Phylogenetic Patterns in the Trochilidae

The Auk 115(1):105-118, 1998

PHYLOGENETIC PATTERNS IN THE TROCHILIDAE

JOHN A. GERWIN•'2 AND ROBERTM. ZINK L3 •Museumof NaturalScience and Department of Zoologyand Physiology, LouisianaState University, Baton Rouge, Louisiana 70803, USA

ABSTRACT.--Althoughmany aspectsof hummingbirdbiology have been studied, few re- centanalyses of higher-levelsystematic relationships exist. Based on morphology, it hasbeen hypothesizedthat the Trochilidaeincludes six majorclades. We used starch-gel electropho- resisto constructand testphylogenetic hypotheses for representativesof the six clades,using two speciesof swifts(Apodidae) as outgroups. Of 45 loci scored,38 werepolymorphic. The averageNei's geneticdistance (D) among14 hummingbirdtaxa was 0.625;D averaged 1.61between the swiftsand hummingbirds.These distances are large,and are consistent with othernonpasserine groups, suggesting that hummingbird taxa are relativelyold. Phy- logeneticanalyses generally were consistentwith the hypothesisthat hermitsare a sister groupto all othertrochilines. Within the Trochilinae, two broad groups are recognized, here calledtrochiline-A and B, whichcorrespond to the morphologicallydetermined "primitive" and"advanced" trochiline groups of Zusiand Bentz (1982). Androdon aequatorialis is genet- ically distinctbut generallyaligns with the trochiline-Agroup. Within the trochiline-B group,four radiationshypothesized by Zusi (pers.comm.), here called Bee, Amazilia ("Em- eralds"),Andean, and High Andean,were corroboratedby our analyses.Our distanceanal- ysissuggests a phylogeneticpattern consistent with that derivedfrom Sibley and Ahlquists' (1990)and Bleiweisset al.'s(1997) DNA-DNA hybridizationstudies. Received 31 October1996, accepted20 June1997.

THE HUMMINGBIRDS(TROCHILIDAE) form one DNA-DNA hybridization(Sibley and Ahlquist of thelargest bird families, with approximately 1990;Bleiweiss et al. 1994,1997). Most of these 325 species.Although many aspectsof hum- studieswere not comprehensiveat the levelof mingbirdbiology have been studied, few high- the family.However, the studyby Bleiweisset er-levelstudies of systematicrelationships ex- al. (1997),which involved 26 species,provided ist. Early taxonomists(Gould 1861, Boucard a molecularphylogenetic hypothesis for the 1895, Simon 1921, Peters 1945, Zimmer 1950- majorclades in the family.We report an allo- 1953) providedspecies descriptions and gen- zymic surveyof major groupsof humming- eraldetails of geographicvariation in mostspe- birdsdesigned to testphylogenetic hypotheses cies,but they did not explicitlyidentify system- derived from previous morphologicaland aticrelationships. Recently, several approaches DNA-DNA hybridizationanalyses. havebeen used to identifyphylogenetic pat- Our study was based in part on Zusi and ternswithin parts of this family: comparative Bentz's (1982) hypothesis of higher-level analysisof vocalizationsand matingbehavior groupsin the Trochilidae.They identified four (Schuchmannunpubl. data), external morphol- major groupings:hermits (Phaethorninae), ogy and ecology (Graves 1980, 1986; Stiles "primitive"trochilines, and two groupsof "ad- 1983, 1996; Hinklemann 1989), comparative vanced" trochilines.The terms primitive and myology (Zusi and Bentz 1982), protein elec- advanced are reserved for discussion of char- trophoresis(Gerwin and Zink 1989, Gill and actersand are misleadingwithout an in-depth Gerwin 1989), mitochondrialDNA sequences phylogeneticanalysis; we refer to the two (Hernandez-Banoset al. unpubl. data), and groupsas trochiline-A (Zusi and Bentz's [1982] primitive group) and trochiline-B(their ad- 2 Present address: North Carolina State Museum of vancedgroup). The widespread notion that the NaturalSciences, P.O. Box 29555, Raleigh, North Car- hermitsrepresent the "primitive" humming- olina 27626,USA. E-mail:[email protected] birds is incorrect in one sense: if these are sister 3Address correspondence to this author.Present address:James Ford Bell Museum, University of clades,each is by definitionthe sameage. Phy- Minnesota, 100 Ecology Building, St. Paul, Minne- logeneticanalysis is requiredto showthat the sota 55108, USA. E-mail: [email protected] hermitclade retains a disproportionatenumber

