Phylogeny of the Falconidae Inferred from Molecular and Morphological Data

Home , Caracara (subfamily), Falconidae, Falconinae, Polihierax

The Auk 116(1):116-130, 1999

PHYLOGENY OF THE FALCONIDAE INFERRED FROM MOLECULAR AND MORPHOLOGICAL DATA

CAROLE S. GRIFFITFIS • Departmentof Ornithology,American Museum of NaturalHistory, Central Park West at 79thStreet, New York, New York 10024, USA

ABSTR^CT.--Moleculardata and variationin syringealmorphology were used to infer a phylogenyfor the family Falconidaeand to addressthree issues currently of interestin systematics:(1) the treatmentof multiple data setsin phylogeneticanalysis, (2) a priorianalysis and differentialweighting of moleculardata, and (3) the reliabilityof molecularversus morphologicaldata in phylogeneticanalysis. Problems in recoveringphylogenetic signal caused by rapidly changingsites in the moleculardata were not solvedby combiningdata sets. Differentiallyweighting saturated partitions of the sequencedata, prior to phylogenetic analysis,provided a phylogenycongruent with morphologicalanalysis. Molecular data provide substantiallymore informative characters than morphological data. However, morphologicaldata provide a higherproportion of unreversedsynapomorphies. A reclassification of the family basedon the phylogenyresults in two subfamilies:(1) the Herpetotherinae, (forest-falcons[Micrastur] and LaughingFalcon [Herpetotheres cachinnans]); and (2) the Fal- coninae,which includes the tribesFalconini (Spot-winged Falconet [Spiziapteryx circumcinc- tus],pygmy-falcons [Polihierax], falconets [Microhierax], and the genusFalco) and Caracarini (caracaras).The phylogenyalso indicates that two genera,Daptrius and Polihierax, are polyphyletic,and thesetwo are split.Finally, a biogeographichypothesis derived from tl•ephy- logenyimplies that the origin and earlydiversification of thefamily occurred in SouthAmer- ica. Received12 November1997, accepted18 June1998.

THE FAI•C¸NIDAE,one of three families in the genera:the four caracara genera, Micrastur (for- orderFalconiformes (the diurnal birds of prey), est-falcons),Herpetotheres (Laughing Falcon) includessome of the fastestand mostspectac- and Spiziapteryx(Spot-winged Falconet). The ular birds in the world (Brown and Amadon relationshipsof the latterthree genera histori- 1968).Although the familyis renownedfor the callyhave been problematic (Sharpe 1874, Gur- huntingability of its species(e.g. the Peregrine ney 1894,Swann 1922, Peters 1931). The place- Falcon[Falco peregrinus]), foraging habits with- ment of thesegenera with the caracaras(Ama- in thefamily are diverse. Thus, the Neotropical don and Bull 1988) is in conflictwith results caracaras(Daptrius, Milvago, Polyborus,and from four cladisticanalyses: (1) an osteological Phalcoboenus)are scavengersthat subsistmain- study (Becker1987; Fig. 1A), (2) a preliminary ly on carrion or invertebrates. morphologicalanalysis (Kemp and Crowe In additionto adaptationsrelated to foraging 1990;Fig. 1A), (3) an analysisbased on syrin- habits, the four caracaragenera share other gealmorphology (Griffiths 1994a; Fig. lB), and morphologicaltraits (Friedmann 1950) and his- (4) an analysisbased on cytochrome-bsequenc- toricallyhave been considered to be closelyre- es (Griffiths 1997;Fig. 1C). Basedon Griffiths lated (Sharpe1874, Gurney 1894, Swann 1922, (1994a),the AOU (1998)reclassified the family, Peters1931). Classification of the othergenera placing Herpetotheresand Spiziapteryxwithin in the familyhas been less stable (Sharpe 1874, the Falconinae and Micrastur in its own sub- Gurney 1894, Swann 1922, Peters1931). Ama- family. don and Bull (1988)allocated the 10 currently In thispaper, I reanalyzethe syringealdata, recognizedgenera into two subfamilies.The incorporatingnew information.Having collect- Falconinaeincluded the speciosegenus Falco ed both morphologicaland moleculardata, I and the smallestdiurnal raptors, Polihierax then examinea controversialissue in phylo- (pygmy-falcons)and Microhierax(falconets). geneticinference, the appropriateanalysis of The Polyborinaeincluded seven Neotropical multiple data sets.This controversyspans a continuum of ideas. At one end is the idea that E-mail:[email protected] data setsshould be analyzedseparately. The

116 January1999] FalconidaePhylogeny 117

A nentsof totalevidence propose giving all char- Microhierax and Polihierax actersequal weight initially in phylogenetic • Falco analysis.Assessments of characterreliability Caracaraspecies derivedfrom phylogeneticinference (i.e. suc-

Micrastur cessiveapproximations; Farris 1969, 1989; Car- penter1988) can thenbe usedto weightchar- t HorpotothorosSpiziaptoryx actersdifferentially (a posterioriweighting). A prioridifferential weighting is criticizedfor in- B • Falco curring unwarrantedassumptions about pro- Microhiorax cessor requiringpartitions of datafor weight- Polihioraxsomitorquatus Spiziaptoryx ing that maybe arbitrary(Eernisse and Kluge Polihieraxinsignis 1993, Brewer and DeSalle 1994, Chippindale Horpotothoros and Wiens 1994). Caracaraspecies As part of the empiricalexploration of the Micrasturspecies analysisof multipledata sets, I examinethe ef- Outgroups fectsof differentialweighting on phylogenetic inference.I then comparethe morphological and moleculardata. The phylogenyinferred c • Falco from thesedata is usedto reclassifythe Falcon-

