THE CONDOR

JOURNAL OF THE COOPER ORNITHOLOGICAL SOCIETY

Volume 88 Number 4 November 1986

The Condor X8:409-420 0 The CooperOrnithological Society 1986

GENETIC RELATIONSHIPS OF NORTH AMERICAN CARDUELINE

JILL A. MARTENS AND NED K. JOHNSON Museum of VertebrateZoology and Department of Zoology, Universityof Cahfornia, Berkeley, CA 94720

Abstract. Starch gel electrophoresiswas used to examine variation at 33 genetic loci in 19 taxa (15 speciesin 6 genera) of cardueline finches (family Fringillidae). Levels of heterozygosityand genetic distanceswere comparable to those reported from surveys of other avian taxa. Twenty- three loci (70%) were polymorphic within taxa and/or were fixed at alternative alleles among taxa. Rogers’ genetic distances were used to construct phenograms, distance Wagner trees, and F-M trees;these provided hypothesesfor the evolutionary relationshipsof taxa. The geneticdata indicate that: (1) Coccothraustes,Pinicola, Leucosticte,Carpodacus, , and Loxia are distinctive generathat vary in estimated age (as measuredfrom nearestbranch point) from approximately 14 MY (Coccothraustes)to 5 MY (Loxia); (2) speciestreated by the AOU (1983) as congenerswithin Carpodacus,Carduelis, and Loxia are correctly classified to genus; (3) the subgeneraAcanthis, Astragalinus., and Carduelis,within the genusCarduelis, are recognizable;(4) the (Loxia) are most closelyallied to Carduelisamong the generaexamined; (5) Carpodacuspurpureus and C. cassiniiare closely related sister specieswhereas C. mexicanusis very distinct; (6) Loxia curvirostraand L. Zeucopteraare moderately different electrophoretically;(7) in contrast, the red- polls, Carduelisflammea and C. hornemanniexilipes, are similar genetically;(8) most speciation events in North American carduelines range from mid-late Pliocene (4 MY) to mid-Pleistocene (500,000 years)in age;but (9) subspeciesdiverged in the late Pleistocene.A phylogenyofcardueline genera derived from these electrophoretic data agrees in major respectswith one proposed by Raikow on the basis of hindlimb myology. The sequenceof appearanceof older taxa is still not resolvedwith certainty, however, becauseofpartially conflictingmolecular and morphologicresults. Key words: Carduelinefinches; Coccothraustes; Pinicola; Leucosticte;Carpodacus; Carduelis; Loxia; allozymes;phylogenetic inference; genetic distance.

INTRODUCTION Short (1970) began their sequence of North Modern studies of the relationships of nine- American generawith Carpodacus,and placed primaried oscines agree that the cardueline Leucostictenext to last, Raikow (1978), in an finches (subfamily Carduelinae, family Frin- analysis of limb myology, deemed Leucosticte gillidae) are closely related; they share a rea- to be the most primitive cardueline of the gen- sonably consistent combination of similar era he examined. The AOU (1983) agreedwith morphologic and behavioral traits (Tordoff Raikow and began their sequence with Leu- 1954a, 1954b; Bock 1960, Raikow 1978). This costicte.The treatment of Howell et al. (1968) consensusis reflected in the classificationsof contrastswith all three of the aforementioned Howell et al. (1968), Mayr and Short (1970), sequences.Considering only the North Amer- and the AOU (1983). Within the cardueline ican generain their world list, they placed Leu- finches, however, relationships are still poorly costictebetween Acanthis and Carpodacus,ap- understood.Limits of generaare controversial, proximately one-third of the way from the particularly among Old World forms. The beginning. proper sequenceof taxa in systematic lists is Because previous systematic approaches another continuing problem. The position of have failed to reconcile the differing views on the genusLeucosticte, for example, illustrates the internal classification of carduelines, the the disagreement often encountered when se- group seemed eminently suitable for the fresh quences are compared. Although Mayr and perspective offered by biochemical methods. Accordingly, we electrophoretically compared 19 taxa, representing 15 speciesin 6 genera, I Received 29 November 1985. Final acceptance 10 most of which are native to North America. February 1986. * Presentaddress: St. Mary’s Hospital and Medical Cen- The analysis includes all speciesin the AOU ter, 450 Stanyan Street, San Francisco, CA 94 117. (1983) except vagrants and introduced forms.