105 106 GERWINAND ZINK [Auk, Vol. 115 of plesiomorphiccharacters relative to the sis- (1997)refer to the Beegroup and the hermits.Blei- ter group of hummingbirdsbefore it would be weiss et al. (1997) further referred to a Mountain "primitive" in any sense.Also, within trochi- Gemgroup, of whichwe had no representatives.For lines,if thereare onlytwo groups,neither can simplicity,we useZusi's names, except for usingEm- erald in place Amazilia, becauseconfusion results be primitive in a phylogenetictree. From his fromnaming a groupafter one of thegenera in it. We work on comparativemyology, using the swifts includedtwo speciesof swifts (Apodidae;Reinarda (Apodidae) as the outgroup, Zusi (pers. squamata,Chaetura cinereiventris) that servedas out- comm.)suggested as a workinghypothesis: (1) groups(see Appendix 1). Althoughinferences based that the trochiline-Agroup included Androdon, on low numbersof specimenshave been criticized Schistes,Colibri, Doryfera,and close relatives; (Archieet al. 1989),in our study geneticdistances and (2) that four major cladesexist within the weresufficiently high that our patterns likely are ro- trochiline-Bgroup: Bee, High Andean,Andean, bust to small samplesizes of individuals.Most spec- and Amazilia ("Emeralds").By selectingat imens were collectedover severalyears during ex- leasttwo representativesof eachmajor group peditionsto variousregions of the Neotropics.Sam- ples of tissueswere preservedin liquid nitrogenin to reflect hummingbird diversity,our phylo- the field until transferred to the Louisiana State Uni- geneticanalysis of allozymevariation provides versity Museum of Natural Sciences(LSUMNS), a higher-levelphylogenetic hypothesis that can wherethey were storedat -70øC (seeJohnson et al. be comparedwith thosegenerated from other [1984] for details of collection and preservation data sets. methods).Voucher specimens (study skins and tis- Most modern molecularstudies of phyloge- suesamples) are housedat the LSUMNS. ny rely on DNA sequences(Hillis et al. 1996). Samplesof pectoraland heart muscle and liver (to- However,phylogenetic hypotheses based on al- tal volumeof tissuewas approximately0.5 cc)were lozymesand mitochondrialDNA oftenare to- placedin 0.8 mL of grindingbuffer consistingof 10 pologically consistent(Zink and Avise 1990, rng NADP and 100 }xLof 2-mercaptoethanolin 100 Zink and Dittmann 1991, Zink et al. 1991, Zink mL distilled water (Richardsonet al. 1986) and ground for 10 to 15 s using a Tekmar Tissuemizer. and Blackwell1996). Thus, althoughthere are Thesecrude homogenates were then centrifugedat reasonsfor preferringDNA data for phyloge- 36,000 x g for 30 min, and the supernatantwas netic inferences(Hillis et al. 1996), allozyme- storedat -70øC.Six aliquots of 20 }xLof eachsample basedphylogenetic inferences are valuablein werestored separately and usedfor the first sixgels, studies of congruenceof tree topologiesin- and the rest of the homogenatewas storedin indi- ferred from different data sets(Swofford et al. vidual vials. Electrophoreticprocedures followed 1996).In our study,we comparetree topologies Selanderet al. (1971),Harris and Hopkinson(1976), from allozymes and DNA-DNA hybridization Johnsonet al. (1984) and Richardsonet al. (1986) (Bleiweisset al. 1997).Although both are based with slight modifications(available from author). on informationfrom the nucleargenome, each Variousgel-buffer combinations were used to opti- mize the resolutionof bandingpatterns (Appendix is likely an independentestimate of phyloge- 2). netic relationships.Hence, congruence of tree Forty-five presumptivegenetic loci were scored topologiescan be used as a measureof confi- (Table1), with allelescoded in referenceto their mo- dencein the phylogenetichypothesis. bility from the origin.The most cathodal alleles were coded"a,"with subsequentlyfaster alleles coded as MATERIALS AND METHODS "b," "c," and so on. Multiple isozymesat a locus were alsocoded by mobility.The mostanodally ap- We used standardhorizontal starch-gel electro- pearingisozymes were codedas "1," with the more phoresisto investigatepatterns of geneticvariation cathodalisozymes coded as "2," "3," and soon. Ac- among 14 taxa (Zusi pers. comm.):Glaucis and Phae- ronymsfor loci follow Harris and Hopkinson(1976). thornisrepresent the hermit, or phaethornineline; We enteredindividual genotypes into the computer Androdon,Colibri, Schistes, and Doryferathe trochi- program BIOSYS-1 (Swofford and Selander 1981), line-A group; Acestruraand Selasphorusthe "Bee" which produceda table of allelic frequencies,Nei's line; Amazilia and Campylopterusthe "Amazilia" (1972, 1978) and Rogers' (1972) geneticdistances, (Emerald) group; Aglaeactisand Coeligenathe "An- Cavalli-Sforzaand Edwards' (1967) chord distance, dean" clade; and Metallura, and Oreotrochilusthe and four UPGMA phenograms(Sheath and Sokal "high Andean"clade. Bleiweiss et al. (1997)renamed 1973) derived using thesefour distancemeasures. these clades Mangoes (Trochiline-A), Emeralds Becausethere is no consensuson tree-buildingmeth- (Amazilia group), Brilliants (Andean), and Co- ods, we usedboth distanceand discrete-locusap- quettes(high Andean);both Zusi andBleiweiss et al. proachesand comparedresults. Distance-Wagner January1998] TrochilidaePhylogeny 107

TABLE1. Allelicfrequencies for variableloci. Numbers in parenthesesare frequencies of allelesnot fixed for that locus.Allelic designations by letterindicate fixation at that locus.

Taxon a

Locus 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

AB-5 B B B B B B B C B B B B B B A A ACON-2 B A B B A B A A A A A A A A A C ACP-2 A A A A A A A A A A A A A A A B ADA D E C D D B C B B B C D C C C A AK-1 F E D D(.50) D C D D D D D D D A B B E(.50) AK-2 B A A A A A A A A A A A A A A A ALD-2 A A D D D D D D C D E E F D B B CK-1 B B F E D E E E E E E E E E A C CK-2 A A A A A A A C A A A A A A D B EAP C C D D D D D D D D D D D D B A ENOL C C D C B B C C C C C C C A D D EST-D F D F B(.50) G E C C E E E E F E A E E(.50) FUM B B B C B C B B C B B C C B A D GDA F F B A C D C C C C C C C C H G GLUD B B B B B B B B B B B B B B C A A-GPD C E E B D D D D D B D D C C A B G6-PDH D D F B D D D D D D D D A D E C GOT-1 B B B B B B B B B B B B B B A A GPI E B D(.50) D D D A A A A D D C C H G F(.50) GR C C B C C C D C C C A C C C C C IDH-1 B B A B B B E E C(.50) E E E E E C C F(.50) IDH-2 C B C B B B B B B B B B B B A A LA H F H J(.50) J K B C A A D D G G E I K(.50) LDH-1 A B A A A A A A A A A A A A A A LDH-2 B B B B A B B B B B B B B B C C LGG L(.75) K H F(.25) H I A A D E C C G(.75) B F F M(.25) H(.75) J(.25) MDH-1 B B B B B B B B B B B B B B A B MDH-2 B E D D B E B B B B B B E B A A ME-1 E D F G(.75) H G J J G G G G I I A(.50) B H(.25) C(.50) ME-2 B A B B B B B B B B B B B B B B MPI G(.75) C B F F F F F F F E E H H D A •(.25) NP F B H A C D K K N(.25) E(.25) O L M O J J P(.75) G(.50) I(.25) 6-PGD I H D D(.25) D B D D B(.50) B(.25) D D E A C B F(.25) D(.50) D(.50) G(.50) PGM-2 D F D D D F E E B E C E E E D A Phe-Pro B B E E A E F F H(.50) E D C F F G G I(.50) PK-1 B B B B A A B B B B B B B B B B PK-2 A A A A A A A A A A A A A A B B SORDH C(.25) C D D A D F F F F E E B E E E G(.75) al, Glaucishirsuta; 2, Phaethornisphilippii; 3, Androdonaequatorialis; 4, Colibri coruscans; 5, Schistes geoffroyi; 6, Doryferajohannae; 7, Acestrura mulsant;8, Selasphorussasis; 9, Amaziliaviridicauda; 10, Campylopteruslargipennis; 11, Aglaeactiscastelnaudii; 12, Coeligenaviolifer; 13, Metallura tyrianthina;14, Oreotrochilusestella; 15, Reinardasquamata; 16, Chaeturacinereiventris. 108 GERWINAND ZINK [Auk,Vol. 115 trees (Farris 1972, 1981; Swofford 1981) were pro- duced using Rogers' (1972) distancemeasure. Dis- tance-Wagnertrees were generatedby specifyingthe Multiple Addition Criterionand allowingfor 30 partial networksto be usedduring each successive step. BothPrager and Wilson's"F" value (1976)and the Fitch and Margoliash (1967) percent standarddevi- ation (% SD) were used to determinewhich partial networksto save.Distance-Wagner trees were rooted by designatingthe two swift taxa as outgroups.To complementthe distance-Wagnertrees, we used PHYLIP to performa Fitch-Margoliash(1967; F-M) distanceanalysis according to the followingparam- eters: global search option, no negative branch lengths,and 10 random addition sequences.Both distance-Wagnerand F-M analysespermit variable evolutionaryrates whereasUPGMA assumesrate constancy.We are awareof the criticismsof distance approaches(Farris 1981, 1985, 1986; Nei et al. 1983; Felsenstein1986); however, in ourexperience there is often little difference between distance and discrete- character(locus) approaches. We conducteda cladisticanalysis using loci as characters and alleles as unordered character states (Baverstocket al. 1979;Patton and Avise 1983;Buth 1984).For polymorphisms,the most frequentallele was consideredthe state,an approachthat ignores frequencyinformation; if there were two allelesat 0.50 frequencyat a locus,we assignedthe statethat matchedother taxa, if at all (othermethods of coding did not alter conclusions).These data were analyzed with the computerprogram PAUP 3.1.1 (Swofford 1993),using heuristic searches with 50 randomad- dition replicates.Multiple equally parsimonious trees were summarized as strict consensus trees. A bootstrapanalysis (Felsenstein 1985) was performed using heuristicsearches and 100 replications.The computerprogram MacClade (Maddison and Mad- dison 1992) was used to evaluate alternative tree to- pologiesderived from the literature.

