The Condor 91:319-396 Q The Cooper Ornithological Society 1989
GENETIC EVIDENCE FOR THE ORIGIN AND RELATIONSHIPS OF HAWAIIAN HONEYCREEPERS (AVES: FRINGILLIDAE)’
NED K. JOHNSON Museum of VertebrateZoology and Department of Zoology, Universityof Caltfornia, Berkeley, CA 94720
JILL A. MARTEN Department of Pathology,Pacific PresbyterianMedical Center, Clay at BuchananStreet, San Francisco,CA 94120
C. JOHN RALPH U.S.D.A. Forest Service,Redwood Sciences Laboratory, 1700 Bayview Drive, Arcata, CA 95521
Dedication. It is a pleasure to dedicate this paper to Dr. Dean Amadon, pioneering investigator of Hawaiian honeycreepers,on the occasionof his 76th birthday. Abstract. Using starchgel electrophoresisof proteins,we examined variation at 36 genetic loci in nine species(eight genera) of Hawaiian honeycreepers(Class Aves; Family Fringil- lidae; Subfamily Drepanidinae). Two speciesof carduelinefinches and two emberizids served as outgrouptaxa. Twenty-three loci (64%) were either polymorphic within taxa and/or were fixed at alternative alleles among taxa. In seven of nine species,low levels of mean Hobs (0.015), percentageof polymorphic loci (4.16) and averagenumber of alleles per polymor- phic locus (2.03) may reflect population bottlenecks that occurred either during or after initial colonization. Phenograms,distance Wagner trees, F-M trees, and a cladistic analysis provided hypothesesfor the evolutionary relationships of taxa and suggestthat: (1) The drepanidinesare monophyletic. (2) The Hawaiian honeycreepersare more similar genetically to the two speciesof emberizids than to the two speciesof carduelines,a result that conflicts with a recent consensusof opinion based on morphologic and other biochemical data. (3) The speciesancestral to modem drepanidines colonized the Hawaiian Archipelago at an estimated 7-8 million years before present (MYBP). This date agreesgenerally with the timing of emergence of Nihoa, now largely submerged, but antedates the appearance of Kauai (5 MYBP), the oldest of the present “high” Hawaiian Islands. (4) The creepers Oreomystisand Paroreomyza represent the oldest and most divergent lineage of living drepanidines. (5) The youngestlineages are representedby the nectar feeders (Himatione and Vestiaria),the thick-billed “finch types” (Loxioidesand Telespiza),and a diverse array of other forms (Loxops and Hemignathus). (6) Hemignathus “virens” stejnegeriis a full species,possibly allied to Loxops coccineus.Our genetic data conflict with the two major phylogenetichypotheses that have been proposed for the radiation of the drepanidines:(1) origin from tubular-tongued,nectar-feeding ancestors; and (2) origin from thick-billed and thick-tongued,seed- and fruit-eating ancestors.Instead, the evidence suggeststhat the earliest Hawaiian honeycreepershad generalizedbills, tongues,and diets. This ancestralgroup gave rise to the lineagesthat eventually led to both (1) modern Paroreomyzaand Oreomystisand (2) a complex group of (a) nectar feeders (Himatione, Vestiaria,and relatives); (b) seedand fruit eaters (Loxioides, Telespiza,and relatives); and (c) a diverse group of speciesthat feed on both arthropods and nectar (Loxops,Hemignathus, and relatives). We speculatethat the most immediate ancestorof all of the heavy-billed specieswas a thin-billed, tubular-tongued, nectarivorous form and that this major morphologic shift was expedited by the alteration of developmental patterns and rates. Key words: Hawaiian honeycreepers;Drepanidinae; allozymes; insular colonizationphy- logeneticinference.
INTRODUCTION Hawaiian Archipelago is unsurpassed. Among As a showcase for the products of evolutionary the endemic birds, the honeycreepers are espe- diversification following colonization, the cially renowned. The profound divergence in bill shape and ecology shown by the 28 living or very ’ ’ Received 12 August 1988. Final acceptance12 De- recently extinct species has long served as a text- cember 1988. book example of adaptive radiation. In recent
13791 380 N. K. JOHNSON, J. A. MARTEN AND C. J. RALPH
decadesa stream of publications has dealt with heart tissue followed Johnson et al. (1984). Tis- the distribution, morphology, ecology, system- sue homogenates(a combination of liver, heart, atics, and evolution of these unique birds (Ama- and an equal volume of de-ionized water) were don 1947,1950,1986; Baldwin 1953; Bock 1970; centrifuged at 4°C and 15,000 rpm for 40 min. Sibley 1970; Richards and Bock 1973; Raikow The aqueous protein extracts were then stored 1974, 1976, 1977a, 1977b, 1978; Zusi 1978; at - 76°C for later electrophoretic analysis. Thir- Berger 1981; Olson and James 1982a, 1982b, ty-six presumptive genetic loci were examined 1988; Sibley and Ahlquist 1982; James et al. by horizontal starch gel electrophoresis using 1987; Pratt et al. 1987; Bledsoe 1988a). Most of standard procedures(Selander et al. 197 1, Yang the foregoingauthors agreethat the drepanidines and Patton 198 1). Protein assayswere prepared are monophyletic, having evolved from a single according to Harris and Hopkinson (1976) and ancestral speciesthat colonized the Hawaiian Is- Selander et al. (197 1). Our specificbuffer system- lands from either Asia or the Americas. How- assay combinations are available upon request. ever, current literature reveals little agreement Alleles (=electromorphs) at each locus were des- on relationships of speciesor on generic limits ignated alphabetically in decreasingorder of mo- within the group. bility. For multiple isozymes of proteins the most Using starchgel electrophoresisof proteins, we anodal locus was identified as 1; more cathodal compared 15 populations of nine speciesof drep- loci were indicated by progressivelyhigher num- anidines. Although 18 living speciesare known, bers. Hemoglobin was the only cathodal protein six of the nine forms omitted are rare and en- detected on LiOH gels. From banding patterns dangered and, hence, difficult to obtain. Se- on gels (presumptive individual genotypes), we quenceand nomenclature of taxa follow the AOU derived a table of allelic frequencies (Table 2). (1983). Our use of this standard reference does Heterozygosity levels were determined by di- not imply agreementwith either the classification rect count. Allelic frequencies were converted to or nomenclature therein because as Olson and genetic distances(Table 3) using the methods of James (1988) have argued recently the AOU no- Rogers (1972) and Nei (1978). To compare pat- menclature for both the Kauai population of the terns of population and/or species similarity Common Amakihi (Hemignathus “virens” or relatedness, phenograms (UPGMA and stejnegeri)and the Kauai Akialoa (Hemignathus WPGMA, Sneath and Sokal 1973) and phylo- procerus)is incorrect. Olson and James’ propos- genetic trees (F-M trees, Fitch and Margoliash als are currently under study by the Committee 1967; distance Wagner trees, Farris 1972) were on Classification and Nomenclature of the AOU. constructed from Rogers’ genetic distance val- From the perspective offered by the genetic ues. data set, we addressthe following issues:(1) the Using the program PAUP (Swofford 1985) we origin of the Drepanidinae; (2) the timing of col- also conducted a cladistic analysis in which loci onization of the Hawaiian Archipelago; (3) the were charactersand alleles at a given locus were evidence for population bottlenecks and founder character states. Becausea cladistic analysis us- effects;(4) the phylogenetic relationships of mod- ing alleles as charactersmay yield intermediate ern species;and (5) the taxonomic implications taxa with no alleles (Buth 1984) and becausedata of the genetic results. reduction by “presence/absence” coding is un- desirable for other reasons (Swofford and Ber- MATERIALS AND METHODS lecher 1987) we did not attempt this method. Sixty-two specimens of honeycreepers repre- For the cladistic analysis by locus, alleles were senting eight genera, nine species,and 15 pop- lettered consecutively among taxa. When poly- ulations were collected in the spring of 1977 and morphic at a locus, a taxon was assignedthe state 1978 in the Hawaiian Islands. Two species of for its commonest allele. If two alleles were Emberizidae, the Western Tanager (Piranga lu- equally common (e.g., ICD-2, H. s. sanguinea doviciana) and Saffron Finch (Sicalis flaveola) [Maui]), the first allele was chosen. Becausethe and two speciesof carduelines, the Purple Finch direction of character state transformation was (Carpodacuspurpureus)and Red Crossbill (LOX- unknown, character states(alleles at each locus) ia curvirostra), were selected as outgroup taxa. were not ordered on input. The addition se- Taxa studied, sample sizes, and geographic quence, CLOSEST, the branch swapping option, sourcesof specimens are listed in Table 1. Pro- ALT, the rooting procedure, OUTGROUP, and ceduresfor the collection and storageof liver and the method of detecting all equally parsimonious GENETICS OF HAWAIIAN HONEYCREEPERS 381
TABLE 1. Taxa studied,asample sizes, sourcesof specimens,and intraspecificgenetic variation.
