The Auk 111(1):178-184, 1994

LOSS OF FUNCTION IN TERRITORIAL SONG: COMPARISON OF ISLAND AND MAINLAND POPULATIONS OF THE SINGING (MELIPHAGA VIRESCENS)

MYRON CHARLES BAKER • ZoologyDepartment, University of WesternAustralia, Nedlands, Perth, 6009

ABSTRACT.--Istudied vocal recognitionin populationsof Singing (Meliphaga virescens)on Rottnest Island and the adjacent mainland of Western Australia through song- playbackexperiments. Morphological evidence suggestedthe island population had differ- entiated. ! addressedthe question of whether there was divergence in the vocal signaling system.! found significant loss of function of the territorial advertising song by comparing responsesof the Rottnest subjectsand those of subjectsfrom the mainland Nedlands popu- lation. The low level of responseby Rottnest to songsfrom the mainland was similar to the responseby Rottnestbirds to songsof the co-occurringGolden Whistler (Pachycephala pectoralis).On the mainland, Nedlands birds responded to the songs of another mainland population,located the samedistance away as RottnestIsland, at an elevatedlevel similar to their responseto songsof their own Nedlands population. I conclude that there are popu- lation-specificvocal featuresin the songsof RottnestSinging Honeyeatersthat result from isolation. The isolation causedby the water barrier between RottnestIsland and the mainland is greater than isolation by distance through continuous populations on the mainland. I speculate that vocal recognition was probably lost as a result of an early colonization of Rottnestand the subsequentestablishment of a new vocaltradition in the island birds. Received 31 July 1992, accepted24 April 1993.

THE STUDY OF the differentiation of island of a new vocal tradition through founder effect forms from mainland source populations has (Thielcke 1973, Baker and Cunningham 1985). contributed much to our understanding of evo- This could have consequencesfor the isolation lution in birds (Lack 1947, Bond 1948, Amadon of gene pools (Carson 1971, Baker and Marler 1950, Raikow 1977). Numerous traits have been 1980, Kaneshiro 1980). shown to diverge as a result of isolation on Previous work indicated that Singing Hon- islands. Traits most commonly documented as eyeaters(Meliphaga virescens) of the Rottnest Is- diverging are morphological,and include land (WesternAustralia) population on average length, body size, wing length, and plumage have a mass that is about 20% more than the coloration (Murphy and Chapin 1929, Amadon adjacent mainland form (Wooller et al. 1985). 1953, Van Valen 1965, Grant 1965, 1968). Experienced birders also have noted that Rott- Rarely have isolated island populations of nestSinging Honeyeatersare somewhatdarker, birds been examined for divergence in com- with more streaked plumage on the underside municative signals.Although the classicalcases of the body, than mainland individuals (Saun- of adaptive radiation in birds are satisfactorily ders and deRebeira 1985). I used tape recording explainedby referenceto ecologicaldivergence and playback experiments to determine wheth- (Amadon 1950, Grant 1986), the initial phase of er there was any loss of communicative func- differentiation and, particularly, the effectsof tion in the vocal signalsof Singing Honeyeaters subsequent new colonizations or recoloniza- from the Rottnest Island and mainland popu- tions by the sameor other forms may be influ- lations. enced by reproductive signals involving terri- tory defense and mate choice. Becausemany MATERIALS AND METHODS speciesof songbirdslearn their songs,it might be expected that in such speciesthe coloniza- The is one of 67 speciesbe- tion of an island would lead to establishment longing to the Meliphagidae,one of the mostspeciose families of Australian (Thompson 1964). Many honeyeater speciesare abundant. Most feed ' Present address: Biology Department, Colorado principally on nectar. The Singing Honeyeater (20- State University, Fort Collins, Colorado 80523, USA. 30 g) is the most widely distributed honeyeater in

