E da Silva,M. F. Smith,and S. processesof diversification:Speciation and his- Simpson.Funding for differentparts of this project torical congruencein some Neotropical birds. was provided by: an NSF dissertationimprovement Evolution 42:603-620. July1998] SystematicsofPiranga 633

DESJARDINS,P., AND R. MORAIS.1990. Sequenceand JOHNSON,N. K., ANDJ. A. MARTEN.1992. Macrogeo- geneorganization of the chickenmitochondrial graphic patternsof morphometricand genetic genome.Journal of MolecularBiology 212:599- variationin the SageSparrow complex. Condor 634. 94:1-19. EDWARDS,S. V., P. ARCTANDER,AND g. C. WILSON. KIMURA,M. 1980. A simple method for estimating 1991.Mitochondrial resolution of a deepbranch evolutionaryrate of basesubstitutions through in the genealogicaltree for perchingbirds. Pro- comparative studies of nucleotide sequences. ceedingsof the RoyalSociety of Londonseries B Journal of Molecular Evolution 16:111-120. 243:99-107 KISHINO, H., AND g. HASEGAWA.1989. Evaluation of FELSENSTEIN,J. 1985. Confidence limits on phyloge- the maximum likelihood estimate of the evolu- nies:An approachusing the bootstrap.Evolu- tionary tree topologiesfrom DNA sequence tion 39:783-791. data, and the branchingorder in Hominoidea. FELSENSTEIN,J. 1993. PHYLIP (PhylogenyInference Journal of Molecular Evolution 29:170-179. Package)version 3.5c. Distributed by the author KORNEGAY,J. R., T. D. KOCHER,L. A. WILLIAMS,AND Departmentof Genetics,University of Washing- A. C. WILSON.1993. Pathways of lysozymeevo- ton, Seattle. lution inferred from the sequencesof cyto- FREEMAN,S., ANDR. M. ZINK. 1995.A phylogenetic chrome b in birds. Journal of Molecular Evolu- studyof the blackbirdsbased on variationin mi- tion 37:367-379. tochondrialDNA restrictionsites. Systematic Bi- MCCORMICK,L. M. 1893.A hybridtanager. Auk 10: ology 44:409-420. 302-303. GAUTHIER,J., A. G. KLUGE,AND t. ROWE.1988. Am- MCCRACKEN, K. G., AND E H. SHELDON. 1997. Avian niote phylogenyand the importanceof fossils. vocalizationsand phylogeneticsignal. Proceed- Cladistics 4:105-210. ings of the NationalAcademy of SciencesUSA HACKETT,S. J. 1996. Molecular phylogeneticsand 94:3833-3836. biogeographyof tanagersin the genusRampho- McDONALD,M. A. 1988.The significanceof hetero- celus(Aves). Molecular Phylogenetics and Evo- chronyto the evolutionof Hispaniolanpalm- lution 5:368-382. tanagers,genus Phaenicophilus: Behavioral, mor- HACKETT,S. J., C. S. GRIFFITHS,J. M. BATES,AND N. phologicaland geneticcorrelates. Ph.D. disser- K. KLEIN.1995. A commentaryon the use of se- tation, University of Florida, Gainesville. quencedata for phylogenyreconstruction. Mo- MENeEL,R. M. 1963.A secondprobable hybrid be- lecularPhylogenetics and Evolution4:350-353. tween the Scarletand Westerntanagers. Wilson HACKETT,S. J., AND K. V. ROSENBERG.1990. Compar- Bulletin 75:201-203. isonof phenotypicand geneticdifferentiation in MENGEL,R. M. 1964.The probablehistory of species South American antwrens (Formicariidae).Auk formation in somenorthern wood warblers (Pa- 107:473-489. rulidae).Living Bird 3:9-43. HAMMER,J. 1985. Avian zoogeographyof the Neo- MILLER, S. A., D. D. DYKES, AND H. E POLESKY.1988. tropicallowlands. Pages 113-146 in Neotropical A simplesalting procedure for extractingDNA ornithology(E A. Buckley,M. S. Foster,E. S. from human nucleated cells. Nucleic Acids Re- Morton, R. S. Ridgely,and E G. Buckley,Eds.). search 16:215. OrnithologicalMonographs No. 36. HARVEY,P. H., ANDg.D. PAGEL.1991. The compar- MOORE, W. S., J. H. GRAHAM, AND J. T. PRICE. 