Intergeneric Phylogenetic Relationships of Swallows Estimated by Dna-Dna Hybridization

Home , Andean swallow, Atticora, Blue swallow, Cecropis, Hirundo, Notiochelidon

The Auk 110(4):798-824, 1993

INTERGENERIC PHYLOGENETIC RELATIONSHIPS OF SWALLOWS ESTIMATED BY DNA-DNA HYBRIDIZATION

FREDERICK H. SHELDON 1 AND DAVID W. WINKLER 2 •Departmentof Ornithology,Academy of Natural Sciences, 1900Benjamin Franklin Parkway, Philadelphia, Pennsylvania 19103, USA; and 2Sectionof Ecologyand Systetnatics,Division of BiologicalSciences, CorsonHall, CornellUniversity, Ithaca, New York 14853, USA

ABSTRACT.--Thephylogeny of the subfamily Hirundininae was estimatedby hybridizing single-copynuclear DNAs of 21 swallow species,representing 19 former and current genera, and a Tufted Titmouse(Parus bicolor) as outgroup.The phylogeny, which was unusually well resolved, consisted of three fundamental clades: Hirundo and allies, core martins, and African sawwings.The clade of Hirundoand allies comprisedHirundo rustica, Ptyonoprogne fuligula, Delichonurbica, Cecropis semirufa, Petrochelidon pyrrhonota, and P. spilodera.The sister-groupof Hirundoand allies was the core martin clade, which consistedlargely of endemic New World taxa (Pygochelidoncyanoleuca, Neochelidon tibialis, Atticora fasciata, Phaeoprogne tapera, Progne chalybea,Haplochelidon andecola, Stelgidopteryx ruficollis, and Tachycinetabicolor) and somebasally branchingOld World groups(Riparia riparia, R. cincta,Phedina borbonica, Pseudhirundo griseopyga, and Cheramoecaleucosternus). The African sawwings (represented by Psalidoprocneholomelas) formed the sister group of the core martins and Hirundoand allies. Among some interesting discoveries,we found a close relationship between the monotypic African and Australian generaPseudhirundo and Cheramoeca.We also found that Delichon,which has persistedin the nomenclatureas a genusseparate from Hirundo,is monophyleticwith taxathat are commonly considered to be members of Hirundo. On the other hand, Haplochelidonandecola, which is often considered to be a Hirundo or Petrochelidon,is not closely related to those genera, but instead lies among the New World membersof the core martin clade. Received1 July 1992, accepted25 November1992.

PERHAPSNO FAMILY of passerinesis asuniform titude of small, seemingly equally divergent morphologicallyand diverse genericallyas the genera had been defined mainly on the basis swallows (Hirundinidae). All swallow species of external morphological characters (e.g. conform to a fundamental body plan that in- plumage color, nasal form, fusion of skin be- cludeslong and pointedwings, mediumlength tween toes,tarsal feathering, and size).To bring tails, short legs, and bills that are short and someorder and logic to this confusingarray of wide. This uniformity is the likely resultof ad- names, Mayr and Bond grouped swallow gen- aptation to a strictly aerial insectivorouslife- era into units using what they consideredto be style. Apparently becauseof this uniformity, conservative, phylogenetically informative systematistshave been loath to attempt a phy- characteristics,in particular those of nesting logeneticreconstruction of the swallow family habits and plumage color patterns. In the pro- as a whole. While there have been many clas- cess,they outlined a rough scenarioof swallow sificationsof the Hirundinidae (e.g. Sharpe1885, evolution based on nesting strategy and to a Peters 1960, Turner and Rose 19•9, Sibley and lesser extent geography. They postulated that Monroe 1990)and many discussionsof the sys- the most "primitive" species were those that tematics of individual speciesor small groups nested in natural cavities or on ledges. These of taxa, only one publishedpaper has consid- were followed in evolutionary sequenceby spe- ered the familywide relationshipsof swallows ciesthat excavatednest holes and finally species based on evolutionary or phylogenetic logic. that built nests from mud. Old World groups This is the 50-year-oldstudy of Mayr and Bond were judged to be more primitive than New (1943). World taxa because Africa, as the continent with At the time of Mayr and Bond's (1943) study, the most swallow species,was perceived as the some 30 to 35 generic nameshad been applied center of their origin. In addition to behavioral to the 75 to 80 speciesof swallows. This mul- and geographic criteria, some features of exter- 798 October1993] SwallowPhylogeny 799

T^nLE1. List of commonly used generic names of portance of historical components to modern swallows and examples of synonymy. Current behaviorand ecology(e.g. Ricklefs1987, Brooks names taken from Sibley and Monroe (1990). Syn- onyms from Roberts(1922), Hellmayr (1935), Peters and McLennan 1991), further understanding of (1960), Brooke (1974), AOU (1983), and Turner and the phylogeny of this group seems especially Rose (1989). warranted. In the present paper, we reexamine the re- Alternative or subsumed lationships among genera of the Hirundininae Current name general names using genetic information derived from DNA- Tachycineta Iridoprocne,Callichelidon, Kalo- DNA hybridization. This independent source chelidon,Lamprochelidon of genealogical information has generated a Phaeoprogne -- Progne Phaeoprogne phylogenetic hypothesisthat can be used to test Notiochelidon Orochelidon,Pygochelidon the schemeof Mayr and Bond (1943), as well as Atti•:ora Diplochelidon serve as a template for analyses of character Neochelidon -- evolution in this popular group. Stelgidopteryx Alopochelidon Cheramoeca -- Riparia Neophedina,Cheimonornis METHODS Phedina Phedinopsis Hirundo Pseudhirundo,Ptyonoprogne, Selectionof taxa.--The speciesused in our study and Cecropis,Petrochelidon, Nata- sources of tissues for DNA extraction are listed in lornis, Charitochelidon,Phoeni- Table 2. While the taxonomic order of this list follows chelidon,Baruwaia, Haplo- Sibley and Monroe (1990), we have employed gen- chelidon erally older generic namesto clarify our discussion, Delichon particularly of taxa that are now commonly lumped Psalidoprocne in Hirundo. All of the specieslisted in Table 2 are members of the subfamily Hirundininae (true swallows). The other swallow subfamily Pseudochelidoni- nal morphology (e.g. wing serrations of saw- nae (river-martins) consistsof only two species,one wing or rough-winged swallows) were used to from central Africa (Pseudochelidoneurystomina) and define evolutionary advancement. one (possibly extinct) from Thailand (P. sirintarae). Unfortunately, we did not have DNA from either of Although openly preliminary, Mayr and these species. Bond's (1943) treatment remains the most in- In choosingtaxa for comparison,we tried to include fluential effort to date, supplying the rationale representativesof as many previously recognized behind the linear arrangementof swallow spe- genera as possible, especially those whose relation- ciesin mostclassifications. Under its logic, many ships remain obscure.We also included representa- of the old generic names have been subsumed tives of all the groups defined by Mayr and Bond (e.g. see Table 1), so that only 10 to 20 genera (1943). Here we summarizetheir groupsand list the are commonly recognized today (Brooke 1974, speciesin each that we compared. Mayr and Bond Bock and Farrand 1980, AOU 1983, Turner and (1943) divided Hirundointo three subgenera:(1) sub- Rose 1989, Sibley and Monroe 1990). However, genus Hbundo, cup-nesting barn swallows (Hirundo rustica)and crag martins (Ptyonoprognefuligula); (2) we still have no idea how thesemodern genera subgenusLillia, rufous-rumped, retort-nest builders fit together phylogenetically, or whether Mayr (Cecropissemirufa); and (3) subgenusPetrochelidon, cliff and Bond's evolutionary |ogic and choice of swallows (Petrochelidonpyrrhonota and P. spilodera). characters have merit. House martins (Delichon urbica) were viewed as "an The uncertainty of relationships among hi- obvious descendant of Petrochelidon"(p. 335). Sepa- rundines could well serve as sufficient justifi- rate from Hirundo and allies, Mayr and Bond distin- cation for further systematic research on the guisheda seriesof hole-nestinggroups: (1) bank-mar- group. The inherent interestin their genealogy tins (Ripariariparia, R. cincta),basically an Old World is strengthened because swallows are among group of four speciesthat excavatetheir own burrows; the most popular subjectsfor behavioral and (2) rough-winged swallow group (Stelgidopteryxruff- coilis,Neochelidon tibialis ), generally dull-colored,New ecologicalresearch in ornithology. Every year, World birds that nest in crevices or burrows excavated dozens of papers are published on the field bi- by otherspecies; (3) the Atticoragroup (Atticora fasciata, ology of several North Temperate species, es- Pygochelidoncyanoleuca), generally blue Neotropical pecially Hirundorustica, Petrochelidon pyrrhonota, taxa that nest mainly in crevices; (4) tree-swallow Delichonurbica, Tachycineta bicolor, and Progne group (Tachycinetabicolor), New World tree-hole subis.Given the growing awarenessof the im- adopters;and (5) purple-martin group (Prognesubis, 800 SHELDONAND WINKLER [Auk, Vol. 110 October1993] SwallowPhylogeny 801