Average Percent no. alleles POlY- Per POlY- morphic TZWl n Source of specimens Ho,, + SE loci “;“Z
Family Emberizidae Subfamily Thraupinae Western Tanager (Piranga ludoviciana) 1 California 0.083 8.3 2.00 Subfamily Emberizinae Saffron Finch (Sicalisflaveola pelzelm] 1 Paraguay 0.028 2.8 2.00 Family Fringillidae Subfamily Carduelinae Purple Finch (Carpodacuspurpureus cal$ornicus) 1 California 0.083 8.3 2.00 Red Crossbill (Loxia curvirostragrinnellr] 1 California 0.0 0.0 1.00 Subfamily Drepanidinae-Tribe Psittirostrini Laysan Finch pily/es@z cantanscantans) 4 Laysan Island 0.007 k 0.008 2.8 2.00
(Loxioidesbailleul] 1 Hawaii 0.0 0.0 - Tribe Hemignathini Common Amakihi (Hemignathusvirens virens) 12 Hawaii 0.039 k 0.009 22.0 2.25 (H. v. wilsom] 4 Maui 0.048 f 0.020 11.1 2.25 (H. v. chloris) 3 Oahu 0.046 k 0.030 8.3 2.30 (H. v. stejnegerz] 8 Kauai 0.059 t- 0.013 16.7 2.17 Anianiau (Hemignathusparvus) 6 Kauai 0.014 * 0.010 8.3 2.00 Kauai Creeper (Oreomystisbairdc] 1 Kauai 0.028 2.8 2.00 Maui Creeper (Paroreomyza montana) 4 Maui 0.049 * 0.015 11.1 2.25 Akepa (Loxops coccineuscaeruleirostris) 1 Kauai 0.0 0.0 - Tribe Drepanidini Iiwi (Vestiaria coccinea) 3 Hawaii 0.019 k 0.011 2.8 2.00 5 Kauai 0.22 -c 0.012 5.5 2.00 Apapane (Himatione sanguineasanguinea) 6 Hawaii 0.005 * 0.005 5.5 2.00 1 Maui 0.028 2.8 2.00 3 Kauai 0.028 i- 0.019 11.1 2.25 Total and meansb 62 0.026 7.4 2.11
a Nomenclature and sequence of species follow AOU (1983). h Drepanidines only: four outgroup species excluded. These values were unweigbted by sample size. trees, MULPARS, were specified. Equally par- mentation, was computed with the modifications simonious trees were input into CONTREE to of Wright (1978) for small sample size and of produce a strict consensus tree. Nei (1975) for multiple alleles. The FITCH option of the computer program PHYLIP (version 2.8, by J. Felsenstein) was used RESULTS to determine the lowest least-squares network for VARIATION AT LOCI AND HETEROZYGOSITY several proposed phylogenies. For species rep- resented by two or more sampled populations, Of the 36 loci scored, 23 (64%) were variable in Wright’s (1965) F,,, a measure of gene pool frag- that they showed either one or more heterozy-
384 N. K. JOHNSON, J. A. MARTEN AND C. J. RALPH
13
D
D D a
D P -0
D D 0‘
D D -0
D D v
D D v
9 D v
D P
386 N. K. JOHNSON, J. A. MARTEN AND C. J. RALPH
TABLE 4. Mean genetic distances(Nei 1978) among samplesfrom populations differentiated at several taxo- nomic levels.