Microhiorax idaeand two generawithin thatfamily. Finally, Polihioraxsomitorquatus the phylogenyprovides a frameworkfor dis- Spiziapteryx cussingthe biogeographyof the family. Caracaraspecies MATERIALS AND METHODS MicrasturHorpotothorosspecies Accipiter Morphologicaldata.--Characters were derived from variation in syringeal supporting elements,mem- FIC. 1. Phylogenetichypotheses inferred from branes,and muscles(Griffiths 1994a). Supporting el- cladisticanalyses of the Falconidae.(A) Phylogeny ementsinclude ringlike elements on the tracheaand inferred from osteologicaldata (Becker1987) and a bronchi(A and B elements;Ames 1971 ), thepessulus, preliminaryanalysis of skinsand characters from lit- and accessorycartilaginous structures. Hemology of erature(Kemp and Crowe1990). (B) Phylogenyin- the ringlikeelements was traditionallybased on the ferredfrom syringealmorphology (Griffiths 1994a). relativeposition of thesestructures to the tracheo- (C) Phylogenyinferred from differentially weighted bronchialjunction (King 1989). This analysis postu- cytochrome-bsequences (Griffiths 1997). lateshomologous elements following Ames' (1971) definitionsof syringealstructures (Griffiths 1994b). Dataon variationin syringealmorphology for two congruenceof phylogeniesinferred from each specieswere addedto the originaldata of Griffiths dataset (taxonomiccongruence) then provides (1994a).The first, Accipiterstriatus (Sharp-shinned a measureof the reliabilityof the phylogenies Hawk; AMNH 18761), was substituted for Gampso- (Lanyon1993, Miyamoto and Fitch 1995).At nyx swainsoniito be consistentwith the molecular the oppositeend is the ideathat multipledata data.The second,Herpetotheres cachinnans (LSUMNS setsshould be combinedfor phylogeneticanal- 123200),recently became available. Because the Her- ysis (total evidenceor charactercongruence; petotheressyrinx used to codecharacters for theorig- Kluge 1989). inal analysis(Griffiths 1994a) had been damaged, Between these two extremes, proposed this new specimenwas examined and the resulting methodsfor analyzingmultiple data setsin- information used to recede two of the characters. Character16, the modificationof the ends of the B1 clude: (1) conditional combinationof data sets, element,changed from one to zero.Character 3, dor- i.e. combinationonly if thedata sets are not sig- sal fusionof the tympanum,changed from one to nificantlyheterogeneous (Bull et al. 1993,Huel- two. In addition,three characters (characters 5, 6 and senbecket al. 1996); or (2) unconditionalcom- 7) that were originallycoded as binary representa- binationof data setswith differentialweight- tions of alternative character states were receded as ing of charactersto accommodateheterogene- one unordered multistate character. ity (Chippindaleand Wiens1994). Differential Syringealcharacter descriptions with thesemod- weightingis, in itself, controversial.Propo- ificationsare presented in Appendix1. Thefinal data 118 CAROLES. GRIFFITHS [Auk, Vol. 116

TABLE1. Speciesincluded in the phylogeneticanalyses.

Species Combinedanalysis Morphological Molecular Milvago chimachima * * * M. chimango no * no Polyborusplancus * * * Phalcoboenus australis * * * Daptriusater * * * D. americanus * * * Spiziapteryxcircumcinctus * * * Micrasturgilvicollis * * * M. semitorquatus * * * Polihieraxinsignis no * no P. semitorquatus * * * Microhieraxerythrogonys * * * Herpetotherescachinnans * * * Falcoberigora no * no E sparverius * * * E mexicanus no * no E peregrinus * * * E rufigularis no * no E biarmicus no * no E columbarius no * no E cenchroides no * no E vespertinus * no * E femoralis * * * Accipiterstriatus * * * Otus asio no * no Pelecanus onocrotalus no * no

matrixfor the reanalysisconsisted of 25 taxaand 23 able.All 10 currentlyrecognized genera were sam- characters,six of which were multistate(Appendix pled for both molecularand syringealcharacters (Ta- 2). Transformation series for multistate characters ble 1). However,taxon sampling was not identicalbe- were proposedif adjacentderived states were simi- tweenthe two datasets. Some species sampled with- lar, and eachsucceeding state was a modificationof in the genusFalco varied betweenthe two, and no the previousstate; that is, the derivedstates formed tissuesamples of Polihieraxinsignis (White-rumped a nestedset of synapomorphies.Multistate charac- Pygmy-Falcon)were available. An initial analysis ters were coded as unordered if states were alterna- was run that includedall species.The numberof tive variations of a character. missingcharacters in the morphologicaldata set re- Moleculardata.--The completesequences for cyto- suited in conflict in relationshipsamong the Falco chromeb were collectedfor speciesfrom eachof the speciesand, therefore,thousands of most-parsimo- 10 generain the family (Griffiths1997; Genbank ac- nioustrees. The topology of thestrict consensus tree cessionnumbers U83305 to U83320). Details of DNA fromthat analysis was the same as the topology pro- extraction,primers used, PCR and sequencingpro- ducedwhen specieswith missingdata were deleted tocols, and data analysisare describedin Griffiths from the analysis.Because this studywas not an at- (1997).In that study,molecular data were differen- temptto resolverelationships within the genus Falco, tially weightedprior to phylogeneticanalysis by fil- speciesnot includedin the molecularanalysis were tering saturatedsubsets of data. These partitions were identified in an assessment of saturation of sub- pruned from the combineddata matrix for the phy- stitutions within domains of cytochromeb (the logenetictests. transmembrane,561 basepairs; two extra-membrane Accipiterstriatus (Accipitridae), the sistertaxon to regionsof 240and 342 base pairs; Griffiths 1997). The the Falconidae(Griffiths 1994b),was used to root the filteredpartitions were first- andthird-position tran- tree. The rationalefor using one outgroupwas de- sitions,and first-positiontransversions and second- tailed by Griffiths (1997).Basically, several species positiontransitions in theextra-membrane domains. within the Accipitridaewere included initially to de- Combineddata.--The syringeal morphological data termine the root (Smith 1994).In the final analysis, were combinedwith cytochrome-bsequences (Grif- only oneoutgroup species was used. This rootedthe fiths 1997). Osteologicalcharacters (Becker 1987) cladogramin the sameplace and producedthe iden- were not included because the data matrix for that tical phylogenyas the analysiswith multiple out- analysisand details of the analysiswere not avail- groups.The additionaloutgroup species were delet- January1999] FalconidaePhylogeny 119 ed to facilitatethe detailedexamination of changein Accipiter Herpetotheres molecularcharacters within the family. 23 To test the differentmethods proposed for com- 2 9* bining data for phylogeneticinference, two setsof 5'12' E M,icrastur semitorquatusgilvicollis phylogeneticanalyses were performed.In the first Daptriusamericanus set (i.e. the total evidencetest), all characterswere Polyborus weightedequally for an initial phylogeneticanalysis. 6 4* 8 Data were thenweighted again to assessthe effects • Daptriusater of a posterioridifferential weighting. Weights were Milvagochirnachirna -- Milvagochimango derivedfrom the fit of the charactersto a phyloge- 22 netichypothesis, using two differentmethods. In the Phalcoboenus 2 23 first method, the maximum value of the rescaledcon- PolihieraXinsignis 16 3 sistencyindex was used to reweightcharacters (suc- Spiziapteryx cessiveapproximations; Farris 1969, 1989;Carpenter 3 Microhierax 1988). In the second method, the characterswere Polihieraxsernitorquatus weightedconcurrently with the analysisbased on 7 the homoplasyimplied by eachtree (Goloboff1993). • Falco cenchroides The weightingfunction used to evaluatetrees was a • Falcobedgora F modificationof the consistencyindex. • Falco colurnbarius The secondset of analysesexamined the effectof Falcorufigularis combiningdata using a prioridifferential weighting Falco femoralis of charactersto accommodateheterogeneity of data Falcosparverius sets (unconditionalcombination; Chippindale and Wiens 1994). Molecular data were differentially Falco biarrnicus weightedas in Griffiths(1997). Because the molec- •8' Falco mexicanus ular phylogenyinferred in thatstudy was congruent Falcoperegdnus _ __ with the morphologicalphylogeny, the combined analysiswas performed to determinethe effectof the FIG.2. Phylogenyof the Falconidaeinferred from addition of the morphologicaldata on supportfor syringealmorphology; F = Falconinae,C = Polybor- nodes.Bootstrap analyses were performedon the inae.Cladogram represents the strictconsensus tree molecular data and then on the combined data to of 54 most-parsimonioustrees of length50; CI = 0.59 and RI = 0.81. Characters(Appendix 2) supporting provide heuristic measuresfor comparingnodal each node are labeled. * = character with CI = 1.0. support. Characters3, 4, 5, 6, 16, and 19 are multistate. Withinthese sets of analyses,the effectof ordering four of the morphologicalcharacters was examined. Ordering effectivelyweights character-statetrans- formationsby forcingintermediate steps in the tran- (1) caracaras;and (2) Falcospecies, including sition from the first to the last character state. Thus, the two small falconets (Microhieraxand Poli- in eachset, morphological characters were first treat- hierax)and Spiziapteryx.Herpetotheres and Mi- ed as unordered, and then four of the six multistate crasturform a dichotomybasal to thesetwo characters were ordered. To determine the effect of clades.This phylogenyis congruentwith the ordering morphologicalchanges on support for osteologicaldata (Becker1987), the molecular nodes, bootstrap analyseson the combineddata were performedwith morphologicalcharacters or- data (Griffiths 1997),the morphologicaldata dered and then unordered. (Kemp and Crowe 1990), and the original sy- All phylogeneticanalyses were conductedusing ringeal data in: (1) establishingthese two the testversion 4.0d54 of PAUP*,written by David L. clades, (2) placing Micrastur basal to these Swofford.Each phylogenetic analysis employed the clades,and (3) placingSpiziapteryx in the Fal- heuristic algorithm, which included 100 replicates coninae clade rather than with the caracaras. that randomlyvaried the order in which taxa were The positionof Herpetotheresin this tree differs added. from the originalsyringeal analysis (Griffiths 1994a;Fig. lB). RESULTS As with Griffiths(1994a), this phylogenyin- dicatesthat two genera,Polihierax and Daptrius, Morphologicaldata.--Phylogenetic analysis of are polyphyletic.Differences in morphology, thesedata resultedin 54 most-parsimoniousbehavior, and habitatuse between the two spe- trees, which are summarized in the strict con- ciesin eachof thesegenera have been described sensustree (Fig. 2). Thereare two majorclades: previously,and placingthe speciesin eachge- 120 CAROLES. GRIFFITHS [Auk, Vol. 116

