Quick viewing(Text Mode)

P0737-P0748.Pdf

P0737-P0748.Pdf

The Auk 116(3):737-748, 1999

COMPARATIVE ANALYSIS OF PERCH-COO VOCALIZATIONS IN DOVES

' HANS SLABBEKOORN,• SELVINO DE KORT, AND CARELTEN CATE BehavioralBiology, Institute of Evolutionaryand EcologicalSciences, Leiden University, P.O. Box 9516, 2300 RA Leiden, The Netherlands

ABSTRACT.--Someof the 16 dovespecies in the genusStreptopelia are very similarwith respectto plumagepattern, but all seemto producespecies-specific "perch-coo" vocaliza- tions.Here, we describevariation in perch-coorecordings of all 16 .All individuals couldbe correctlyclassified by speciesin a discriminantfunction analysis, which means that the overlapin inter- and intraspecificvariation was limited for the total set of acousticpa- rameters.Interspecific similarity in acousticparameters was compared with taxonomicclas- sification,based upon qualitative morphological characteristics and geographic distribution (Goodwin1983). The clusteringof speciesusing the acousticdata set showed little congru- encewith taxonomicclustering. This indicatesthat differentiationin plumagepattern does not necessarilycoincide with differentiationin acousticcharacteristics. However, our anal- ysisdid not completelycontradict the subdivisioninto four taxonomicgroups based on mor- phologyand distribution.Two of the four putativegroups differed significantly from the othergroups in oneof the componentsof a principalcomponents analysis. Vocal differen- tiation in Streptopeliadoves was strongestin temporalcomponents, which is in line with expectationsbased on the evolutionarilyconservative syringeal constraints. Received 29 Sep- tember1997, accepted 4 January 1999.

DIFFERENTSPECIES of a particular taxonomic Perch-coovocalizations of dovesare regard- groupof birdsoften share vocal characteristics. ed as functionallysimilar to advertisingsongs At the same time, some vocal characteristics in songbirds;they servein male-maleconflict serve as accurateindicators of speciesstatus. and for female attraction (Jacksonand Basket For this reason,song comparisons have been 1964;Davies 1970, 1974;Goodwin 1983;Cramp widely usedto evaluaterelationships among 1985;Baptista 1996).Sexual selection may lead species(see Payne 1986). Comparative analyses to intraspecificexpansion of variancein the of vocalizations have been used to evaluate in- acousticparameters of sucha signal.However, tra- and interspecificvariation in severalavian the samesignal also may servein speciesrec- genera(e.g. Zann 1974,Miller 1978,Collins and ognition(Becker 1982, Nelson 1989). In this Goldsmith1998) and haveled to suggestions context,the negativeconsequences of hybrid- for taxonomicsubdivisions within somegenera izationmay lead to intraspecificcontraction of of doves(, Harrison 1969;Zenaida, Bap- varianceand interspecificdivergence (see Mil- tista et al. 1983). Other studieshave evaluated ler 1982).Sexual selection may shapeacoustic vocalvariation among species in the contextof parameterssuch as frequencyrange or repeti- phylogeneticreconstruction (e.g. G•ttinger tion patternof vocalizations,but acousticlimits 1970, Miller et al. 1988, and Crawford areaffected by environmentaland phylogenet- 1996, Miller 1996, McCracken and Sheldon ic constraints(Ryan and Brenowitz 1985). Phy- 1997).In this study,we use a comparativeap- logeneticconstraints make it likelythat closely proachto matchvocal similarity among Strep- related speciesare more similar in vocal pa- topeliadoves with their taxonomicclassifica- rameters than are distantly related species. tion. We usedphenetic analyses of the "perch- Featureslike body size (Wallschl•iger1980, Tu- coo" vocalization, and, as our taxonomic ref- baro and Mahler 1998)and syrinxmorphology erence, we used a clustering suggestedby (Gaunt 1988, Podos 1997) may impose limita- Goodwin (1983),based on the qualitativein- tions on soundproduction, and the auditory terpretationsof differencesamong speciesin sensitivityof receiversmay restrictthe char- morphologyand distribution. acteristics of communicative sounds to those of the species-specifichearing range (Dooling E-mail: [email protected]. nl 1982,Ryan et al. 1990).In addition,the evolu-

737 738 SLABBEKOORN,DE KORT,AND TEN CATE [Auk, Vol. 116 tionaryand ontogenetic flexibility of theneural pathwaysinvolved in vocal controland audi- hypopyrrha tory processingmay play an important role, orientalis potentiallyaffecting acoustic character diver- bitorquata gence and convergencevia constraints on sound productionand perception(Bass and Baker 1991, Smith 1994). For selection to act on behavioral traits, var- Streptopefia-- d•cipiens sernitorquata iance in the traits must have a heritable basis. capicola Dovesdevelop vocal characters independent of decaocto learning(Nottebohm and Nottebohm1971). Darwin (1868)reported that RockDoves (Co- PI picturata lumbalivia) could be artificially selectedfor -- cs - chinonsis temporalqualities of their coos(see also Bap- senegalensis tistaand Abs 1983).Vocalizations of hybridsof F•G. 1. Dendrogramreconstructed after Good- variousStreptopelia doves either resembled one win (1983) on the classificationof 16 Streptopeliaspe- parentaltype, were intermediatebetween the cies.The classificationis basedupon qualitativemor- parents,or resembledneither of the parents; phologicaland distributionalcharacteristics. TD = the degreeof disruptionin coocharacteristics turtle-dove group, RD = ringed dovegroup, PI S. increasedas the relatednessof the parental picturata,and CS S. chinensis/ S. senegalensis group. formsdecreased (Lade and Thorpe1964, Da- vies 1970, Baptista 1996). The aim of this paper is to examinea repre- necessarilyreflect common ancestry, but it is of sentativeset of recordingsfor all 16 speciesof interestto seewhether they show the same pat- Streptopeliadoves and to give a descriptive tern of clustering.Insight into how - overviewof theirperch-coo characteristics. The ary pathwaysled to thesimilarities and dissim- coos of Streptopeliaare relatively simple and ilaritiesin suchfeatures has to await a phylo- well suited for extensivequantitative analyses, geneticanalysis of the genusStreptopelia based as was shown for S. decaocto(ten Cate 1992;ten on an independentdata set. Cateand Ballintijn1996; Ballintijn and ten Cate 1997a,b). Speciesof Streptopeliashow little di- METHODS vergencein allometry,and their syrinx appears Subjects.--The16 speciesof Streptopelia,commonly to be anatomicallyrestricted in producingvar- referred to as "turtle-doves," are slender,relatively iation in frequencyrange or frequencymodu- long-tailed,gray or brownpigeons (Goodwin 1983). lation patterns(Warner 1972, Ballintijn et al. Neck or collarpatterns are the mostconspicuous ex- 1995).Gaunt (1988) remarkedthat evenbirds ternal plumagecharacteristics. The turtle dovesare with a relativelysimple syrinx morphology can native to temperateand tropicalregions of , produce rather differently structuredvocali~ Asia, and . Someof the speciesare expanding zationsby varyingthe rate and patternof air- their distribution in Pacific, Australian, and Nearctic flow. If this has affected vocal evolution in regionswhere they were introducedrelatively re- cently.Most speciesinhabit open woodlands and a Streptopeliaspecies, then onemight expectvo- wide rangeof humansettlements. They feed mainly cal differentiationin temporal featuresin par- on grainsand ,although the forest-dwelling ticular. speciesfrequently eat berries and other small fruits. We compared interspecific similarity in Streptopeliadoves are usually abundant within their perch-cooswith the taxonomicclustering pro- range.They are territorialbut oftenform large for- videdby Goodwin(1983). This is theprime ref- agingor sleepinggroups, particularly outside of the erencefor the taxonomyof doves(Howard and breedingseason. All speciesare strongfliers, and Moore 1991,Baptista et al. 1997) and is based someare long-distancemigrants. Goodwin (1983) divided the 16 speciesinto four on the qualitativeinterpretation of morpholog- groupsbased on appearanceand geographicdistri- ical characteristicsand geographicdistribution bution (seeTable 1). The four speciesof typical "tur- (seeFig. 1). Pheneticclassifications, like those tle doves"are darkerthan the otherspecies and have basedon plumagepattern (Goodwin 1983) or a mottledpattern on the backof their wingsand ei- acousticcharacteristics (this study), need not ther a black-and-whitestriped patch or a completely July1999] Perch-coosin Doves 739