Number of Level of comparison COlIlpUiSOllS lj + SE Range
Local populations* 4 0.0015 k 0.0006 0.0-0.003 Intraspecific (subspecieQb 6 0.0443 k 0.0123 0.002-0.087 (excludingH. v. stejnegeri) 3 0.0203 k 0.0093 0.0024.032 Interspecific congener@ 4 0.0540 k 0.0106 0.032-0.080 Intergeneric confamilial 91 0.1683 k 0.0089 0.05 l-O.377 Intratribal 29 0.1543 k 0.0173 0.032-0.297 Intertribal 66 0.1675 k 0.0104 0.053-0.377 Intersubfamiliap 30 0.4669 k 0.0105 0.350-0.576 Interfamilialf 30 0.3862 +- 0.0083 0.296-0.456
= Vestrana coccinea and Hmatione sanguinea only. b Hemignathw virens only. LH. “Y.” stenegeri’ is a probable species. * Hemignat x us vutm vs. H. panus only. cDrepanidines vs. carduelines. i Drepanidines vs. embaizids (Piranga ludovrciana and Sicalisflaveola). gotes(15 loci) or were fixed at alternative alleles milial. In general, Nei’s D increases at succes- among species,including outgroup taxa (8). The sively higher taxonomic levels of the AOU (1983) 13 monomorphic loci were: EST-l, GDA, LAP, classification:local populations are differentiated LDH- 1, PT- 1, AB-2 (hemoglobin), ACON, GOT- at D of 0.002, subspeciesat 0.044, speciesof the 2, ALD, GLUD, MDH-1, MDH-2, and G-6- same genus at 0.054, speciesof different confa- PDH. Six additional loci could not be scored: milial genera at 0.168, speciesof different tribes ACP, SOD-2 (absent), PK, LDH-2, (X-3, and at 0.168, and speciesof different families at 0.386. ME. The value for subspeciescan be refined. Because Levels ofgenetic variation within taxa are pro- Hemignathus“ v.” stejnegeriis genetically well- vided in Table 1. For the honeycreepers, ob- differentiated, we regard it as a species(see be- served heterozygositiesare not correlated with yond). Removal of stejnegerifrom the calcula- sample size (r = 0.3671; P > 0.05). Hobsranged tion ofD for subspeciesresults in a figure of 0.020 from 0.0 (in Loxioidesbailleui and Loxopscoc- (Table 4). Both the intertribal and interfamilial cineus)to 0.059(in Hemignathus“virens ”stejne- level values also require comment. Becausethe geri). Average Hobsover all drepanidines was intergenericconfamilial and intertribal values are 0.026, a value 39.5% lower than the average of virtually identical (0.1683 vs. 0.1675), we cal- 0.043 reported for birds in general (Barrow- culatedthe value for intratribalcomparisons. This clough 1980) and a value 51% lower than the value, D = 0.154, is again very similar to the average of 0.053 reported for large single breed- previous two, a result which indicates that the ing populations of 30 species (summarized by electrophoretic data for the speciesassayed do Barrowclough 1983:228-229). A reduction in not support the classification of Hawaiian hon- observed heterozygosity below typical levels is eycreepersinto formal tribes. Unexpectedly, the apparent in all drepanidines except the four pop- value obtained when speciesof different subfam- ulations of Hemignathusvirens and Paroreo- ilies of the same family were compared (n = myzamontana. Mean Hobsfor the 10 populations 0.4669) was largerthan that found when the hon- of the remaining seven specieswas 0.0 15. None eycreeperswere compared with the two species of the 15 population samples departed signifi- of Emberizidae (0.3862). cantly (P < 0.05; x2 test) from Hardy-Weinberg It is apparent that the genetic distancesdo not expectations. always change from one taxonomic level to the next according to expectation. This may result from the fact that the AOU classification of hon- GENETIC DISTANCES eycreepersis “evolutionary” or eclectic and not In Table 4 we summarize averageNei ’s D among cladistic.Thus, we cannot necessarilyassume that sample sets differentiated at increasingly inclu- the levels therein represent hierarchical phylo- sive levels, from local populational to interfa- genetic levels and, therefore, paraphyly cannot GENETICS OF HAWAIIAN HONEYCREEPERS 387
UPGMA Vestiaria coccinea (Hawaii) I Vestiaria coccinea (Kauai) Cophenetic - Himatione sanguinea (Hawaii) Correlation = 0.969 r”L Himatione sanguinea (Maui) Himatione sanguinea (Kauai) ‘2.4 - Hemignathus virens virens (Hawaii) oT”‘EHemignathus virens wilsoni (Maui) r - Hemignathus virens chloris (Oahu) L1.9 Hemignathus parvus yi‘.o- Telespiza cantans cantans L 1.7 - Loxioides bailleui 5:6 I Hemignathus virens stejnegeri (KaW -1.8 II Loxops coccineus caeruleirostris (Kauai) 716MYBP Paroreomyza montana -6.0 rl Oreomystis bairdi
Carpodacus purpureus californicus
1 I I I I I 0.40 0.30 0.20 0.10 0.0 Distance FIGURE 1. Phenogrambased on RogersD’ valuesand derivedby the UPGMA method.The highcophenetic correlationcoefficient (TV,) indicates excellent agreement between the distancesshown in the phenogramand the originaldata matrix. Numbersat branchingpoints provide crude estimates, in millions of years,for the timing of cladogenesisof taxa (seetext).
be ruled out (Bledsoe 1988b, p. 6-7; R. J. Rai- the two populations compared (Hawaii and kow, pers. comm.). Kauai). GENETIC POPULATION STRUCTURE BRANCHING DIAGRAMS Three of the speciesconsidered here are repre- The results of the four analyses, UPGMA, sented by two or more samples, permitting the WPGMA, F-M Tree, and Distance Wagner Tree, assessment of genetic population structure were broadly similar. The arrangement of branch through the calculation of Wright’s (19 5 1) F,,. At tips was especially concordant. In the UPGMA any locus, an F,, value of 1 indicates fixation of (Fig. l), seven major clustersare evident: (1) the alternative alleles between populations and a val- two populations of Vestiaria coccinea and the ue of 0 indicates panmixis. In Hemignathus vi- three populations of Himatione sanguinea; (2) rem, F,, averagedacross subspecies equals 0.3 107; Hemignathus virens virens, H. v. wilsoni, and H. in Himatione sanguinea, 0.0525; and in Ves- v. chloris; (3) Hemignathus parvus, Telespiza tiaria coccinea, 0.0 111. The value for H. virens cantans, and Loxioides bailleui; (4) Hemignathus points to clear population subdivision. This is “virens” stejnegeri and Loxops coccineus caeru- largely a result of the inclusion of the genetically leirostris; (5) Paroreomyza montana and Oreo- divergent taxon, H. “v. ” stejnegeri, which we re- mystis bairdi; (6) the two emberizids, Sicalisflav- gard as a species-level differentiate. The value eola and Piranga ludoviciana and (7) the two for H. sanguinea indicates modest genetic struc- carduelines, Loxia curvirostra and Carpodacus turing in that species.The low F,, for Vestiaria purpureus. Clusters 2 and 3 are sistergroups and coccinea suggestsslight subdivision, in keeping thesetwo clustersform a sistergroup with cluster with the very low Nei’s D (0.001) recorded for 1. In turn, the first three clusters form a sister 388 N. K. JOHNSON, J. A. MARTEN AND C. J. RALPH
Distance Wagner
’ ’ Carj,odac”s ~“r~ureus F-M TREE ca,ior”ic”*‘ Rogers’ D x 1000