[4091 410 JILL A. MARTEN AND NED K. JOHNSON

Because the greatest diversity within the BIOSYS- 1 (Swofford and Selander 198 1) was subfamily occurs in the Old World (where all usedto compute expectedheterozygosity (H_,) of the 18 genera listed by Howell et al. [ 19681 per sample, percentage polymorphic loci, av- occur and where 11 [6 l%] of the total genera eragenumber of alleles per polymorphic locus, are endemic), we do not attempt to interpret Nei’s (1978) and Rogers’ (1972) genetic dis- relationshipsbeyond the North American taxa. tances(Table 3), UPGMA and WPGMA phe- Avian electrophoretic research is still in an nograms (Sneath and Sokal 1973) and dis- expanding phase (see reviews of existing stud- tance Wagner trees (Farris 1972, 198 1; ies and citations in Barrowclough 1983, Cor- Swofford 198 1). The distribution of observed bin 1983, Matson 1984, Zink and Johnson and expected number of heterozygotes (Table 1984, and Barrowclough et al. 1985). There- l), over all loci in a sample was examined for fore, the new information on carduelinesadds departure from Hardy-Weinberg expectation to the gradually growing base of avian genetic (Hart1 198 1) with a x2 test (Barrowclough 1980). data. Finally, because we make genetic com- Fitch-Margoliash (F-M) trees (Fitch and Mar- parisons at the familial level, the cardueline goliash 1967) were generated with the com- data are also of “macrotaxonomic” interest. puter program EVOLVE. The various branch- As Barrowclough has noted (1983:228), re- ing diagrams portray patterns of genetic newed investigation of the utility of electro- similarity and provide estimates, under differ- phoresisin the systematicsof higher categories ing assumptions,of the evolutionary relation- of is just beginning. Only a few recent ships among taxa (Felsenstein 1983, 1985). examplesexist of studiesin which geneticcom- parisons have been made above the generic RESULTS level; these include Barrowclough et al.( 981) VARIATION AT LOCI AND HETEROZYGOSITY on the Procellariiformes, Gutierrez et al. ( 983) Of the 33 loci scored, 14 (42.4%) showed at on Galliformes, and Johnson and Zink ( 985) least a single heterozygote. At nine other loci on the Vireonidae. (Glud, Eap, Got- 1, Ald, Acon, Ck- 1, Gda, Ck-2 and Ldh- 1) the species(including the outgroup MATERIALS AND METHODS taxon) were fixed at alternative alleles. Thus, Using starch gel electrophoresis,we am._11 yzed we regard 23 (70%) of the total loci as being tissuefrom 96 specimensrepresenting 1Y taxa variable within the taxa surveyed. Allelic fre- (15 speciesof 6 genera) of cardueline finches. quenciesat the polymorphic loci are listed by All but one of theseforms, the European Gold- taxon in Table 2. The 10 monomorphic loci (Carduelis carduelis), are native to North were: Icd-2, Sod-2, Got-2, Mdh-I, Mdh-2, America. A single specimen of the Sage Spar- Ldh-2, Ab-hemoglobin, Pgm-2, Ab- 1 and Ab- row (Amphispiza belli; subfamily Emberizi- 2. We also attempted to analyze five additional nae, family Emberizidae) was used as an out- loci, Gpt, La-2, Est-D, Gsr and Acp, but these group. Taxa studied, sample sizes and proved to be unscorable. geographic sourcesof specimens are listed in Levels of genetic variation within taxa are Table 1. Nomenclature follows the most recent shown in Table 1. Hobs.ranged from 0.0 (in Check-list of North American birds (AOU Carpodacus p. purpureus,Loxia c. grinnelli and 1983). Carduelis lawrencei) to 0.103 (in Loxia 1. leu- Procedures for the collection and storageof coptera).Average Hobs.over all taxa was 0.034, tissue sampleshave been described elsewhere a value 21% lower than the average of 0.043 (Johnsonet al. 1984). Electrophoretic methods reported for birds in general (Barrowclough essentially followed Selander et al. (197 1) and 1980) and a value 36% lower than the average Yang and Patton (198 l), with the slight mod- of 0.053 reported for large singlebreeding pop- ifications outlined by Johnson et al. (1984). ulations of 30 species(summarized in Barrow- Thirty-three presumptive genetic loci were clough 1983:228-229). Few patterns were ob- scored. Alleles at a locus were coded by their served;however, we note that the three species mobility from the origin. The most anodal lo- of North American goldfinches (subgenusAs- cus was designated as “a,” with successively tragalinus) all have low observed heterozy- slower alleles denoted as “b,” “c,” etc. Iso- gosities (X = 0.012). Percentage of polymor- zyme nomenclature follows Yang and Patton phic loci ranged from 0.0 (again involving C. (198 1). From banding patterns on gels (pre- p. purpureus, L. c. grinnelli and C. lawrencei) sumptive individual genotypes), we derived a to 20.7 (in Loxia 1. leucoptera), with a mean table of allelic frequencies(Table 2). Observed of 8.97. The average number of alleles per heterozygosity (Hobs.)was determined by direct polymorphic locus ranged from 1.00 in the count for each specimen and then averaged three monomorphic forms already cited to 1.45 (f SE) for each sample. The computer program (in Carpodacus m. frontalis), with a mean of GENETIC RELATIONSHIPS OF CARDUELINE FINCHES 411

TABLE 1. Taxa studied, sample sizes, sourcesof specimens,and intraspecificgenetic variation.

No. Percent- all&s w AVW- at poly- POlY- at%+ mor- mor- number 7;;; phic of TaXOIl n Sample region* H,, + SE H,., + SE locib ?lll&S’

Rosy Finch (Leucosticte arctoa littoralis) 3 Idaho 26 0.034 & 0.019 0.034 * 0.019 10.34 1.10 Pine (Pinicola enucleator alascensis) 4 Alaska 27 0.034 + 0.014 0.047 + 0.023 13.79 1.14 (P. e. leucura) 3 Minnesota 28 0.069 ? 0.034 0.064 + 0.027 17.24 1.17 (Carpodacus purpureus purpu- 4 Michigan 23 0.0 0.0 0.0 1.00 reus) (C. p. calijknicus) 15 Califomiad 29 0.034 * 0.009 0.041 * 0.021 13.79 1.21 Cassin’s Finch (Carpodacus cassinit) 12 Califomiad (1 1), 31 0.026 f 0.007 0.038 f 0.014 13.79 1.28 Montanad (1) (Carpodacus mexicanus fronta- 19 Califomiad 36 0.040 * 0.009 0.056 + 0.022 20.69 1.45 lis) Red (Loxia curvirostra grinnellz) 1 Califomiad 23 0.0 0.0 0.0 1.oo White-winged Crossbill (Loxia leucoptera leucoptera) 3 Alaska 31 0.103 f 0.034 0.094 ? 0.038 20.69 1.28 Common (Carduelisjlammeaflammea) 5 Alaska 24 0.021 + 0.008 0.016 f 0.016 3.45 1.03 Hoary Redpoll (Carduelis hornemanni exilipes) 2 Alaska 26 0.052 & 0.018 0.052 * 0.029 10.34 1.10 (Carduelis pinus pinus) 6 Califomiad (3), 28 0.035 k 0.015 0.033 k 0.016 13.79 1.17 Minnesota (3) (Carduelis psaltria hesperophilus) 4 Califomiad 25 0.017 + 0.010 0.016 + 0.016 3.45 1.07 Lawrence’s Goldfinch (Carduelis lawrencet) 3 Califomiad 23 0.0 0.0 0.0 1.00 (Carduelis tristis tristis) 5 Michigan 24 0.014 & 0.008 0.012 Ifr 0.012 3.45 1.03 (C. t. salicamans) 2 Califomiad 24 0.017 * 0.017 0.017 z!z0.017 3.45 1.03 (Carduelis carduelis) 1 Australia 25 0.069 0.069 2 0.048 6.90 1.07 (cagebird) (Coccothraustes vespertinus ves- 2 Minnesota 25 0.052 -+ 0.018 0.040 + 0.028 6.90 1.07 pertinus) (C. v. brooksz) 2 Oregot+ 25 0.036 + 0.034 0.034 + 0.024 6.90 1.07 SageSparrow (Amphispiza belli nevadensis) 1 Nevadad 24 0.034 0.034 Ik 0.034 3.45 1.03 Total and meanse 91 0.034 0.035 8.97 1.12

a Exact localities available from authors. b Frequency of most common allele 5 0.95. r Per locus. d From breeding grounds. r Unweighted by sample size.