Phaeoprognetapera), another group of New World hole (HAP) columnsplaced in a thermal-elution device adopters.Mayr and Bondalso identified severalprob- similar to those of Sibley and Ahlquist (1981) and lematic taxa that did not fit well into any of these Kirschet al. (1990).Fractions of swallow/swallowhy- categories.Of these, we comparedthe Australian brids were taken at 60øC and 72 ø to 94øC in 2.0øC White-backed Swallow (Cheramoecaleucosternus), Af- incrementsby pumping 4 ml of 0.12 M sodiumphos- ricanGrey-rumped Swallow (Pseudhirundo griseopyga), phate buffer through the HAP columns. Becauseof African sawwings (Psalidoprocneholomelas), Mascar- the greater genetic distancesinvolved, fractions of ene Martin (Phedinaborbonica), and Andean Swallow swallow/Paruscomparisons were taken over a wider (Haplochelidonandecola ). range (60ø and 68ø-94øC),but still in 2.0øCincrements. As an outgroup we used Tufted Titmouse (Parus Experimentaldesign and data analysis.--Each fraction- bicolor).This specieswas selected for two reasons. ation "experiment" consisted of a maximum of 35 PreliminaryDNA-DNA hybridizationstudies (Sibley hybrids.These hybrids were of two main types:ho- and Ahlquist 1990:fig.380) indicated that titmice are moduplexes(i.e. controlhybrids formed from labeled about as closegenetically to swallowsas any other and unlabeled DNA of a single individual); and het- group of sylvioid passerines.Second, we have been eroduplexes(i.e. hybrids of DNA from two different studying titmouse phylogeny using DNA-DNA hy- individualsof the sameor different species).To con- bridization (Sheldon et al. 1992). trol for sample-preparationbias, we hybridized a se- Our selectionof taxa for comparisonswas con- ries of different DNA preparationsof each species strained by the availability of tissuesfor DNA ex- (Table 2). traction. We did not have DNA of someparticularly To constructa complete matrix among 17 species interestingswallows, including Blue Swallow (Hirun- of swallows and Parus bicolor, we ran two kinds of do atrocaerulea),Brazza's Martin (Phedina["Phedinop- experiments. Ingroup experiments included only sis"]brazzae), Tawny-headed Swallow (Stelgidopteryx swallowsand comprised2 homoduplex,1 intraspe- ["Alopochelidon"]fucata) and, asmentioned above, riv- cific heteroduplex,and 32 interspecificheteroduplex er-martins. We were also limited in the number of hybrids. This design permitted the measurementof taxa that could be comparedin this study. For the two replicatesof 17 speciesof swallow per experi- most reliable results, DNA-DNA hybridization re- ment. To obtain a total of four replicates for each quires replicate pairwise measurementsamong all pairwisecomparison, we labeledall 17 swallow spe- species(Springer and Krajewski 1989a, Bledsoeand ciesand ran two experimentsfor eachlabeled sample. Sheldon 1990, Lanyon 1992), but as the number of In outgroup experiments,P. bicolorwas hybridized speciesunder examination grows, the number of re- with swallows.The reason for separatingingroup from quired comparisonsincreases geometrically. This is outgroup experiments was that outgroup compari- the "n 2 problem" describedby Barrowclough(1992). sonsrequired more temperaturefractions and, thus, Moreover, our DNA-hybrid fractionating machine were more expensiveto run than ingroup compari- could processonly 35 hybrids at a time. Thus, to sons.The three speciesthat were not part of the com- achieve an adequatenumber of replicates,we were plete 18 x 18 matrix (Cheramoecaleucosternus, Phedina limited to a matrix of 17 taxa (i.e. two replicates borbonica,and Haplochelidonandecola) were compared per machine run). In the end, we constructed a com- with all other species,but usually only as radio la- plete set of pairwisecomparisons among 17 species beled tracer. of swallowsand Parusbicolor. We alsocompleted a set The hybrid indexes(Tin, T50H, modified Fermi-Dirac of one-waycomparisons among these taxa and three mode, •Tm, •T5oH, •mode, and normalized percent additional species (Cheramoecaleucosternus, Phedina reassociation[NPR]) were calculated as described in borbonica,and Haplochelidonandecola), tissues of which Sheldonand Bledsoe(1989). Delta valuesare genetic were received late in our study. dissimilarities(distances) computed by subtracting Biochemistry.--Ourmethod of DNA-DNA hybrid- heteroduplexvalues from the averagehomoduplex izationwas basedon that of Sibleyand Ahlquist (1990) value. Uncorrected mean values for indexes and dis- and Sheldon et al. (1992). DNA was obtained from tancesare provided in the Appendix. When compil- frozen tissues(-80øC) or erythrocytesstored in 10% ing these means, we excluded data from entire ex- EDTA, extracted with pronase/phenol/chloroform, periments if we had major mechanicalfailure with RNAsed,and sonifiedto about500-base-pair lengths. the fractionator or if homoduplex Tinswere low Single-copyDNA was preparedto Cot1000 (Werman (<82øC), becausesuch low values suggestshort- et al. 1990)and oligo-labeledwith tritium (Feinberg strandednessof labeled DNA (Springer and Kirsch and Vogelstein 1983). Hybrids were formed with 1991).We excludedindividual hybrids if leakage,me- 20,000 to 50,000 disintegrationsper minute of radio- chanicalproblems, or possiblemisidentification or labeled single-copynuclear DNA tracer (i.e. about mixing of specimenswere discoveredin the course 0.002 #g) and 20 to 30 #g of nonlabeledtarget DNA of their preparationor fractionation.Because leakage (tracer:target about 1:10,000) and incubated at 60øC during incubationcaused unusually low percentsof in 0.48 M phosphate buffer to Cotof 22,000 or more. hybridization,we excludedall swallow/swallowhy- They were fractionated on 1.0-ml hydroxlapatite brids with less than 70% NPR, even if we did not 802 SHELDONAND WINKLER [Auk, Vol. 110