A era in the family (Brownand Amadon1968). In i_•j.• caracaraspecies addition, the node supportingmonophyly of I • Spiziapteryx theMicrastur species is oneof themost strongly I [•---Falcospecies supportedin the morphologicalphylogeny I --•Mlcrohierax (Fig. 2), with four synapomorphies,three of these with a CI of 1.0. The alternative result, in- L...... Z;;[;;;Td. TvZ;u'usferred using data weighted differentially prior to phylogeneticanalysis, is congruentwith the morphologicalhypotheses; i.e. the Micrastur ...... Accipitor speciesare monophyletic. Combineddata: Total evidence.--Combining equallyweighted molecular data and morpho- B Caracaraspecies logical characters(either ordered or unor-

Spiziapteryx dered)produced five most-parsimonioustrees. The strictconsensus of thesefive (Fig. 3B) is Falco species lessresolved than the tree inferredfrom only •J Microhierax moleculardata (Fig. 3A), reflectingthe conflict Polihieraxsemitorquatus betweenthe moleculardata and the morpho- ...... Micrasturgilvicollis logicaldata. Support exists for threeclades: (1) Herpetotheres the Falcospecies, (2) the caracaraspecies and ...... Micrastursemitorquatus Spiziapteryx,and (3) Microhieraxand Polihierax. Accipiter Thereis no resolutionof relationshipsamong theseclades nor of the relationshipsof the two FIG. 3. (A) Phylogenyinferred from equally Micrasturspecies and Herpetotheres. weightedmolecular data (Griffiths 1997). (B) Phylog- A posterioridifferential weighting, using as- eny inferredfrom combineddata setswith equally weightedmolecular data. Strict consensustree of sessmentsof characterreliability derived from five most-parsimoniouscladograms (length = 1,089, the phylogeneticanalysis (i.e. successiveap- CI = 0.46,RI = 0.45),morphological characters un- proximations),or weightsderived during phy- ordered.The samecladograms are inferred if mor- logenetic analysis, produced one cladogram phologicalcharacters are treatedas ordered. Dashed similarto that inferredfrom equallyweighted lines connectspecies in the genusMicrastur. moleculardata (Fig. 3A). Resultsare the same with the morphologicalcharacters ordered or nusin eitherdifferent subgenera or generahas unordered.Spiziapteryx is sister taxon to theca- been advocated(Brown and Amadon 1968). racaras,and the genusMicrastur is not mono- Analysisof moleculardata.--Results from the phyletic. analysisof variationin cytochrome-bsequences Combined data: Unconditional combination.- are describedin detail in Griffiths (1997). A Combiningdata setswith a priori differential summaryis presentedto serveas background weightingof moleculardata (i.e. unconditional for the combinedanalysis. Equal weighting combination)produced two most-parsimoni- (Fig.3A) anddifferential weighting (Fig. 1C)of ous cladograms.These differed only in reso- sequencedata provided two different results. lution within the caracaraclade (Fig. 4). Be- In the phylogenyinferred from the equally causestep matrices were used for the molecu- weightedmolecular data (Fig. 3A), Spiziapteryx lar characters,summary statistics were not pro- is the sister taxon to the caracaras,a result in- ducedfor thesecladograms. As expected,this congruentwith that inferredfrom osteological phylogenyis basicallycongruent with the sep- and morphologicaldata (Figs. 1A, lB, 2). A arateanalyses (a prioriweighting of molecular more anomalousresult is that the genusMi- data [Fig. 1C] and syringealmorphological crasturis not monophyletic.Contrary to the re- data [Fig. 2]). Differencesbetween the com- suitsfor Daptriusand Polihierax,polyphyly of bined and separateanalyses appear in two ar- the Micrasturspecies has not beensuggested easof the tree:(1) the sistertaxa relationship of previously.The specieswithin this clade are Micrasturand Herpetotheres is unresolved in the similar in morphologyand habitat use and morphologicalanalysis (Fig. 2); and (2) in one havebeen considered distinct from othergen- of the two trees,Polyborus and Daptriusameri- January1999] FalconidaePhylogeny 121