%S.D. = 13.22 \ Loxia curvirosfra l__/--F grinnelli FIGURE 2. Branching diagram derived by the pro- c”r”i,oslra grinne,,i cedureof Fitch and Margoliash (1967). Branchlengths, FIGURE 3. Optimized distance Wagner tree rooted in units of Rogers’ D ( x 1,000) are drawn proportion- at the composite outgroup.This analysisproduced no ately. The tree is rooted at the composite outgroup.A negative branches. negative branch of miniscule length (-0.7) between the branch of length 6.9 and the bifurcation leading to Telespizacantans cantans and Hemignathusparvus is not plotted. Of 21 F-M trees examined, the one illus- groupings of speciesas revealed by the previous trated best summarized the original genetic distance two analyses. Indeed, the F-M tree is essentially matrix basedon the fewest(two) negativebranches and identical to the UPGMA phenogram: Himatione the lowest percentagestandard deviation. and Vestiaria are sister groupsjoined to a large cluster comprised of Hemignathus virens virens, group with cluster 4, the first four clusters form H. v. wilsoni, H. v. chloris, Hemignathus parvus, a sister group with cluster 5, and so on. Telespiza,and Loxioides; Hemignathus “virens” The WPGMA (not shown) differed from the stejnegeriand Loxops coccineusare sister species UPGMA in only two respects: cluster 4 (see linked to all the previously mentioned forms; above) is combined with cluster 2 in an unre- Paroreomyza montana and Oreomystis bairdi solved trichotomy and this enlargedcluster forms form a clade tied to all of the previous species; a sister group with clusters 1 and 3 which are in and the closest speciesto the drepanidines are turn sister groups. Thus, the main ambiguity the two emberizids rather than the two cardue- concerns the placement of the species couplet, lines. H. b.‘ ” stejnegeri and Loxops coccineus;other- Finally, the distance Wagner tree (Fig. 3) ex- wise both analyses lead to extremely similar pressesa structure which is again very similar to branching schemes.Finally, both UPGMA and that of the F-M tree. In the Wagner tree, how- WPGMA show the drepanidines closer to the ever, Loxops coccineusand H. “v. ” stejnegeri emberizids than to the carduelines, a result that branch off independently rather than being sister is contrary to expectation. taxa. Moreover, the three remaining subspecies The F-M tree (Fig. 2) identified the same major of Hemignathus virens form a cluster that also GENETICS OF HAWAIIAN HONEYCREEPERS 389
7 Telespiza cantans cantans
Vestiaria coccinea (Hawaii)
- Vestiaria coccinea (Kauai) Loxioides bailleui Oreomys tis bairdi Paroreomyza montana
Hemignathus virens virens (Hawaii) Hemignathus virens wilsoni (Maui)
Hemignathus virens chloris (Oahu)
Hemignathus virens stejnegeri (Kauai) Loxops coccineus caeruleirostris Wauai) Hemignathus parvus Himatione sanguinea (Hawaii)
Himatione sanguinea (Maui) Himatione sanguinea (Kauai) FIGURE 4. Strict consensustree resultingfrom a cladisticanalysis by locus.Although 186 equallyparsimonious treeswith 5 5 constructionsteps were produced,we stoppedthe output at 83 treesfor production of the consensus tree shown. Statisticsderived from this tree are as follows: Consensusfork index (component count) = 10, CF (normalized) = 0.588, Term information = 3 1, and Total information = 41. branches off the main stem independently, al- less, in most essential features the cladistic anal- though still in close proximity to the cluster ysis by locus (Fig. 4) suggestsrelationships that formed by H. parvus, Telespiza, and Loxioides. supportthose indicated by the UPGMA, the F-M We interpret the overall congruence and the tree, and the distance Wagner network: (a) the few discrepanciesof the four branching diagrams composite outgroup of four speciesforms a ma- to indicate substantial robustness in their por- jor cluster distinct from that containing all of the trayal of the relationships of taxa as expressed drepanidines (only in distance Wagner and cla- by Rogers’ D. The minor differencesamong the distic analyses);(b) Oreomystisbairdi clusterswith four topologies could potentially be resolved Paroreomyza montana; (c) H. “v. ” stejnegerial- through the construction of a “consensus tree.” lies with L. c. caeruleirostris rather than with However, for reasonspresented elsewhere (John- the three other forms of H. virens; (d) the two son and Marten 1988: 185) we do not advocate populations of l? coccineaform a cluster as do the use of such trees for determination of con- the three populations of H. sanguinea. In two gruence among branching patterns derived by important respects,however, the topology of Fig- different methodologies. ure 4 differs from any distance analysis, namely, in that Vestiaria does not group near Himatione CLADISTIC ANALYSIS and in that the two thick-billed drepanidine taxa, Because cladistic approaches have often provid- Telespiza and Loxioides, although part of the ed ambiguous results in other avian studies (Av- same major cluster are not portrayed as sister ise et al. 1980, Zink and Johnson 1984, Johnson groups. Furthermore, the basal relationships of et al. 1988) we did not anticipate congruence many of the speciesare unresolved in the con- with the various distance techniques. Nonethe- sensustree. 390 N. K. JOHNSON, J. A. MARTEN AND C. J. RALPH
DATING OF MAJOR CLADOGENETIC EVENTS ily Fringillidae. Recent findings from morphol- By applying Marten and Johnson’s (1986) mod- ogy (Bock 1970; Richards and Bock 1973; Rai- ification of the calibration of Gutierrez et al. kow 1976, 1977a, 1977b, 1978; Zusi 1978) and (1983) to Nei’s D, we can obtain estimates of the biochemistry (Sibley 1970, Sibley and Ahlquist timing of past cladogeneticevents. Thus, t = 19.7 1982, Bledsoe 1988a) have strongly supported x 106D, where t is the time since divergence and Sushkin’s view. D is Nei’s (1978) genetic distance. This exercise Our results clearly suggestthat the drepani- assumesthe operation of a molecular clock (Wil- dines are genetically closer to the two speciesof son et al. 1977, Thorpe 1982) in which allelic emberizids than to the two speciesof carduelines differencesaccumulate uniformly over time. The examined. As shown in Table 3, this general- report of Barrowclough et al. (1985), that genie ization is violated for only two of 60 Nei’s D variation in birds generally agreeswith the pre- values (in both Loxopscoccineus and in the Kauai dictions of Kimura’s (1979, 1982) neutral, mu- population of Vestiaria coccinea, Nei’s D was tation-drift model, supports the hypothesis of a larger in the comparison with Sicalis jlaveola clock and encouragesthe use of Nei’s D for the [0.456] than in that with Carpodacuspurpureus estimation of divergence times. However, be- [0.383]) and for only one of 60 Roger’s D values cause significant problems attend the determi- (L. coccineusvs. S. jlaveola is larger than in the nation of any calibration value, including ours comparison of the same specieswith C. purpu- (Marten and Johnson 1986:416-417), and be- reus). causegenetic distancesare accompanied by sub- Although this conflict is not easily resolved, stantial standard errors, we caution that the fig- four possibleexplanations come to mind. (1) Our ures for the timing of cladogeneticevents are best study examined insufficient loci to reveal true regarded as exploratory, blunt estimates. These relationships. This possibility seems unlikely in figureshave been added at the appropriate nodes view of the fact that the number of loci analyzed in Figure 1. here (36) is at the high end of the range typically