TABLE1. Overviewof thegenus Streptopelia. A description of theneck pattern is givenbecause it is themost conspicuousfield characteristic.We alsolist brief descriptionsof geographicdistribution based on Good- win (1983) and Howard and Moore (1991).

Species Groupa Sizeb Neck pattern Distribution S. TD 26 to 28 Stripedpatch Europe,W Asia, N Africa S. lugens TD 28 to 31 Blackpatch E Africa, SW Arabia S.hypopyrrha TD 29 to 31 Blackpatch E Nigeria,W Cameroon S. orientalis TD 33 to 35 Stripedpatch Centralto E S. bitorquata RD 29 to 31 Blackcollar Philippines, S.decaocto RD 31 to 33 Blackcollar Europeto E China,India S. roseogrisea RD 29 to 30 Blackcollar SahelianAfrica S. reichenowi RD 26 to 28 Blackcollar S. Somalia,NE Kenya S. decipiens RD 28 to 30 Blackcollar Africa S of S. semitorquata RD 33 to 36 Blackcollar Africa S of Sahara S. capicola RD 25 to 28 Blackcollar S and E Africa S. vinacea RD 24 to 26 Blackcollar Africa S of Saharaand N of equator S. tranquebarica RD 22 to 24 Blackcollar SE Asia S. picturata PI 27 to 33 Side spots Madagascar S. chinensis CS 27 to 33 Spottedpatch SE Asia S. senegalensis CS 25 to 27 Throatspots Africa,,SW Asia "TD - turtle-dovegroup, RD - ringed dove group, PI = S. picturata,and CS - S. chinensis/S.senegalensis group. Basedon Goodwin (1983). bHead to-tail length(cm).

blackpatch at either sideof their neck.The largest 1997a).Variation in the perch-cooof individual Eur- group within the , the "ring-neckedturtle asianCollared-Doves (S. decaocto) is lessthan that be- doves" (or "ring doves"), consistsof nine species, tween individuals (ten Cate 1992), and differences eachof whichhas a moreor lessuniform gray plum- between individuals are consistent for the same ageand a characteristicblack collar around the neck. adult malesover time (Ballintijnand ten Cate 1997b). The other two groupsconsist of, respectively,one Cooassignment.--A "coo" bout consistsof a series species(S. picturata,Madagascar Turtle-Dove) and of soundelements that vary in frequencyand in tem- two species(S. chinensis[] and S. sene- poral and structuralfeatures. In sucha bout, we de- galensis[]). These speciesdiffer in fineda species-specificcoo as the smalleststereotyp- severalmorphological features such as pat- ic repetitionof similarelement sequences. Addition- tern and relativelength of the tail and wings.Good- al criteriawere (1) pausesbetween coos typically are win (1983)also suggested that within theturtle-dove longerthan pauses within coos;(2) the firstelement group,S. lugensand S.hypopyrrha are the productof of a boutis likelyto bethe firstelement of a coo;and the mostrecent species split, and theymay be more (3) the last elementof a bout is likely to be the last closelyrelated to S. turtur thanto S.orientalis. In the elementof a coo.These criteria were not always ap- ringed group, S. vinaceaand S. capicolaare assumed plicablesimultaneously, e.g. because the bout ended to be monophyletic,as are S. bitorquata,S. decaocto, in the middle of a coo,leading to contradictingcri- and S. roseogrisea. teria. However, all dubiouscoo assignmentswere re- Recordings.--Mostof the recordingswere made in solved after analysesof multiple bouts and multiple privateaviaries and zoologicalgardens in the Neth- individualsper species.One cooin the middle of a erlandsand Belgium.These were supplementedby bout was chosen for detailed measurements. This is recordingsof free-livingdoves in the Netherlands, the least-variablepart of a bout in S. decaocto,for ,Uganda, Cameroon, and Ghana.The final which the perch-coohas been studiedextensively collectionwas completed with somerecordings from (M. R. Ballintijn pers. comm.). the National Sound Archive in London. The record- Measurements.--Therecordings were digitized at a ingsof dovesin aviarieswere madewith a Sennheis- sample rate of 11 kHz using Canary 1.1 software on er MKH P48 microphoneand a SonyTC-D5 Pro re- a Macintosh computer.We applied Fast Fourier corder,using TDK AD90 tapes.Some of the record- transformationsto createsonagrams for quantitative ings of free-living doveswere made with the same measurementswith a temporalresolution of 1.44ms equipment,and somewith a SonyF-V9 microphone and a frequencyresolution of 10.87 Hz. First, the with a SonyTCM-S66V recorder. We assumed that all number of coos in a bout and the number of elements individuals recorded were males, because males are in a coowere determined. Deviations in generalbout the mostvocal sex,and femalesproduce distinctive structure were marked. Duration was determined for coosthat differ from thoseof malesin variousways all elements(El, E2, E3, etc.) and pausesbetween el- (Goodwin 1983,Cramp 1985,Ballintijn and ten Cate ements(P1, P2, P3, etc.). Other temporalmeasure- 740 $LABBEKOORN,DEKORT, AND TEN CATE [Auk, Vol. 116