Falco sparveriu$ -- -- ment of different data sets. Detailed reviews of ß Falcovespergnus this issuehave been presented previously (Hil- -- Falco femoralis lis 1987, de Queiroz et al. 1995,Miyamoto and Falcoperegrinus F Fitch 1995,Huelsenbeck et al. 1996), and only Polihieraxsernitorquatu$ main pointsof the argumentwill be discussed. Microhierax The two main questionsto be answeredare: (1) Spiziapteryx Shoulddata setsbe combinedprior to any phylogeneticanalysis? and (2) If not combinedini- tially, shoulddata setsbe combinedafter initial analyses?The initial combinationof data sets __ Phalcob•enus•C has beenjustified philosophically (i.e. the hy- DaptriusamericanuZ pothesis with the greatestexplanatory power is provided by the concurrentanalysis of all rel- Micrasturgilvicollis evant evidence;Kluge 1989). Analyzing data M. semitorquatus (•.•oo) setsseparately is disputedbecause the creation HerpetoEleres of separatedata sets assumes that there are natural classesof data, an assumptionthat is not FIC. 4. Phylogenyinferred from combineddata justifiedscientifically (Kluge 1989). This argu- setswith a prioridifferential weighting of molecular ment carriesover into the appropriatemethod data, and morphologicalcharacters ordered or un- of weighting characters.A priori character ordered. F = Falconinae,C Polyborinae(caracaras). weighting requires partitioning data and Strictconsensus tree of two most-parsimoniousclad- would thereforebe unjustifiedby thosewho ograms.Bootstrap values at nodesare shownfor the believe that data should be combined before molecular data set and for the combined data sets; numbers above the branches are for the molecular analysis.A posterioriweighting, determined ei- data only,numbers below the branchesare for the ther iterativelyor directly,is appropriatebe- combineddata sets. Range of numbersfor the com- causeweights are inferred through phyloge- bined data sets are for ordered and unordered mor- netic analysis. phologicalcharacters (left = unordered,right = or- Separateanalysis of multipledata sets is sup- dered). ported by the argument that congruence amonghypotheses from differentdata setsis a canusare sistertaxa, a result congruentwith powerfultool for assessingaccuracy of the hy- the morphologicalanalysis. In the secondtree, potheses(Miyamoto and Fitch 1995). An as- Polyborusis basalto the othercaracaras. sumptionof this methodologyis that dividing Bootstrapanalysis of the combineddata in- data into classescan be justifiedscientifically dicatesadditional support for key nodescom- (e.g. different codon positions).Advocates of paredwith the analysisof moleculardata alone the initial separateanalysis of data setsdiffer (Fig.4). The positionof Spiziapteryxin the Falco on the final treatment of multiple data sets. cladeis not supportedin the majorityof boot- Opinionsrange from alwayskeeping data sets strap replicateswith molecular data alone. apart (Miyamoto and Fitch 1995) to always When morphologicaldata are added, that re- combiningdata sets,after any conflictsare ac- lationshipis supportedin 61 to 71% of the rep- commodatedthrough differential character licates.The supportfor monophylyof Micrastur weighting(Chippindale and Wiens 1994).An increasesfrom 72% to 94 to 97%, and the sup- intermediateposition is to combineonly con- port for the sister-taxarelationship of the ca- gruent data sets,using either biological(e.g. racara and Falco clade increases from 65 to 78 genetrees for a species;de Queiroz et al. 1995) to 82%. Orderingthe four multistatemorpho- or statistical (significant differences among logicalcharacters does not changethe topolo- data sets;Huelsenbeck et al. 1996) criteria for gy, but it provides strongersupport for six of determiningincongruence. One problemwith sevennodes on the cladogram(Fig. 4). this approachis that the two criteria are not

DISCUSSION equivalent;data setsresponding to differentbi- ologicalprocesses may or may not differ sig- Analysisof multiple data sets.--One of themost nificantly.In addition,although various signif- contentiousissues in systematicsis the treat- icancetests have been proposed (bootstrap [de 122 CAROLES. GRIFFITHS [Auk, Vol. 116

Queiroz1993], tree lengths[Farris et al. 1995], stateat nodes(Kim 1996).These kinds of char- and likelihood ratios [Huelsenbeckand Bull acters occur in the molecular data in this re- 1996]), the power and sensitivityto the as- search.For thesedata, ratesof changediffer sumptionsof thesetests are unknown. among partitions of sequences,with several If the goalof phylogeneticanalysis is to pro- partitions containingcharacters with high vide the hypothesiswith the greatestexplana- rates of change(Griffiths 1997).Substitutions tory power,then comparingthe variousmeth- are saturatedin thesepartitions, and recover- ods is unnecessarybecause only a combined ing phylogeneticsignal is problematic(Grif- analysismeets that goal.If the goalis accuracy fiths1997). When a prioridifferential weighting of phylogeneticinference, then empiricaltests is usedto correctfor the high ratesof change comparingthese methods are useful(de Quei- in these partitions, the phylogenyindicates roz et al. 1995).One implicit assumption of the monophylyof Micrastur,a result congruent total evidenceapproach is that the sameun- with the morphologicalanalyses. derlying signalis containedin eachdata set Theseresults support the intermediatepo- (Hillis et al. 1996).Thus, any potentialconflict sition,the initial separateanalysis of multiple amongthe hypothesesinferred from individ- data sets,and differentialweighting to accom- ual data sets will be overcome when data sets modateheterogeneity in ratesof change.After are combined,circumventing the need to ad- these processeswere performed,combining dress conflict among data sets (Brower and the congruentdata setsincreased the support DeSalle 1994). for nodesin the phylogeny(Fig. 4). This is a testablehypothesis and is directly Moleculesand morphology.--Although the ini- relatedto the idea of consistency;i.e. a consis- tial useof biochemicaldata in systematicswas tent methodologywill resultin convergenceto accompaniedwith assertionsthat molecular the correcttopology as data are added(Felsen- evidencewas more reliable than morphological stein 1978, Penny et al. 1992). In this study, evidence(e.g. Sibleyand Ahlquist 1987),the combiningmorphological and moleculardata value of both morphologicaland molecular produceda phylogenyindicating polyphyly of charactersfor systematicsis now generallyac- the genusMicrastur. This resultis inconsistent knowledged(Donoghue and Sanderson1992, with cladisticmorphological analyses and with de Queiroz et al. 1995;but seeHedges and Sib~ generallyaccepted conclusions based on mor- ley 1994).For this analysis,each method has phological,behavioral, and geographicdata. advantages and disadvantages.Informative Using assessmentsof characterreliability to molecularcharacters were an order of magni- reweightthe characters(a posterioridifferential tude greater in number than morphological weighting)also resultedin polyphyly of Mi- characters(328 vs. 23). The relativepaucity of crastur.This is not surprising;searches using a charactersis the mostsubstantial disadvantage posterJoriweighting are dependenton the val- of the syringealdata. For this analysis,there uesobtained from the startingtree and may be- are 31 character states for 23 characters. There cometrapped in local optima (Swoffordet al. are few characterssupporting most nodes, and 1996). the relationshipof Herpetotheresand Micrastur All phylogeneticmethods can be inconsis- is unresolved. tent when the assumptionsof the methodare Theadvantage of the syringealdata set is the violated (e.g. Debry 1992, Huelsenbeck1995, proportionof unreversedsynapomorphies; 15 Swofford et al. 1996, Sullivan and Swofford of 31 charactersstates have a C! of 1.0, and 40% 1997). Theseviolations can be accommodated of the characterschange only oncein the phy- throughthe useof a moreappropriate model logeny(Figs. 5A, 6A). Homoplasyis moreprev- or by differentiallyweighting data (e.g.Steel et alent in the molecular data. Of the 328 charac- al. 1993, Sullivan and Swofford 1997). The ters, only 10% (33) have a C! of 1.0, and only problem,then, lies in recognizingthe condi- 5% changeone time (Figs.5B, 6B).Homoplasy tionsin a data setunder which a phylogenetic obscuresthe signalin the morebasal branches method is inconsistent. connectingthe monotypicgenera Herpetotheres Inconsistencycan be causedby characters and Spiziapteryx;differential weighting cor- with high ratesof changeand, subsequently,a rectsthese problems. It is possiblethat more se- low probabilityof recoveringthe homologous quencedata would increase the signalat these January1999] FalconidaePhylogeny 123