Average Nei’s Ds separating the drepanidines assayed in avian studies. (2) Because both the from the carduelines and emberizids are 0.4669 emberizids and carduelinesare comprisedof large and 0.3862, respectively. These distances trans- clustersof speciesofwidely varying age (note the late into divergence times of 9.2 and 7.6 million substantial Nei’s genetic distances [Table 31 be- years before present (MYBP). The next major tween the two emberizids [0.212] and between split, that of the remainder of the drepanidines the two carduelines [0.321]), comparison of the from the creepersin the generaParoreomyza and drepanidines with speciesof those groups other Oreomystis, occurred at approximately 5.6 than the four chosenhere might provide a picture MYBP. Generic and specific divergence times of relationships more in agreement with the cur- ranged from approximately 2.4-1.0 MYBP, that rent consensusof opinion. Genetic comparison is, from the late Pliocene to the mid-Pleistocene. of drepanidines with Asiatic speciesof cardue- Subspecific divergences, as calculated from lines and emberizids would be especially perti- Hemignathus virensexclusive of H. “v.” stejne- nent in view of the possibility that the original geri (a probable species),occurred from approx- colonist arrived from that region rather than from imately 580,000 to 39,400 YBP (0.6-0.04 North America, as suggestedby Berger (198 1). MYBP), into the late Pleistocene. (3) High levels of homoplasy (parallelism and/ / or convergence)in allozyme expressionpatterns \\ ~&DISCvSS&N have obscuredtrue patterns of relationship. (The report of Mindell and Sites [1987] of frequent OkIFIN-&HE HAWAIIAN HONEYCREEPERS homoplasy and consequent slight phylogenetic Most earlier workers (Gadow 189 1, 1899; Ama- information in isozyme patterns, when higher don 1950; Baldwin 1953; Beecher 1953) believed categoriesof birds were compared, would not be that the honeycreeperscame from New World pertinent here.) Homoplasy in allozyme expres- nine-primaried oscine stock, specifically either sion has been suspectedin other avian studies the thraupines or coerebines, groups currently (Zink and Johnson 1984). Although this possi- placed in the Family Emberizidae. In contrast, bility is difficult to exclude completely, we note Sushkin (1929) proposed that the drepanidines that the drepanidines unambiguously allied more evolved from the cardueline finches of the Fam- closely with the emberizids than with the car- GENETICS OF HAWAIIAN HONEYCREEPERS 391 duelines. No vaguenesswas shown in this alli- Sibley and Ahlquist that the split of the drepan- ance as might be expected if either parallelism idines from its ancestral species probably oc- or convergence played a role. (4) The Hawaiian curred at the time of colonization. Furthermore, honeycreepers shared a more recent common as foreshadowedby Amadon (1947) and stated ancestor with the emberizids than with the car- explicitly by Sibley and Ahlquist (1982), the fact duelines. A. H. Bledsoe (in litt. 21 November that the honeycreepers captured so many eco- 1987) has suggestedtwo additional possibilities: logical niches provides excellent evidence that (5) “An increase in the rate of amino acid sub- their ancestor was the first passerine speciesto stitution in the carduelines over the rate typical become established in the islands. Finally, the of the other taxa could yield the pattern observed genetic evidence supports Sibley and Ahlquist’s in Table 3” (for additional comments on the pos- view (1982:138) that “most of the adaptive ra- sibility of variable rates see Bledsoe 1987a, diation that produced the 22 known speciesof 1987b), and (6) “that the drepanidines and car- Hawaiian honeycreepers probably occurred duelines are sister groups but that the effect of a within the past 5 million yr on the ‘high’ islands colonization bottleneck was to remove from the from Kauai to Hawaii.” drepanidines rare alleles evolved in the stem lin- eageleading to the cardueline-drepanidine clade. EVIDENCE FOR POPULATION BOTTLENECKS Such alleles would remain among the carduelines AND FOUNDER EFFECTS but be lacking in both the emberizids and the Insular colonization models typically assumethat drepanidines. The effect, like an increase in rate the initial pioneers were few in number and, in carduelines, would be to make the drepani- therefore, simply by chance introduced an un- dines and emberizines more similar to one representative sample of the parental gene pool another than the drepanidines are to the cardue- to the island (“founder effect,” Mayr 1942:237). lines.” In theory, passagethrough sucha population bot- All but the first of these possibilities seem rea- tleneck, both during and for a significant period sonable. Nonetheless,we seeno easyway at pres- after colonization, should be reflected in dimin- ent to judge their relative merits. This dilemma ished levels of genetic variation as demonstrated notwithstanding, the allozyme data should at least by low values of observed heterozygosity, per- re-open the apparently long-settled question of centageof polymorphic loci, and averagenumber drepanidine relationships and expose the need of alleles per polymorphic locus (Table I). Low for extensive comparisons, molecular and oth- heterozygosities are also expected if the total erwise, among a wide array of potentially related population of a species exists currently at low drepanidine, cardueline, and emberizid taxa. densities (Nei et al. 1975). The reduced genetic variability recorded in TIMING OF COLONIZATION several taxa can be explained by either or both Sibley and Ahlquist (1982) proposed that the of the above phenomena. For one species, the ancestor of the drepanidines colonized a now- Laysan Finch, clear evidence for a past bottle- submerged island of the Hawaiian chain at 20- neck is available. Although encountered com- 15 MYBP and that “it was the colonization event monly all over Laysan Island in 19 12-l 9 13, hab- that caused the dichotomy between the cardue- itat destructionby rabbits reducedthe population line and drepaninine lineages.” Our data indicate to an estimated few dozen individuals in 1923. a much more recent arrival, at approximately By 1938 the population had recovered to at least 7.6 MYBP. This younger date agrees generally 1,000. Present numbers of the speciesfluctuate with that published for the emergence of Nihoa between 7,500 and 14,800 (Berger 198 1). Other (7 MYBP, Dalrymple et al. 1974) which is now forms which are presently common, but which nearly submerged to the northwest of the main have low heterozygosities(e.g., Vestiariacocci- Hawaiian Islands. Furthermore, because of the nea and Himatione sanguinea)perhaps reflect imprecision of the 7.6 MYBP estimate, we can- past bottleneck events. Three currently scarce not rule out colonization of the ancestral species species,Loxioides bailleui (total population es- at Kauai, the oldest of the present main islands, timated at 1,600 individuals in 1975 [van Riper when it formed over the hot spot in the ocean et al. 19781) Oreomystisbairdi (total population floor at an estimated 5 MYBP (McDougall 1979). estimated at 6,800 -+ 1,900 individuals in 1968- Despite our conflicting dates, we agree with 1973, but declining [Scott et al. 19861) and 392 N. K. JOHNSON, J. A. MARTEN AND C. J. RALPH