RESULTS

Geneticvariation.--Of the 45 loci scored, 38 exhibitedsome polymorphism (at leasttwo alleles found acrossall taxa; Table 1). Sevenloci were monomorphicand fixed for the sameal- lelein all taxa(Gapdh, Got-2, HK, Lap,Odh-1, SOD-I,2).Of the 38 polymorphicloci, five were nearly monomorphicand exhibitedonly two alleles: AcP-2, AK-2, Ldh-1, Mdh-1, ME-2. Threeloci (Glud, Got-1,PK-2) were monomor- phic and fixed in the Trochilidaefor an allele thatdiffered from the outgroup (Apodidae). Geneticdistances.--Nei's (1978) genetic dis- tanceaveraged 0.625 + SD of 0.215 (Table2) amongthe 14hummingbird taxa. Values range January1998] TrochilidaePhylogeny 109

Glaucis hirsuta

Phaethomisphilipipi Androdonaequatodalis

Colibri coru$cans

Schistesgeoffroyi Doryferajohannae

-- Acestrura mulsant

-- Selasphorussasin Amazilia viridicauda

Campylopteruslargipennis Aglaeactis castelnaudii Coeligenaviolifer Metalluratyrianthina

Oreotrochilus estella

Reinarda squamata

Chaetura cinereiventris

0.00 0.10 0.20 0.30 0.40 ROGERS' GENETIC DISTANCE Fic. 1. Distance-Wagnertree rooted by the outgroupmethod (Farris 1972). Units are in Rogers'(1972) D. The % SD equals13.635 (unoptirnized). from 0.137 (Selasphorussasin vs. Acestruramul- biascaused by usinga phenogramwhen rates sant)to 1.086(Phaethornis philippii vs. Androdon vary (Felsenstein1986). aequatorialis).The averagegenetic distance be- The hermits (Glaucis,Phaethornis) were most tweenthe families(two speciesin the Apodi- similarto thetrochilines. The placement of An- dae,14 in the Trochilidae)was 1.61 + 0.180(n drodondiffered between the UPGMA pheno- = 28). Interfamilial values range from 1.348 gram(not shown) and distance-Wagnertree. In (Oreotrochilusestella vs. Reinardasquamata) to the UPGMA phenogramit is a sistergroup to 2.054 (Phaethornisphilippii vs. Chaeturacinereiv- thetrochilines, whereas in thedistance-Wagner entris). Each swift speciesshows a similar analysisit is placedbasally within the trochi- range of valuesto the hummingbirds(1.4 to line-A assemblage,although it was relatively 2.0, C. cinereiventris;1.4 to 1.9,R. squamata).The divergent.The F-M tree (not shown) differs geneticdistance between the two swifts was from the distance-Wagnertree only in suggest- 0.655. ing that Androdonwas sisterto all otherhum- Distanceanalyses.--Because most distance mingbirds,and in thatthe remainder of thetro- analyses use metric measures (Farris 1972, chiline-Agroup was paraphyletic, but placed Sneathand Sokal 1973; but see Nei 1987), we between the hermits and the rest of the trochi- used Rogers'(1972) distances.UPGMA phen- lines (as in the distance-Wagnertopology). ograms (not shown) generatedwith various Thus,the F-M treesuggests more than two ma- distance measures (Cavalli-Sforza and Ed- jor lineagesof hummingbirds.The trochiline- wards1967; Nei 1972,1978; Rogers 1972) yield- A assemblagealso includes Colibri and Schistes ed the sametopology and differed only in as sistertaxa, with Doryferaa distantmember; branchlengths. These phenograms exhibited a however,in the F-M tree,Doryfera and Schistes nearly identicaltopology to the "best" dis- were sistertaxa. The distanceanalyses are in- tance-Wagnertree (Fig. 1), as judged by the conclusiveconcerning the monophylyand re- minimumvalue of %SD(60 trees) or Pragerand lationshipsof the trochiline-Agroup. Wilson's "F" (15 trees). Low values of each of The othermajor branch leads to fourpairs of thesegoodness-of-fit measures indicate that the taxacomprising the trochiline-B group, three of dendrogramfaithfully portraysthe distances which are characterizedby long branch in the originalmatrix. We emphasizethe dis- lengths.The length of thesebranches, however, tance-Wagnertrees becauseof the potential could be a function of the low number of taxa 110 GERWINAND ZINK [Auk, Vol. 115