• • • • • 100 - 2o

• 0 0 I 76 80 84 88 92 76 80 84 88 92

Temperature ]Fig.1. Examplesof differential and cumulativeDNA-hybrid melting curves:(open circles)Petrochelidon spilodera;(closed circles) P. pyrrhonota;(open triangles) Hirundomstica; (closed triangles) Tachycinetabicolor; and (open squares)Pseudhirundo griseopyga.

detect leakage.Similarly, we excludedall swallow/ More importantly,AT• is better characterizedthan Parushybrids with NPR lessthan 50%.We also ex- the mode, having been calibrated numerous times cludedall hybridsthat showedbimodality in their againstsequences of knowndivergence (e.g. Caccone melting distributionssuch that the fractioncount of et al. 1988,Springer et al. 1992).While knownin some the secondarypeak was more than 100disintegrations casesto yield remarkablyaccurate estimates of long per minute larger than countsat adjacentfraction- geneticdistance (Goodman et al. 1990),ATsoH is gen- ation temperatures.In the end, about 10%of the hy- erally not as reliable and conservativea measureas brids were excluded. either ATtoand Amode when applied to short genetic We used AT• as the principal measureof phylog- distances(e.g. Sheldon and Bledsoe 1989, Schmid and eny.AT,, and Amodeyield virtuallyidentical results Marks 1990).It confoundstwo measures(Tin and NPR), and are highly correlated(Figs. 1 and 2), but ATtois and the latter carriesa large error (seeTable 3). the more robust of the two distances. It relies on data To constructour final estimateof swallow phylog- distributed over the entire melting curve, whereas eny, we correctedATto to ATmCin a two-stepprocess. Amodeis influencedmost significantly by datain the First, ATtowas converted to percent sequencediver- immediatevicinity of the melting-distributionpeak. gence(d) usingthe factorof Springeret al. (1992): d = 1.18 (ATe/100). (1) Then d wasadjusted for multiplemutations at single

lO basesites with the equationof Jukesand Cantor(1969), assuminga 60:40AT:GC base pair ratio (Arthur and

8 Straus1978, Epplen et al. 1978,Swofford and Olsen 1990):

6 ATmC= (100)(-0.74)1n(1 - 1.35[d]). (2) Correction to AT•C was intended to increase the ad- ditivity of the distancevalues, thereby improving the adherenceof the datato the additivityassumption of 4.,/ tree-buildingalgorithms (Springer and Krajewski 1989a, b).

2 ½ 6 8 10 12 To fit distancesto a tree-branchingpattern, we used the "Fitch" additive-distanceprogram of PHYLIP 3.4 AMode (Felsenstein1989). Options were setso that treeswere built by unweightedleast-squares regression (Cav- Fig. 2. Plot of Amode versusATto. Points represent alli-Sforza and Edwards1967) and the input order of comparisonsof mean values from Appendix. the taxa was random. "Fitch" does not assume a con- October1993] SwallowPhylogeny 803

T^m,E3. Summary of matricessubjected to bootstrapanalysis. Labeled Cheramoecaleucosternus, Phedina borbonica,and Haplochelidonandecola were not included in bootstrappedmatrices becausecomplete sets of pairwise comparisonswere lacking for these species.

No. No. Matrix size Distancetype bootstraps Descriptionof taxa and notes 1 18 x 18 ATmC 1,000 See Figure 4 2 18 x 18 AT• 100 3 18 x 18 Amode 100 4 18 x 18 AT m 1,000 Using "symboot"a 5 18 x 18 Amode 1,000 As in no. 4 6 17 x 17 ATto 100 Parus omitted 7 17 x 17 Amode 100 As in no. 6 8 7 x 7 ATto 100 Including Hirundo,Petrochelidon (2 species),Cecropis, Ptyonoprogne, Delichon, Pseudhirundo 9 7 x 7 Amode 100 As in no. 8 10 10 x 10 AT., 100 Omitting Hirundo,Petrochelidon (2 species),Cecropis, Ptyonoprogne, Delichon, Psalidoprocne,Parus 11 10 x 10 Amode 100 As in no. 10 12 10 x 10 AT. 100 As in no. 10, but with Petrochelidonspi- loderareplacing Pseudhirundo 13 10 x 10 Amode 100 As in no. 12 14 9 x 9 AT. 100 As in no. 10, but also omitting Progne 15 9 x 9 Amode 100 As in no. 14

""Symboot"is programof A. Dickerman(pers. comm.) that smoothesreciprocal discrepancies via methodof Sarichand Cronin(1976; e.g. see Springer and Kirsch 1989).

stant or monotonic rate of evolution. Unweighted RESULTS least squares is the appropriate method when vari- ance doesnot increasewith geneticdistance, as is the casein this and mostDNA-DNA hybridizationdata sets(e.g. Sheldon 1987a,Werman et al. 1990). Characterizationof data.--A total of 1,566 DNA The robustnessof branchingpatterns was tested by hybrids consistingof 20,516melting-curve frac- bootstrapping the 18 x 18 matrix and various of its tions contributed to the final phylogenetic es- subsets (see Table 3). Trees were then constructed timate. DNA of 80 individuals of swallows and from the bootstrappseudomatrices with "Fitch" and, 6 of P. bicolorwere used in the comparisons. fromthese trees, a majority-rulephylogeny was formed Mean values of the melting-curve indexes(T•, with PHYLIP's "Consense"program (Krajewskiand mode, and TsoH),NPR, and genetic distances Dickerman 1990, Dickerman 1991). We also jack- (ATto,Amode, and AT50H) are listed in the Ap- knifed our databy systematicallyremoving one taxon at a time and reestimatingthe tree (Lanyon 1985).As pendix. Table 4 summarizesstandard deviations notedby Krajewskiand Dickerman(1990), bootstrap- for all indexes.As is usually the case,mode and ping and jackknifing complementone another; jack- T,,are lessvariable than NPR and its dependent knifing teststree stability at the level of matrix col- Ts0H(Sheldon 1987a, b, Sarich et al. 1989, Shel- umns and rows,and bootstrappingtests at the level don and Bledsoe 1989; but see Marshall and of matrix cells. Swift 1992). Reciprocal measurement discrep-

TABLE4. Summarystatistics of hybrid stability,reassociation, and distancevalues, and of matrixasymmetry.

Index Tm AT• Mode Amode TsoH AT5oH NPR Mean SD 0.31 0.23 0.27 0.21 0.56 0.52 4.82 Matrix asymmetrya -- 4.45 -- 3.62 -- 12.02 -- Mean percentnonreciprocity for 18 x 18 matrix(Sarich and Cronin 1976,Springer and Kirsch1989). 804 SHELDONAND WINKLER [Auk,Vol. 110 October1993] SwallowPhylogeny 805

r•_•Petrochelidonpyrrhonota* 0.99

Hitundoandallles • 1.00 Dellchon urbica

Hitundo rustica

Ptyonoprognefuligula

1.00 [---Phaeoprognetapera* L-prognechalybea*

1.00 1.0r•_•A#icørafasciata* 0.511.00 , •Neochelidon tibia#s* Pygochelidoncyanoleuca*

Stelgidopteryxruficol#• *

Tachycinetabicolor*

Ripa•fa •fpa•fa cor1.00[ 0.99 Riparia cincta martins[ Pseudhirundogriseopyga

Africansawwings Psalidoprocneholomelas

,Parusbicolor (outgroup)