Loxops coccineuscaeruleirostris (total popula- with one branch producing the lineage of thick- tion estimated at 5,100 -t 1,700 for Kauai in billed forms and the other giving rise in succes- 1968-l 973 [Scottet al. 19861)may illustrate both sion to Paroreomyza, Hemignathus, Loxops, past bottlenecks and the loss of genetic variabil- Palmeria, Himatione, Vestiaria, and Drepanis. ity which is expected to accompany small pop- In the absenceof material for detailed dissection, ulation size. Reduced levels of observed hetero- he excluded Ciridops from the phylogeny. zygosity have been reported for other insular taxa The aforementioned three phylogenies were of birds (Zink et al. 1987, Johnson and Marten derived from morphologic and ecologicdata. Our 1988). phylogeny based on allozymes differs fundamen- tally. First, neither the specialized nectarivorous PHYLOGENY OF THE DREPANIDINAE forms nor the thick-billed speciesare segregated In recent decades,three explicit phylogenies for into distinctive lineages.Thus, no genie evidence the Hawaiian honeycreepershave been offered. supportsthe views that either Himatione or the Amadon (1950:23 l-233), following Gadow finch-like forms are primitive. Instead, the creep- (189 1, 1899) envisioned a nectar-feeding, coe- ers (Paroreomyza and Oreomystis)divide basally rebid-like (=thin-billed thraupine) ancestor. from the remaining species.The fact that the flat- Rather early it split into two principal stocks. tongued, insectivorous creepers represent the The first lineage(the “Psittirostrinae”) led through modem descendantsof the earliest major branch intermediate forms, suchas Loxops (which under of drepanidines encouragesspeculation that the Amadon’s classification included H. virens, H. common ancestor of both the creepersand the parvus, P. montana, and 0. bairdi) and He- remaining modem drepanidines could have pos- mignathus, to culminate in thick-billed species, sesseda somewhat similar, generalized bill form, like Pseudonesterand Psittirostra (which then in- tongue morphology, and diet. Furthermore, it is cluded T. cantans and L. bailleui). The second plausible to envision subsequentradiation from lineage (the “Drepaniinae”) included specialized such an ancestor of separate lineages that cul- nectar feeders and led through Himatione and minated in modem speciesrepresenting (a) both Vestiaria to Drepanis and Ciridops. Thus, Ama- generalized (e.g., H. virens) and specialized (L. don viewed Loxops (sensu latu) and Himatione coccineus)omnivores, (b) specializedthick-billed to be the least specialized as well as earliest and thick-tongued finch taxa (e.g., Loxioides and evolved members of each major lineage. Baldwin Telespiza), and (c) specialized decurved-billed (1953:386-388) agreedwith the essenceof Ama- and tubular-tongued nectarivorous taxa (e.g., don’s proposed phylogeny. Himatione and Vestiaria). Although we do not Richards and Bock (1973: 122-128) also hy- propose that any modem speciesevolved from pothesizedan ancestorwith a tubular tongue but another, that possibility cannot be excluded. chosea Ciridops-like speciesfor this role. They, The aforementioned scenario requires the der- too, proposed a rather rapid cladogenetic event ivation of finch-billed forms with thick, flat soon after colonization in which two major lin- tongues from tubular-tongued ancestors, a pos- eagesarose from the Ciridops-like ancestor. The sibility first proposed by Amadon (1947). Such first lineage (the “Drepanidinae”) evolved into a route presents no serious difficulty. (Note the the stock that produced definitive Ciridops,then close relationship of thick-billed Telespiza and Himatione and the other advanced nectarivo- Loxioides with thin-billed Hemignathus sug- rous species. The other stock (the “Psittiros- gested by Fig. l-3.) Even an advanced tubular trinae”) passedthrough an ancestral “Loxops vi- tongue possessesa moderately thick base. Phy- rens-like” stage, then bifurcated into a branch logenetic loss through truncation of the rolled leading to Hemignathus and the thick-billed distal portion and subsequent thickening of the forms and a branch leading to modern Loxops basal portion of the tongue could be accom- coccineusand “L.” maculata (=Paroreomyza and plished through the alteration of developmental Oreomystis), “L.” parva, and “L.” virens. patterns and rates. Concomitant change in bill The third recently-published phylogeny is that shape also seems reasonable. As Alberch et al. of Raikow (1977a: 113-l 15). Like Sushkin (1929), (1979:315) wrote, “We know that slight pertur- he believed that a primitive cardueline finch col- bation in the ontogenetic trajectory of an organ onized the Hawaiian Islands and evolved into can be amplified through time, by the dynamics the drepanidines. He envisioned an initial split of growth and tissue interactions, to produce an GENETICS OF HAWAIIAN HONEYCREEPERS 393
adult phenotype drastically different from that no disadvantage to the birds . . . . That your of the ancestor. Clearly, major morphological hypothesiswould require the reappearanceof the changescan be the product of minor genetic mu- muscle would not, I think, be a strong argument tations.” Gould (1977) provides elaborate dis- against your view. Of course,it offersno support cussion of the crucially important relationship either.” We conclude therefore that the presence of ontogeny to phylogeny and Olson (1973:31- of M. plantaris in the thick-billed species does 36) describesan example in rails. not preclude the possibility that they were im- The presenceor absenceofM. plantaris, a small mediately derived from an ancestor that lacked muscle of the shank, has figured prominently in this muscle. recent discussions of the phylogeny and rela- The FITCH option of the program PHYLIP tionships of the drepanidines (Raikow 1977a). allows an objective, quantitative comparison of Many forms (e.g., “Loxops” virens, Himatione, phylogenies.Although the phylogenies proposed Vestiaria, and Palmeria) have lost this muscle, by Amadon (1950) and by Richards and Bock a condition consideredto be derived becausethe (1973) are too generalized to permit analysis by muscle is present in many passerine and non- this method, that of Raikow is explicit enough passerine families (Raikow 197822-23). Be- to allow comparison with our distance Wagner cause each of three thick-billed taxa (including results. Becauseour distance Wagner phylogeny Telespiza) examined by Raikow (1978) possess (Sumofsquares= 3.278,SD=9.819)hasalower M. plantaris, he cites this as evidence for the least squaresnetwork and lower standard devia- relative primitiveness of these taxa within the tion than Raikow’s phylogeny (SS = 5.454, SD drepanidines. However, several lines of evidence = 12.666) we conclude that it more accurately suggestthat the presenceof this muscle does not portrays the relationships inherent in the original argue persuasively for a basal position for the genetic distance matrix. Bledsoe (1987a) dis- thick-billed forms. First, as Raikow’s (1978:21) cussesthe use of the Fitch option