S. decipiens

S. lugens S, semitorquata

0.7 1.4 2.1 2.8 3.5 S. hypopyrrha S. capicola

S. orientaris S. vinacea

S. bitorquata S. tranquebarica

S. decaocto S. picturata

S. roseogrisea S. chinensis

1.5- S. reichenowi S. senega•ns• 1.0- 0.54- 0.0 0.5 1.0 1.5 2.0 2.5 015 1.0 1.5 2.0 2.5 Time (s) F•c. 2. Sonagramsof an exampleof one perch-coofor eachspecies. The sonagramswere generatedby FastFourier Transformations using Canary 1.1 software.The frequencyresolution is 10.87Hz and the tem- poral resolutionis 1.4 ms. mentswere: total coolength (COLE), definedas the quencyrange (S. turtur in Fig. 2). The absoluteam- durationfrom the start of the first element(E 1) to the plitude was not used as a parameterbecause of var- end of the final element (EX); coo interval (INT), de- iation in recordingdistance and orientationtoward fined as the duration from the end of EX to the start the microphone. of thenext El; andsound percentage (SOPE), defined Statisticalanalyses.--Phenetic analyses comprise asthe summedduration of the separateelements di- methodsto form groupsbased upon overall similar- vided by COLE.Frequency measurements included ity amongentities, in contrastto phylogeneticanal- overallpeak frequency (PEFR), which is thefrequen- ysisthat are based on evolutionary relationships (see cy with the highestamplitude in the coo;maximum Felsenstein 1982, Sokal 1986, de Queiroz and Good frequency(MAFR), whichis the highestfrequency 1997).The methodsin usefor pheneticanalyses can present in the coo; minimum frequency (MIFR), be classifiedinto ordinationand clusteranalysis. Or- whichis the lowestfrequency present; and peakfre- dination arranges entities in a continuousmultidi- quenciesfor the individual elements(PF1, PF2, PF3, mensionalspace defined by the measuredparame- etc.). Furthermore, the structure of the elementsof a ters, and clusteranalysis assigns entities to groups coo was categorizedas a "bow,.... noise," or "trill" (seeJames and McCulloch1990). We usedboth meth- type. A "bow" structurewas definedas a relatively odsin this study. tonal elementwith a gradualincrease in frequency We applied a canonical discriminant function in the firsthalf and a gradualdecrease in frequency analysis(DFA) on the acousticparameters of indi- in the secondhalf of the element (see S. decaoctoin viduals. Theseparameters were averagesfor all the Fig. 2). A "noise"structure was definedas an irreg- bouts of each individual. The individual measures ular patternof intensitychanges in a relativelybroad were used to test whether the individuals could be frequencyrange, which leadsto a hoarsesound (S. correctlyclassified by species.Subsets of acoustic orientalisin Fig. 2). A "trill" structurewas definedas characterswere analyzedto determinewhether fre- a regular pattern of rapidly alternating phasesof quencyaspects and temporalfeatures classified the high and low amplitudefor a relativelybroad fre- species'averages and individual measuresequally July1999] Perch-coosin Doves 741

TABLE2. Generaldescription of perch-coosfor 66 individualsof 16 speciesof Streptopelia.N = numberof individualsrecorded; Nr = reducedsample size for measurementsof EX for turtur,lugens, and chinensis and of E1 for senegalensis;n = numberof boutsanalyzed (range 1 to 5 per individual); BL = averagebout length;NOEL = averagenumber of elementsper coo;structure = categoricaldescription of elementstruc- turespresent. S. risoriais the domesticatedform of S. roseogrisea.

Species N N, n BL NOEL Structure S. turtur 5 2 21 4.0 3.4 Trill S. lugens 4 3 4 5.0 3.7 Noise S.hypopyrrha 1 1 3 5.7 2.0 Trill S. orientalis 5 5 16 4.8 4.0 Noise S. bitorquata 1 1 2 15.5 3.0 Trill S. decaocto 5 5 25 6.4 3.0 Bow S. risoria 5 5 21 5.4 2.0 Bow and trill S. reichenowi 1 1 5 7.4 2.0 Bow and trill a S. decipiens 5 5 17 9.4 3.0 Bowb S. semitorquata 5 5 18 7.0 6.0 Bow S. capicola 5 5 10 13.6 3.0 Bow and trill S. vinacea 5 5 20 18.2 3.0 Bow S. tranquebarica 5 5 22 8.3 3.8 Noise S. picturata 4 4 8 4.6 2.0 Noise S. chinensis 5 2 21 3.3 3.4 Bow and trill c S. senegalensis 5 2 16 12.2 6.2 Bow Coo characteristics show a switch in the bout. •Bout startswith a cry. Coosproduced independently (see text for furtherexplanation). well. The parameterset usedfor the DFA consisted gramwith hierarchicalnonoverlapping groups. The of continuous variables that could be measured in all magnitudeof the distortionof theoriginal similarity species.Measurements concerning the coo as a relationshipsby the clusteringprocedure is reflected whole were incorporated:COLE, INT, SOPE,PEFR, in the copheneticcorrelation coefficient (Sokal and MAFR, and MIFR togetherwith additionalmeasure- Rohlf 1962).The parameterset used for clusteranal- ments on the first and last element of the coo: El, ysisconsisted of thecontinuous variables used in the PF1, EX and PFX. The measurements of other ele- DFA and PCA supplementedby the numberof ele- ments were not incorporatedbecause interspecific mentsin a coo (a multivariate discretevariable) and comparisonof an elementsuch as E3 makesno sense the binary-codedcategorical variables noise, bow, if onespecies usually utters three elements, whereas and trill. All parameterswere Z-transformedto the otherusually utters six elements.Measurements makethem contribute equally to the similaritycom- of the first and last elementswere incorporatedbe- parison.Cluster analyses were performedwith the causestart and end of the coo are unambiguous total set of 16 species,with the four groups(using acrossspecies. We further explored the data set using group averagesgenerated via species'averages), a principalcomponents analysis (PCA) of the spe- with only the turtle-dovesubset of four species,and cies'averages to reducethe set of 10 measurements with only the ringeddove subset of nine species.All into uncorrelatedmultivariate components. We used analyseswere performedusing SPSS/PC+ software. ANOVA and post-hoccomparisons to test whether the speciesfrom Goodwin's(1983) putative taxo- nomicgroups (turtle-doves, ringed doves,S. pictur- RESULTS ata,and S.chinensis / S. senegalensis ) differed from one anotherin thesecomponents. Wegathered recordings for all 16species (Ta- Subsequently,we used cluster analysesfor de- ble 2). The total setof recordingsanalyzed con- tailed comparisonof interspecificacoustic similarity sisted of 229 bouts from 66 individuals. For 11 with the similarity-basedtaxonomic classification of of the 16 species,data were obtainedfor five Goodwin(1983). A hierarchicalcluster analysis was different individuals. The number of recorded performed using an average linkage method boutsper individual rangedfrom one to five. (UPGMA),based on a dissimilaritymatrix, for which we usedEuclidean distances (Romesburg 1984). The Some spedes were less common in captivity, which led to a lower number of recorded indi- degree of similarity betweenperch-coos is repre- sentedby the proximityof the horizontalbranches viduals.We recordedonly two individualsof S. in the dendrogram,with the mostsimilar coos clos- roseogriseabut collecteddata on five individu- est to one another The methodproduces a dendro- als of S. risoria, which is the domesticated form 742 SLABBEKOORN,DEKOI•T, AtqI• TEN CATE [Auk, Vol. 116