1991. Mitochondrial DNA variation in the Northern ative method in evolutionarybiology. Oxford UniversityPress, Oxford. Flicker (Colapresauratus). Molecular Biology and HOWELL,S. N. G., ANDS. WEI313.1995. A guideto the Evolution 8:327-344. birds of Mexico and northern Central America. MORSE,R. J., AND G. MONSON. 1985. Flame-colored Oxford University Press,Oxford. Tanagerin Arizona.American Birds 39:843-844. HUBBARD,J.P. 1973.Avian evolutionin the aridlands MORTON,E. S. 1975.Ecological sources of selection of North America. Living Bird 12:155-196. on avian sounds. American Naturalist 109:17- HUDON,J. 1990.Unusual carotenoid use by the West- 34. ern Tanager(Piranga ludoviciana) and its evolu- NEI, M. 1987. Molecular evolutionary genetics.Co- tionary implications.Canadian Journal of Zool- lumbiaUniversity Press, New York. ogy 69:2311-2320. PAYNE,R. B. 1986.Bird songsand avian systematics. IRWIN,R. E. 1994.The evolutionof plumagedichro- Current Ornithology3:87-126. matism in the New World blackbirds: Social se- RAND,A. L. 1948. Glaciation,an isolating factor in lectionon femalebrightness? American Natu- speciation.Evolution 2:314-321. ralist 144:890-907. RIDGELY,R. S., AND G. TUDOR. 1989. The birds of ISLER,g. L., ANDP.g. ISLER,1987. The tanagers:Nat- South America, vol. 1 The oscinepasserines. ural history, distribution, and identification. University of TexasPress, Austin. SmithsonianInstitution Press, Washington, D.C. RIDGWAY,R. 1902. The birds of North and Middle 634 KEVINJ. BURNS [Auk, Vol. 115

America,part 2. Bulletin of the United States for PCR-based typing from forensic material. National Museum No. 50. Biotechniques10:506-513. RO•ILF,E J.1988. NTSYS-pc numerical and WELLER,S. J., t. 1•. FRIEDLANDER,J. A. MARTIN, AND multivariate analysissystem, version 1.50. Exe- D. 1•. i•ASHLEY.1992. Phylogenetic studies of ri- ter Publications, Setauket, New York. bosomal RNA variation in higher moths and S•¾, E. 1983.The relationof geographicalvariation butterflies (Lepidoptera: Ditrysia). Molecular in songto habitatcharacteristics and body size Phylogeneticsand Evolution1:312-337. in North Americantanagers (Thraupinae: Piran- ZIMMER,J. T. 1929.A study of the Tooth-billedRed ga).Behavioral Ecology and Sociobiology12:71- Tanager,Piranga fiava. Zoological Series of Field 76. Museum of Natural History 7:169-219. S•¾, E. 1984a.Habitat shift and geographicalvaria- ZiN•C,R. M. 1994.The geographyof mitochondrial tion in North Americantanagers (Thraupinae: DNA variation,population structure, hybridiza- Piranga).Oecologia 63:281-285. tion,and specieslimits in the FoxSparrow (Pas- S•¾, E. 1984b. The structure of song and its geo- serellailiaca). Evolution 48:96-111. graphicalvariation in the ScarletTanager (Piran- ZINK, g.M., AND