Fig. 3. Majority-ruleconsensus tree of 18 x 18 ATmCmatrix computed by "Consense"program of PHYLIP 3.4 from 1,000 bootstrap"Fitch" trees. ancy values (Table 5) are also summarized in group or ingroup species,the branchingpattern Table 4. might change(e.g. Lanyon 1985).To test such Ratesof evolution.--Ratesof swallow evolu- taxic effects given the data at hand, we built tion were measured by the outgroup relative trees using subsetsof the species(Table 3). In rate test of Sarich and Wilson (1967). Distances doing so, we made assumptionsabout mono- to and from Parusbicolor and the swallow species phyly of groupsand designationsof outgroups. displayedsignificant differences (Kruskal-Wal- For example, we assumed,based on traditional lis one-way ANOVA chi-squareapproximation classificationand their derived state of pure = 37.68, P = 0.007), which suggestsvariability mud-nest building, that Hirundo,Petrochelidon, in rates. All rate tests of DNA hybridization Cecropis,Delichon, and Ptyonoprognerepresented data, however, suffer because of the noninde- a monophyletic group. We then used one of pendence of the distances;that is, sets of dis- these taxa as outgroup in estimating trees for tance measurementsare linked by common nonmud-nestingswallows, and vice versa.All homoduplex references.There is the possibili- such tests,as summarized in Table 3, yielded ty, therefore, of measurement errors causing trees that were consistent with the branching systematicpatterns that appear asrate variation. of the 18 x 18 consensustree (Fig. 3), if nodes Phylogeny.--Figure3 is a majority-rule con- that are supportedby fewer than 89% of the sensustree built after bootstrappingthe 18 x bootstrapsare collapsed. In addition, a strict- 18 ATmCmatrix 1,000 times. This tree is an un- consensustree producedby jackknifing the 18 usually well-resolvedmolecular estimate of avi- x 18 matrix (Lanyon 1985)was identical to the an phylogeny(e.g. Lanyon 1985, Sheldon 1987b, 89%majority-rule bootstraptree. These analy- Edwards et al. 1991, Sheldon et al. 1992), but its ses indicate that the DNA-DNA hybridization degreeof resolutionrelates only to the taxawe phylogenyis remarkablyinsensitive to changes sampled. If we were to compare different out- in outgroup and ingroup taxa. We were con- 806 SHELDONAND WINKLER [Auk,Vol. 110

ATmC which is the sister taxon of Hirundo and allies and includes all the endemic New World genera, as well as some Holarctic, African, Asian, and Australian taxa; and (3) the African saw- •L• Petrochelidonpyrrhonota* wings (representedby Psalidoprocneholomelas), Petrochelidonspilodera which appear as the sister taxon to the rest of -C•cropissemirufa the hirundinines. Delichonurbica Within the clade of Hirundo and allies, Hirun- Hirundorustica doand Ptyonoprogneare basallybranching taxa; Ptyonoprognefu#gula Delichon,Cecropis, and Petrochelidonare more re- cently derived. The preciserelationships of Hi- _•PrognePhaeoprognetapera* rundoand Ptyonoprogneto one another and to chalybea* the other membersof this group have not been resolved. In about 40% of the bootstraps,Hirun- • Atticorafasciata* do and Ptyonoprognewere linked together on a Neochelidontibialis* single branch;at other times they switched back Haplochelidonandecola* and forth as the sistertaxon to Delichon,Cecropis, Pygochelidoncyanoleuca* and Petrochelidon.In contrast,the relationships of the three latter genera are clearer. The Af- I .Stelgidopteryxruficollis*rican and American cliff swallows, P. spilodera Tachycinetabicolor* and P. pyrrhonota,are almostcertainly sister taxa, Ripadaripatfa despite their markedly disjunct distribution. They occur on the same branch in all of the jackknife trees and in most bootstraptrees. In

Cheramoeca leucosternus the bootstrapanalyses listed in Table 3, they co- occur the following percentagesof the time: I• Pseudhirundogriseopyga(analysis 1) 89%, (2) 89%, (3) 99% (4) 81%, (5) Psalidoprocneholomelas 98% (6) 87%, (7) 98%, (8) 93%, and (9) 96%. / The core martins contain a well-resolved ter- ,,/ Parusbicolor (outgroup) minal clade of New World endemics,Pygochel- idon,Neochelidon, Atticora, Haplochelidon, Phaeo- Fig. 4. Best-fit least-squares tree computed by "Fitch" program of PHYLIP 3.4 from ATmCvalues progne,Progne, and Stelgidopteryx.Among these using 21 labeledtaxa. Negative branchesnot allowed, taxa, only the positionsof Stelgidopteryxand the P set to 0, and sum of squaresequal to 139.72. Progne/Phaeoprogneclade are ambivalent.In most analyses,Stelgidopteryx occurs as the outgroup to the other taxa (e.g. in 51% of the 18 x 18 cerned particularly about the selection of Parus bootstraptrees and 16 of 17 jackknife trees).The bicoloras outgroup for the study. This choice, nextoutgroup of thisterminal New World group however, while possiblyaffecting the branch- is also unresolved. The Holarctic Ripariariparia ing pattern of the three major clades (see be- and the New World endemic Tachycinetabicolor low), clearly did not affect branching within occur either as sister taxa to the clade of New lower-level clades. World endemics or to one another. In all anal- Figure 4 is our best-fit 21-taxon "Fitch" tree, yses,Riparia cincta is the sistertaxon of the New which includesthe (mainly) one-way compar- World endemics(including Tachycineta)plus Ri- isonsof Cheramoeca,Phedina borbonica, and Hap- paria riparia.In the 21-taxon tree (Fig. 4), a short lochelidon.It is consistentwith Figure 3, but the branch links Ripariacincta and Phedinaborbonica, robustnessof the branching pattern could not but the stability of this branch has not been be tested becauseof the lack of a complete set tested.Finally, Pseudhirundoand Cheramoecajoin of reciprocal data. to form a branch that is the sister to all the other Three fundamentalgroups are definedin Fig- core martins. ures 3 and 4: (1) Hirundoand allies, represented Psalidoprocneholomelas always appears as the by Hirundorustica, Petrochelidon spilodera, P. pyr- outgroup to the other hirundinines in "Fitch" rhonota,Cecropis semirufa, Delichon urbica, and tree analyses.However, in some "Kitsch" anal- Ptyonoprognefuligula; (2) the "core martins," yses,which assumea constantrate of evolution, October1993] SwallowPhylogeny 807