TABLE3. Temporaland frequencymeasurements of completeperch-coos (• +_SD). COLE = coolength; INT = interval to start of next coo;SOPE = soundpercentage (summed duration of elementsdivided by COLE); PEFR = peak frequency;MAFR = maximum frequency;MIFR = minimum frequency.See Table 2 for sam- ple sizes.S. risoriais the domesticatedform of S. roseogrisea.

Species COLE (ms) INT (ms) SOPE(%) PEFR(Hz) MAFR (Hz) MIFR (Hz) S. furfur 2,399 -+ 478 643 +- 10 70.8 -+ 6.3 579 _+ 28.5 767 _+ 30.8 334 -+ 46.3 S. lugens 3,438 + 561 616 _+15 67.8 + 5.6 463 -+ 80.1 638 _+67.5 227 + 84.7 S. hypopyrrha 1,986 743 77.4 547 872 283 S. orientalis 2,088 + 197 334 +_ 6 68.1 +_ 5.3 462 +_ 17.3 681 _+ 77.1 241 _+ 12.3 S. bitorquata 1,945 861 44.1 494 655 345 S. decaocto 1,202 -+ 58 419 + 14 70.1 + 3.7 554 +- 32.0 807 +- 32.2 374 -+ 32.8 S. risoria 1,616 _+ 179 879 +- 18 82.6 + 5.5 596 +- 24.5 822 +- 17.5 387 -+ 42.5 S. reichenowi 373 172 81.5 543 785 306 S. decipiens 660 _+122 562 +_9 76.9 +_5.9 467 _+31.0 663 + 46.4 276 +_49.2 S. sernitorquata 1,449 +_49 358 +- 2 52.6 -+ 4.6 489 ñ 22.9 694 + 72.4 315 _+34.6 S. capicola 884 _+74 356 -+ 2 62.6 +_3.0 902 +- 62.5 1,142 -+ 84.0 602 _+65.7 S. vinacea 433 ñ 31 280 ñ 5 75.9 -+ 3.3 795 +- 54.0 1,037 +- 56.6 552 _+26.0 S. tranquebarica 637 + 116 126 +_2 75.8 +- 3.8 521 +- 24.2 694 _+31.0 357 -+ 31.3 S. picturata 1,155 _+409 591 _+15 80.4 +_6.4 367 +_48.3 507 _+63.3 207 _+10.0 S. chinensis 1,141 +_262 2,168 +_ 1,700 70.2 ñ 9.8 715 _+53.1 955 _+48.2 446 +_47.6 S. senegalensis 1,153 _+92 1,438 +_590 65.3 +_7.5 621 +_19.0 868 _+46.7 341 _+42.5 of this species.Because the vocal distinctions speciesare summarizedin Table2. Fourspecies betweenS. roseogriseaand S. risoriawere insig- produced two different element types within nificant,we usedthe setof S. risoriarecordings their coo:S. risoria,S. reichenowi,S. capicola,and for further analyses.For three specieswe ob- S. chinensis;all producedbow and trill type el- tained datafor oneindividual only.Because the ements. Peculiarities in bout structure were data for the other speciesindicated a smaller foundfor S. reichenowiand S. decipiens.Strepto- degree of intraspecificcompared with inter- pelia reichenowiproduced predominantly two- specificvariation, we included these data to element coos but switched to one-element coos provide a completeoverview of all species. at the end of the bout; S. decipiensstarted each Nevertheless,inferences concerning these spe- perch-coobout with a cry from its species-spe- cies should be considered tentative. cific repertoire (also used in other contexts). Descriptionof coo characteristics.--The general Both speciesproduced consecutiveelements perch-coocharacteristics of the 16 Streptopeliathat occasionallywere connectedby a low-fre-

TABLE4. Temporalmeasurements of perch-coosof 16 speciesof Streptopelia.Durations of elementsare cod- ed as E1 to E7, and durations of pausesare coded as P1 to P6 (values are ß _+SD in ms). See Table 2 for samplesizes. S. risoriais the domesticatedform of S. roseogrisea.