50 A 45

40'

35'

30'

25' 15

20' 1

15'

10. O' 1 ' 2 ' 3 ' 4 ' 5 ' 6 5' Numberol • steps

0 0-0.24 0.25-0.49 0.50-0.74 0.75-0.99 1.00 ConsistencyIndex 30 B

25 50' B 45, .• 20

40' (D 10 35' o• 5

o 1 2 3 4 5 6 7 Numberol • steps 20,

15' FIG.6. The proportionof characterswith a given numbersof steps.(A) Morphologicaldata, (B) mo- 10. lecular data.

0 derson 1992, de Queiroz et al. 1995). Differen- 0-0.24 0.25-0,49 0.50-0.74 0.75-0.99 1.00 tially weighted moleculardata stronglysup- ConsistencyIndex port the sisterrelationship of Herpetotheresand FIG.5. The proportionof charactersfor eachval- Micrastur,which is unresolvedin the morpho- ue of the consistencyindex. (A) Morphologicaldata, logical tree. The position of Spiziapteryxhas (B) molecular data. weak supportin the moleculartree. The inclu- sionof Polihieraxinsignis in the morphological analysis(Fig. 2) strengthensthe support for nodes.Alternatively, additional sequencedata Spiziapteryxwithin the Falcoclade. may not resolvethese branches if the additional In this study,the moleculardata were more datahave the samepatterns of noiseand signal homoplastic,thus disputingthe assertionthat as the presentdata (Brunset al. 1992)or if the moleculardata are more reliablethan morpho- data do not fit the assumptionsof the model logical data for phylogeneticinference (Sibley used (Sullivan and Swofford 1997). Because and Ahlquist 1987, Hedgesand Sibley1994). these genera are monotypic,increased sam- The occurrenceof saturateddata partitionsin pling of taxawithin the family wouldprobably the falconid sequencesfollows a pattern illus- not resolvethe problem. trated in studiesof cytochromeb and othermi- The resultsof my studysupport the general tochondrialprotein-coding genes (e.g. Korne- conclusionthat both typesof data are valuable gay et al. 1993,Hackett 1996, Yoder et al. 1996). for phylogeneticinference (Donoghue and San- The valueof differentialweighting for rapidly 124 CAROLES.GRIFFITHS [Auk,Vol. 116 evolvingpartitions of molecularcharacters, as GENUS Falco shown in this research,has also been demon- GENUS Microhierax stratedpreviously (e.g. Simon et al. 1994,Sul- GENUS Polihierax livan and Swofford1997). For morphological Tribe Caracarini(all generasedis data,the levelof homoplasymay alsovary in mutabilis) partitionsof data (e.g. in cranialversus post- GENUSPo[yborus cranial osteologyin birds; Livezey 1986).Fi- GENUSIbycter nally,homoplasy may not be as prevalentin GENUSMilvago somemorphological data setsbecause charac- GENUS Phalcoboenus tersin thesedata are generallyfiltered before GENUSDaptrius phylogeneticanalysis (Livezey 1986, Novacek and Wheeler1992, Griffiths 1994b). To minimize the numberof higher taxa in Phylogeneticrelationships in the Falconidae.-- this classification,taxa of equalrank within a Cladogramsinferred from variationin syrin- monophyletichigher category are listedin or- gealmorphology and cytochrome-bsequences der; eachtaxon is the sistergroup to the re- bothprovide evidence that the two speciesin mainingtaxa at that level(Raikow 1985). Prob- thegenus Daptrius are not sister taxa. The sug- lematic relationshipsare indicatedby the gestionthat thesetwo be split hasbeen made placementof a taxonas sedismutabilis (order previouslybased on differencesin habitat use within groupis ambiguousand interchange- and foraging (Brown and Amadon 1968).The able)or incertaesedis (uncertain position) with- nameIbycter Vieillot 1816(type, by monotypy, in thehigher category (Raikow 1985). Falcoamericanus Boddaert 1783) is resurrected Thisclassification divides the family into two for Daptriusamericanus. The name Daptrius subfamilies.Herpetotheres is removed from the Vieillot1816 (type, by monotypy,Daptrius ater Falconinaeand placedin a subfamilywith Mi- Vieillot 1816)is retainedfor Daptriusater. crastur.Herpetotherinae (Lesson, 1843, type I alsorecommend that the two speciesof Po- HerpetotheresVieillot, 1816) has priority over lihieraxbe splitbased on polyphylyof this ge- Micrasturinae and is resurrected for that clade. nus revealedby syringealcharacters. The sub- Thesubfamily name Falconinae is retainedand stantial differencesof these speciesin size, includes the tribes Falconini and Caracarini. plumagecolor, and tail shapehave been noted The compositionof the Falconini is not previously,with the suggestionthat they be changed.Within that tribe, the relationship of placed in different subgenera(Brown and Neohieraxinsignis is unresolved.Polihierax sem- Amadon 1968). The name NeohieraxSwann itorquatusand Microhieraxare sister taxa, a re- 1922 (type, by original designation,Polihierax lationshipsupported by molecularsequences, insignisWalden 1872) is resurrectedfor Poli- syringealmorphology, and a morphometric hieraxinsignis. The namePolihierax Kaup 1847 analysis(Kemp and Crowe 1993).The Caracar- (type, by monotypy, Falco semitorquatusA. ini is a new tribe created to include the five ca- Smith 1836) is retained for Polihieraxsemitor- racaragenera formerly placed in the Caracari- quatus. nae.Relationships of thegenera in thistribe are Therelationships revealed in theanalyses of unresolved.Although there is supportfor the syringealmorphology (Fig. 2), DNA sequence cladecontaining Daptrius ater, Phalcoboenus and data (Fig. 1C), and osteology(Fig. 1A; Becker Milvago,further resolution of the relationships 1987)are reflectedin the phylogeneticclassifi- amongthese three genera, and of this cladeto cationpresented below, which is a revisionof the othertwo genera,await a detailedphylo- the classificationof AOU (1998): geneticanalysis of speciesand subspeciesin that tribe. FAMILY Falconidae Biogeography.