178 January1994] SingingHoneyeater Vocal Recognition 179

Australia (Blakers et al. 1984) and seems to be es- wheneverthere was a changeof distancethat resulted pecially successfulin the abundantly flowered gar- from movement initiated by the . In addition, a dens of suburban areas.There are no published rec- synthetic variable, sumbeh, was created by summing ords of this species'vocalizations. vocalizations,hops and flightsto providean index of The SingingHoneyeater has a discretedawn chorus the total activity. The distanceestimates were aided that commencesabout 20 to 30 rain beforelight, and by pacingbetween key featuresof the habitatin the lasts about 40 to 60 min. Typically, the male roosts vicinity of the speakerafter the test was completed. overnight in a tree and sings from that location in For analysis,the distanceestimates were averagedfor the pre-dawn period. This songis heard infrequently the trial to give a single value of distance for that at other times of the day. Along a residential street subject'sresponse to the stimulus.Hops occurredusu- it is commonto find a singing male nearly every 50 ally where a subjectremained within a single bush m in the treesalong the vergeand in the front gardens or tree for a period of time and changedlocations of the homes.A territory is defended physicallyand without flying. Flights were usuallybetween treesor vocally, defenseoften being centeredaround nectar bushes,and were frequently to and from the bush sources. Access to these nectar sources is contested near the playbackspeaker. All types of vocalizations frequently by the (Anthochaeracarun were summed in a single category.There was con- culata)and the (Lichmerai•distinc- siderablevariety in the vocal performanceof subjects ta). Vocal defense during daytime is usually not by during tests,but the complexitywas too great to make songbut by a number of different calls.These appear reliable categoriesof vocalizations.This would re- to indicate a range of motivational conditions from quire a separate study. The kinds of vocalizations I an initial "distant alarm," given when an intruder heardduring playbacktests were the sameas I heard first enters the spaceof another male, to intense chat- during my observationsof the aforementionedtwo tering and continuousmultiple chirping when two territorial males during their normal day's interac- birds are closetogether. Observations I made on two tions with intruder neighborsand other species. vocally distinct malesover a period of severalweeks In the playback tests on Rottnest, subjectsheard suggestthat the birds are monogamous.Most breed- songsof Singing Honeyeatersfrom Rottnestor from ing commencesin the Australasianspring (August- Nedlandsor heard GoldenWhistler (Pachycephalapec- September) and ends in December-January.Males toralis;Gws) songsfrom Rottnest.The Golden Whis- were singing persistently when this study was ini- tler stimuli were used to determine if the response tiatedon 1 October,but by mid- to lateJanuary sing- to Nedlands stimuli was equivalent to the level of ing had decreasedconsiderably. Songs were heard responsegiven to heterospecificsongs. The Golden occasionallyon into March. Whistler occurstogether with Singing Honeyeaters Males in three populationswere tape recorded.One in many locations on Rottnest Island. Two different population was located in suburbanNedlands (Ned), Golden Whistler songswere used as stimuli. There- near Perth, Western Australia. A second population fore, the statisticaldesign consistedof three treat- was located 25 km west of Nedlands on the island of ments (Rot, Ned, Gws) with five different stimulus Rottnest (Rot) in the Indian Ocean. Rottnest was iso- songswithin eachof Rot and Ned treatmentsand two lated about 6,500 yearsago as the sea level rose near different stimulus songswithin the Gws treatment. the end of the Holocene (Playford 1983). A third pop- Forty subjectswere tested(15 with Rot, 15 with Ned, ulation was located 25 km east of Nedlands in the 10 with Gws). suburbanareas on the western edge of John Forrest When evaluatingthe mainland population of Ned- National Park (Jfp). Songswere recorded during the lands, subjectsheard the samefive Nedlands stimuli dawn chorususing a Marantz cassetterecorder (PMD and five Rottneststimuli as did the Rottnestsubjects. 201), Sennheiser microphone (MD 402-K) mounted In addition, I used five Jfp songsas stimuli. Forty- in a 40-cm parabolicreflector, and TDK Type I tape. eight subjectswere tested(16 eachwith Ned, Rot,and Songsof five different maleswere randomly se- Jfp). lected from each population and transferred to 20-s Data were analyzed by nested ANOVA (stimuli continuous-loop cassettes.These served as playback nested within treatments). Fisher's LSD (Carruer and stimuli (Fig. 1), which were delivered at the rate of Swanson1973) was used to make multiple compari- one song per 10 s for 5 rain during each playback sonsbetween pairs of treatments. trial. A trial was initiated by locating a vocalizing Singing Honeyeaterand placingthe playbackspeaker RESULTS (Perma Power S-610) within 5 m of the individual and under or beside a tree or bush. I activated the stimulus and retreated 5 to 10 m to make a record of Rottnestsubjects.--The experimental birds on events for 5 rain on another cassette recorder. RottnestIsland respondedto the playback tests I recorded the occurrence of vocalizations (vocals), with behavioral reactions that indicated they flights, and hops as discretecountable events; also, I clearly discriminatedbetween classesof stimuli estimatedthe distancefrom the speakerto the subject presentedto them. Songsrepresenting their own 1 80 MYRONCHARLES BAKER [Auk, Vol. 111