Species E1 P1 E2 P2 E3 P3 S. turtur 674 _+ 182 320 _+ 60 352 +_ 126 314 + 178 505 _+ 28 217 _+ 143 S. lugens 779 +_62 352 +_33 627 +_55 481 _+121 592 ñ 104 430 +_12 S. hypopyrrha 688 437 849 ------S. orientalis 248 _+ 160 180 +_ 59 392 + 140 257 +_ 88 333 _+ 14 237 + 102 S. bitorquata 216 379 477 702 165 -- S. decaocto 248 -+ 25 79 -+ 19 403 + 27 286 +_ 47 192 + 59 -- S. risoria 215 _+ 67 276 _+88 1,123 _+ 171 ------S. reichenowi 71 63 233 ------S. decipiens 143 + 31 98 +_13 148 _+40 53 ñ 35 226 _+79 -- S. sernitorquata 113 ñ 43 66 +- 31 150 -+ 44 176 _+124 94 _+20 88 ñ 26 S. capicola 97 +- 19 272 _+42 399 ñ 49 65 +- 32 57 -+ 13 -- S. vinacea 137 ñ 19 63 +_ 4 101 _+ 15 30 _+ 7 91 + 20 -- S. tranquebarica 159 + 25 66 _+10 122 ñ 25 47 -+ 25 80 +- 88 45 -+ 1 S. picturata 142 _+41 252 -+ 105 768 -+ 266 ------S. chinensis 73 + 13 152 + 72 186 + 22 139 _+ 35 485 _+ 72 263 _+ 53 S. senegalensis 92 _+5 48 +_49 109 +_28 181 _+36 151 _+68 73 _+18 July1999] Perch-coosin Doves 743 quency sound. We based their temporal and the first principal component(PC1) accounted frequencymeasurements on the separateele- for 53.3% of the variance,representing mostly ments to keep the measurementscompatible the five frequencymeasures. The temporalfea- with otherbouts of the sameor other conspe- tures were representedin three uncorrelated cifics.An overviewof the species-typicalperch- components;PC2 was determined mostly by coosis presentedin Figure2, and detailedmea- the duration measurements(COLE, El, and to surementsof temporal and spectralstructure a lesserdegree EX), PC3mostly by the duration of the coos are listed in Table 3. Measurements of elementsrelative to the durationof the paus- of the duration of individual elements are listed es between them (SOPE and EX), and PC4 in Table4, and peak frequenciesfor individual mostly by intervals between coos (INT). PC2, elements are in Table 5. For all measurements, PC3, and PC4 accounted for an additional 17.3, we calculated the individual averagesfirst, 14.2,and 10.2%of the variance.The 16 species whichwere then used to calculateaverages and are depictedin two bivariate plots that show standarddeviations for eachspecies their respectivevalues for PC1and PC2 (Fig. 3), Discriminantand principalcomponents analy- and PC3 and PC4 (Fig. 4). We usedANOVA to ses.--TheDFA using the individualaverages of test whether the taxonomicgroups differed in the 10 measurements led to a 100% correct clas- any of the four components.The groupsdif- sificationof the individuals into species.The fered for PC2 (P = 0.0014) and PC4 (P = first of the five discriminant functions(DF1) ac- 0.0011),and both were significantafter Bonfer- counted for 64.8% of the variation, and the sec- roni correction.The turtle-dovegroup differed ond (DF2) accountedfor 20.4%of the variation. significantly(P < 0.05) from the ringed dove Classificationbased solely on the five frequen- and the S. chinensis/S.senegalensis groups for cy parametersled to 64.3%correct assignment, PC2 (Tukey'spost-hoc comparisons). The chi- whereas classification based on the five tem- nensis/senegalensisgroup differed significantly poral featuresled to 94.6%correct assignment. (P • 0.05)from the turtle-doveand ringed dove Some individuals showed variation in the num- groupsfor PC4.The S.picturata group consists ber of elements.In particular, someS. senega- of only one species,which makesstatistical lensis and S. chinensis lacked the first or last el- testing impossible;however, S. picturatawas ement,respectively. This led to missingvalues, well outsidethe range of all other speciesfor reducingthe samplesize to 56 individuals. PC1. The results of the PCA on the 10 measure- Clusteranalysis.--The cluster procedure led ments are summarized in Table 6. Four com- to the dendrogramdepicted in Figure 5A. ponentswith eigenvalues>1 were extracted Someof the speciesthat are closelyrelated ac- from the data set. After varimax rotation to cording to Goodwin (1983) are clustered as maximize the correlationamong parameters, nearestneighbors (S. capicolaand S. vinacea).

TABLE 4. Extended.

E4 P4 E5 P5 E6 P6 E7

395 • 41 ...... 469 + 30 ......

441 ñ 83 ......

105 + 25 267 • 130 155 + 58 91 ñ 26 144 + 49 -- --

141 ñ 19 ......

188 ñ 25 ...... 151 ñ 56 39 ñ 24 149 ñ 26 73 ñ 13 112 ñ 21 79 ñ 30 101 ñ 25 744 SLABBEKOORN,DEKORT, AND TEN CATE [Auk, Vol. 116

TABLE5. Frequencymeasurements of perch-coosof 16 speciesof Streptopelia.Peak frequencies of elements are codedas PF1 to PF7 (œ+ SD in Hz). SeeTable 2 for samplesizes. S. risoriais the domesticatedform of S. roseogrisea.

Species PF1 PF2 PF3 PF4 PF5 PF6 PF7 S. turtur 542 +_ 12.8 576 + 21.8 582 + 26.1 560 ñ 38.0 ------S. lugens 431 _+38.2 424 + 39.2 464 + 80.1 500 _+107.6 ------S. hypopyrrha 547 540 ..... S. orientalis 480 ñ 30.5 478 + 22.2 421 + 59.5 451 _+ 21.5 ------S. bitorquata 494 489 494 ñ 7.7 .... S. decaocto 579 _+ 39.5 541 ñ 29.6 530 + 38.4 .... S. risoria 607 ñ 30.2 591 + 16.5 ..... S. reichenowi 585 543 ..... S. decipiens 482 ñ 56.7 460 + 39.7 468 _+43.5 .... S. semitorquata 489 ñ 17.2 500 + 31.3 477 ñ 22.7 482 + 32.5 472 ñ 54.6 472 + 72.0 -- S. capicola 911 ñ 56.1 881 + 70.0 814 ñ 107.5 .... S. vinacea 819 ñ 62.5 808 ñ 57.7 775 + 41.3 .... S. tranquebarica525 ñ 24.6 522 + 24.4 512 + 32.9 515 + 28.5 ------S. picturata 266 + 28.8 368 ñ 48.3 ..... S. chinensis 638 ñ 78.5 719 + 58.2 689 ñ 48.9 678 + 36.5 ------S. senegalensis 522 ñ 46.1 456 ñ 89.7 577 + 61.0 615 ñ 32.9 632 +_19.5 635 ñ 23.6 595 + 47.2