--Iused the phylogeneticclas- SubfamilyHerpetotherinae sificationas a frameworkfor examiningthe bio- GENUSHerpetotheres geographyof the Falconidae.Two approaches GENUS Micrastur basedon cladisticmethodology can be usedto SubfamilyFalconinae infer biogeographichistory. Cladistic bioge- Tribe Falconini ography(i.e. vicariancebiogeography), the GENUSNeohierax (incertae sedis) firstapproach, was developed as an alternative GENUSSpiziapteryx to "center of origin" and "improbable"dis- January1999] FalconidaePhylogeny 125 persalscenarios prevalent in the 1970s(Nelson Falco World-wide India to the 1978,Bremer 1992). Vicariance biogeography is Microhiorw(Philippines an attempt to recoverthe biogeographicpat- ß Polihiorax Aftica tern of an entirebiota throughthe congruence Spiziapteryx-- of multiplecladistic hypotheses. For this meth- • Daptriu$ od, an area cladogramrepresenting the distri- Phalcoboenu$ bution of one group is interestingonly to the • Miivago South extentthat it confirmsor conflictswith a gen- Ibycter America eral pattern.The alternativeapproach consid- Polyborus ers the historicalbiogeography of an individ- ual groupto be an importantpart of its natural • Micrastur history(Bremer 1992). Two methods,ancestral- Horpotothoro$ area (Bremer 1992) and dispersal-vicariance analysis(Ronquist 1997), have been developed F•G.7. Area cladogramof the Falconidae.Distri- thatuse character optimization to interpretthe butionof falconidgenera mapped onto the phylog- biogeographicpattern of an area cladogram. eny of the Falconidae. The ancestral-areamethodology estimates the originaldistribution of a groupby compar- ly diversificationof the Falconidaeoccurred in ing the numberof gainsversus losses of areas the Neotropics. under two models,forward (all gains) or re- verse(all losses)Camin-Sokal parsimony (Bre- ACKNOWLEDGMENTS mer 1992).For example, if thereare moregains than losses,the most-parsimoniousinterpre- I thank G. Barrowcloughand R. Rockwellfor ad- tationis that the area waspart of the ancestral viceand support throughout the course of thisstudy. distribution. J. Bates,K. Burns, N. Caithness,T. Chesser,J. Groth, Dispersal-vicarianceanalysis uses Fitch op- S. Hackett,and N. Klein provideduseful conversa- tion and thoughtfulcriticisms. B. Livezey and C. timizationto reconstructthe history of bioge- Crajewskiprovided valuablecomments on the man- ographyby minimizing the dispersal-extinc-uscript. I thank the followingcurators and institution eventson the area cladogram.A stepma- tions for providing tissuesused in this research:G. trix, whichspecifies the costof combinationsof Barrowclough(AMNH), E Gill (ANSP), E Sheldon ancestraland descendantdistributions (the dif- (LSUMNS), and M. Braun (USNM). This work was ference between the distribution of the ancestor supportedby grantsfrom the Frank M. Chapman andeach daughter area), is usedin a three-pass Memorial Fund of the American Museum of Natural History, a Frank M. ChapmanPredoctoral Fellow- algorithm to find the optimal reconstruction ship, NSE DissertationImprovement Grant BSR (Ronquist 1997). 9321486,and City University of New York Disserta- I used the basicprinciples of the secondap- tion ImprovementGrants. The researchreported proachto inferthe biogeography of the Falcon- hereis a contributionfrom the LewisB. andDorothy idae.The area cladogram,illustrating the dis- Cullman ResearchFacility at the AmericanMuseum tributionsof the falconidgenera mapped onto of Natural History and has receivedgenerous sup- the phylogeny,is shownin Figure7. The pat- port from the Lewis B. and DorothyCullman Pro- gram for MolecularSystematics Studies, a joint ini- tern of distributionof thesegenera enables a tiative of the New York Botanical Garden and the simpleinterpretation of biogeographichistory American Museum of Natural History. for the family. Distributionof speciesin two of the three lineages,and the basalbranch in the LITERATURE CITED third, is limited to South America. Distribution of generaoutside the Neotropicsoccurs only in AMADON, D., AND J. BULL. 1988. Hawks and owls of the Falconinae.The sister taxa, Microhieraxand the world: A distributional and taxonomic list. Polihieraxsemitorquatus, occur in SoutheastAsia Proceedingsof the WesternFoundation of Ver- tebrateZoology 3:294-357. and Africa, and the genusFalco contains spe- AMERICAN ORNITHOLOGISTS' UNION. 1998. Check- ciesoccurring in all regionsof the world. The list of North American birds, 7th ed. American most-parsimoniousexplanation of this distri- Ornithologists'Union, Washington,D.C. butionpattern suggests that the origin and ear- AMES,P. L. 1971. The morphologyof the syrinx in 126 CAROLES. GRIFFITHS [Auk, Vol. 116