d/Reinarda squamata resentativesof the Emeraldgroup form a clade / • Chaeturacinereiventris in only 90% of the 40 trees;there is no strong ,G/ashiua cladisticsupport for sister-grouprelationships / •'• Phaethornisphilippii in the trochiline-Bgroup, although the Beeand Emeraldgroups are a cladein 90% of the 40 / • •Androdonaequatorialis trees. Bootstrap analysis (not shown)supports / / (at ->70%)the samegroupings in thetrochiline- A group,and Androdonplus hermits,but no othermajor groupings. Overly distant outgroupscan bias tree to- • / • Oraotmch#usostolla pology and interpretationby alteringthe in- grouptopology or by placingthe root random- ly alongthe longestbranch within the ingroup • A•stmramulsant (Smith 1994).Because of the relativelygreat • • Se•asphorus•sin distancefrom the outgroupsto the humming- • Am•iliaviridi•uda birds,we excludedthe swifts and recomputed • Campyloptoruslargiponnis maximum parsimony trees. We found 102 equallyparsimonious trees (length 127, CI = FIC. 2. Strictconsensus of 40 equallyparsimoni- 0.84, R! = 0.68), the strict consensusof which ous trees. supportsthe two trochilinegroups, but places Androdonsister to them.However, this topology wouldresult from excluding the swifts from used per clade.The two membersof eachof Fig. 2 and rerootingat one of the hermits.In- Zusi's (pers. comm.) four cladesgrouped to- clusionof the swifts(Fig. 2) thereforedid not gether,and topologies did notdiffer between F- alterthe phylogenetic patterns within the hum- M and distance-Wagnertrees. The high An- mingbirds,but the placementof theroot could dean taxa (Metallura,Oreotrochilus) were basal be problematicowing to thehigh level of swift- followedby the Andeangroup (Aglazactis, Coe- hummingbirddivergence. Thus, cladistic anal- ligena)and the remainingtwo lines,the Emer- yses do not support a hermit-Trochilinedi- alds(Amazilia, Campylopterus) and Bee (Acestru- chotomy. ra, Selasphorus)groups, the lattertwo of which aresister-groups. All distanceapproaches pro- ducedcongruent topologies for the trochiline- DISCUSSION B group. Parsimonyanalysis.--Various alleles (Table 1) Levelsof geneticdivergence.--Protein electro- support phylogeneticgroupings: Hermits phoresisfound widespreadapplication in avi- (ALD-2, CK-1, EAP, Phe-Pro,SORDH), trochi- an systematics,although it was appliedinfre- lines (EAP), Emeralds (LA), Bee (EST-D, LGG, quentlyat higher taxonomic levels (Dittmann et ME-1, NP), Andean (ALD-2, LA, LGG, MPI), al. 1989).The major emphasiswas on conge- and High Andean(GPI, LA, ME-1, MPI). We neric passerines,and in particular Nearcftc found 40 equally parsimonioustrees (length passerines(Zink 1991). These studiesfound 178, CI = 0.87, RI = 0.71). Using PAUP,we that birds were lessdifferentiated at compara- found that the shortest of 1,000 random trees ble taxonomiclevels than many other verte- was204, and the negativeg•-statistic (-0.88) is brates (Avise 1994). However, non-passerine significant,suggesting signal in the data set groupsand tropical passerines showed greater (Hillis 1991,Kallersjo et al. 1992).The strict levelsof divergence(Guti&rrez et al. 1983;John- consensustree (Fig.2) indicatesthat: Androdon sonand Zink 1983;Lanyon and Zink 1987;Ger- is a sistertaxon to the hermitsand theseplus win and Zink 1989; Gill and Gerwin 1989; the trochiline-Agroup are a sisterclade to the Hackett and Rosenberg1990; Christidis et al. trochiline-Bgroup. No clearpattern of sister- 1991; Randi et al. 1991, 1992; Hackett 1995; group relationshipsexists within the trochi- Brumfield and Capparella 1996). Levels of al- line-A group. Within the trochiline-Bgroup, lozymicdifferentiation in hummingbirdswere the Bee,Andean, and High Andeangroups high, 0.625(- 0.215)within familiesand 1.61 eachare monophyletic,whereas the two rep- (_+ 0.180) between families, consistentwith January1998] TrochilidaePhylogeny 111 other nonpasserines(e.g. Randi et al. 1991, logical variation,as well as the recentDNA- 1992). DNA hybridization study by Bleiweisset al. The absoluteage of hummingbirds is un- (1997).Because our distanceand cladisticanal- known: "There is no fossil record of the Tro- ysesdiffer, below we point out discrepancies. chilidaeother than of modernspecies from a Phaethorninae.--Distanceanalyses (Fig. 1) few Quaternarycave deposits,mostly in the suggestthat the hermits(Glaucis and Phaethor- WestIndies..." (Olson 1985).True swifts ap- his)are a sistergroup to otherhummingbirds, pear in the fossilrecord in the early Mioceneof which is consistentwith other evidence(Gould France (Olson 1985). If the Trochilidaeis the 1861;Zusi and Bentz1982; Sibley and Ahlquist sistergroup of swifts (as found by Sibleyand 1990; Bleiweiss et al. 1994, 1997; Hernandez-Ba- Ahlquist 1990),then hummingbirdsare of at nos et al. unpubl. data). Surprisingly,cladistic leastthis vintage.Based on DNA-DNA hybrid- analyses(Fig. 2) did not support the long- ization studies, the split between the two standingdivision into hermitsand trochilines. groups is an ancient one, occurringapproxi- Althoughour use of the swiftsapparently did mately 95 million yearsago (Sibleyet al. 1990). not biasingroup relationships, a traditionalto- Applying a calibrationof geneticdistances sug- pology (Fig. 3) required 180 steps,only two gestedby Marten and Johnson(1986), namely stepslonger than the mostparsimonious trees one unit of Nei's (1978) 1D equals20 million (178).Lack of a clearsignal for the basalrela- years (MY), the hummingbird-swiftsplit was tionshipslikely contributesto the different 32 MY ago;whereas, if we use the calibration placementof the hermitsin the cladisticanal- of Gutierrez et al. (1983), 1D = 26.3 MY, this ysisin whichrelatively few characterssupport split occurred43 MY ago (range 36 to 55 MY major groupings(Fig. 3). Basedon a 433-bp ago). Conversely,if one assumesa split of 95 segmentof mitochondrialcytochrome-b, Her- millionyears for swiftsand hummingbirds, the nandez-Banoset al. (unpubl.data) found that correspondingallozymic rate calibrationfor the two subfamilies of hummingbirds were hummingbirdswould be 1D = ca. 50 million only slightlymore distant than some of the tro- years, one of the slowestsuggested rates for chilines were from each other. Nonetheless, we vertebrates (Avise 1994). These rate calibra- concludethat the weight of evidencefavors a tions obviously conflict, and additional data hermit-trochilinedichotomy. are neededto resolvethem. We agreewith Av- Trochilinegroup A: Problemsof classifyingmor- ise (1994) that rate calibrationsmust be inter- phologicallycomplex taxa.--An exampleof com- preted cautiously. plex morphologicalpatterns obscuring phylo- geneticaffinities involves the placementof An- INTRAFAMILIAL PHYLOGENETIC RELATIONSHIPS drodonand Doryfera.Androdon and Doryferaare usually placedat the beginningof the hum- The lack of previousexplicit hypotheses of mingbird sectionof checklists(Peters 1945, higher-level relationshipsin hummingbirds Meyerde Schauensee1966, Morony et al. 1975), might be due to the complexnature of mor- which might simply representuncertainty phologicalvariation. Morphological patterns of abouttheir relationships.Our resultsand those variation,especially plumage patterns and col- of others (Zusi and Bentz 1982; Sibley and oration,may be inadequateindicators of phy- Ahlquist 1990, Bleiweisset al. 1994, 1997; logeneticrelationships because of convergence Schuchmannpers. comm.) indicate that Andro- or parallelism(homoplasy), sexual selection, or donand Doryferado not belongin the hermit extremeanagenesis (Sibley and Ahlquist 1990), group.Proteins of Androdonand Doryferawere factors probably widespread in humming- analyzedelectrophoretically along with several birds. Forexample, approximately 60 monotyp- Phaethornisspecies, both speciesof Eutoxeres ic generaoccur in the Trochilidae,a likely sig- and Threnetes, and Glaucis hirsuta, and were nal of taxonomicuncertainty in general(Plat- found to be distinct from phaethorninetaxa nick 1977) and amonghummingbirds in par- (Gill and Gerwin 1989). ticular (Gill and Gerwin 1989),owing to a lack We suspectthat the placementof Androdon of synapomorphiesat higherlevels. Our work near hermits in traditional checklists is the re- permitsevaluation of previoushypotheses of suit of several factors. It is morphologically hummingbirdrelationships based on morpho- similar to the hermits in two ways. Androdon 112 GERWINAND ZINK [Auk, Vol. 115