However,most of the suggestedintragroup re- efficientof 0.8 or higheris consideredto indi- lationships,and even the intergroupsegrega- catea usefultree, resembling the real dissimi- tion, are absentin the dendrogrambased on larities (Romesburg1984). The coefficientsfor acousticsimilarities. The dendrogramof group the dendrogramsof all 16 species,of the four averagesshows ringed doves and the S. chilen- groups,of the four turtle-dovespecies, and of sis/S. senegalensisgroup as nearestneighbors, the nine ringed dove specieswere 0.60, 0.74, which are subsequentlymore similar to turtle- 0.93,and 0.94,respectively. dovesthan to S.picturata (Fig. 5B).Cluster anal- ysesbased on subsetsof parameters(either fre- DISCUSSION quencyor temporal features)led to identical linkagepatterns. The within-groupsimilarities The 16 speciesof dovesin the genusStrep- for turtle-dovesand ringed dovesled to the topelia,some of whichare very similarin plum- dendrogramsin Figures5C and D. Cophenetic age pattern, all produce distinctivespecies- correlation coefficients were calculated for all specificperch-coos. All individualswere cor- four dendrograms.This is an indexfor how ac- rectly classifiedinto speciesin a DFA, which curatelythe original data in the dissimilarity meansthat intraspecificvariation was limited matrix is representedby the dendrogram.A co- relative to interspecificvariation for the total

TABLE6. Factorloadings of the 10 acousticparameters on the four principalcomponents after varimaxro- tation to maximizethe correlationamong the parameters.Eigenvalues and amountof varianceexplained by the respectivecomponents are given at bottom of table.

Acousticparameter PC1 PC2 PC3 PC4 Highestfrequency 0.968 -0.104 0.015 0.113 Lowestfrequency 0.929 -0.245 -0.148 0.003 Overall peak frequency 0.977 -0.153 0.066 0.097 Peakfrequency of element1 0.972 -0.126 -0.068 -0.121 Peak frequencyof element9 0.985 - 0.100 - 0.032 0.090 Total coolength 0.267 0.929 -0.142 0.110 Duration of element 1 -0.097 0.936 0.147 -0.095 Duration of element 9 0.263 0.467 0.726 0.141 Coo interval 0.093 0.009 -0.031 0.989 Soundpercentage 0.016 -0.155 0.946 -0.112 Eigenvalue 5.328 1.728 1.421 1.024 % Varianceexplained 53.3 17.3 14.2 10.2 July 1999] Perch-coosin Doves 745

2.5- A 0 5 10 15 20 25

turtur 1.5- hypopyrrha orientalis

picturata PC 2 .5- orienlalis oria sernitorquata

0.0 - cap,cola bitorquatasenegalensis seml[orquata turtur hypopyrrha -.5- p.cturata vinacae dsofia -1.0 - dec•p•ens lugens reichenowi capicolareichenowi -1.5 chinensisvinacea PC 1

F•G.3. Plotof the 16 speciesof $trcptopctiadoves B o 5 10 15 20 25

in two-dimensionalspace defined by two prindpal RD components.PC1 is determinedmostly by the five frequencymeasurements, and PC2 mostly by dura- TDcs I J PI tionmeasurements (COLE, E1 andto a lesserdegree EX). SeeTable 6 for factorloadings. Polygons connect and encloseall specieswithin a taxonomicgroup. 0 5 10 15 20 25

tu•ur

set of acousticparameters. The clusteringof hypopyrrhalugens J J speciesusing the acoustic data set showed little oden•lis -•J congruencewith the taxonomicclustering of Goodwin(1983), which is based on morpholog- ical characteristicsand geographicdistribu- D 0 5 10 15 20 25 tion.This indicates that evolutionary differen- decipiensJ tiationin plumagepattern does not necessarily reichenowJ coincide with differentiation in acoustic char- sernitorquata tranquebaffca acteristics.However, our analysisdoes not con- bitorquata fisoda capicola

3 FIG. 5. Dendrogramscreated by hierarchical clusteranalyses of thefour Streptopelia groups using the UPGMA average-linkagemethod, based on Eu- clidiandistances. Shown are (A) all 16 species;(B) 1- rlsona group averages;(C) turtle-dove group; and (D) PC 4 ringeddove group. RD = ringeddove group, CS = 0 S. chinensis/S.senegalensis group, TD = turtle-dove semRorqu decipiens group, and PI = S. picturata. -1 tranquebarica tradict the subdivision into four taxonomic