1.12. These values are all very dependent on none of the 20 population samplesdeparts sig- sample size. nificantly (P > 0.05) from Hardy-Weinberg For all comparisons(Table l), IJobs.and He._,. expectations. are similar. The greatest difference occurs in Carpodacuscassinii, in which Hobs,is approx- GENET1CD*STANCES imately 32% less than Hexp..However, chi- Genetic distances(Nei ’s D 1978) between sam- square tests reveal that genetic variation in ples differentiated at several taxonomic levels 412 JILL A. MARTEN AND NED K. JOHNSON

TABLE 2. Allelic frequencies for polymorphic loci. Numbers in parenthesesare frequencies for alleles (coded as letters), when a particular allele was not fixed. Abbreviations for proteins follow Harris and Hopkinson (1976).

L. a. P. e. P. e. c. p.pur- c. p. C. c. m. L. c. L. 1. C.f LOCUS littoralis alascensis Ieucura pureus californicus cassinir frontalis grinnelli leucoptera flammea

Mpi e C C b b b d (0.97) b b b e (0.03) Gpd b(0.17) C C C C a (0.04) b (0.03) b b C c (0.83) c (0.88) c (0.97) d (o.osj Icd- 1 e b (0.25) b(0.17) b b c (0.08) c (0.05) b a (0.17) a (0.30) e (0.75) e (0.83) e (0.92) e (0.95) c (0.17) c (0.70) e (0.66) Adh e a a f f (0.83) c (0.18) b b f g (0.17) f (0.82) Glud C b C C C C Pgm- 1 d d : (0.96) d (0.97) b a(0.17) b e (0.04) e (0.03) b (0.83) Eap b d d b b e a a e Sdh C d d e C a (0.03) C C C c (0.97) Got- 1 a b b d d e C C b NP b C d d d b b e Gpi b e i(O.17) c (0.10) e (0.96) e (0.03) C C C e (0.83) e (0.90) f (0.04) f (0.97) Lgg e c (0.25) d (0.83) C b(O.lO) c (0.96) a (0.05) C a (0.33) c d (0.75) f (0.17) c (0.67) d (0.04) c (0.95) c (0.33) d (0.23) d (0.34)

Est a (0.17) C C C C C a C a(0.17) c c (0.83) c (0.83) Ada g (0.83) h h C a (0.07) f i (0.94) e C C 1 (0.17) b (0.07) k (0.03) c (0.86) 1 (0.03) La-l e b(0.13) b(0.17) f c (0.13) d (0.08) a (0.05) f c(O.17) f f (0.87) f (0.83) f (0.87) f (0.92) b (0.40) f (0.83) c (0.55) Ald a a a a a a a a a a

Acon C a a C C C C C C C 6-Pgd C c (0.87) c (0.33) b b b a (0.11) b a(0.17) b d(0.13) d (0.67) b (0.89) b (0.83)

Ck- 1 C C C C C Sod- 1 b b C b b Gda b b b b b Ck-2 b b b b b Ldh- 1 a a a a a

C. h. c. p. c. f. c. 1. C. c. Y. c. Y. A. b. LOCUS exilrpes C. p. pinus hesperophilusC. lawrencei tnstis SllhC~PIllPIS carduelis vespertim brooksi nevadensis

Mpi b b b C b b a b (0.50) d(0.75) f e (0.50) e (0.25) Gpd C C C C C C C b b c (0.50) e (0.50) Icd- 1 e c (0.08) b b a (0.20) c (0.75) f d d e e (0.92) c (0.80) e (0.25) Adh f f f f f f f f f d Glud C C C C C C C C a Pgm-1 b E(O.17) d d d d c (0.50) d d d d (0.83) d (0.50) Eap e e e Sdh C C C

Got-l b d e NP d (0.25) b b e (0.75)

Gpi C a (0,.08) C C C C e e e e c (0.92) GENETIC RELATIONSHIPS OF CARDUELINE FINCHES 413

TABLE 2. Continued.

C. h. c. p. c. f. c. f. C. c. v. c. Y. A. b. LOCUS exrlipes C. p. pma hesperophilus C. lawrencei lt-ISliS salicamans carduelk vesperlinus brooks; nevadensis

J4Z a (0.25) c a(0.13) c c C c (0.50) c C C c (0.75) c (0.74) g (0.50) d(0.13) Est C C C C C C C a a b