to evaluate dissections show, this muscle is not uniformly phylogenetic estimates of distance data. present in either the emberizids or the cardue- Hopefully, the continuing researchprogram on lines, groups that have been considered as an- fossil depranidines (Olson and James 1982a, cestral to the drepanidines. For example, M. 1982b) will eventually shed more direct light on plantaris is absent in Diglossa, Cyanerpes, Car- the ancestors of the thick-billed taxa. Unfortu- podacus,Serinus (present in S. mozambicus but nately, fossils presently available are modern in absent in S. serinus), Carduelis, and Loxia. appearance(S. L. Olson, pers. comm.) and much Therefore, it is by no means certain that the an- too young (James et al. 1987) to assist in phy- cestraldrepanidine, whether of emberizid or car- logenetic reconstruction. dueline origin, possessedM. plantaris. But if it Our genetic data offer the most convincing evi- were present in the ancestor (and its presencein dence yet for monophyly of the drepanidines. the creeperParoreomyza arguesfor this), Raikow Doubts regarding monophyly have persisted in (1975) and Raikow et al. (1979) have described the literature up to the presentdecade (Bock 1970, a means through which the M. plantaris could Berger 198 1). In particular, Pratt (in Berger 198 1: have reappeared in the thick-billed forms even 149) felt that members of Paroreomyza were du- though absent from their most immediate ances- biously drepanidines based on a combination of tors. In brief, they review evidence suggesting behavioral, morphologic, and other evidence. that the phenotypic expression of particular ge- Although the creepers are clearly the phyloge- netic information can be interrupted for signifi- netically most distant taxa of drepanidines con- cant periods of time. Later, reactivation of rel- sidered here, they obviously ally with the other evant regulatory genes allows the “lost” feature drepanidines, rather than with either the embe- to reappear (Raikow et al. 1979:206). Recently, rizid or cardueline outgroup species.The refined Raikow (in litt., 27 January 1988) has kindly morphologic comparisons of Richards and Bock commented on this hypothesis with respect to (1973) and ofRaikow (1976, 1977a, 1977b) also the drepanidines: “With regard to the possibility indicate that the creepersare drepanidine. How- that M. plantaris could be lost and subsequently ever, demonstration of monophyly in the nine regained, I see no theoretical impediment. It is specieswe studied by no means rules out the a small muscle of dubious functional signifi- possibility that unstudied taxa could point to cance, and seems to be easily dispensed with at polyphyly. Specifically, the rare and little-known 394 N. K. JOHNSON, J. A. MARTEN AND C. J. RALPH
Melamprosopsphaeosoma is suspected of being and DEB-79-09807 to N. K. Johnson).Permits for the nondrepanidine (Pratt, in Berger 198 1: 170). collection of specimenswere provided by the Hawaii Department of Land and Natural Resources,Division TAXONOMIC IMPLICATIONS OF THE of Fish and Game, and by the United States Fish and GENETIC RESULTS Wildlife ServiceOffice of Rare and EndangeredSpecies. We wish to expressour sincere appreciation to all of In several instances the electrophoretic data the above individuals and agencies. pointed to relationships at variance with those of the most widely used current classifications LITERATURE CITED (e.g., Bock 1970, Berger 1981, AOU 1983). Dif- ALBERCH, P., S. J. GOULD, G. F. OSTER, AND D. B. ferences are evident from the subspecific to tribal WAKE. 1979. Size and shape in ontogeny and levels. The genie information suggests that: (1) phylogeny. Paleobioloav 5:296-3 17. Hemignathus “virens” stejnegeriis a full species AMA&N, D. 1947. Ecologyand the evolution of some possibly allied to Loxops coccineuscaeruleiros- Hawaiian birds. Evolution 1:63-68. AMADON, D. 1950. The Hawaiian honeycreepers tris; (2) Telespiza cantans and Loxioides bailleui (Aves, Drepaniidae). Bull. Am. Mus. Nat. Hist. are sister taxa in the distance Wagner treatment 95:151-262. and are perhaps congeneric; (3) the relationships AMADON,D. 1986. The Hawaiian honeycreepersre- of H. parvus are unresolved, (4) Vestiaria coc- visited. ‘Elepaio 46:83-84. AMERICANORNITHOLOGISTS UNION. 1983. Check- cinea and Himatione sanguinea are sister taxa list of North American birds. 6th ed. American in several analyses and are possibly congeneric; Ornithologists’ Union, Washington, DC. (5) Paroreomyza and Oreomystisare valid genera AVISE,J. C., J. C. PATTON,AND C. F.AQUADRO. 1980. and, although distantly related, are each other’s Evolutionary genetics of birds. I. Relationshins closest relatives among the drepanidines we among North American thrushes and allies. Auk 97:135-147. studied; and (6) among the species of Drepanid- BALDWIN,P. H. 1953. Annual cycle, environment inae examined, only two tribes are justified, one and evolution in the Hawaiian honeycreepers for Paroreomyza plus Oreomystis and another (Aves: Drepaniidae). Univ. Calif. Publ. Zool. 52: for the remaining species. We recommend that 285-398. BARROWCLOUGH,G. F. 1980. Genetic and uhenotvn- these suggested taxonomic changes be formally __ ic differentiation in a wood warbler (Genus Den- considered by the Committee on Classification droica)hvbrid zone. Auk 971655-668. and Nomenclature of the AOU. BARROWCL&GH,G. F. 1983. Biochemical studiesof microevolutionary processes,p. 223-261. In A. ACKNOWLEDGMENTS H. Brush and G. A. Clark, Jr. [eds.], Perspectives in ornithology. Cambridge Univ. Press, Cam- For assistancein collecting specimenswe are grateful bridge. to Helen F. James, Storm L. Olson, H. Douglas Pratt, BARROWCLOUGH,G. F., N. K. JOHNSON,AND R. M. Carol PearsonRalph, and CharlesVan Riper, III. Rob- ZINK. 1985. On the nature of genie variation in ert M. Zink provided helpful assistancein data analysis birds, p. 135-154. In R. F. Johnston[ed.], Current and useful discussion. Preliminary electrophoretic ornithology. Vol. 2. Plenum Press, New York. analysisof someof the sampleswas conductedby Rich- BEECHER,W. J. 1953. A phylogeny of the oscines. ard Sage, who graciouslyshared his findings with us. Auk 70:270-333. David Cannatella introduced NKJ to PHYLIP. Carla BERGER,A. J. 1981. Hawaiian birdlife. 2nd ed. The Cicero conducted the PAUP analysis. Drafts of the University Press of Hawaii, Honolulu. manuscript were read by D. Amadon, A. H. Bledsoe, BLEDSOE,A. H. 1987a. Estimation of phylogenyfrom P. F. Cannell. H. F. James. S. L. Olson. R. J. Raikow. moleculardistance data: the issueof variable rates. C. G. Sibley, and R. M. Zink, whosecollective insights Auk 104:563-565. greatly improved the final version. E. Prager, V.-M. BLEDSOE,A. H. 1987b. DNA evolutionary rates in Sarich, A. C. Wilson. and other members of the Mo- nine-primaried passerinebirds. Mol. Biol. Evol. lecular Evolution Seminar, Department of Biochem- 4:559-571. istry, University of California, Berkeley contributed BLEDSOE,A. H. 1988a. Nuclear DNA evolution and useful discussion.Although none of these colleagues phylogeny of the New World nine-primaried os- agreedwith all of our conclusions,each demonstrated tines. Auk 105:504-515. a commendable interest in what, rather than who, was BLEDSOE,A. H. 1988b. A hybrid Oporornisphila- correct. J. R. Northern of the Bird and Mammal Col- delphia x Geothlypistrichas, with comments on lection, Department of Biology, University of Califor- the taxonomic interpretation and evolutionary sig- nia, Los Angeles, kindly sent-copiesof D.-R. Dickey’s nificanceofintergeneric hybridization. Wilson Bull. 1923 field notes from Lavsan Island. The illustrations 100:l-8. were prepared by Karen l&z. Some analysesrelevant BOCK,W. J. 1970. Microevolutionary sequencesas to this researchwere conductedwith support from the a fundamental concept in macroevolutionary National Science Foundation (Grants DEB-79-20694 models. Evolution 24~704-722. GENETICS OF HAWAIIAN HONEYCREEPERS 395