passerinebirds. Bulletinof the PeabodyMuse- FARR1S,J. S. 1989. The retention index and the re- um of Natural History 37:1-194. scaledconsistency index. Cladistics5:417-419. BECKER,J. 1987. Revision of "Falco"ramenta Wetmore FARRIS, J. S., M. KALLERSJO,A. G. KLUGE, AND C. and the Neogeneevolution of the Falconidae. BULT. 1995. Testing significance of incongru- Auk 104:270-276. ence. Cladistics 10:315-319. BREMER,K. 1992. Ancestral areas: A cladistic rein- FELSENSTE1N,J. 1978. Cases in which parsimonyand terpretationof the centerof origin concept.Sys- compatibilitywill be positivelymisleading. Sys- tematicBiology 41:436-445. tematicZoology 27:401-410. BROWER,A., AND R. DESALLE. 1994. Practical and FRIEDMANN,H. 1950. The birds of North and Middle theoretical considerations for choice of a DNA America. Part XI. Smithsonian Institution Bul- sequenceregion in insectmolecular systematics, letin 50:543-755. with a shortreview of publishedstudies using GOLOBOFF,P. A. 1993.Estimating character weights nucleargene regions. Annals of the Entomolog- duringtree search.Cladistics 9:83-92. ical Societyof America 87:702-716. GR1FF1THS,C. S. 1994a.Syringeal morphology and BROWN,L., AND D. AMIDON. 1968. Eagles,hawks thephylogeny of the Falconidae.Condor 96:127- and falconsof the world. Hamlyn House,Fel- 140. tham, United Kingdom. GRIFFITHS,C. S. 1994b.Monophyly of the Falconifor- BRUNS,t. D., R. V1LGALYS,S. M. BARNS,D. GONZA- mesbased on syringealmorphology. Auk 111: LEZ, D. S. HIBBETT,D. J. LINE, L. SIMON, S. STICK- 787-805. EL, T. M. SZARO,W. G. WEISBURN,AND M. L. SO- GR1FF1THS,C. S. 1997. Correlation of functional do- tIN. 1992.Evolutionary relationships within the mainsand ratesof nucleotidesubstitution in cy- fungi: Analysesof nuclearsmall subunitrRNA tochromeb. MolecularPhylogenetics and Evo- sequences.Molecular Phylogenetics and Evolu- lution 7:353-365. tion 1:231-241. GURNEY,J. H. 1894. Catalogueof the birds of prey. BULL,J. J., J.p. HUELSENBECK,C. C. CUNNINGHAM,D. R. H. Porter, London. L. SWOFFORD,AND P. J. WADDELL. 1993. Parti- HACKETT,S. 1996.Molecular phylogenetics and bio- tioning and combiningdata in phylogenetic geographyof tanagersin the genusRamphocelus. analysis.Systematic Biology 42:384-397. MolecularPhylogenetics and Evolution5:368- CARPENTER,J.1988. Choosing among multiple equal- 382. ly parsimoniouscladograms. Cladistics 4:291- HEDGES,S. g., AND C. G. SIBLEY.1994. Molecules vs. 291. morphologyin avianevolution: The caseof the CH1PPINDALE,ET., ANDJ. J. W1ENS. 1994. Weighting, "pelecaniform"birds. Proceedingsof the Na- partitioning,and combiningcharacters in phy- tional Academyof SciencesUSA 91:9861-9865. logeneticanalysis. Systematic Biology 43:278- Hll•l•l$,D. M. 1987.Molecular versus morphological 287. approachesto systematics.Annual Review of DEBRY,R. W. 1992.The consistencyof severalphy- Ecologyand Systematics18:23-42. logeny-inferencemethods under varying evolu- HILLIS,D. M., B. K. MABEL,AND C. MORITZ.1996. Ap- tionary rates.Molecular Biology and Evolution plicationsof molecularsystematics. Pages 515- 9:537-551. 544 in Molecularsystematics, 2nd ed. (D. M. Hil- DE QUEIROZ, A. 1993. For consensus (sometimes). lis, C. Moritz and B. K. Mable, Eds.). SinauerAs- SystematicBiology 43:368-372. sociates, Sunderland, Massachusetts. DE QUE1ROZ,A., g. J. DONOGHUE, AND J. KIM. 1995. HUELSENBECK,J.P. 1995. Performanceof phyloge- Separateversus combined analysis of phylogeneticmethods in simulation.Systematic Biology netic evidence.Annual Review of Ecologyand 44:17-48. Systematics26:675-681. HUELSENBECK,J.P., ANDJ. J. BULL.1996. A likelihood DONOGHUE,M. J., AND M. J. SANDERSON.1992. The ratio test for detectionof conflictingphyloge- suitabilityof molecularand morphologicalevi- netic signal.Systematic Biology 45:92-98. dencein reconstructingplant phylogeny.Pages HUELSENBECK,J, P., J. J. BULL,AND C. W. CUNNING- 340-368 in Molecularsystematics of plants(P.S. HAM. 1996. Combining data in phylogenetic Soltis,D.E. Soltisand J. J. Doyle, Eds.). Chapman analysis.Trends in Ecologyand Evolution11: and Hall, New York. 152-158. EERN1SSE,D. J., AND A. G. KLUGE. 1993. Taxonomic KEMP,A. C., ANDT. M. CROWE.1990. A preliminary congruenceversus total evidence,and amniote phylogeneticand biogeographicanalysis of the phylogenyinferred from fossils,molecules and generaof diurnal raptors.Pages 161-175 in Pro- morphology.Molecular Biology and Evolution ceedingsof the internationalsymposium on ver- 10:1170-1195. tebratebiogeography and systematicsin the FARRIS,J. S. 1969. A successiveapproximations ap- tropics(R. Hutterer and G Peters,Eds.). Buseam proachto characterweighting. Systematic Zo- A. Koening,Bonn, Germany. ology 18:374-385. KEMP,A. C., ANDt. M. CROWE.1993. Morphometrics January1999] FalconidaePhylogeny 127

of falconetsand huntingbehaviour of the Black- world, vol. 1. Harvard University Press,Cam- thighed FalconetMicrohierax fringillarius. Ibis bridge,Massachusetts. 136:44-49. RAIKOW, R. 1985. Problems in avian classification. KIM, J. 1996. General inconsistencyconditions for Current Ornithology2:187-212. maximumparsimony: Effects of branchlengths RONQUIST,E 1997. Dispersal-vicarianceanalysis: A and increasingnumbers of taxa. SystematicBi- new approachto the quantificationof historical ology 45:363-374. biogeography.Systematic Biology 46:195-203. KING,A. S. 1989.Functional anatomy of the syrinx. SHARPE,g. B. 1874.Catalogue of birds. BritishMu- seum, London. Pages 105-192 in Form and function in birds, SIBLEY,C. G., ANDJ. E. AHLQUIST.1987. Avian phy- vol. 4 (A. S. King and J. McLelland, Eds.). Aca- demic Press, New York. logenyreconstructed from comparisonsof the geneticmaterial, DNA. Pages95-121 in Mole- KLVCE,A. G. 1989.A concernfor evidenceand a phy- culesand morphologyin evolution:Conflict or logenetichypothesis of relationshipsamong Ep- compromise?(C. Patterson, Ed.). Cambridge icrates(Boidae, Serpentes). Systematic Zoology University Press,Cambridge, United Kingdom. 38:7-25. SIMON, C., E FRATI, A. BECKENBACK,g. CRESPI,H. KORNEGAY,J. R., KOCHER,t. D., WILLIAMS,L. A. AND LIU, AND P. FLOOK.1994. Evolution, weighting, A. C. WILSON.1993. Pathways of lysozymeevo- and phylogeneticutility of mitochondrialgene lution inferred from the sequencesof cyto- sequencesand a compilationof conservedpoly- chrome b in birds. Journal of Molecular Evolu- roerasechain reaction primers. Annals of theEn- tion 37:367-379. tomologicalSociety of America87:651-701. LANYON,S. g. 1993.Phylogenetic frameworks: To- SMITH,A. B. 1994.Rooting molecular trees: Problems wards a firmer foundationfor the comparative and strategies.Biological Journal of the Linnean approach.Biological Journal of the LinneanSo- Society51:279-292. ciety 49:45-61. STEEL,M. E., M.D. HENDY, AND D. PENNY. 1993. Par- LIVEZEY,B.C. 1986.A phylogeneticanalysis of re- simonycan be consistent!Systematic Biology 42: cent anseriformgenera using morphological 581-87. characters. Auk 103:737-754. SULLIVAN,J., AND D. L. SWOFFORD.1997. Are guinea MIYAMOTO,M. M., AND W .M. FITCH.1995. Testing pigsrodents? The importanceof adequatemod- elsin molecularphylogenetics. Journal of Mam- speciesphylogenies and phylogeneticmethods malian Evolution 4:77-86. with congruence.Systematic Biology 44:64-76. SWANN,H. K. 1922. A synopsisof the Accipitres. NELSON, G. 1978. From Candolle to Croizat: Com- Wheldonand Wesley,London. mentson thehistory of biogeography.Journal of SWOFFORD,D. L, G. J. OLSEN, P. J. WADDELL, AND D. the Historyof Biology11:269-305. M. HILLIS.1996. Phylogeneticinference. Pages NOVACEK, M., AND Q. WHEELER. 1992. Extinct taxa: 407-514 in Molecular systematics,2nd ed. (D. M. Accountingfor 99.999%of theearth's biota. Pag- Hillis, C. Moritz and B. K. Mable, Eds.). Sinauer es 1-16 in Extinctionand phylogeny(M. Nova- Associates,Sunderland, Massachusetts. cekand Q. Wheeler,Eds.). Columbia University YODER, A.D., R. VILGALYS, AND g. RUVOLO. 1996. Press, New York. Molecularevolutionary dynamics of cytochrome PENNY, D., M.D. HENDY, AND M. E STEEL.1992. Pro- b in strepsirrhineprimates: The phylogenetic gress with methods for evolutionarytrees. significanceof third-positiontransversions. Mo- Trends in Ecologyand Evolution7:73-79. lecularBiology and Evolution13:1339-1350. PETERS,J. L. 1931. Check-list of the birds of the Associate Editor: R. M. Zink 128 CAROLES. GRIFFITHS [Auk, Vol. 116