Ada C C d d d d C j j f La-l f f f f f f f C C f Ald a a a a a a a b b C Acon c C b C C C C d d C 6-Pgd b(0.75) b (0.84) b b b b b e d (0.25) g c (0.25) c (0.08) e (0.75) f (0.08) Ck-1 c C C C C C C b b a Sod-l b b b b b b b a(0.75) a d e (0.25) Gda b b b b b b b b b C Ck-2 b b b b b b b b b a Ldh-1 a a a a a a a a a b are summarized in Table 4. In general, average A few examples of shared alleles that were D increaseswith increasing taxonomic level. possibly derived (synapomorphies) were un- Thus, subspeciesare differentiated at D of covered;L. curvirostraand L. leucopterashared 0.0048, species of the same genus differ at Adh, b; Eap, a; and Got-l, c. Pgm-1, b, united 0.1739, speciesfrom different genera are dif- L. curvirostra,L. leucoptera,C. jlammea, C. ferentiated at 0.5209, and speciesfrom differ- hornemanni,and C. pinus.Sod-l, c, occurred ent families have average Nei’s D of 0.9239. only in the three species of Carpodacus.C. purpureusand C. cassiniiwere synapomorphic CLADISTIC ANALYSIS at Gda, a. All speciesof Carpodacus,Loxia, In addition to comparisonsofgenetic distances and Carduelisshared 6-Pgd, b. Otherwise, pat- between taxa, distancescomputed from allelic terns of allelic occurrence among taxa did not frequencies, it is of interest to examine the clearly agree with clusters defined by genetic occurrenceof allelesfrom a cladistic viewpoint distances. The latter are of course influenced (Hennig 1966, Avise et al. 1980a, Nelson and both by frequencies and occurrence of alleles. Platnick 198 1, Matson 1984). Of special con- Avise et al. (1980a, b), Patton and Avise (1983), cern are unique alleles(autapomorphies), those and Zink and Johnson (1984) have reported confined to particular taxa, and shared-derived similar results for other avian taxa. alleles (synapomorphies), those alleles that by Finally, the outgroup, Amphispizabelli ne- their pattern of occurrence define clusters of vadensis,was fixed at the following 13 alleles species. that were not represented in any of the car- We observed unique alleles in 10 of the 15 dueline finches: Mpi, c Adh, d; Glud, a; Eap, speciesof carduelines:L. a. littoralis(6 auta- c; Np, a; Est, b; Ald, c; 6-Pgd, g; Ck- 1, a; Sod- pomorphies, of which 5 [Adh, allele e; Got-l, 1, d; Gda, c; Ck-2, a; and Ldh-1, b. From a a; Gpi, b; Lgg, e; La-l, e] are fixed), P. enu- cladistic viewpoint, such alleles in the out- cleator(8, of which 6 [Adh, a; Eap, d; Sdh, d; group taxon could represent the primitive or Np, c; Ada, h; Acon, a] are fixed), C. purpureus pleisiomorphic character state for each of the (5) C. cassinii(5) C. mexicanus(5) L. cur- loci. One way to examine this possibility would virostru(l), L. leucopteru(l), C. pinus(2), C. be to survey additional outgroups. carduelis(4) and C. vespertinus(8, of which 6 [Icd-1, d; Sdh, b; Np, c Ada, j; Ald, b; Acon, d] are fixed). The scarcity or absenceof unique BRANCHING DIAGRAMS alleles in the seven speciesof Curduelisis no- Relationships among taxa, based on genetic table. Only C. pinusand C, carduelispossessed distance values, were analyzed with four autapomorphies; C. jlammea, C. hornemanni, branching methods: distance Wagner (Fig. l), C. psaltria,C. lawrencei,and C. tristislacked F-M tree (Fig. 2), WPGMA (Fig. 3), and unique alleles. The degree to which the occur- UPGMA. We considerthe four branching pro- rence of autapomorphies ‘is sample-size de- tocols to represent different hypotheseson the pendent deservesfurther study in these taxa. phylogenetic relationships of the taxa and equate congruenceamong them with relative robustnessof results. Thus, we interpret any difference in topology among the trees to imply ambiguity in the resolution of relationshipsand order of evolutionary descent of taxa, ambi- guity resulting from the nature of evolution of alleles at allozyme loci and/or different as- sumptions of the tree-constructingalgorithms. For discussionof the likelihood that these var- ious branching methodologiesreveal real phy- logenetic relationships, see the differing views of Felsenstein (1984) and Farris (198 1). The structure of the UPGMA dendrogram (not shown)was similar to that of the WPGMA tree, with two important differences. First, C. mexicanusdid not link with its congeners.In- stead, it formed a sister group with a cluster that included both forms of Loxia and all species of Carduelis.Second, Pinicola, Leu- costicte,and the stem of a large cluster com- prised of all speciesof Carpodacus,Loxia, and Carduelisformed an unresolved trichotomy. Other differences between the UPGMA and WPGMA dendrograms were trivial. Based on the foregoing premises, the branching diagrams consistently support the following results: (1) the named subspeciesof P. enucleator,C. purpureus,C. tristis,and C, vespertinuscluster together within their re- spective speciesand therefore are very closely related; (2) in the genus Carduelis,species within the subgeneraAcanthis and Astragali- nus group “properly” according to subgenus and the monotypic subgeneraSpinus (C. pi- nus)and Carduelis(C. carduelis)stand some- what apart; therefore all the subgeneramain- tain their integrity; (3) traditional genericlimits are supported within Carpodacus,Loxia, and Cardueliswith two exceptions, both involving the House Finch (in the distance Wagner tree, C. mexicanusclusters with C. vespertinusand in the UPGMA dendrogram, C. mexicanus clusterswith Loxia-Carduelis);(4) within Car- podacus,C. purpureusand C. cassinii are closely-related sister species,distinct from C. mexicanus;(5) within Loxia, L. ct.&rostraand L. leucopteraare moderately different geneti- cally; (6) in contrast, the two forms of , C. jlammea and C. hornemanniare similar genetically; (7) the closestally of the crossbills (Loxia) in all four analyses is the cluster of species that comprises Carduelis;(8) radia- tions of species in Loxia and Carduelisoc- curred over several million years, starting 5 MYBP (million years before present; see be- yond); (9) in contrast, genesisof speciesin Car- podacuswas prolonged, with C. mexicanus splitting from the lineage leading to its con- geners, C. purpureusand C. cassinii,at least several million years prior to the split of the GENETIC RELATIONSHIPS OF CARDUELINE FINCHES 415

TABLE 4. Mean genetic distances(Nei 1978) among samples from populations differentiated at several taxonomic levels.

Number of compari- Taxonomic level sons D k SE Range D + SE for other birds - Intraspecific (subspecies) 4 0.0048 * 0.0021 0.000-0.010 0.0048 f 0.0005 Interspecific congeners 33 0.1739 k 0.0206 0.028-0.527 0.0440 + 0.0026 Within Carpodacus 5 0.3424 f 0.0839 0.1494.527 Within Loxia 1 0.047 Within Carduelis 27 0.1465 f 0.0138 0.028-0.256 Same subgenusof Carduelis 6 0.0600 t 0.0066 0.028-0.073 Different subgeneraof Carduelis 21 0.1712 + 0.0132 0.051-0.256 Intergeneric confamilial 134 0.5209 k 0.0143 0.193-0.883 0.2136 f 0.0141 Interfamilial: 19 0.9239 k 0.0303 0.622-1.144 0.6829 + 0.0304 Amphispizabelli (Emberizidae) vs. all 19 taxa of Carduelines(Frin- gillidae)

a Data from Barrowclough (1980:661), who summarized values for Nei’s D from a variety of bxds (mostly emberizids and stumids) latter taxa at approximately 3 MYBP; (10) quence of their cladogenesisis tentative; (2) Coccothraustes, Pinicola, Leucosticte, and the phylogenetic relationships and age of Car- Carpodacusare older than either Loxia or Car- podacusare unclear, although the sum of the duelis; and (11) all six genera considered in results suggestsan intermediate position for this study are clearly defined by the new genetic this genus,between Leucosticteand Carduelis; data. (3) despite the alliance of the crossbills(Loxia) The following resultsare lesscertain. (1) Al- to the redpolls in the distanceWagner analysis, though the F-M tree and WPGMA analyses accordingto the structure of both the F-M tree suggestthat the three oldest genera appeared and the WPGMA dendrogram the relationship in the order Coccothraustes,Pinicola, and of crossbillsto particular speciesor subgenera Leucosticte,the rather deviant pattern shown of Carduelis is unresolved; and (4) none of the by the Wagner tree, the close placement of trees suggesta monophyletic relationship be- branching nodes in the WPGMA analysis,and tween any two of the three species of gold- the miniscule branch lengths separating these finches(C. psaltria, C. lawrencei, and C. tristis) taxa on the F-M tree indicate that this se- in the subgenusAstragalinus.