Ned

i i Rot

z Jf:

i • 0.5s 4-

Fi•. •. Sou• s•ectto•ta•so• t•o exa•[es o• eachc[ass o• so•s use•as •[aybac• stimuli. •o•u[atio• sourceso•so•s ate:SJ•i• •o•eyeatets•to• •ed[ands(•ed), •ottnest Island (•ot), John Forrest Nationa[ •at• (]•); GoldenWhistlers (Gws) •tom •ott•est Island.

Rottnestpopulation elicited heightenedlevels the Nedlandspopulation on the mainlandalso of behavioral activity commonly seen in natu- reactedto the playbacksin waysthat indicated rally occurringterritorial interactions. All the discrimination between the different kinds of dependentvariables exhibited significant treat- songsused as stimuli. In general,Nedlands sub- menteffects and no significanteffect of stimulus jectsshowed elevated responseto songsrep- within treatment (Table 1). Multiple-compari- resentative of their own Nedlands population son tests indicated that for the vocals, flights, and songsof the Jfp populationin comparison hopsand sumbehvariables Rottnest males re- to a lower responseto songsfrom RottnestIs- spondedmore to Rot songsthan to Ned or Gws land. Significanttreatment effects were present songs(Fig. 2; P < 0.05).There was no difference for the variables vocals, sumbeh and distance betweenthe responsesto Ned and Gws.For the but not for hopsor flights.There were no sig- distancemeasure, response to Rot (27= 3.3 ñ SE nificant effects of stimuli within treatments for of 0.4 m) was greater (approachedcloser to anyvariable (Table 1). Multiple-comparison tests speaker)than to Ned (27= 12.2ñ 3.1;P < 0.05), indicatedthat for vocals,hops and sumbehthe but the responsesto Rotversus Gws, and to Ned subjectsresponded more to Ned than to Rot versus Gws (œ= 7.8 +_ 1.7 m) did not differ. stimuli(Fig. 2; P < 0.05).There were no differ- Nedlandssubjects.--The experimental birds in encesbetween Rot and Jfp or between Ned and January1994] SingingHoneyeater Vocal Recognition 181

TABLE1. Nestedanalysis of varianceresults for play- < 0.05). The distance measure did not differ for back tests conductedon Singing Honeyeaterson Ned versusJfp. Rottnest Island and in Nedlands, Western Austra- lia DISCUSSION Stimulus within Treatment treatment I concludethat there are population-specific Variable F-value P F-value P songfeatures (or species-specificfeatures in the

Rottnest Island case of the response to Gws) that are used as Vocals 7.35 0.003 2.01 0.076 cues by Singing Honeyeaters for judging the Hops 6.39 0.005 1.64 0.151 appropriatenessof an aggressiveterritorial re- Flights 13.16 0.0001 0.33 0.958 sponse.In 8 of 10 comparisons,there were sig- Sumbeh 10.86 0.0003 1.92 0.090 nificant alterationsin the behavioralresponse Distance 5.18 0.012 1.37 0.246 of the playback subjectscaused by the three Nedlands stimulus treatments. In none of the 10 was there Vocals 4.12 0.025 0.88 0.569 significant variability caused by the different Hops 3.00 0.063 1.28 0.271 stimulus songswithin a treatment (Table 1). Flights 1.43 0.254 0.57 0.848 Sumbeh 5.32 0.009 0.81 0.639 Subjectsin both the mainland Nedlands pop- Distance 5.48 0.008 0.75 0.693 ulation and the RottnestIsland population re-

' df = 2 for all treatments.df = 9 and 12, respectively,for evaluations spondedconsistently more to songsof their own of stimuli within treatments for Rottnest Island and Nedlands. populationthan to songsrepresentative of the other's population. This was true for the mea- suresof vocals,hops, sumbeh and dist.The only Jfp. For flights, there were no significantdif- exceptionwas for the variable "flights" in the ferencesin any of the three pairwise compari- Nedlandspopulation. Therefore, I concludethat sons. For the distance measure, subjectsre- the songsfrom the Nedlandspopulation have sponded more (approached closer) to the Ned lost significant function as territorial releasers (2 = 1.4 _+ 0.3 m) and Jfp (2 = 1.6 _+ 0.4 m) for birdsof the RottnestIsland population, and stimuli than to Rot (2 = 4.8 _+ 1.2 m) stimuli (P Rottnestsongs function poorly in the Nedlands