FIG.3. Topologysuggested by Sibleyand Ahlquist (1990) and distance-Wagnerapproach (Fig. 1). Numbers indicateunambiguous synapomorphies along branches. Note the relatively weak support for majorclades. hasa long (45 mm) and straightor slightlyup- sampling of taxa is insufficient to determine turnedbill similarto somespecies in thegenus preciselythe coevolutionaryrelationship be- Phaethornis.On its underparts(throat, breast tween this behavior and bill serrations, but our and abdomen),Androdon shows a pattern of tree agreeswith Ornelas(1994:708) in suggest- dull gray-white with dark brown/black ing that "the complexof featuresof the bill for streaks,similar to the hermit genusEutoxeres. nectar robbery has evolvedmore than oncein Conversely,Androdon is similar to many tro- birds with suchmorphology." Thus, bill ser- chilines in dorsal color (green), tail shape rationsin taxa suchas Androdonmight indeed (rounded)and pattern (broadwhite tips), al- be homoplasiousand unlikely to providea re- thoughall three charactersare shownby some liable phylogeneticsignal. Phaethorninae. The bills of Androdon and Rarn- In the distance analyses, Androdonwas phodonare uniqueamong trochilids in possess- placedin oneof four positions:(1) entirelyout- ing small,comb-like serrations along both tom- side the other trochilids;(2) as a sistertaxon to ia, and in beinghooked (Ornelas 1994). A num- the hermits;(3) as a separatelineage between ber of otherspecies have minute bill serrations, the hermitsand trochiline-Agroup; and (4) as involvingeither the maxilla or tomia,and some a sistergroup to the othertrochiline-A mem- additionallyhave a hook at the end of the bill; bers(as in Fig. 1). Maximumparsimony (Fig. 2) however,no otherspecies share the Androdon- consistentlyplaces Androdon as a sistertaxon to Ramphodonpattern. In addition to serratetom- the hermits. BecauseZusi and Bentz (1982), ia, Androdonand Rarnphodon share a similarpat- Sibley and Ahlquist (1990), Bleiweisset al. tern of streakingunderneath. Thus, morpho- (1994,1997) and our distance-Wagnertree (Fig. logicallyAndrodon shares traits with different 1) suggestthat Androdonis a basalmember of groups of hummingbirds,and its placement the trochiline-Agroup, we favorthis placement dependson which set of traits is emphasized. pendingan analysisthat includesmore taxa. Ornelas(1994) hypothesized that bill serra- Androdonand Doryferaappear to sharechar- tions aid in nectarrobbing. Unfortunately, our acteristicswith severalhummingbird groups, January1998] TrochilidaePhylogeny 113 which has no doubt contributed to taxonomic gestthe following ordering of taxabelonging to uncertainty.The placementof Doryferanear the familyTrochilidae (they did not includea Androdonin mostchecklists is presumablybe- putativeBee representative):(Phaethorninae, causeit resemblesAndrodon in bill morphology (trochilinae-Aplus Androdon),(High Andean, (long (25-35 ram) and straightor slightlyup- (Andean,Emerald)))). With theexception of the turned).Doryfera, however, lacks serrations on missing Bee group, this topology exactly thebill (Ornelas1994) and possesses a uniform matchesour distanceanalysis (Fig. 1), and re- greento darkoverall plumage with a glittering quires only four more stepsin our cladistic frontlet(green or violet);neither characteristic analysis;our distanceanalysis unites the Bee resemblesAndrodon nor Phaethornis.The posi- and Emerald groups as sister clades.Placing tion of Doryferavaries in our distancetrees, al- the Bee group basal within the trochiline-B thoughit is clearlyassociated with othertaxa groupresults in a total tree lengthof 182,and in the trochiline-Agroup. It is placedbetween movingit to the trochiline-Agroup (suggested Androdon and Schistes/Colibri within the tro- as a possibilityby R. Zusi) results in a tree chiline-Aassemblage (Fig. 1). However,in the length of 184. The study of Bleiweisset al. F-M treeDoryfera groups with Colibriand in our (1997),which unlike the Sibley / Ahlquiststudy cladisticanalyses it groupswith Schistes.Zusi includeda complete,reciprocal matrix of hy- and Bentz (1982) and Sibley and Ahlquist bridizationdistances, yields a topology(their (1990)placed Doryfera and Colibriwithin the figure2) that alsomatches that of ourdistance- trochiline-A assemblage.Our data agree in Wagnertree (Fig. 1). However,Bleiweiss et al. suggestingthat four of thetaxa we studied(An- (1997)favored their figure 3, whichdepicts the drodon,Doryfera, Colibri, and Schistes)are part Andean clade (Brilliants)as basalin the Tro- of this group,but discrepanciesbetween dis- chiline-Bassemblage. The distinctionbetween tance and cladisticanalyses do not suggesta treesin figures2 and 3 of Bleiweisset al. (1997) clearphylogenetic pattern (Figs. 1, 2).Bleiweiss involvedvery shortinternodes, and their rea- et al. (1997)found that Eulampis holosericeus and sonsfor favoringtheir figure 3 involvedas- sumptionsused in the analysisof DNA-DNA Heliothryxbarroti grouped with Doryfera,Colibri hybridizationdata. Althoughthey concluded and Androdon.Other taxa are consideredpart that additional study would likely favor the of this group (Zusi and Bentz1982), and fur- Andean (Brilliant) cladeas basal, our data sug- ther analysesare requiredto assuremonophy- gestthat the high Andeanclade is basal. ly of the group. Of the 25 generaand 26 speciesused by Blei- Trochilinegroup B.--Zusi and Bentz (1982) weisset al. (1997),eight genera and onlythree studiedhummingbird and swift musculature, specieswere commonbetween our studies. particularlythe tensorpatagii brevis (TPB) mus- Thus,different groups of specieswere used to cle,including the samegenera and mostof the representthe majorlineages. Although differ- same speciesincluded in this study. They entnumbers of specieswere used in eachstudy, foundtwo muscletypes among the trochiline- it is exceedinglyunlikely that our topologies B clade,and membersof the Beegroup (Selas- wouldmatch by chance.An obviousreason for phorus,Acestrura) have one or the other;unfor- the matchis thatboth treesrecover phylogeny; tunately no phylogeneticconclusion can be in theory,another reason might be that each drawn concerningthe monophylyof the Bee analysisis biasedby homoplasyin the same groupnor the relationshipsof the fourgroups. way. Although we found conflictingsupport Our datasupport the four groups suggested by for the relationshipsin Figure 1 in our alter- Zusi (pers.comm.), although support for the native analyses,we suggestthat the congru- Emerald clade was relatively weak (< 50% ence between the DNA-DNA hybridization bootstrap support). Hernandez-Banoset al. (Bleiweisset al. 1997:figure2) and our allo- (unpubl. data) suggestedthat Amazilialikely zymedistance tree (Fig. 1) indicatesthat the re- was not monophyletic,although the two Bee lationshipsof the majorclades of humming- hummingbirdsformed a clade.Thus, addition- birds are nearly resolved. al samplingis neededto assuremonophyly of the four groups,especially the Emeralds. ACKNOWLEDGMENTS Congruenceof allozymesand DNA-DNA hybrid- We thank numerouscolleagues at the LSUMNS ization.--Sibleyand Ahlquist (1990:846)sug- who collectedtissue samples used in thisstudy. We 114 GERWINAND ZINK [Auk, Vol. 115 thankthe lateJ. S. McIlhenny and the late B. S. Odum (Aves:Psittaciformes): Protein evidence.Condor for providingfunds for thoseexpeditions. We thank 93:302-317. E B. Gill and M. B. Robbins,of the Philadelphia DITTMANN, D. L., R. M. ZINK, AND J. A. GERWIN. Academyof Natural Sciences,for providing tissue 1989. Evolutionarygenetics of phalaropes.Auk samples.We thank E B. Gill, K.-L. Schuchmann,C. 106:326-331. G. Sibley,and R. Zusi for theuse of unpublisheddata FARRIS,J. S. 1972. Estimatingphylogenetic trees and for fruitful conversations.S. Lanyon,G. Graves, from distance matrices. American Naturalist R. Zusi, J.Klicka, K. Winker,J. Weckstein, K. Johnson, 106:645-668. A. Fry, and three anonymousreviewers provided FARRIS,J. S. 1981. Distancedata in phylogenetic useful commentson the manuscript.We thank C. analysis.Pages 3-24 in Advancesin cladistics(V. White,N. Tague,and S. J. Hackettfor preparingthe A. Funk and D. R. Brooks, Eds.). New York Bo- figures.J. V. Remsenand M. S. Hafner provided use- tanical Gardens, New York. ful commentson early drafts of the manuscript, FARRIS,J. S. 1985. Distance data revisited. Cladistics which formeda part of a M.S. thesisby Gerwin in 1:67-85. the Departmentof Zoologyand Physiologyat LSU. FARRIS,J. S. 1986. Distancesand statistics.Cladistics Financialsupport was provided by the American 2:144-157. Museumof Natural History (ChapmanFund), and FELSENSTEIN,J. 1985. Confidencelimits on phylog- the LSUMNS. enies:An approachusing the bootstrap.Evolu- tion 79:783-791. LITERATURE CITED FELSENSTEIN,J. 1986. Distancemethods: A reply to Farris. Cladistics 2:130-143. ARCHIE, J. W., C. SIMON, AND A. MARTIN. 1989. FITCH, W. M. AND E. MARGOLIASH. 1967. Construc- Smallsample size does decrease the stabilityof tion of phylogenetictrees. Science 155:279-284. dendrograms calculated from allozyme-fre- GERWIN,J. g., ANDR. g. ZINK. 1989. Phylogenetic quencydata. Evolution43:678-683. patternsin the genusHeliodoxa (Aves: Trochili- AVISE,J. C. 1994. Molecular markers, natural histo- dae):An allozymicperspective. Wilson Bulletin ry, andevolution. Chapman and Hall, NewYork. 101:525-544. AVISE,J. C., AND C. E AQUADRO.1982. A compara- GILL, F. B., AND J. A. GERWIN. 1989. Protein relation- tive summaryof geneticdistances in the verte- ships among Hermit Hummingbirds.Proceed- brates: Patternsand correlations.Evolutionary ingsof the Academyof Natural SciencesPhila- Biology15:151-185. delphia141:409-421. BAVERSTOCK,P. R., S. R. COLE,B. J. RICHARDSON,AND GOULD,J. 1861. An introductionto the Trochilidae, C. H. S. WATTS.1979. Electrophoresisand cla- or family of humming-birds.Published by the distics.Systematic Zoology 28:214-219. author, London. BLEIWEISS,g., J. g. W. KIRSCH, AND J. C. MATHEUS. GRAVES,G. R. 1980.A new speciesof metaltailhum- 1994. DNA-DNA hybridization evidencefor mingbird from northernPeru. WilsonBulletin subfamilystructure among hummingbirds. Auk 92:1-7. 111:8-19. GRAVES,G. R. 1986. Systematicsof the gorgeted BLEIWEISS,g., J. g. W. KIRSCH,AND J. C. MATHEUS. woodstars (Aves:Trochilidae:Acestrura). Pro- 1997. DNA hybridizationevidence for the prin- ceedingsof the BiologicalSociety of Washington cipal lineagesof hummingbirds(Aves: Trochil- 99:218-224. idae). MolecularBiology and Evolution14:325- GUTII•RREZ,R. J., R. M. ZINK, AND S. Y. YANG. 1983. 343. Genicvariation, systematic, and biDgeographic BOUCARD,A. 1895.Genera of hummingbirds.Pardy relationshipsof somegalliform birds. Auk 100: and Son, London. 33-47. BRUMFIELD,R. t., AND A. P. CAPPARELLA.1996. His- HACKETT,S. J. 1995. Molecularsystematics and zo- torical diversification of birds in northwestern ogeographyof fiowerpiercersin the Diglossabar- SouthAmerica: A molecularperspective on the itula complex.Auk 112:156-170. role of vicariant events. Evolution 50:1607-1624. HACKETT, S. J., AND K. V. ROSENBERG.1990. Evolu- BUTH,D. G. 1984.The applicationof electrophoretic tion of South American antwrens (Formicari- datain systematicstudies. Annual ReviewEcol- idae): Comparisonof phenotypicand genetic ogy and Systematics15:501-522. differentiation. Auk 107:473-489. CAVALLI-SFORZA,L. L., AND g. W. E EDWARDS. 1967. HARRIS,H., AND D. A. HOPKINSON. 1976. Handbook Phylogeneticanalysis: Models and estimation of enzyme electrophoresisin human genetics. procedures.American JournalHuman Genetics North Holland PublishingCompany, Amster- 19:233-257. dam. CHRISTIDIS, L., R. SCHODDE, D. D. SHAW, AND S. E HILLIS,D. M. 1991. Discriminatingbetween phylo- MAYNES.1991. Relationshipsamong the Aus- genetic signal and random noise in DNA se- tralo-Papuan parrots, lorikeets, and cockatoos quences.Pages 278-294 in Phylogeneticanalysis January1998] TrochilidaePhylogeny 115