Wagner Tree Cophenetx Correlation=0.989 % Standard Deviation = 6.95

5.1 7 CorpodocusOcossinii

Pinicola enuclealor alascensis

Corduelis f flommeus C hornemonni exilipes

C psoltria hesperophilus

,3 C tristis salicamans C. corduelis _ Amphispiza belli 37 I I 1 I 1 I 00 0.1I 0.22 0.33 0.44

Distance from Root FIGURE 1. Distance Wagner tree rooted at the outgroup, Amphispiza belli. This analysis produced no negative branches. 416 JILL A. MARTEN AND NED K. JOHNSON

However, recent evaluation of the circum- stancessurrounding the dating of C. cooki has revealed an ambiguity that suggeststhat the conversion figure of 26.3 offered by Gutierrez et al. (1983) could be too large by a factor of two. For this clarification we are indebted to Carl Swisher, Department of Paleontology, University of California, Berkeley. Regarding the type specimen of C. cooki (Amer. Mus. Nat. Hist. No. 8301; ~011.by Harold Cook [HC No. 6471 in 1933 at Aphelops Draw, Sioux Co., Nebraska), Swisher wrote (undated letter received by NKJ on April 15, 1985): “Ac- cording to Morris Skinner [et al.] (1977) and Cook’s notes, the ‘specimen came from the Evening Grosbeak upper Sheep Creek beds, above the heavy ash layer, western exposures.’ Skinner states that C. cooki came from Aphelops Draw. The geo- logic sectionprovided by Skinner includesonly one prominent ash bed, the Sheep Creek Ash FIGURE 2. Branching diagram derived by the proce- (#3). This is the most prominent ash in the dure of Fitch and Margoliash (1967). Branch lengths are Sheep Creek and is with little doubt the ash in units of Rogers’ D (x 100). The tree is rooted (seeFarris Cook referred to. Unfortunately, Skinner points 1972) at Amphispiza belli. Of 5 F-M trees examined, the one illustrated best summarized the original matrix based out that all Sheep Creek localities are below on the fewest (2) negative branchesand lowest percentage this ash bed. Fossil localities above this ash standard deviation. are present in Aphelops Draw and very fos- siliferous, but are much younger. Skinner says DISCUSSION the fossil is from the Sheep Creek Fm which would be below the ash. This gets more com- TIMING OF CLADOGENETIC EVENTS plicated . . . the ash has been dated by K-Ar Nei (1975), Sarich (1977), ‘Yang and Patton methods at 15.1 MY. If type came from below (198 1) and Gutierrez et al. (1983), among oth- the ash, I feel it is fair to say it is 15 to 16 MY, ers, proposed calibrations applied to Nei’s D but could be as old as 17 MY (L. Early Miocene values among existing speciesin attempting to to Early-Middle Miocene). If Cook was right date phyletic divergence. The calibration of and it came from above the ash bed, then it Gutierrez et al. (1983) basedon galliform taxa, would be as young as 9 to 7 MY (or Late Mio- is the only such estimate available so far spe- cene).” cifically for birds. They calculatedthat one unit Because the collector of the fossil stated of Nei’s (1978) D accumulated over approxi- clearly that it was found “above the heavy ash mately 26.3 MY, a figure arrived at by dividing layer,” we believe that there is sufficient prob- the age of a fossil quail, C. cooki (Wetmore ability that it came from the younger upper 1934), which is presumed to be mid-Miocene stratum, approximately 8 MY in age. But be- (16 million years before the present [MYBP]) cause of the uncertainties alluded to above, by the average Nei’s D (= 0.609) between C. and in view of the stratigraphicdata of Skinner montezumaeand its Odontophorine sistertaxa. et al. (1977), we propose that a compromise Thus, t = 26.3 x 106D, where t is the time figure of 12 MYBP would be the most appro- since divergence and D is Nei’s (1978) genetic priate estimated age of C. cooki. The formula distance. Such calibrations assume the oper- for the calibration would then read: t = 19.7 x ation of a molecular clock (Wilson et al. 1977, 106D. (If an age of 8 MYBP is applied to C. Thorpe 1982) whereby allelic differences cooki, then the formula would read: t = 13.1 x among populations accruerandomly in a more 106D. or less steady, time-dependent manner. The The scaleof Figure 1 is based upon the con- report of Barrowclough et al. (1985), that pat- version factor of 19.7. In interpreting this fig- terns of genetic divergence in a variety of birds ure it should be kept in mind that becauselarge agree with the predictions of Kimura’s (1979, standard errors accompany genetic distance 1982) neutral, mutation-drift model, supports values and because of uncertainty regarding both the notion of a clock and the attempts to the calibration, the following dates proposed derive calibrations based on the magnitude of for the splitting of lineages can be only very Nei’s D. gross approximations. We feel that the se- GENETIC RELATIONSHIPS OF CARDUELINE FINCHES 417

Millions of years I8 14.3 10.6 8.4 5.4 4 2 I I 1 I I 1 1

Corpodocus cossinii c p. purpureus WPGMA 6: p. californicus Cophenetlc Correlation = 0.961 C mexicanus frontolis L oxia curvirostra grinnelli L I. leucoptero Carduelis f flammeus C bornemanni exilipes Cp pinus C. psaltrio besperophilus C lawrencei C t tristis C t salicamans C carduelis Leucosticte a. littoralis Pinicola enucleator alascensis P e leucura Coccothraustes Y vesper//flus C v brooksi Amphispiza belli