RottnestSinging Honeyeaters Nedlands Singing Honeyeaters

140t ß[]FIights Vocals 140-ß co120t []Hops co120- '• 1001 • 100- O O ß '• 80- • 80- "• Q) • 60' • 60' 7 7 4o- 4o- • 20- 20

o o Rot Ned Gws Rot Ned Jfp StimulusSong StimulusSong Fig. 2. Responsesof Rottnestand Nedlands subjectsto playbackstimuli. Each histogrambar is stackof three separateresponse categories (vocals, flights, hops),and total bar height is their sum (sumbeh).Whisker indicatesSE. Horizontal linesabove graph connect pairs of stimuli that were significantlydifferent. Responses of Rottnestbirds to Rotstimuli were significantlygreater than responsesto Ned and Gwsfor all threeresponse measuresseparately (vocals, flights, hops),as well as for compositesumbeh measure. Responses of Nedlands birds to Ned stimuli were significantlygreater than responsesto Rot for measuresof vocals,hops and sumbeh, but not for flights. 182 MYRONCHARLES BAKER [Auk, Vol. 111 population. For Rottnest birds, the degree of circumstantial evidence of a role for imitation recognitionof Nedlands songswas no different in song development (Kroodsma 1982). from the degree of recognition of Golden Whis- Other comparisonsof island-island or island- tler songs according to the nondistance mea- mainland vocal recognition in birds are rare in sures of response.Thus, Rottnest birds treated the literature. A number of studies compared songs of Nedlands birds as if they were those song structure of island and mainland popu- of another species.The closeapproach to speak- lations (e.g. Lack and Southern 1949, Marler ers by Rottnest birds, however, suggeststhat and Boatman 1951, Marler 1960), but did not the Golden Whistler song was recognized to a conduct experiments to determine if the differ- degree intermediate between Rottnest and ences were discriminated by the birds (re- Nedlands stimuli. Thus, Golden Whistler song viewed in Miller 1982). An exception is the is- may containsome features that tended to attract land-islandcomparison of responsesto playback Rottnest Singing Honeyeaters, but once the by Galapagosfinches. Five speciesof Geospiza subjectsapproached the stimulus they did not were tested in seven between-island compari- increasevocalizations, hops or flights. sonsof responseto songplayback (Ratcliffe and The nondistance measures of response by Grant 1985). The general pattern of results Nedlands subjectssuggest that the songsof Jfp showed that in most comparisonsthe subjects elicited an intermediate degree of recognition exhibited some degree of discrimination; they compared to Nedlands and Rottnest songs.Al- responded more to songsfrom their local pop- though Jfp songs elicited as much responseas ulation than to songs representing another is- Nedlands songs,Jfp songsalso did not elicit a land population (Ratcliffe and Grant 1985). responsesignificantly different from that given Finding a significant loss of recognition of to Rottnest songs.The distance measure,how- conspecificsongs raises questions about the ever, suggeststhat Jfp songswere asstimulating possibleconsequences for population process- to Nedlands subjectsas Nedlands songs were, es. In particular, the immigration of a Rottnest and that both these classesof songs were dis- male into the Nedlands population, or vice ver- criminated from Rottnestsongs. Therefore, the sa, presumablywould not be met with any ex- data from the mainland populations indicate cessaggression by a resident male. Instead, it that the 25-km distance between Nedlands and might initially be ignored.The immigrant'ssong John Forrest Park does not lead to a major loss would not be an effective keep-out signal for of songfunction. Together with the resultsfrom territory establishment.Morphologically, the the Rottnestsubjects, I conclude that isolation birds of Rottnest and Nedlands are similar by distancethrough continuouspopulations of enoughthat it seemslikely an immigrant, when Singing Honeyeaters is not as effective in caus- seen by the resident, would elicit an aggressive ing loss of vocal recognition as is isolation by reaction. In turn, this response might trigger the water barrier between Rottnest Island and an adoption of the local song traits by the im- the adjacent mainland of Western Australia. migrant or the acquisition of perceptual rec- There are no experimental data concerning ognition by the residentof the immigrant'ssong the method by which song features are trans- features.The former hypothesiscould be tested mitted acrossgenerations of Singing Honey- by experimentson captive birds and the latter eaters.Previously, the only evidencethat learn- hypothesis might be tested by extended-time ing is involved in the acquisitionof songin any playback experiments on territorial residents. speciesin the Meliphagidae restedupon an an- Suchoccurrences of immigration,however, may ecdotalreport of the ParsonBird (Prosthemaderabe quite rare. On nearby Carnac Island, 8 km novaezealandiae)mimicking nonavian sounds from the mainland, 70 daysof continuousmon- (Gilllard 1958). My observations(Baker 1993) itoring of immigrants revealed no Singing suggestvocal learning is involved in songde- Honeyeaters, and only one immigrant Singing velopment in Singing Honeyeatersbecause of Honeyeater was found during the period 1934- the existenceof local song neighborhoods.In 1977 (Abbott 1978). thesesong neighborhoods, a few neighboring Lossof recognition by males could imply a maleshave very similar songsand these differ similar lossby femalesand this suggestsa fur- markedlyfrom other nearbygroups of neigh- ther consequencefor gene flow. Perhapsan im- bors.This type of vocalconvergence at the mi- migrant male would have trouble attracting a crogeographicscale is usually taken as strong mate, assuming that the song of the Singing January1994] SingingHoneyeater Vocal Recognition 183