of DNA sequences(M. M. Miyamotoand J.Cra- vol. 5. Harvard University Press,Cambridge, craft, Eds.). Academic Press,New York. Massachusetts. HILLIS, D. M., C. MORITZ, AND B. K. MABLE.(Eds). PLATNICK,N. I. 1977. Monotypyand the origin of 1996. Molecular systematics,2nd ed. Sinauer highertaxa: A reply to E. O. Wiley. Systematic and Associates, Sunderland, Massachusetts. Zoology 26:355-357. HINKLEMANN,C. 1989. Taxonomic,geographische PRAGER, E. M., AND A. C. WILSON. 1976. Construc- variationund Biogeographicder GattungPhac- tion of phylogenetictrees for proteinsand nu- thornis(Aves, Trochilidae). Dissertation, Rhein- cleic acids: Empirical evaluationof alternative ische Friedrich-Wilhelms-Universitat, Bonn, matrix methods.Journal of Molecular Evolution Germany. 11:129-142. JOHNSON,N. K., ANDR. g. ZINK. 1983. Speciationin RANDI, E., G. Fusco, R. LORENZINI, AND E SPINA. sapsuckers(Sphyrapicus): I. Genetic differentia- 1991. Allozyme divergenceand phylogenetic tion. Auk 100:871-884. relationshipswithin the Strigiformes.Condor JOHNSON,N. K., g. M. ZINK, G. E BARROWCLOUGH, 93:295-301. ANDJ. A. MARTEN.1984. Suggestedtechniques RANDI, E., A. MERIGGI,g. LORENZINI,G. Fusco, AND for modern avian systematics.Wilson Bulletin E U. ALKON. 1992. Biochemicalanalysis of re- 96:543-560. lationshipsof MediterraneanAlectoris partridg- KALLERSJO,M., J. S. FARRIS,A. G. KLUGE, AND C. es. Auk 109:358-367. BOLT.1992. Skewnessand permutation.Cladis- RICHARDSON,B. J., P. R. BAVERSTOCK,AND M. AD- tics 8:275-287. AMS.1986. Allozymeelectrophoresis. Academic LANYON, S. g., AND R. g. ZINK. 1987. Genetic vari- Press,Sydney, Australia. ation in piciformbirds: Monophylyand generic ROGERS,J. S. 1972. Measuresof genetic similarity and familial relationships.Auk 104:724-732. and genetic distance.Studies in geneticsVII. MADDISON, W. P., AND D. R. MADDISON. 1992. University of TexasPublications 7213:145-153. MacClade, version 3. Sinauer and Associates, SELANDER,R. K., M. H. SMITH, S. Y. YANG, W. E. Sunderland, Massachusetts. JOHNSON,AND J. B.GENTRY.1971. Biochemical MARTEN, J. A., AND N. K. JOHNSON. 1986. Genetic polymorphismand systematicsin the genus relationshipsof North American cardueline Peromyscus.I. Variation in the old-field mouse finches. Condor 88:409-420. (Peromyscuspolionotus). Studies in geneticsVI. MEYERDE SCHAUENSEE, R. 1966. The speciesof birds University of TexasPublications 7103:49-90. of South America with their distribution. Liv- SIBLEY,C. G., ANDJ. E. AHLQUIST.1990. Phylogeny ingstonPublishing Company, Narbeth, Pennsyl- and classificationof birds. A study in molecular vania. evolution. Yale University Press, New Haven, MORONY,J. t., W. J. BOCK,AND J. FARRAND,JR. 1975. Connecticut. Reference list of the birds of the world. Ameri- SIMON, E. 1921. Histoire naturelie des Trochilidae canMuseum of Natural History,New York. (Synopsiset catalogue).Encyclopedie Rorer, NEI, g. 1972. Genetic distancebetween popula- Paris. tions. American Naturalist 106:283-292. SMITH,A. B. 1994. Rootingmolecular trees: Prob- NEI, M. 1978.Estimation of averageheterozygosity lems and strategies.Biological Journal of the and geneticdistance from a smallernumber of Linnean Society51:279-292. individuals. Genetics 89:583-590. SNEATH, P. H. A., AND R. R. SOKAL. 1973. Numerical NEI, M. 1987. Molecular evolutionarygenetics. Co- taxonomy.W. H. Freeman and Company,San lumbia UniversityPress, New York. Francisco. NEI, g., F. TAJIMA,AND Y. TATENO.1983. Accuracy STILES,E G. 1983.Systematics of the southernforms of estimatedphylogenetic trees from molecular of Selasphorus(Trochilidae). Auk 100:311-325. data. II. Genefrequency data. Journal of Molec- STILES,E G. 1996. A new speciesof emeraldhum- ular Evolution 19:153-170. mingbird (Trochilidae,Chlorostilbon) from the OIlSON,S. L. 1985. The fossilrecord of birds. Pages Sierra de Chiribiquete,southeastern Colombia, 80-218 in Avian biology,vol. 8 (D. S. Farner,J. R. with a reviewof the C. mellisuguscomplex. Wil- King, and K. C. Parkes,Eds.). AcademicPress, son Bulletin 108:1-27. New York. SWOFFORD,D. L. 1981. Utility of the distance-Wag- ORNELAS,J. E 1994. Serratetomia: An adaptationfor ner procedure.Pages 25-44 in Advancesin cla- nectarrobbing in hummingbirds?Auk 111:703- distics,vol. 1 (V. A. Funk and D. R. Brooks,Eds.). 710. Annals of the New York Botanical Gardens, New PATTON,J. C., AND J. C. AVISE. 1983. An empirical York. evaluationof qualitative Hennigian analysesof SWOFFORD,D. L. 1993. PAUP:Phylogenetic analysis proteinelectrophoretic data. Journal of Molecu- using parsimony,version 3.1.1. Illinois Natural lar Evolution 19:244-254. HistorySurvey, Champaign. PETERS,J. L. 1945. Check-list of birds of the world, SWOFFORD,D. L., G. J. OLSEN,P. J. WADDELL,AND D. 116 GERWINAND ZINK [Auk, Vol. 115