P. holomelasswitches position with the clade adopters (Pygochelidon,Neochelidon, Atticora, comprising Cheramoecaand Pseudhirundo. Phaeoprogne,Progne, Stelgidopteryx, and probably Tachycineta)have arisen independently from hole excavators. See Winkler and Sheldon (1993) DISCUSSION for a formal analysis of nest-type evolution. ASSESSMENTOF MAYR AND BOND (1943) Mayr and Bondalso implied that mud nesting progressedevolutionarily from simple cups to The groups of genera outlined by Mayr and more ornate structures. Our data suggest that Bond (1943) basedon nesting habits and plum- the complexity of nests is tied to phylogeny. age patterns conform to our clades to a remark- Hirundoand Ptyonoprogne,which build cup nests, able degree.The only discrepancyis their place- branched earlier than Delichon, which builds an ment of Neochelidontibialis. They groupedit with enclosed nest. Delichon, in turn, branched ear- Stelgidopteryx,and we found it to be more close- lier than Cecropisand Petrochelidon,which build ly related to Pygochelidonand Atticora.Our re- retort nestswith entrance tunnels. By itself, our suits even support several of Mayr and Bond's phylogeny does not indicate a progression more subtle assertionsor suggestions.For ex- through time; the same number of character- ample, they thought that the Australian Cher- state changesis required to achieve the distri- amoecaand African Pseudhirundomight be dis- bution of nest types on the phylogeny whether tantly related to one another based on their the primitive nest was cup- or retort-shaped. "tunneling" behavior and color patterns. We However, the observation in living birds that have found just such a relationship between the constructionof cupsprecedes enclosure into these highly disjunct,monotypic genera. Mayr closedcups or retorts, provides developmental and Bond indicated that the generic allocation evidence in support of Mayr and Bond's view of Haplochelidonhinged on its nesting habits;if of an evolutionary progressionfrom simple to it were found not to build a mud nest, then it complex nests (Winkler and Sheldon 1993). would belong with the New World endemic genera rather than in Petrochelidon.We now TAXONOMIC HIGHLIGHTS know that it does not build a mud nest and that it is more closely related to Pygochelidon,Atti- Barn Swallows and allies.--In most modern cora, and Neochelidon than to Hirundo and allies classifications,Hirundo encompasses the former (for a thorough discussion,see Parkes 1993). genera Petrochelidon,Cecropis, and Ptyonoprogne, Mayr and Bond also remarked on the difficulty but excludes Delichon.Mayr and Bond (1943) of allocating such taxa as Phedina,Cheramoeca/ and subsequentclassifiers have suggested,how- Pseudhirundo,and Psalidoprocneto larger groups. ever, that Delichon and Hirundo should be In light of our molecular data, it is clear that lumped. Wolters (1952) for example, like Mayr these taxa were difficult to place becausethey and Bond (1943), discounted the importance of are relatively highly divergedand without close tarsalfeathering in Delichonas a distinguishing relatives.Finally, Mayr and Bond divided swal- characteristic and emphasized instead the oc- lows into Old and New World groups,and our currenceof interspecifichybridization between results concur; the New World endemic taxa speciesof Hirundo and Delichonas evidence of are confined to the terminal branches of the their closerelationship. Our resultssupport this core martin clade. view, asDelichon, Petrochelidon, and Cecropishave Our resultsdo not supportthe rationale used been found to be monophyletic. Interestingly, by Mayr and Bond, however, to describe the Delichonreplaces the two latter genera geo- relative evolutionary advancement of major graphically (and probably adaptively) in Eu- swallow groups. Mayr and Bond regarded the rope and northern Asia (Turner and Rose 1989). adoption of natural hollows as the most likely Woltersalso suggested that Hirundo(sensu stric- primitive nesting strategy,followed by the ex- to) and Ptyonoprognemight be closerto one an- cavation of holes and finally building of mud other than to Delichon,Cecropis, and Petrocheli- nests. Instead, we found hole excavators (Psal- don. Our data, unfortunately, do not indicate idoprocne,Pseudhirundo, and Riparia) to be the the precisesister-group relationships of Hirundo most basally branching ("primitive") swallows rustica and Ptyonoprognefuligula vis-a-vis the and that mud nesters(Petrochelidon, Cecropis, De- other three genera. Even if nest-type data are lichon, Hirundo, and Ptyonoprogne)and hole mapped onto the phylogeny, a clearer picture 808 SHELDONAND WII•IKLER [Auk, Vol. 110 of the rusticaand fuligulasister relationships does minor importance and merged the two genera not emerge, because their cup nesting is the into Progne.Our results indicate a close rela- apparent primitive state. tionship between Prognechalybea and Phaeo- EndemicNew World genera.--The interrela- prognetapera. The distance between them was tionships of Pygochelidon,Neochelidon, Atticora, the shortestof any two genera we compared Stelgidopteryx,and other endemic New World (ATto0.75) and typical of closecongeners (e.g. taxaare particularly poorly known. On the basis Stegidopteryxruficollis to S. serripennisis 0.64;Pet- of plumage, Brooke (1974) grouped Notiocheli- rochelidonspilodera to P. pyrrhonotais 1.47; see don, Pygochelidon,Neochelidon, and Atticora to- Appendix). Given the monophyly and close- gether in an "Atticora"group. Turner and Rose ness of Progneand Phaeoprogne,we agree with (1989) presenteda somewhat different arrange- Turner and Rose(1989) that thesegenera should ment. They grouped the four speciesthat had probably be lumped. been distributed among Notiochelidon,Orochel- Tachycineta,represented in our phylogeny by idon,and Pygochelidoninto Notiochelidon,citing T. bicolor,was the only endemic New World similaritiesin their body proportions,plumage, genus that did not group unambiguouslywith and nesting. They placed the enlarged Notio- the other New World endemics. However, if its chelidon next to Atticora in their linear classifi- nesting behavior (viz. hole adoption) is taken cation and separated these two genera from into accountand treated as a synapomorphy, Neochelidon,which they believed to be closely then Tachycinetabecomes the sister taxon of all related to Stelgidopteryxand Alopochelidon.Their the other New World endemic genera (Winkler arrangementof taxaresembles Mayr and Bond's and Sheldon 1993). Further genetic work on the (1943) in that it distinguishesthe nest-adopting other Tachycinetaspecies and their inter- and species with marked amounts of blue in their intragenericrelationships is underway. plumages (e.g. Pygochelidonand Atticora)from Other taxa.--With the exceptionof Ripariari- the browner nest adopters(e.g. Neochelidonand paria, which is Holarctic in distribution, the re- Stelgidopteryx).Our results are more in accord maining species in our study are Old World with Brooke's (1974) arrangement. We found taxa. Of the relationshipsfound among them, that there is no subdivision of these taxa based the most unexpectedin terms of traditional clas- on plumage color (i.e. the dull-colored Neo- sificationis the paraphyly of Ripariariparia and chelidonappears to be more closely related to R. cincta.Both speciesexcavate nest cavities in the blue Atticoraand Pygochelidonthan to the sand banks and are quite similar in plumage. dull Stelgidopteryx). However, cinctadiffers from all other Riparia A speciesthat has often been placed in Hi- speciesin overall size, bill and nostril shape, rundoor Petrochelidon(e.g. Peters 1960, Turner possessionof sharploral bristles,and the extent and Rose 1989, Sibley and Monroe 1990), but of its coloniality. On the basisof these differ- which is clearly more closelyrelated to the en- ences, Roberts (1922) placed cinctain a mono- demic New World genera, is the Andean Swal- typic genus,Neophedina, and it is for this reason low (Haplochelidonandecola). This dull-colored we included it in our study. In addition to these blue-backedhole adopter is the sister taxon of morphological and behavioral differences, we Pygochelidonin our tree. As mentioned above, have found cinctato be much more highly di- Mayr and Bond (1943) indicated that the affin- verged from riparia than is another congener, ities of this specieswould be clarified once its the Plain Martin (R. paludicola):AT• of 2.88 ver- nesting was described. Parkes (1993) has re- sus 1.4, respectively (unpubl. data). Given the viewed the taxonomic issues in detail. apparent paraphyly of ripariaand cincta,which The genera Progneand Phaeoprognehave al- is supported circumstantially by the other ev- ways been thought to be closely related. The idence of substantial divergence, we are in- monotypic Phaeoprognewas originally erected clined at this point to agreewith Roberts(1922) to emphasizeits differencesfrom Progne:similar and recognizeNeophedina for cincta. sexes,lack of metallic blue in its plumage, a In Figure 4, Ripariacincta and Phedinaborbonica more slender bill, a lessdeeply forked tail with appear as sister taxa. Despite the coincidental broader plumes, weaker feet, and more exten- nature of their alternativegeneric names (Phedi- sive feathering on the inner side of the upper na and Neophedina),their relationship to one tarsus (Zimmer 1955). Turner and Rose (1989), another is not well established. The robustness however, believed these characteristics to be of of the branch linking the two specieswas not October1993] SwallowPhylogeny 809 tested becausewe were unable to completea TABLE6. Comparison of distances from this study set of reciprocalcomparisons of borbonica.In ad- and from Sibley and Ahlquist (1982). dition, borbonica does not resemble cincta in Sibley and This studyb plumage or nesting behavior. Indeed, borbonica Ahlquist• is a most unusual swallow in several respects. Taxa (Ts0H) T, T50H In particular,its nest is apparentlyunique among Labeled Riparia riparia swallows.It is an open cup placed in a nook or Prognesubis 0.7 3.26 3.97 crevice (e.g. on beamsin buildings or among Hirundo rustica 2.4 3.30 3.50 rocksby the coast)and apparentlycontains no Petrochelidonspilodera 2.5 3.69 4.18 mud,but insteadis constructedentirely of veg- Notiochelidon murina 2.5 -- -- etable matter (Langrand 1990;R. K. Brookeand Stelgidopteryxruficollis 3.0 2.54 3.67 Parus atricapillus 10.8 -- -- T. Schulenbergpers. comm.). The nest doesnot even resemble that of borbonica'ssole putative Labeled Hirundo rustica congener, Brazza'sMartin (Phedina["Phedinop- Petrochelidonspilodera 0.5 2.53 3.00 Riparia riparia 2.1 3.41 4.36 sis"] brazzae),which is placed in a (self-exca- Parus bicolor 11.2 8.78 9.33 vated?) tunnel (Turner and Rose 1989). Because Eachvalue representsa single hybrid measurement. of its unusual nest and general plainness,bor- Eachvalue representsa mean distance(see Appendix). bonicahas been describedas the "least special- ized"or "mostprimitive" of all swallows(Mayr and Bond1943, R. K. Brookepers. comm.). Until DNA of brazzae becomes available and more Ahlquist's T5oHdistances are shorter than ours. complete comparisonscan be made between it, To someextent, this shortnessmay be an artifact borbonica,and the other basallybranching core of differencesin methodology.In the caseof martins,the compositionand phylogeneticpo- labeledH. rustica,it almostcertainly results from sition of Phedina remain uncertain. a low-homoduplex melting temperature and Psalidoprocne,the genusof African sawwings, consequentdistance compression in the Sibley is a relatively large group of about 12 taxa. We and Ahlquist data (C. Sibley pers.comm.). (The have assumedthe monophylyof Psalidoprocne,causes and consequencesof DNA-hybrid dis- basedon the synapomorphicserration on the tance compressionare describedin Sheldon and outerweb of the first primary and other plum- Bledsoe[1989] and Springer and Kirsch [1991].) age and behavioral similarities. The inclusion Although distracting,this compressiondoes not of more speciesof Psalidoprocnein the study, changethe ranking of their distancesfrom ours however,would undoubtedlyhave helped to (e.g.H. rusticais closer to P. spiloderathan to solidifythe positionof thisgenus in the family, Riparia).On the other hand, their labeledRiparia aswell asreveal interesting phylogenetic struc- distancestell a different story. Their Riparia/ ture among the congeners. Prognedistance is muchshorter than ours(AT•r/ 0.7 versus3.97), and their Riparia/Stelgidopteryx COMPARISON WITH SIBLEY AND distanceis longer than the Riparia/Hirundoand AHLQUIST'S(1982) RESULTS Riparia/Petrochelidondistances. The disparityin distanceranking led Sibleyand Ahlquist (1982, The only othermolecular study that bearson 1990) to a branching pattern that differs mark- the generic-levelrelationships of swallowsis edly from ours. Their tree placesStelgidopteryx the DNA-DNA hybridizationwork of Sibley as the outgroup to a clade comprisingHirundo, and Ahlquist (1982, 1990).They labeled two Petrochelidon,Riparia, and Progne. speciesof swallow, Ripariariparia and Hirundo Our resultsare likely to be moretrustworthy rustica,and hybridized them with a series of than Sibley and Ahlquist'sfor severalreasons. swallowand nonswallowspecies in a studyde- We comparedmany more swallowsthan they signedprimarily to determinethe extrafamilial andreplicated pairwise measurements (usually) relationshipsof the Hirundinidae. Their results among several individuals of each speciesto were summarizedas lists of TsoHsand a phen- achieverelatively robustestimates of distances ogram. We have extractedpertinent distances among those species,as well as among major from their tablesand juxtaposed them with ours clades.Based as they were on nonreplicated in Table 6. comparisons,Sibley and Ahlquist's distances For all swallow/swallowhybrids, Sibley and between specieswere not averages,but single 810 SHELDON^ND WIN•CLER [Auk,Vol. 110 points subject to greater perturbation or error ACKNOWLEDGMENTS (e.g. as causedby mismeasurement,misidenti- In particular, we thank Maureen Kinnarney and ficationof species,or mix-upsof DNA samples). Marija Helt for technicalassistance. Others who helped Further, by comparing so few species,Sibley in the laboratory are L. Hyatt, K. Kahn, S. Mitra, B. and Ahlquist did not produce replicate esti- Slikas, and T. Tufts. Tissueswere kindly donated by mates of distancesamong major clades,a pro- the collectionslisted in Table 2. For help obtaining cessthat helps to fortify internodes(again by materials for DNA extraction, we thank F. Catzeflis, averaging).Finally, unlike Sibleyand Ahlquist, L. Christidis, S. Goodman, L. Kiff, S. Lanyon, M. Mar- we completeda matrix of pairwise comparisons. in, E. Morton, M. Robbins,T. Schulenberg,F. Sibley, Doing so allowed us to avoid phenetic cluster- and R. Zink. While collecting in South Africa, we ing methods and the consequentassumptions received a great deal of help from Alan Kemp and the staff of the Transvaal Museum and also benefited of a molecularclock and monophyly of taxathat from the help and hospitalityof T. Crowe, R. Earl•, they require (e.g. Bledsoeand Sheldon 1990, D. and H. Hall, A. Harris, C. Moloney, P. Ryan, and Lanyon 1992, Sheldon and Bledsoe1993). the Doyer and Neser families. We owe specialthanks to T. and R. Cassidy,W. Neser, and J. McCarty for the successof our trip. A. Bledsoe,R. Brooke, T. Crowe, ISSUES OF GENERIC CLASSIFICATION and R. Prum commentedon the manuscript.Funding was supplied by NSF BSR-8806890,BSR-9020183, and As mentioned at the outset, there is a mul- a grant from the American PhilosophicalSociety. titude of poorly defined swallow genera. Even with the enthusiasticlumping of the last 50 years, 10 to 20 genera still remain in common LITERATURE CITED use. This number is in marked contrast to some other sylvioid families (sensuSibley and Mon- AMEmCAN ORNITHOLOGISTS' UNION. 1983. Check-list roe 1990). For example, Paridae (titmice and of North American birds, 5th ed. Am. Ornithol. chickadees)consists of only a singlegenus, even Union, Washington, D.C. though it exhibitsessentially the sameamount ARTHUR, R. R., AND N. A. STRAUS. 1978. DNA-se- of genetic divergence as Hirundinidae among quence organization in the genome of the do- mestic chicken (Gallus domesticus).Can. J. Blo- its most distantly related species(Sheldon et al. chem. 56:257-263. 1992). BARROWCLOUGH,G.F. 1992. Biochemical studies of It is our intention eventually to revise the the higher level systematicsof birds. Bull. Br, classification of the Hirundinidae so that swal- Ornithol. Club Supp1,112A:39-52. low taxa are monophyleticand the linear ar- BLEDSOE,A. H., AND F. H. SHELDON. 1990, Molecular rangement of speciesreflects phylogeny (e.g. homologyand DNA hybridization.J, Mol. Evol. Wiley 1981).A genericrevision with theseob- 30:425-433. jectiveswould require, for example,that Pseud- BOCK,W. J., AND J. FARRAND. 1980. The number of hirundogriseopyga and Haplochelidonandecola be speciesand genera of recentbirds: A contribution split and separatedfrom Hirundoand allies,with to comparativesystematics. Am. Mus. Novit. 2703: 1-29. which they have often been associatedbut are BROOKE, R. K. 1974. Review of the classification of not monophyletic.We also believe, in lieu of the Hirundinidae (Aves:Passeriformes) with par- namingeach bifurcation, that the family should ticular reference to those of the Old World. M.S. be divided into genericgroups that reflectbasic dissertation,Univ. Natal, Pietermaritzburg,South biological characteristics.Thus, we are inclined Africa. to maintain the genera Petrochelidon,Cecropis, BROOKS,D. R., ANDD. A. MCLENNAN.1991. Phylog- and Delichon,based on the characteristicsof their eny, ecology,and behavior. Univ. ChicagoPress, nesting.Such splitting would simplify the de- Chicago. scription of distinct, coherent,and ecologically CACCONE,A., R. DESALLE,AND J. R. POWELL. 1988. significantgroups. However, before we can un- Calibrationof the changein thermal stabilityof dertake this revision, we need to learn a great DNA duplexesand degree of basepair mismatch. J. Mol. Evol. 27:212-216. deal more about the phylogeny of many taxa, CAVALLI-SFoRZA, L. L., AND A. W. F. EDWARDS. 1967. especially the speciosegenera Hirundo(sensu Phylogeneticanalysis: Models and estimation lato) and Psalidoprocne,and the poorly under- procedures. Evolution 21:550-570. stood complex of New World Neotropical en~ DICKERMAN,A.W. 1991, Among-rUnartifacts in DNA demics. hybridization. Syst. Zool. 40:494-499. October1993] SwallowPhylogeny 811