BUTH, D. G. 1984. The application ofelectrophoretic MWDELL, D. P., AND J. W. SITES,JR. 1987. Tissue data in systematicstudies. Annu. Rev. Ecol. Syst. expressionpatterns of avian isozymes: a prelim- 15:501-522. inary studyofphylogenetic applications. Syst. Zool. DALRYMPLE,G. B., M. A. LANPHERE,AND E. D. 36:137-152. JACKSON.1974. Contributions to the petrogra- NEI, M. 1975. Molecular population genetics and phy and geochronologyof volcanic rocks from the evolution. North Holland Publ. Co., Amsterdam. Leeward Hawaiian Islands. Bull. Geol. Sot. Am. NEI, M. 1978. Estimation of average heterozygosity 851727-738. and genetic distance from a small number of in- FARRIS,J. S. 1972. Estimatingphylogenetic trees from dividuals. Genetics 89:583-590. distance matrices. Am. Nat. 106:645-668. NEI, M., T. MARWAMA, ANDR. CHAKRABORTY.1975. FITCH,W. M., ANDE. MARGOLIASH.1967. Construc- The bottleneck effect and genetic variability in tion of phylogenetictrees. Science 155:279-284. populations. Evolution 29:1-10. GADOW,H. 1891. Remarks on the structureofcertain OLSON,S. L. 1973. Evolution of the rails of the South Hawaiian birds, with referenceto their systematic Atlantic Islands (Aves: Rallidae). Smithson. Con- position, p. 219-24 1. In S. B. Wilson and A. H. trib. Zool. 152:1-53. Evans [eds.], Aves Hawaiienses: the birds of the OLSON,S. L., AND H. F. JAMES. 1982a. Fossil birds Sandwich Islands. R. H. Porter, London. from the Hawaiian Islands:evidence for wholesale GAD~W. H. 1899. Further remarks on the relation- extinction before Western contact. Science 217: ships of the Drepanididae, p. 243-249. In S. B. 633-635. Wilson and A. H:Evans [eds.],Aves Hawaiienses: OLSON,S. L., AND H. F. JAMES. 1982b. Prodromus the birds of the Sandwich Islands. R. H. Porter. of the fossil avifauna of the Hawaiian Islands. London. Smithson. Contrib. Zool. 365:i-vi+59. GOULD,S. J. 1977. Ontogenyand phylogeny.Belknap OLSON, S. L., AND H. F. JAMES. 1988. Nomenclature Press, Cambridge, MA. of the Kauai Amakihi and Kauai Akialoa (Dre- GUTI~RREZ,R. J., R. M. ZINK, ANDS. Y. YANG. 1983. panidini). ‘Elepaio 48: 13-l 4. Genie variation, systematic,and biogeographicre- PRATT,H. D., P. L. BRUNER,AND D. G. BERRETT. lationships of some galliform birds. Auk 100:33- 1987. A field guide to the birds of Hawaii and 47. the tropical Pacific. Princeton Univ. Press,Prince- HARRIS,H., ANDD. A. HOPKINSON.1976. Handbook ton, NJ. of enzyme electrophoresis in human genetics. RAIKOW,R. J. 1974. Species-specificforaging behav- North-Holland Publishing Co., Amsterdam. ior in some Hawaiian honeycreepers (Loxops). JAMES,H. F., T. W. STAFFORD,JR., D. W. STEADMAN, Wilson Bull. 86:471474. S. L. OLSON. P. S. MARTIN. A.J.T. JULL, AND P. RAIKOW,R. J. 1975. The evolutionary reappearance C. MCCOY. ’ 1987. Radiocarbon dates on bones of ancestralmuscles as developmental anomalies of extinct birds from Hawaii. Proc. Natl. Acad. in two speciesof birds. Condor 77~514-5 17. Sci. USA 84:2350-2354. RAIKOW,R. J. 1976. Pelvic appendagemyology of JOHNSON,N. K., AND J. A. MARTEN. 1988. Evolu- the Hawaiian honeycreepers(Drepanididae). Auk tionary geneticsof flycatchers.II. Differentiation 93~774-792. in the Empidonax d$ficiliscomplex. Auk 105:177- RAIKOW,R. J. 1977a. The origin and evolution of 191. the Hawaiian honeycreepers(Drepanididae). Liv- JOHNSON,N. K., R. M. ZINK, G. F. BARROWCLOUGH, ing Bird 15:95-l 17. ANDJ. A. MARTEN. 1984. Suggestedtechniques RAIKOW,R. J. 1977b. Pectoral appendagemyology for modem avian systematics.Wilson Bull. 96: of the Hawaiian honeycreepers (Drepanididae). 543-560. Auk 94:33 l-342. JOHNSON,N. K., R. M. ZINK, AND J. A. MARTEN. 1988. RAIKOW, R. J. 1978. Appendicular myology and re- Genetic evidence for relationships in the avian lationships of the New World nine-primaried os- family Vireonidae. Condor 90:428-445. tines (Aves: Passeriformes).Bull. Carnegie Mus. K~MURA,M. 1979. The neutral theory of molecular Nat. Hist. 7: l-43. evolution. Sci. Am. 241:98-126. _ RAIKOW,R. J., S. R. BORECKY,AND S. L. BERMAN. KIMURA. M. 1982. The neutral theorv as a basis for 1979. The evolutionary re-establishmentof a lost understanding the mechanism of
SELANDER, R. K., M. H. SMITH, S. Y. YANG, W. E. VAN RIPER,C., III, J. M. SCOTT,AND D. M. WOODSIDE. JOHNSON,AND J. B. GENTRY. 1971. Biochemical 1978. Distribution and abundancepatterns of the polymorphism and systematicsin the genusPero- palila on Mauna Kea, Hawaii. Auk 95:5 18-527. myscus. I. Variation in the old-field mouse (Pero- WILSON,A. C., S. S. CARLSON,AND T. J. WHITE. 1977. myscus polionotus). Univ. Texas Publ. 7 103:49- Biochemical evolution. Annu. Rev. Biochem. 46: 90. 573-639. SIBLEY,C. G. 1970. A comparative study of the egg- WRIGHT, S. 1951. The genetical structure of popu- white proteins of passerinebirds. Peabody Mus. lations. Ann. Eugen. 15:323-354. Nat. Hist. Bull. 32:1-131. WRIGHT, S. 1965. The interpretation of population SIBLEY,C. G., AND J. E. AHLQUIST. 1982. The rela- structureby F-statistics with specialregard to sys- tionships of the Hawaiian honeycreepers (Dre- tems of mating. Evolution 19:395-420. paninini) as indicated by DNA-DNA hybridiza- WRIGHT,S. 1978. Evolution and the geneticsof pop- tion. Auk 99: 130-140. ulations,Vol. 4. Variabilitv within and amona nat- SNEATH,P.H.A., ANDR. R. SOKAL. 1973. Numerical Ural populations. Univ. Chicago Press, Ch&go. taxonomy. W. H. Freeman and Co., San Francis- YANG. S. Y.. AND J. L. PATTON. 1981. Genie vari- co. ability and differentiation in the Galapagosfinch- SUSHKIN,P. P. 1929. On the systematic position of es. Auk 98:230-242. the Drepanidae.Verhandl. VI Int. Omithol. Congr. ZINK, R. M., ANDN. K. JOHNSON.1984. Evolutionary (1926):379-381. geneticsof flycatchers.I. Sibling speciesin the gen- SWOFFORD,D. L. 1985. Phylogenetic analysis using era Empidonaxand Contopus. Syst.Zool. 33:205- parsimony, version 2.4. Illinois Nat. Hist. Surv., 216. Champaign, IL. ZINK, R. M., D. F. LQTT, AND D. W. ANDERSON.198 7. SWOFFORD.D. L.. AND S. H. BERLOCHER.1987. In- Genetic variation, population structure,and evo- ferring evolutionarytrees from genefrequency data lution of California Quail. Condor 89:395405. under the principle of maximum parsimony. Syst. ZUSI, R. L. 1978. The interorbital septum in cardue- Zool. 36:293-325. line finches. Bull. Br. Omithol. Club 98:5-10. THORPE,J. P. 1982. The molecular clock hypothesis: biochemical evolution, geneticdifferentiation and systematics.Annu. Rev. Ecol. Syst. 13:139-168.