Honeyeater has a role in this process.Results anismsin a hybridizing pair of species.Evolution of female preferencetests in severalspecies sug- 44:332-338. gest that lossof recognition by females might BAKER,g.C., AND g. g. CUNNINGHAM. 1985. The occur (Baker et al. 1987a, b, Balaban 1988, Baker biology of bird song dialects. Behav. Brain Sci. 8:85-133. and Baker 1990),although experimentson Sing- BAKER,M. C., AND P. MARLER. 1980. Behavioral ad- ing Honeyeatersare needed to examine this aptationsthat constrainthe gene pool in verte- possibility. brates.Pages 59-80 in Evolution of social behav- Conceivably, there is sufficient genomic di- ior: Hypotheses and empirical tests (H. Markl, vergence between Rottnestand mainland Sing- Ed.). Dahlem Konferenzen, Verlag Chemie, ing Honeyeatersto causehybrid breakdown in Weinheim, Basel. the event of matingsbetween the populations. BAKER,M. C., BJERKE,T., LAMPE, H., AND Y. ESPMARK. This speculationis based upon the morpholog- 1987a. Sexual response of female Yellowham- ical divergence of the Rottnest population mers to differencesin regional songdialects and (Saunders and deRebeira 1985, Woollet et al. repertoire sizes. Anim. Behav. 35:395-401. BAKER,M. C., P. K. MCGREGOR,AND J. R. KREBS. 1987b. 1985)that earlier led Milligan (1911) to describe Sexualresponse of female Great Tits to local and the Rottnest form as a new species.Increased distant songs.Ornis Scand. 18:186-188. understanding of this problem could be BALAI3AN,E. 1988. Cultural and genetic variation in achieved by a description of the genetic struc- SwampSparrows (Melospiza georgiana). II. Behav- ture of the populations, as well as by further ioral sallericeof geographic song variants. Be- experimentation on the reproductive commu- haviour 105:292-322. nication systemof these birds. BEAKERS,M., S. J. J. F. DAVIES,AND P. N. REILLY. 1984. The atlas of Australian birds. Melbourne Univ. Press, Melbourne. ACKNOWLEDGMENTS BOND,J. 1948. Origin of the bird fauna of the West Indies. Wilson Bull. 60:207-229. I thank the faculty and the staff of the Zoology Department of the University of Western Australia CARMER, S.G., AND M. R. SWANSON. 1973. An eval- for their hospitality. I owe special thanks to Mike uation of ten pairwisemultiple comparisonspro- Johnston,Win Bailey, BobBlack, Dale Roberts,Darren cedures by Monte Carlo methods. J. Am. Star. Murphy, Perry deRebeira, and Mike Craig for their Assoc. 68:66-74. helpful advice and the loan of equipment. Ron John- CARSON,H. L. 1971. Speciation and the founder stone provided natural-history information on Sing- principle. Stadler Genet. Symp. 3:51-70. ing Honeyeaters.The general nature of the problem GILLIARD,E.T. 1958. Living birds of the world. Ham- addressedin this study was suggestedby Ron Wool- ish Hamilton, London. ler, who also reviewed the manuscript.Work on Rott- GRANT,P.R. 1965. The adaptive significanceof some nest Island was made possible by Don Edward and size trends in island birds. Evolution 19:355-367. Charles Hansen. Financial assistancewas through GRANT,P.R. 1968. Bill size, body size, and the eco- grants from the National Science Foundation (BNS logical adaptationsof bird speciesto competitive 87-06526, INT 9113640). situations on islands. Syst. Zool. 17:319-333. GRANT,P. R. 1986. Ecology and evolution of Dar- win's finches. Princeton Univ. Press, Princeton,