EDWARDS,S. V., P. ARCTANDER,AND A. C. WILSON. roles of local and regional processes.Science 235: 1991. Mitochondrial resolution of a deep branch 167-171. in the genealogicaltree for perchingbirds. Proc. ROBERTS,A. 1922. Review of the nomenclature of R. Soc. Lond. B 243:99-107. South African birds. Ann. Transvaal Mus. 8:187- EPPLEN,J. T., M. LEIPOLDT,W. ENGEL,AND J. SCHMIDTICE. 272 1978. DNA sequenceorganization in avian ge- SARICH, V. M., AND J.E. CRONIN. 1976. Molecular nomes. Chromosoma 69:307-321. systematicsof the primates.Pages 141-170 in Mo- FEINnERC,A. P., AND B. VOGELSTEIN. 1983. A tech- lecular anthropology (M. Goodman and R. E. nique for radiolabeling DNA restriction endo- Tashian, Eds.). Plenum Press, New York. nucleasefragments to high specificactivity. Anal. SARlCH,V. M., C. W. SCHMID, AND J. MARKS. 1989. Blochem. 132:6-13. DNA hybridization as a guide to phylogenies:A FELSENSTEIN,J. 1989. PHYLiP--Phylogenyinference critical analysis.Cladistics 5:3-32. package(version 3.2). Cladistics5:164-166. SARlCH, V. M., AND A. C. WILSON. 1967. Immuno- GOODMAN,M., D. A. TAGLE,D. H. A. FITCH, W. BAILEY, logical time scale for hominoid evolution. Sci- J. CZELUSN•aC,B. F. KOOP, P. BI•SON, mad J. L. ence 158:1200-1203. SLIGHTOM. 1990. Primate evolution at the DNA SCHMID,C. W., AND J. MARKS. 1990. DNA hybrid- level and a classification of hominoids. J. Mol. ization as a guide to phylogeny:Chemical and Evol. 30:260-266. physical limits. J. Mol. Evol. 30:237-246. HELLMAYR,C. H. 1935. Catalogue of birds of the SHARPE,R. B. 1885. Catalogue of the birds in the Americasand the adjacentislands, part 8. Field British Museum, vol. 10. Trustees British Muse- Mus. Nat. Hist. Zool. Ser. 13, part 8. um, London. JUKES,T. H., AND C. R. CANTOR. 1969. Evolution of SHELDON,F. H. 1987a. Rates of single-copy DNA protein molecules. Pages 21-123 in Mammalian evolution in herons. Mol. Biol. Evol. 4:56-69. protein metabolism (H. N. Munro, Ed.). Academ- SHELDON,F. H. 1987b. Phylogeny of herons esti- ic Press, New York. mated from DNA-DNA hybridization data. Auk KIRSCH,J. A. W., R. J. GANJE,K. G. OLESEN,D. W. 104:97-108. HOFFMAN, AND A. H. BLEDSOL. 1990. TED, an SHELDON, F. H., AND A. H. BLEDSOL. 1989. Indexes improved thermal elution devise for the simul- to the reassociationand stability of solutionDNA taneous hydroxyapatite chromatographyof so- hybrids. J. Mol. Evol. 29:328-343. lution DNA/DNA hybrids.BioTechniques 8:505- SHELDON, F. H., AND A. H. BLEDSOL. 1993. Avian 506. molecular systematics,1970s to 1990s.Annu. Rev. KRAIEWSKI,C., AND A. W. DICKERMAN. 1990. Boot- Ecol. Syst. 24:243-278. strapanalysis of phylogenetictrees derived from SHELDON,F. H., B. SLIKAS,M. KINNARNEY, F. B. GILL, DNA hybridizationdistances. Syst. Zool. 39:383- E. ZHAO,AND B. SILVERIN.1992. DNA-DNA hy- 390. bridization evidence of phylogenetic relation- LANGRAND, O. 1990. Guide to the birds of Mada- ships among major lineages of Parus.Auk 109: gascar. Yale Univ. Press, New Haven, Connect- 173-185. icut. SIBLEY,C. G., ANDJ.E. AHLQUIST.1981. The phylog- LANYON,S.M. 1985. Detecting internal inconsisten- eny and relationshipsof the ratite birds as in- cies in distance data. Syst. Zool. 34:397-403. dicatedby DNA-DNA hybridization. Pages301- LANYON,S. M. 1992. Interspecificbrood parasitism 335 in Evolution today (G. G. E. Scudderand J. in blackbirds (Icterinae): A phylogenetic per- L. Reveal, Eds.). Carnegie-Mellon Univ., Pitts- spective. Science 255:77-79. burgh. MARSHALL, C. R., AND H. SWIFT. 1992. DNA-DNA SIBLEY,C. G., AND J.E. AHLQUIST. 1982. The rela- hybridization phylogeny of sand dollars and tionshipsof the swallows(Hirundinidae). J. Ya- highly reproduciblehomologous extent of hy- mashina Inst. Ornithol. 14:122-130. bridization values. J. Mol. Evol. 34:31-44. SIBLEY,C. G., AND J.E. AHLQUIST.1990. Phylogeny MAYR,E., ANDJ. BOND. 1943. Notes on the generic and classification of birds. Yale Univ. Press, New classification of the swallows, Hirundinidae. Ibis Haven, Connecticut. 85:334-341. SIBLEY,C. G., AND B. L. MONROE. 1990. Distribution PARKES,K. C. 1993. Systematicsand nomenclature and taxonomy of birds of the world. Yale Univ. of the Andean Swallow, "Petrochelidon" andecola. Press, New Haven, Connecticut. Auk 110:947-950. SPRINGER,M. S., E. H. DAVIDSON, AND R. J. BRITTEN. PETERS,J. L. 1960. Family Hirundinidae. Pages80- 1992. Calculation of sequencedivergence from 129 in Check-list of birds of the world, vol. 9 (E. the thermal stability of DNA heteroduplexes.J. Mayr and J. C. Greenway, Jr., Eds.). Museum of Mol. Evol. 34:379-382. Comparative Zoology, Cambridge, Massachu- SPRINGER,M. S., AND J. A. W. KIRSCH. 1989. Rates of setts. single-copy DNA evolution in phalangeriform RICKLEFS,R.E. 1987. Community diversity: Relative marsupials.Mol. Biol. Evol. 6:331-341. 812 SHELDONAND WINKLER [Auk, Vol. 110