g. C. BLACKWELL.1996. Patterns of ga olivacea).American Midland Naturalist 112: allozyme,mitochondrial DNA, and morphomet- 119-130. ric variation in four sparrow genera. Auk 113: S•¾, E. 1984c. Sympatry and allopatry of song in 59-67. North Americantanagers. Behavioral Ecology ZINK, g. g., AND S. J. HACKETT.1988. Historical bio- and Sociobiology15:189-195. geographicpatterns in the avifauna of North SIBLEY,C. G., AND B. L. MONROE, JR. 1990. Distri- America. Pages2574-2580 in Acta XIX Congres- butionand taxonomyof the birds of the world. sus InternationalisOrnithologici (H. Ouellet, Yale University Press,New Haven, Connecticut. Ed.). Ottawa, Ontario, 1986. National Museum SMIT•, A. B. 1994.Rooting molecular trees: Problems of Natural Sciences, Ottawa. and strategies.Biological Journal of the Linnean ZINK, g.M., AND M. C. McKIT•ICK. 1995. The debate Society51:279-292 over speciesconcepts and its implicationsfor or- SMITH,g. F., AND J. L. PATTON.1993. The diversifi- nithology.Auk 112:701-719. cation of South American murid rodents: Evi- AssociateEditor: A. J. Baker dencefrom mitochondrialDNA sequencedata for the akodontinetribe. Biological Journal of the Linnean Society50:149-177. APPENDIX.Voucher numbers of specimensused in SNEATH,P. H. A., AND g. g. SOKAL.1973. Numerical morphologicalanalyses. See Table 1 for institu- taxonomy.W. H. Freeman, San Francisco. tional abbreviations. Abbreviations after MZFC re- STORER,R. W. 1970.Subfamilies Thraupinae and Ter- fer to collector's initials. sininae,Pages 246-409 in Check-listof birds of the world, vol. 13 (R. A. Paynter,Jr., Ed.). Mu- AMNH: 198831, 36413, 40857, 46739, 63942, 65943, 6678Z 66789, 66790, 66791, 66792, 71972, 110274, seumof ComparativeZoology, Cambridge, Mas- 110469, 113084, 154512, 176342, 180863, 183006, sachusetts. 235319, 235320, 235321, 235322, 364065, 364094, SWAINSON,W. 1827.A synopsisof the birds discov- 364106, 364114, 387961, 392574, 392578, 392586, ered in Mexico by W. Bullock and Mr. William 510309, 510350, 510355, 510356, 510357, 808804 Bullock.Philosophical Magazine (New Series)1: MVZ: 294, 5950, 13415, 13425, 14935, 16130, 36713, 364-369, 433-442. 36715, 40110, 41619, 54355, 76563, 84139, 86476, SWOFFORD,D. L. 1993.PAUP. Phylogenetic analysis 86478, 86481, 86483, 96637, 105072, 105076, using parsimony,version 3.1.1. Illinois Natural 105077, 105078, 108844, 114467, 114469, 114470, HistorySurvey, Champaign. 114475, 121933, 124536, 126289, 126290, 133086, TORDOFF,H. B. 1950. A hybrid tanagerfrom Min- 135780, 139019, 139021, 139023, 139024, 142971, nesota. Wilson Bulletin 62:3-4. 147449, 147478, 150084, 153529, 156375, 161190, VAN BUSKIRK,J. 1997.Independent evolution of song 161191, 161192, 161651, 162131, 191198, 105083, 105086, 135776, 139022, 5299, 64883, 78761 structure and note structure in American wood warblers. Proceedingsof the Royal Societyof MZFC: AGNS 0300, AGNS 478, BMM 496, BMM 820, London Series B 264:755-761. FMNH 7901, JEMP 0002, JEMP 044, JEMP 0067, WALSH,P.S., D. A. METZGER,R. HIGUCHI. 1991. Chel- JEMP 0079, JEMP 0128, JEMP 0285, JEMP 0354, JEMP 0364, JEMP 0418, OMVP 552 ex 100 as a medium for simpleextraction of DNA