APPENDIX1. Descriptionsof syringealanatomy and the 23 syringealcharacters; 17 charactersare binary and 6 are multistate. Characters 3, 4, 6, and 16 are ordered multistate; characters5 and 19 are unordered multistate.In all cases,the hypothesizedplesiomorphic state is state0.

Tympanum:Fused A elementsnear the tracheo-bronchialjunction (1) Presenceof a tympanumor trachealdrum [0] No fusion of A elements [1] Ossification and fusion of first A elements (2) Dorsal fusion of tympanum by medial bar [0] Medial bar absent [1] First A elementsfused medially by ossifiedbar (3) Fusionalong edges of A elementsthrough ossification of cartilaginousborders separating elements [0] Ossifiedborders not apparent [1] Threeor four A elementsfused along margins of elements [2] Heavierfusion, margins somewhat obliterated, light suturesapparent (4) At least five A fused entirely along edges [0] Light fusionor fusionabsent [1] First five elementsheavily fused, some sutures apparent [2] Suturesobliterated (5) Degree of ventral fusion of tympanum [0] No ventral fusion of A elements [1] First threeor four A elementsfused partially alongmargins [2] First threeor four A fusedentirely along margins [3] At leastfive A elementsfused entirely along margins, some sutures apparent (6) Shapeof fusedand ossifiedtympanum (partially ordered character; state 3 is a modificationof state2) [0] Tympanum absentor not ossified [1] Tympanum graduated,widens caudally [2] Tympanum cylindrical [3] Tympanumcylindrical; A1 laterallyflattened causing "pinching in" of most-caudalelement A Elements:Occur on the tracheaas singlerings but may extendonto the bronchias paired doublerings; they are ossifiedand flattenedin cross-section (7) Ridge coveringdorsal A1 ends [0] Ridge absent [1] Ridge of ossifiedtissue forms medial ridge over A1 ends,covers dorsal ascent of pessulus (8) Separationof dorsalA1 ends [0] A1 ends separated [1] Endsclose medially, fused by ossifiedtissue (9) Size of A1 ends [0] A1 singleor width of endssimilar to width medial [1] Dorsal endsvery enlarged (10) Flatteningof paired A1 elements [0] A1 singleelement, or roundedhalf rings [1] A1 flatteneddorsoventrally into parentheticalshape; laterally, A1 dorsaland ventral ends protrude (11) Appearanceof A1 on lateral view [0] A1 concaveup medially [1] A1 flattenedmedially B Elements:Occur as paired rings on the bronchiand are generallycartilaginous and D-shapedin cross- section (12) Abrupt changein shapeand orientationof B1 ends [0] B1 endschange gradually or are unchanged [1] B1 endsthick and wide, ascendabruptly in L-shapeto fusewith A1 (13) Changein orientationof B1 ends [0] B1 endschange shape or endsmaintain orientation [1] B1 endsthin, ascendgradually to fusewith A1 (14) Knobbingof B1 ends [0] No knobbing [1] Craniadedges knobbed, fused with A1 ends (15) Gradual changein shapeand orientationof B1 ends [0] B1 endsthin or L-shaped [1] B1 endsthicken; ends are roundedand ascendgradually to fusewith A1 (16) Fusion of B element ends [0] No fusion of B elements [1] B2 ends fused to B1/A1 [2] B3 endsfused forming ridge borderingmedial (internal)membrane January 1999] FalconidaePhylogeny 129

APPENDIX 1. Continued.

Muscles and Membranes (17) Existenceof lateral (external)membrane on bronchi ' [0] Membrane absent [1] Membranepresent and locatedbetween A1 and B1 (18) Presenceof cartilaginousbar on lateral membrane [0] Bar is absent [1] Bar is presentM. tracheolateralisinserts on bar (19) Presenceof membranousextension of lateralmembrane onto which M. tracheolateralisinserts [0] Membranous extension is absent [1] Thick, bulbous membrane on dorsal half of membrane [2] Thick, bulbous membrane coversentire width of membrane (20) Insertionof M. tracheolateralisonto lateralmembrane [0] Membrane absent or modified [1] M. tracheolateralisinserts directly onto dorsalhalf of lateralmembrane (21) Cartilaginousborder on (medial) membrane [0] Border is absent [1] Internal (medial)membrane has some cartilage on caudalborder (22) Complete border on internal membrane from A1 dorsal to A1 ventral ends [0] Borderabsent or incomplete [1] Thick, even, concaveborder on internal membrane from A1 dorsal to A1 ventral ends AccessoryCartilaginous Structures: Occur on internalbronchial membranes (23) Presenceof thin, amorphousaccessory structure extending from border [0] Accessorystructure absent [1] Cartilageforms straight caudal edge from dorsalto ventralends 130 CAROLES. GRIFFITHS [Auk, Vol. 116