SPRINGER,M. S., ,•ND J. A. W. KIRSCH. 1991. DNA 1990. Nucleic acidsI: DNA-DNA hybridization. hybridization, the compressioneffect, and the Pages 204-249 in Molecular systematics(D. M. radiation of diprotodontian marsupials.Syst. Zool. Hillis and C. Moritz, Eds.). Sinauer, Sunderland, 40:131-151. Massachusetts. SPRINGER,M., .•',ID C. KRAJEWSK•.1989a. DNA hy- WILEY, E. O. 1981. Phylogenetics, the theory and bridization in animal taxonomy: A critique from practiceof phylogenetic systematics.Wiley, New first principles. Q. Rev. Biol. 64:291-318. York. SPRINGER,M. S., AND C. KRAJEWSKI.1989b. Additive WINKLER, D. W., AND F. H. SHELDON. 1993. The evo- distances,rate variation, and the perfect-fit the- lution of nest construction in swallows (Hirun- orem. Syst. Zool. 38:371-375. dinidae): A molecular phylogenetic perspective. SWOF•ORDD. L., AND G. J. OLSF•. 1990. Phylogeny Proc. Natl. Acad. Sci. USA 90:5705-5707. reconstruction.Pages 411-501 in Molecular sys- WOL•'EP,S, H. E. 1952. Die Gattungen der westpa- tematics (D. M. Hillis and C. Moritz, Eds.). Sin- laearktischen Sperlingsv6gel (Ordn. Passeri- auer, Sunderland, Massachusetts. formes). Bonn. Zool. Beitr. 3:231-288. TUP,NER, A. T., AND C. ROSE. 1989. Swallows and ZIMMER,J.T. 1955. Studies of Peruvian birds. No. martins. An identification guide and handbook. 66. The swallows (Hirundinidae). Am. Mus. Nov- Houghton Mifflin, Boston. it. 1723:1-35. WERMAN, S. D., M. S. SPRINGER,AND R. J. BKrrrm•. October! 993] SwallowPhylogeny 813