M. HILLIS. 1996. Phylogeneticinference. Pages ZINK, R. M., AND J. C. AVISE. 1990. Patterns of mi- 407-514 in Molecular systematics,second edi- tochondrialDNA and allozymeevolution in the tion (D. M. Hillis, C. Moritz, and B. K. Mable, avian genus Ammodramus.Systematic Zoology Eds.). Sinauerand Associates,Sunderland, Mas- 39:148-161. sachusetts. ZINK, R. M., AND R. C. BLACKWELL.1996. Patterns of SWOFFORD,D. L., AND R. B. SELANDER. 1981. BIOS- allozyme,mitochondrial DNA, and morphomet- YS-I: A FORTRANprogram for the comprehen- ric variation in four sparrow genera.Auk 113: sive analysisof electrophoreticdata in popula- 59-67. tion geneticsand systematics. Journal of Hered- ZINK, R. M., AND D. L. DITTMANN. 1991. Evolution ity 72:281-283. of Brown Towhees: Mitochondrial DNA evi- ZIMMER, J. t. 1950-1953. Studies of Peruvian birds. dence. Condor 93:98-105. American Museumof Natural History Novitates ZINK, R. M., D. L. DITTMANN, AND W. L. ROOTES. Nos. 1449, 1450, 1463, 1474, 1475, 1513, 1540, 1991. Mitochondrial DNA variation and the 1595, 1604. phylogenyof Zonotrichia.Auk 108:578-584. Z•NK,R. M. 1991. Concludingremarks. Pages 629- ZusI, R. L., AND G. D. BENTZ. 1982. Variation of a 636 in Acta XX CongressusInternationalis Or- musclein hummingbirdsand swiftsand its sys- nithologici(B. D. Bell, Ed.). Christchurch,New tematicimplications. Proceedings of the Biolog- Zealand, 1990. New Zealand Ornithological ical Societyof Washington95:412-420. Trust Board,Wellington. AssociateEditor: A. ]. Baker January1998] TrochilidaePhylogeny 117