LITERATURE CITED New Jersey. KANESHIRO,K.Y. 1980. Sexual isolation, speciation ABBOTT,I. 1978. Factorsdetermining the number of and the direction of evolution. Evolution 34:437- land bird specieson islands around south-west- 444. ern Australia. Oecologia33:221-233. KROODSMA,D.E. 1982. Learning and the ontogeny AMADON, D. 1950. The Hawaiian honeycreepers of sound signalsin birds. Pages1-23 in Acoustic (Aves, Drepaniidae). Bull. Am. Mus. Nat. Hist. communication in birds, vol. 2. (D. E. Kroodsma 95:151-262. and E. H. Miller, Eds.). Academic Press,New York. AMADON,D. 1953. Avian systematicsand evolution LACK,D. 1947. Darwin's finches. Cambridge Univ. in the Gulf of Guinea. Bull. Am. Mus. Nat. Hist. Press,Cambridge. 100:397-431. LACK, D., AND H. N. SOUTHERN. 1949. Birds on Tene- BAKER,g.C. 1993. Evidence of intraspecific vocal rife. Ibis 91:607-626. imitation in Singing Honeyeaters (Meliphagi- MARLER,P. 1960. Bird songsand mate selection.Pages dae) and Golden Whistlers (Pachycephalidae). 348-367 in sounds and communication Condor 95:1044-1048. (W. E. Lanyon and W. N. Tavolga, Eds.). Publ. BAKER,M. C., AND A. E. M. BAKER. 1990. Reproduc- no. 7. American Institute of Biological Sciences, tive behavior of female buntings: Isolation mech- Washington, D.C. 184 MYRONCHARLES BAKER [Auk,Vol. 111

MARLER,P., AND D. J. BOATMAN. 1951. Observations III. Male responsesto playback of different song on the birds of Pico, Azores. Ibis 93:90-99. types, dialects and heterospecificsongs. Anita. MILLER, E.H. 1982. Character and variance shift in Behav. 33:290-307. acousticsignals of birds. Pages253-295 in Acous- SAUNDERS, D. A., AND C. P. DEREBEIRA. 1985. The tic communication in birds, vol. 1 (D. E. Kroods- birdlife of Rottnest Island. DAS and CPdeR, Perth, ma and E. H. Miller, Eds.). Academic Press, New Western Australia. York. THIELCKE,G. 1973. On the origin of divergence of MILLIGAN,A. W. 1911. Descriptionof a new Ptilotis. learned signMs (songs) in isolated populations. Emu 11:124-125. Ibis 115:511-516. MURPHY,R. C., AND J. CHAPIN. 1929. A collection of THOMPSON,A. L. 1964. A new dictionary of birds. birds from the Azores. Am. Mus. Novit. 384:1- McGraw-Hill, New York. 23. VAN VALEN,L. 1965. Morphological variation and PLAYFORD,P. E. 1983. Geological research on Rott- the width of the ecologicalniche. Am. Nat. 99: nest Island. J. R. Soc. West. Aust. 66:10-15. 377-390. RAIKOW,R.J. 1977. The origin and evolution of the WOOLLER,R. D., D. A. SAUNDERS,J. S. BRADLEY,AND Hawaiian Honeycreepers(Drepanididae). Living C. P. DEREBEIRA.1985. Geographicalvariation Bird 15:95-117. in size of an Australian honeyeater (Aves: Me- RATCLIFFE,L. M., AND P. R. GRANT. 1985. Species liphagidae):An exampleof Bergman'srule. Biol. recognitionin Darwin's finches(Geospiza, Gould). J. Linn. Soc. 25:355-363.