+1+1+1 +1+1+1+1 +1+1+1 +1+1+1+1+1 +1+1 +1 +1 +1 +1+1 +1+1+1 +1 +1+1 +1+1+1 +1+1 +1+1 +1

+1 •1 •1 +1 +1 +1 •1 •1 +1 •1 •1 +1 •1 +1 •1 +1 •1 •1 +1

• +1+l +1+1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 •

+1+1+1+1+1+[+1+1+1+1+1+1+1+1+1+1+1+1+1 m

+1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 814 $u•rvou Ai•i•WI•CI.œR [Auk,Vol. 110 October1993] SwallowPhylogeny 815 816 SHELDONAND V•rlNKLER [Auk,Vol. 110

+1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1

+1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1

+} +1 +1 +1 +1 +1 +1 +1 +} +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1

+1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +[ +1 +1 +1 +1 +1 +1 +l+l+l+l+l+l+l +1+1+1+1+1+1

..... • .... October1993] SwallowPhylogeny 817

+1 +1 +1+1 +1+1 +1

..•... • oo •o

+1 +1 +1 +1 +1 +1 +1 +1 +1 +1+1 +1+1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1+1

+1 +1 +1 +1 +1 +1 +1

0 818 SHELDONAND WINKLER [Auk,Vol. 110

..• ...... •

c•c• October1993] SwallowPhylogeny 819

+1 +1 +1 +[ +1+1 +1 +1 +1+1 +1 +1 +1+1 +1 +1 +1 +1 +1

+1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1

+1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +l +1 +1 +1 +1 +1

..... •..

+1 +1 +1 +1 +1 +1 +1+1 +1 +1 +1 +l+l +1 +1 820 s.•rvotq AND WIIqKLER [Auk, Vol. 110

+l+l+l+l+l+l+l+l+l+l+l+l+l+l+l+l+l+l+l+l +1+1+1+1+1+1+1+1+1+1+1+1+1+1+1+1

+l +l +l +l +l +l +l +l +1 +l +l +l +l +l +l +l +{ +l +l

+l +l +l +l +1 +l +l +l +l +l +1 +1 +l +l +l +l +l +1 +l +l October1993] SwallowPhylogeny 821

+1 +1 +{ +1+1 +1+1 +1+1+1+1+1 +1+1+1+1 +1+1+1+1+1 +1+1 +1+1+1 +1+1 +1

+1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1

+1 +1+1+1+1 +l+l +1 +1+1 +1 +1+1

+1 +1 +1

+1 +l +l+l +1+1+1 +1+1 +1 +1 +1 +1+1

+1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1+1+1+1 +1 +1+1+1 +1+1 +1 +1

oo oo oooooo +1 +1 +1 +1+1 +1+1 +1+1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +l +1 +1 +1 822 SHELDONAND WlrqKLER [Auk, Vol. 110

+l +l +l +l +l +1 +l+[+l+{+l+l+l+l+l+l+l+l+l +l

+l +{ +l +l +l +1 +• +• +• +• +• +• +• +• +• +• +• +• +• +• +• +• +: +• +l +l +l +l +l +l +l +l +l

+l +l +l +l +l +l +l +l +l +l +l +l +l +l +l +l +l +l +l +l +{ +l +l +l +l +l

+l+l+l • +[+l +[ +l +l +l +l +{+l +l +l +l +{+l +l +l +l +l

+1 +l +l +l +l +l

+l +l +l +l +1 +l +l +l +l +l +l +l +l +l +l +l October1993] SwallowPhylogeny 823

+1 +1 +1 +1 +1 +1 +1 +1 +1 +[ +1 +1 +1 +1 +1 +1 +1 +1 +1+1+1 +1+1 +1+1

+1 +1 +1 +1+1+1 +1+1 +1+1+1 +1+1 +1+1 +1 +1 +1 +1 +[ +1 +1 +1 +1 +1

+1 +1 +1 +1 +1 +l +1 +t +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +[ +1 +1 +1 +1 +1 +1

+1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +I +1 +1 +1 +1 +l+l+l+l+l+l+l 824 SHELDONAND WINKLER [Auk,Vol. 110