0.80 0.60 0.40 Distance FIGURE 3. Phenogram based on Rogers’ D-values and derived by the WPGMA method. The high cophenetic correlation coefficient (Y,, = 0.96 1) indicates excellent agreement between the distancesshown in the phenogram and the original data matrix. The dating scaleis based on a modification of the calibration offered by GutiCrrezet al. (1983); seetext. quenceof appearanceof the various taxa, how- LEVELS OF GENETIC VARIATION ever, is more trustworthy. A major surprise of the study was the com- Using the compromise conversion factor of paratively great genetic divergence of C. mex- 19.7, the split between the emberizids and the icanus from its phenetically very similar con- lineage leading to the modern carduelinescon- geners, C. purpureus and C. cassinii. Wide sidered here occurred at 18 MYBP, the diver- genetic separation of specieswithin one genus gence of C. vespertinustook place at 14.3 is not without precedent, however. In the MYBP, the separation of the ancestorsof Pin- woodpecker genus Sphyrapicus, for example, icola from the rest of its sister taxa happened the phenetically similar S. nuchalisand S. var- at 10.6 MYBP, the predecessorsof Leucosticte ius are separated by a substantial genetic dis- split at 10.5 MYBP, the lineageleading to Car- tance (Johnson and Zink 1983). Another ex- podacus divided from that leading to Loxia- ample occurs in Vireo, in which l? jlavoviridis Carduelis at 8.4 MYBP, and the ancestorsof is apparently fixed at alleles that differ from Loxia diverged from those of Carduelis at 5.4 the predominant allele found in its extremely MYBP. Except for the separationof the lineage similar allopatric relative, V. olivaceusat 6 loci leading to C. mexicanus, which split from the (Johnson and Zink 1983). The opposite situ- clade leading to C. cassinii and C. purpureus ation occursin the goldfinchesof the subgenus at 9.3 MYBP, speciation events within genera Astragalinus, in which the three species ex- occurredgenerally over the period of time from amined, although phenetically distinctive, are the late Pliocene to mid-Pleistocene. For ex- genetically very similar. Furthermore, both in ample, the ancestors of C. carduelis divided certain sapsuckers(S. nuchalis and S. ruber, from those leading to its congeners at 4.2 Johnsonand Zink 198 5) and in warblers (Den- MYBP. The two forms of redpolls, in contrast, droica coronatacomplex, Barrowclough 1980), seem to have split 550,000 years ago. The ap- forms with obviously different plumages also proximate time of divergence of subspeciesis show miniscule genetic differentiation. Clear- illustrated by three comparisons: within C. ly, the relationship between phenotypic diver- purpureus, 78,000 years; Pinicola enucleator, genceand protein divergenceat the near-species 98,500 years;and C. vespertinus,197,000 years, level deservesmuch further study in birds. At all of which would represent late Pleistocene present, too few studies have been conducted events. to support generalization. Our finding of low 418 JILL A. MARTEN AND NED K. JOHNSON

levels of heterozygosity and scarcity or lack of prove to be conspecific,it will still be desirable unique allelesin certain subgeneraof Carduelis to compare the enzyme genes of these forms and in Loxia is also noteworthy. Although low with those of the morphologically different values of H in L. curvirostraare almost cer- Hornemann’s Redpoll (C. hornemannz],which tainly related to sample size, such is probably may be specificallydistinct from theftammea- not the complete explanation for the reduced exilipescomplex (Todd 1963, AOU 1983). genetic variability found in C. jlammea, C. Although we assume that the New World psaltria,C. lawrencei,or C. tristis.Perhaps his- goldfinches, C. tristis,C. psaltria,and C. law- torical patterns of gene flow, drift, and/or fluc- renceiare all near relatives, as is evident from tuations in effective population size have been their placement in the subgenusAstragalinus important in the maintenance of low levels of (AOU 1983), the molecular data do not iden- genetic variation in these species. Such sto- tify any pair of the three speciesas closestrel- chastic phenomena have been invoked in ex- atives. Indeed, conclusions on this point are planations of relative geneticvariability in oth- unwarrantedbecause several Central and South er avian examples (Barrowclough et al. 1985). American speciesof Carduelis,one or more of which could be the nearest relatives of these EVOLUTIONARY RELATIONSHIPS OF species(Mayr and Short 1970) were unrepre- CONGENERIC SPECIES sented in our comparison group. In most instancesthe electrophoreticdata agree with relationships of congeners proposed on EVOLUTIONARY RELATIONSHIPS OF GENERA the basis of conventional taxonomic practice. The most explicit phylogeny of generaof North For example, the protein evidence clearly sup- American carduelines was offered by Raikow ports Mayr and Short’s (1970:79) view that (1978,1985).Othertaxonomicworksinrecent the “Cassin’s Finch is closely related to C. pur- years(Howell et al. 1968, Mayr and Short 1970, pureusand can be considered a sibling species AOU 198 3) have simply listed species,leaving the reader to interpret evolutionary affinities with it . . . .” However, under the account of from the sequencein which taxa were treated. the House Finch they comment, in apparent Raikow’s phylogeny was basedon the precepts conflict with the aforementioned statement, of cladistics(Hennig 1966) applied to an anal- that “Conceivably mexicanusand cassiniirep- ysis of morphologic characters, mostly those resent an older invasion of Carpodacusfrom of appendicular myology. He recognized two Eurasia, and purpureusa more recent entrant major clusters of cardueline finches. The first into North America.” Perhaps the possible cluster included Leucosticteas the oldest genus close relationship of mexicanusand cassinii, and as a sistertaxon to a clade formed of Frin- implied by their proposed association in the gilla, Pinicola,Carpodacus, and same early invasion, was unintentional. In any (=Coccothraustesof the present study). The event, the genie information does support both latter two genera were aligned as sister taxa; the notion of an older entry from the Old World they lack M. plantaris of the hindlimb, the loss of the lineage leading to mexicanusbut not of which is considered to be a derived char- including cassinii,and the close relationship acter state. Raikow’s second major cluster in- of C. purpureusand C. cassinii.Presumably, cluded four genera, , Chloris,Loxia, the lineage leading to the latter two species and Carduelis.These genera sharethe presum- either arrived in the New World from Eurasia ably derived features of the presenceof a tibia1 comparatively recently, as suggestedby Mayr head on M. peroneusbrevis (a trait also shared and Short (1970) or was derived in the New with the Hawaiian honeycreepers)and loss of World from the older mexicanuslineage after a patellar band. Pyrrhulaand Chlorishave also its arrival and subsequentestablishment. lost M. obturatorius dorsalis;Loxia and Car- Several speciesof Carduelishave been iden- duelisboth have lost M. plantaris. Lossof either tified as probable close relatives. For example, muscleis presumably a derived character state. the frequent hybridization and similar mor- The latter two pairs of genera are thus sister phologic features of C. flammea’ and C. horne- groups. manniexilipes point to their possibleconspec- In major featuresthe phylogenyderived from ificity (Mayr and Short 1970, Troy and Brush the protein data agreeswith that based on the 1983, Troy 1985) a statuswith which the pro- morphologic information. Both phylogenies tein evidence would not conflict. However, low divide cardueline genera into an older group genetic distancesalone are unrelated to species comprised of Coccothraustes,Pinicola, Leu- status;perfectly good biologic speciescan have costicte,and Carpodacus,and a more recent genetic distancesapproaching zero (Yang and clade consisting of Carduelisand Loxia. The Patton 198 1, Johnson and Zink 1983). Even latter relationship is also supportedby the dis- if C. jlammea and C. hornemanniexilipes covery of a hybrid between the GENETIC RELATIONSHIPS OF CARDUELINE FINCHES 419