-3 -2 -1 0 1 groups.Two of the four groups(turtle-doves PC 3 andchinensis/senegalensis) differed significant- ly fromthe other groups in oneof thecompo- FIG. 4. Plot of the 16 Streptopeliain two-dimen- sionalspace defined by two principalcomponents. nents of the PCA(PC2 and PC4, respectively). PC3 is determinedmostly by the durationof ele- Thecluster analysis with theturtle-dove subset mentsrelative to theduration of the pausesbetween of four species(Fig. 5C), and with the ringed them(SOPE and EX) andPC4 by intervalsbetween dovesubset of nine species(Fig. 5D), led to coos(INT). SeeTable 6 for factorloadings. Polygons dendrogramswith high copheneticcorrelation connectand encloseall specieswithin a taxonomic coefficients,which indicatesthat they ade- group. quatelyreflect similarities among species. This 746 SLABBEKOORN,DEKORT, AND TEN CATE [Auk,Vol. 116 makes them suitable for future evaluation of ship betweenbody size and vocalizationfre- evolutionarytransitions using a phylogeny quency(Wallschl•iger 1980, Ryan and Brenow- based on an independentdata set. Such an itz 1985, Tubaro and Mahler 1998). On the oth- analysismay shedlight on intriguing situa- er hand, S. tranquebarica,which is the smallest tionssuch as in the turtle-dovegroup. Here, S. species(see Table 1), used a surprisinglylow turtur and S. orientalisare closelymatched in frequencyrange. However,the range of fre- appearance,as are S. hypopyrrhaand S. lugens. quenciesused by all 16 specieswas relatively Nevertheless,acoustically, S. turtur is much small,especially compared with songbirdsfor more similar to S. hypopyrrha,and S. orientalisis which frequency-dependentattenuation may muchmore similarto S. lugens. causedifferential selectionpressures in differ- Vocal differentiation among Streptopeliaent (see Morton 1975, Hunter and dovesconcerns temporal features in particular. Krebs1979, Wiley and Richards1982, Badyaev The successwith whichwe correctlyassigned and Leaf 1997). Whether the variation in tem- individuals into specieswas dramaticallyre- poral and frequencymeasurements of Strepto- duced, only 64.3% correct,when we used the peliavocalizations is meaningful to the doves set of frequencyparameters only. If we used themselvesawaits investigationsbased on the set of temporalfeatures only, the analysis playbackexperiments in a natural context(see still correctlyclassified most (94.6%) of the spe- Slabbekoornand ten Cate 1996, 1997, 1999). cies.The PCA led to one componentmostly representingfrequency measures and three un- ACKNOWLEDGMENTS correlated componentsmostly representing temporalmeasurements. The bivariateplots of We are grateful to Sarah Collins, FrietsonGalis, Marcovan Veller,and StephenRothstein for helpful the four principalcomponents clearly showed commentson earlierversions of themanuscript, and the parametersfor whichthe groupsdiffered, to Martin Brittijn, Gerdie Dijkstra, and Evert Meelis andthey indicated that each group can be char- for technicalassistance. Martin Poot, Jacques van Al- acterizedby distinctivetemporal aspects.The phen,George Sangster, and the NationalSound Ar- DFA and PCA resultsmay be due to anatomical chive,London supplied us with additionalrecord- restrictionsin producingvariation in frequency ings.We thank G. Bruins,A. Cleynen,J. Dekkers,G. range or frequencymodulation patterns (War- Ketelslegers,W. Moed, W. Silkens,and G. van Bael ner 1972,Ballintijn et al. 1995).The relatively from the "Wilde duiven" club, Belgium/theNeth- simplesyringeal morphology of dovesmay be erlands,and Avifauna,Alphen aan de Rijn, who al- lowed us to observeand make recordingsin their bettersuited to varyingthe rate and patternof dovecollections. This study was supported by a SLW airflow (Gaunt 1988, ten Cate and Ballintijn projectgrant (SLW-805.30.043). 1996).The neural basisfor suchvariation in vo- cal controlmight be lessconservative than sy~ LITERATURE CITED ringeal morphology.The acousticparameters that best discriminateamong the StreptopeliaBADYAEV, A. V., AND E. S. LEAF. 1997. asso- dovesare in line with a study of quail (Cotur- ciationsof song characteristicsin Phylloscopus nix) vocalizationsin which temporalparame- and Hippolaiswarblers. Auk 114:40-46. ters also yielded the best discriminationbe- BALLINTIJN,M. g., AND C. TEN CATE. 1997a. Sex dif- ferencesin the vocalizationsand syrinxof the tween species(Collins and Goldsmith 1998). Collared Dove (Streptopeliadecaocto). Auk 114: McCrackenand Sheldon(1997) concluded from 22-39. their comparativestudy of heronvocalizations BALLINTIJN, M. g., AND C. TEN CATE. 1997b. Vocal de- thatthe number of syllables,syllable structure, velopmentand its differentiationin a non-song- and fundamentalfrequency were the mostphy- : The Collared Dove (Streptopeliadecaocto). logeneticallyinformative parameters, whereas Behaviour 134:595-621. frequency measurementsconcerning higher BALLINTIJN,M. g., C. TEN CATE, E W. NUIJENS,AND rangeswere more subject to habitat-dependent H. BERKHOUDT.1995. The syrinxof the Collared Dove (Streptopeliadecaocto): Structure, inter-in- convergence. dividual variation and development.Nether- Thesmaller species (S. vinacea and S. capicola) landsJournal of Zoology45:455-479. usedhigher frequencies, and the largerS. pic- BAPTISTA,L. F. 1996.Nature and its nurturing in avi- turataused relativelylow ones(Tables 3 and 5). an vocal development.Pages 39-60 in Ecology This agreeswith otherstudies on the relation- and evolution of acoustic communication in July1999] Perch-coosin Doves 747