APPENDIX1. Speciesstudied, sample sizes, and localitiesfor specimens.

Species Tissueno. Locality Glaucis hirsuta 103561,103612 PERU: Loreto:S bank Maranon R. along Samiria (Rufous-breastedHermit) R., Est. Biol. Pithecia Phaethornisphilippii 9020,9168 BOLIVIA: depto.Pando; Prov. Nicolas Suarez, (Needle-billedHermit) 12 km by road S Cobija,8 km W on road to Mucden Androdonaequatorialis 1402 PANAMA: Darien;ca. 9 km NW Cana on slopes (Tooth-billedHummingbird) Cerro Pirre Colibri coruscans 5549, 5574 PERU: depto. SanMartin; 15 km NE Jirillo, (SparklingVioletear) 1,350 m Schistesgeoffroyi 8126 PERU:depto. Pasco; Playa Pampa, 8 km NW (Wedge-billedHummingbird) Cushion ChagllaTrail Doryferajohannae 1728 PERU:depto. Pasco; Santa Cruz; ca. 9 km SSE (Blue-frontedLancebill) Oxapampa 5402 PERU: depto. San Martin; 20 km NE Tarapoto, 1,050 m Acestrura rnulsant 6312, 6314 ECUADOR:Pichincha; Yanayacu, N SlopePi- (White-belliedWoodstar) chincha, 3,500 m Selasphorussasin 0142 USA: Louisiana; JeffersonParish; Metairie (Allen'sHummingbird) 5740 USA: Louisiana,E. BatonRouge Parish; Baton Rouge Amazilia viridicauda 8136, 8158 PERU:depto. Pasco; Cushi, ca. 1,800m (Green-and-whiteHumming- bird) Campylopteruslargipennis 4474 PERU: depto. Loreto;Lower Rio Napo, E Bank (Gray-breastedSabrewing) Rio Yanayacu 5577 PERU:depto. San Maritn; ca. 15 km by trail NE Jirillo on trail to Balsapuerto,1,350 m Aglaeactiscastelnaudii 3605,3620 PERU:depto. Huanuco; Quebrada Shugush, 30 (White-tuftedSunbeam) km on Huanuco-La Union road Coeligenaviolifer 3504 PERU:depto. Huanuco; Bosque Potrero, 14 km (Violet-throatedStarfrontlet) W Panao 8218 PERU:depto. Pasco; Millpo, E Tambode Vacas on Pozuzo-ChagllaTrail, 3,450m Metalluratyrianthina 8209,8218 PERU:depto. Pasco; Millpo, E Tambode Vacas (Tyrian Metaltail) on Pozuzo-ChagllaTrail, 3,450m Oreotrochilus estella 103834,103835 PERU:depto. Ayacucho; Pampa Galeras, 25 km (AndeanHillstar) WNW of Puquio,3,850 m Reinardasquamata 5039 PERU: depto. Loreto;S Rio Amazonas,ca. 10 (Fork-tailedPalm-Swift) km SSW Rio Napo Chaetura cinereiventris 9397 BOLIVIA: depto. Pando;Prov. NicolasSuarez, (Gray-rumpedSwift) 12 km by road S Cobija,8 km W on road to Mucden 118 GERWINAND ZINK [Auk,Vol. 115

APPENDIX2. Lociscored, gel typeused and the po- sitionof bandson that particulargel type.For loci with two types listed,both have reproduciblere- suits.

Locus E.C. number Gel-buffer *,b

GP-5 -- Poulik A ACON-2 4.2.1.3 AC C ACP-2 3.1.3.2 PC C ADA 3.5.4.4 TC III A ADH (ODH-1) 1.1.1.1 TM II C AK-1 2.7.4.3 TC III A AK-2 2.7.4.3 TM I A ALD-2 4.1.2.13 AC C CK-1,2 2.7.3.2 Poulik A ACP-1 (EAP) 2.7.3.2 TM I, A TC II ENOL 4.2.1.11 TM I A EST~D 3.1.1.1 Poulik A FUM 4.2.1.2 TC III C GAPDH 1.2.1.12 AC C GDA 3.5.4.3 TC II A ct-GPD 1.1.1.8 AC C G6PDH 1.1.1.49 PC A GPI 5.3.1.9 AC C GR 1.6.4.2 TM I A GLUD 1.4.1.3 TM II A GOT-1,2 2.6.1.1 AC C HK 2.7.1.1 TM I C IDH-1,2 1.1.1.42 AC C PEP~A(LA) 3.4.11 or 13 AC A LDH-1,2 1.1.1.27 Poulik A PEP-B (LGG) 3.4.11 or 13 AC A+C MDH-1 1.1.1.37 TC II A MDH-2 1.1.1.37 TC II, PP C ME-1 1.1.1.40 TC II, A TC 8.5 ME-2 1.1.1.40 TM 7.5 C MPI 5.3.1.8 AC A NP 2.4.2.1 AC A 6PGD 1.1.1.44 AC A PGM-2 2.7.5.1 Poulik, A TC 8.5 PEP-D 3.4.11 or 13 AC A (PHE PRO) SORDH 1.1.1.14 PP C PK-1 2.7.1.40 Poulik A PK-2 2.7.1.40 Poulik A

aAC = Amine-CitratepH 6.1; PC = Phosphate-CitratepH 6.2;PP = PGI-PhosphatepH 6.8;TC II = Tris-CitratepH 8.0;TC III= Tris- Citrate pH 7.0; TC 8.5 = TC lI titrated to pH 8.5 with N-3-aminopro- pylmorpholine;TM I = Tris-MaleatepH 7.5;TM II = Tris-edta-Ma- leate pH 6.5. • A = anodalmobility; C = cathodalmobility.