and the Pine Siskin (Tallman and Zusi 1984). D. Gibson, Jeffrey G. Groth, Fern B. Hall, Richard E. The ability to hybridize provides strong evi- Johnson, Robert E. Jones, Brina Kessel, Josef H. Lind- holm III, Oliver P. Pearson,Robert M. Zink and the staff dence for considerable genetic compatability of the Alexander Lindsay Junior Museum, Walnut Creek, between the taxa involved. California. Monica M. Frelow gave expert advice on data Within the older cluster of genera, however, analysis, especially computer techniques. Peter Marler the two proposed phylogenies are discordant sharedwith us his extensiveknowledge of CarduelineFinch with regard to the sequence of genera. Al- behavior. Carl Swishercalled to our attention the literature relevant to the age assignedto the fossil quail, C. cooki though all are apparently old, distinctive lin- Wetmore, and carefully interureted it for us. Gene M. eages, there is no evidence from the electro- Christman prepared the final versions of the figures. Ms. phoretic results that Leucosticteis the oldest Bonnie Resnick typed several versions of the manuscript. genus. Instead, based on its greater accumu- We are indebted to G. F. Barrowclough, R. J. Raikow, and R. M. Zink, who read a draft of the manuscript and lation of genie differences compared with the offeredimportant suggestionsfor improvement. Some pre- other taxa, Coccothruustesis probably the old- liminary analysesrelevant to this researchwere conducted est genus, followed by Pinicola, Leucosticte, with supportfrom the National ScienceFoundation (Grants and Carpodacus.Another disagreementof the DEB-79-20694 and DEB-79-09807 to N. K. Johnson). protein and morphologic results involves the Expensesfor electrophoreticanalyses were borne by a grant from the Annie M. Alexander Fund of the Museum of linkage of Coccothrausteswith Carpodacusas Vertebrate Zoology. Field expenseswere covered in part sister taxa (Raikow 1978). The genetic data by a grant from the Committee on Research, University maintain Coccothruustesas a well-separated of California, Berkeley. lineage in all but one analysis, the distance Wagner tree. There, a single species of Car- LITERATURE CITED podacus,the House Finch, joins with the Eve- AMERICANORNITHOLOGISTS UNION.’ 1983. Check-list of ning Grosbeak. However, because the other North American birds, 6th ed. American Omithol- two speciesof Carpodacuswere excluded from ogists’ Union, Washington, DC. this clade, and in view of the very short branch AVISE,J. C., J. C. PATTON,AND C. F. AQUADRO. 1980a. Evolutionarygenetics of birds. I. Relationshipsamong lengths separating all of the older genera, we North American thrushesand allies.Auk 97:135-147. consider relationships at that level to be es- AVISE,J. C., J. C. PATTON,AND C. F. AQUADRO. 1980b. sentially unresolved by the Wagner procedure. Evolutionary geneticsof birds. II. Conservative pro- Despite these differences, which we judge to tein evolution in North American sparrowsand rel- atives. Syst. Zool. 29~323-334. be minor, the phylogeniesdeveloped indepen- BARROWCLOUGH,G. F. 1980. Genetic and phenotypic dently from electrophoretic and morphologic differentiation in a wood warbler (Genus Dendroica) information are in basic agreement. The lack hybrid zone. Auk 97:655-668. of identity in sequenceof older genera is not BARROWCLOUGH.G. F. 1983. Biochemical studies of surprisingand should not detract from the fun- microevolutionary processes,p. 223-261. In A. H. Brush and G. A. Clark, Jr. [eds.], Perspectivesin or- damental compatibility of our genetic findings nithology. Cambridge Univ. Press, England. with the earlier myologic results on the car- BARROWCLOUGH,G. F., K. W. CORBIN,AND R. M. ZINK. duelines. The occurrence of incomplete con- 198 1. Genetic differentiation in the procellari- gruence of molecular and morphologic data iformes. Comn. Biochem. Phvsiol. 69:629-632. BARROWCLOUGH,G. F., N. K. JOHNSON,AND R. M. ZINK. sets is becoming increasingly recognized as a 1985. On the nature of genie variation in birds, p. commonplace and thus far rather unyielding 135-154. In R. F. Johnston [ed.], Current omithol- problem in comparative phylogenetics(Lewin ogy. Vol. 2. Plenum Press, New York. 1985). BOCK,%‘. J. 1960. The palatine processof the premaxilla Because this study was restricted to North in the Passeres.Bull. Mus. Comn. Zool. 122:36l-488. CORBIN,K. W. 1983. Genetic structure and avian sys- American taxa (plus C. carduelis), our ability tematics, p. 2 1 l-244. In R. F. Johnston[ed.], Current to interpret phylogeneticrelationships has been ornithology. Vol. 1. Plenum Press, New York. somewhat compromised in view of the pos- FARRIS,J. S. 1972. Estimating phylogenetic trees from sibility that some of the taxa are closely related distance matrices. Am. Nat. 106:645-668. FARRIS,J. S. 1981. Distance data in phylogenetic anal- to forms occurringin other geographicregions. ysis, p. l-23. In V. A. Funk and D. R. Brooks [eds.], Therefore, the larger question of genetic rela- Advancesin cladistics.Vol. 1. ProceedingsWilli Hen- tionships of cardueline taxa worldwide awaits nig Society. New York Botanical Garden, New York. the assembly of tissue of the remaining taxa FELSENSTEIN,J. 1983. Numerical methods for inferring for biochemical study. Many of these forms evolutionary trees. Q. Rev. Biol. 57~379-404. FELSENSTEIN,J. 1984. Distance methods for inferring can be obtained only with great difficulty, if at phylogenies:a justification. Evolution 38: 16-24. all. Hopefully, someone with accessto speci- FELSENSTEIN,J. 1985. Confidencelimits on phylogenies: mens in the Old World will accept this chal- an approach using the bootstrap. Evolution 39:783- lenge. 791. FITCH,W. M., ANDE. MARGOLIASH.1967. Construction ACKNOWLEDGMENTS of phylogenetictrees. Science 155:279-284. GUTIBRREZ,R. J., R. M. ZINK, AND S. Y. YANG. 1983. We are very grateful to the following individuals for ob- Genie variation, systematic,and biogeographicrela- taining specimen material: FrancescaJ. Cuthbert, Daniel tionships of some galliform birds. Auk 100:33-47. 420 JILL A. MARTEN AND NED K. JOHNSON