(D. E. Kroodsma and E. H. Miller, Eds.). ssp.)with referenceto their taxonomicstatus. CornellUniversity Press, Ithaca, New York. Emu 69:67-71. BAPTISTA,L. E, ANDM. ABS.1983. Vocalizations. Pag- HOWARD, R., AND A. MOORE. 1991. A complete es309-325 in Physiologyand behavior of the pi- checklist of the birds of the world, 2nd ed. Ac- geon (M. Abs, Ed.). AcademicPress, New York. ademic Press, London. BAPTISTA, L. E, W. 1. BOARMAN, AND P. KANDIANIDIS. HUNTER,M. L., ANDJ. R. KREBS.1979. Geographical 1983. Behaviorand taxonomicstatus of Gray- variation in the songof the Great Tit (Parusma- son's Dove. Auk 100:907-919. jor) in relation to ecologicalfactors. Journal of BAPTISTA, L. E, P. W. TRAIL, AND H. M. HORBLIT. Ecology48:759-785. 1997. (Pigeons and Doves). ISLAM,K., AND J. A. CRAWFORD.1996. A comparison Pages60-111 in Handbookof the birds of the of four vocalizationsof the genus Tragopan world, vol. 4. Sandgrouseto . (J. del (Aves,). Ethology 102:481-494. Hoyo, A. Elliott, and J. Sargatal,Eds.). Lynx Fd- JACKSON,G. L., AND t. S. BASKETT.1964. Perch-coo- icions,Barcelona, Spain. ing and other aspectsof breedingbehavior of BASS,A., AND R. BAKER.1991. Evolution of homolo- Mourning Doves.Journal of Wildlife Manage- gous vocal control traits. Brain Behaviorand ment 28:293-307. Evolution 38:240-254. JAMES,F. C., AND C. E. MCCULLOCH.1990. Multivar- BECKER,P. H. 1982. The coding of species-specific iate analysisin ecologyand systematics:Pana- characteristicsin bird sounds.Pages 213-252 in ceaor Pandora'sbox? Annual Reviewof Ecology Acousticcommunication in birds, vol. 1 (D. E. and Systematics21:129-166. Kroodsma, E. H. Miller, and H. Ouellet, Eds.). LADE, B. I., AND W. H. THORPE.1964. Dove songs as Academic Press, New York. innately coded patternsof specificbehaviour. COLLINS,S. A., AND A. R. GOLDSMITI-L1998. Individ- Nature 202:366-368. ual and speciesdifferences in quail crows (Co- MCCRACKEN, K. G., AND F. H. SHELDON. 1997. Avian turnix c. japonica,C. c. coturnixand a hybrid). vocalizationsand phylogeneticsignal. Proceed- Ethology 104:977-990. ings of the National Academyof SciencesUSA CRAMP,S. (Ed.). 1985. The birds of the western Pa- 94:3833-3836. learctic, vol. IV. Oxford University Press,Ox- MILLER,D. B. 1978. Species-typicaland individually ford. distinctive acoustic features of crow calls of Red DARWIN,C. 1868. The variation of plants and ani- JungleFowl. Zeitschriftffir Tierpsychologie47: mals under domestication.J. Murray, London. 182-193. DAVIES, S. J. J. E 1970. Patterns of inheritance in the MILLER, E. H. 1982. Character and variance shift in bowing display and associatedbehavior of some acoustic signals of birds. Pages 253-295 in hybrid Streptopeliadoves. Behaviour 36:187-214. Acoustic communication in birds, vol. 1 (D. E. DAVIES,S. J. J. F. 1974. Studies on the three coo-calls Kroodsma, E. H. Miller, and H. Ouellet, Eds.) of the male . Emu 74:18-26. Academic Press, New York. DE QUEIROZ,K., AND D. A. GOOD. 1997. Phenetic MILLER,E. H. 1996.Acoustic differentiation and spe- clusteringin biology: A critique.Quarterly Re- ciationin shorebirds.Pages 241-257 in Ecology view of Biology72:3-30. and evolution of acoustic communication in DOOLING,R. J. 1982. Auditory perceptionin birds. birds (D. E. Kroodsma and E. H. Miller, Eds.). Pages 95-130 in Acoustic communication in Cornell University Press,Ithaca, New York. birds, vol. 1 (D. E. Kroodsma, E. H. Miller, and MILLER, E. H., W. W. H. GUNN, AND B. N. VEPRIN- H. Ouellet, Eds.). Academic Press,New York. TSEV.1988. Breeding vocalizationsof Baird's FELSENSTEIN,J. 1982. Numerical methods for infer- SandpiperCalidris bairdii and related species, ring evolutionarytrees. Quarterly Review of Bi- with remarkson phylogenyand adaptation.Or- ology 57:379-404. nis Scandinavica 19:257-267. GAUNT,A. S. 1988.Interaction of syringealstructure MORTON,E. S. 1975. Ecologicalsources of selection and airflow in avian phonation.Pages 913-924 in on avian sounds. American Naturalist 109:17- Acta XIX CongressusInternationalis Ornitholo- 34. gicus (H. Ouellet, Ed.). Ottawa, Ontario, 1986. NELSON,D. A. 1989.The importanceof invariantand National Museum of Natural Sciences, Ottawa. distinctivefeatures in speciesrecognition of bird GOODWIN,D. 1983.Pigeons and dovesof the world. song. Condor 91:120-130. BritishMuseum of Natural History,London. NOTTEBOHM,E, AND M. E. NOTTEBOHM.1971. Vocal- GO'rTINGER, H.-R. 1970. Zur Evolution von Verhal- izations and breeding behavior of surgically tensweisenund Laut•iuserungenbei Prachtfin- deafenedRing Doves,Streptopelia risoria. Ani- ken (Estrildidae).Zeitschrift ffir Tierpsychologie mal Behaviour 19:313-327. 27:1011-1075. PAYNE,R. B. 1986.Bird songsand avian systematics. HARRISON,C. J. O. 1969. Some comparative notes on Current 3:87-126. the Peacefuland Zebra Doves (Geopeliastriata PODOS,J. 1997.A performanceconstraint on theevo- 748 SLABBEKOORN,DEKORT, AND TEN CATE [Auk, Vol. 116

lution of trilled vocalizationsin a songbirdfam- SOKAL,R. R., ANDR. J.ROHLF. 1962. The comparisons ily (Passeriformes:Emberizidae). Evolution 51: of dendrogramsby objectivemethods. Taxon 11: 537-551. 33-40. ROMESBURG,H. C. 1984. Cluster analysis for re- TENCATE, C. 1992.Coo types in the CollaredDove, searchers.Lifetime learning publications,Bel- Streptopeliadecaocto: One theme,distinctive var- mont, California. iations. Bioacoustics 4:161-183. RYAN,M. J., AND E. •a•.BRENOWITZ. 1985. The role of TEN CATE, C., AND M. R. BALLINTIJN.1996. Dove coos body size,phylogeny, and ambientnoise in the and flashlights: The interruptabilityof "song" evolutionof bird song.American Naturalist 126: in a non-songbird.Journal of ComparativePsy- 87-100. chology110:267-275. RYAN, M. J., J. H. Fox, W. W1LCZINSK1,AND •a•. S. TUBARO, P. L., AND g. MAHLER. 1998. Acoustic fre- RAND.1990. Sexual selection for sensoryexploi- quenciesand body massin New World doves. tationin the frogPhysalaernus pustulosus. Nature Condor 100:54-61. 343:66-67. WALLSCHL•GER,D. 1980. Correlationof song fre- SLABBEKOORN,H., AND C. TENCATE. 1996. Respons- quencyand body weight in passerinebirds. Ex- esof CollaredDoves to playbackof coos.Behav- perientia 36:412. ioural Processes 38:169-174. WARNER,R. W. 1972.The syrinx in the family Col- SLABBEKOORN,H., AND C. TENCATE. 1997. Stronger umbidae.Journal of Zoology(London) 166:385- territorial responsesto frequency modulated 390. coos in Collared Doves. Animal Behaviour 54: WILEY,R. g., AND D. G. RICHARDS.1982. Adaptions 955-965. for acoustic communication in birds: Sound SLABBEKOORN,H., AND C. TENCATE. 1999. Playback transmissionand signal detection.Pages 131- experimentsin Collared Doves:Slaves to the 181 in Acoustic communication in birds, vol. 1 rhythm.Ethology 105:377-391. (D. E. Kroodsma, E. H. Miller, and H. Ouellet, SMITH, K. K. 1994. Are neuromotor systems con- Eds.). Academic Press,New York. served in evolution? Brain Behaviour and Evo- ZANN, R. 1974. Inter- and intraspecificvariation in lution 43:293-305. the callsof three speciesof Grassfinchesof the SOKAL,R. R. 1986.Phenetic : Theory and subgenusPoephila (Gould) (Estrildidae).Zeit- methods.Annual Review Ecological Systematics schriftffir Tierpsychologie39:85-125. 17:423-442. Associate Editor: S. I. Rothstein

Top View