HENNIG, W. 1966. Phylogenetic systematics.Univ. Il- SARICH,V. M. 1977. Rates, sample sizes, and the neu- linois Press, Chicago, IL. trality hypothesisfor electrophoresisin evolutionary HOWELL,T. R., R. A. PAYNTER,JR., AND A. L. RAND. studies.Nature 265:24-28. 1968. Subfamily Carduelinae, Serins, Goldfinches, SELANDER,R. K., M. H. SMITH,S. Y. YANG,W. E. JOHNSON, Linnets, Rose Finches, ,and Allies, p. 207- AND J. B. GENTRY. 1971. Biochemical polymor- 306. In R. A. Paynter, Jr. [ed.], Check-list of birds of phism and systematicsin the genus Peromyscus.I. the world; Vol. XIV. Museum of Comparative Zo- Variation in the old-field mouse (Peromyscuspolio- ology, Cambridge, MA. notus).Univ. Texas Publ. 7 103:49-90. JOHNSON,N. K., AND R. M. ZINK. 1983. Speciation in SKINNER, M. F., S. M. SKINNER, AND R. J. GOORIS. 1977. sausuckers(Sohvvaoicus): I. Genetic differentiation. Stratigraphy and biostratigraphy of L. Cenozoic de- A;k 100:8711884.- ’ positsin C. Sioux Co., Western Nebraska.Bull. Amer. JOHNSON.N. K.. AND R. M. ZINK. 1985. Genetic evi- Mus. Nat. Hist. 158:265-371. dence for relationships among the red-eyed, yellow- SNEATH,P.H.A., AND R. R. SOKAL. 1973. Numerical green, and Chivi vireos. Wilson Bull. 97:421-435. . W. H. Freeman and Co., San Francisco. JOHNSON,N. K., R. M. ZINK, G. F. BARROWCLOUGH,AND SWOFFORD,D. L. 1981. On the utility of the distance J. A. MARTEN. 1984. Suggestedtechniques for mod- Wagner procedure, p. 25-43. In V. A. Funk and D. ern avian systematics.Wilson Bull. 96543-560. R. Brooks [eds.],Advances in cladistics.Vol. 1. Proc. KIMURA,M. 1979. The neutral theory of molecular evo- Willi Hennig Society, New York Botanical Garden, lution. Sci. Am. 241:98-126. New York. KIMURA, M. 1982. The neutral theory as a basis for SWOFFORD,D. L., AND R. B. SELANDER.198 1. A com- understandingthe mechanism of evolution and vari- puter program for the analysis of allelic variation in ation at the molecular level, p. 3-56. In M. Kimura genetics.j. Hered. 72:281-283. [ed.], Molecular evolution, protein polymorphism and TALLMAN.D. A.. AND R. L. ZUSI. 1984. A hvbrid Red the neutral theory. Japan Scientific Societies Press, Crossbill-Pine Siskin (Loxiu curvirostra x ,&rduelis Tokyo, Japan. pinus) and speculationson the evolution of Loxia. LEWIN,R. 1985. Molecules vs. morphology:of mice and Auk 101:155-158. men. Science 229~743-745. THORPE,J. P. 1982. The molecular clock hypothesis: MATSON,R. H. 1984. Applications of electrophoretic biochemical evolution, genetic differentiation and data in avian systemat&. Auk 101:7 17-729: systematics.Annu. Rev. Ecol. Syst. 13:139-l 68. MAYR. E.. ANDL. L. SHORT. 1970. Sueciestaxa of North TODD, W.E.C. 1963. Birds of the Labrador Peninsula. A&e&can birds. A contribution io comparative sys- Univ. Toronto Press, Canada. tematics. Publ. Nuttall Omithol. Club 9:1-127. TORDOFF,H. B. 1954a. A systematicstudy of the avian NEI, M. 1975. Molecular population geneticsand evo- family Fringillidae based on the structureof the skull. lution. North Holland Publ. Co., Amsterdam, Neth- Misc. Publ. Mus. Zool. Univ. Mich. 81:1-42. erlands. TORDOFF,H. B. 1954b. Relationshipsin the New World NEI, M. 1978. Estimation of averageheterozygosity and nine-primaried oscines.Auk 7 1~273-284. genetic distance from a small number of individuals. TROY, D.-M. 1985. A phenetic analysis of the redpolls Genetics 89:583-590. Carduelisflammeajlammea and C. hornemanni ex- NELSON,G., AND N. PLATNICK. 1981. Systematicsand ilipes. Auk 102~82-96. biogeography: cladistics and vicariance. Columbia TROY, D. M., AND A. H. BRUSH. 1983. Pigments and Univ. Press, New York. feather structure of the redpolls, Carduek jlammea PATTON,J. C., AND J. C. AVISE. 1983. An empirical and C. hornemanni. Condor 85:443-446. evaluation of qualitative Hennigian analysesof pro- WETMORE,A. 1934. A fossil quail from Nebraska. Con- tein electrophoreticdata. J. Mol. Evol. 19:244-254. dor 36:30. RAIKOW,R. J. 1978. Appendicular myology and rela- WILSON,A. C., S. S. CARLSON,AND T. J. WHITE. 1977. tionships of the New World nine-primaried oscines Biochemicalevolution. Annu. Rev. Biochem. 46:573- (Aves: Passeriformes).Bull. CarnegieMus. Nat. Hist. 639. 7:1-43. YANG, S. Y., AND J. L. PATTON. 1981. Genie variability RAIKOW,R. J. 1985. Problems in avian classification,p. and differentiation in the Galapagosfinches. Auk 98: 187-212. In R. F. Johnston [ed.], Current omithol- 230-242. ogy. Vol. 2. Plenum Press, New York. ZINK, R. M., AND N. K. JOHNSON.1984. Evolutionary ROGERS,J. S. 1972. Measures of genetic similarity and geneticsof flycatchers.I. Sibling speciesin the genera genetic distance. Univ. Texas Stud. Genet. 7:145- Empidonax and Contopus.Syst. Zool. 33:205-216. 153.