J. FieldOrnithol., 58(2):152-170

SONG DIALECTS AND NEIGHBORHOOD HABITATS IN THE INDIGOBIRDS CHALYBEATA AND V. PURPURASCENS AT LOCHINVAR NATIONAL PARK,

ROBERT B. PAYNE Museumof Zoologyand Departmentof Biology The Universityof Michigan Ann Arbor, Michigan 48109 USA Abstract.--As part of a study of songdialects in the brood-parasiticAfrican indigobirds, habitatsof in different songdialect neighborhoodswere comparedin populationsof V. chalybeataand V. purpurascens.Univariate and multivariateanalyses of vegetationclasses, vegetationstructure, and resourceabundance at eachcall-site revealed no consistenthabitat differencesamong song neighborhoods of V. chalybeata.The sametechniques resolved sig- nificant differencesbetween the two indigobirdspecies. V. chalybeatahad three major song dialect repertoiresin the same area where V. purpurascensall sanga single songdialect repertoire.The lack of consistenthabitat differenceamong song dialect neighborhoods suggeststhat songneighborhoods in thesespecies are not explainedby differencesin the acousticproperties of songsin differenthabitats nor by geneticdifferentiation among local populationsadapted for differenthabitats. Supporting evidence of independenceincludes (1) the consistentdistribution of songneighborhoods over years of changinglocal habitats, (2) one instanceof a songneighborhood replacing another without a correspondingchange in local habitat, (3) the ongoingcultural evolutionof the fine detailsof all songsfrom year to yearin theabsence of any correspondingchange in habitat,(4) movementof birdsbetween songneighborhoods, and (5) a lack of congruenceamong song dialect areas of other speciesin the samehabitats. Observations of socialbehavior on the sameindividual indi- gobirdsprovides support for competitivesocial adaptation as the primary mechanismex- plaining the songsharing among neighbors.

DIALECTOS EN HABITATS VECINOS DE VIDUAS (VIDUA CHALYBEATAY V. PURPURASCENS) EN EL PARQUE NACIONAL LOCHINVAR, DE ZAMBIA Sinopsis.--Comoparte dc un estudiosobrc dialcctos en avesparaslticas, sc cornpar6cl hfbitat de fireasvecinales de poblacionesde Viduachalybeata ¾ V. purpurascens,en donde habladiferencias en dialectos.Anflisis monovariable¾ multivariable del tipo de vegetaci6n, cstructurade csta, ¾ abundanciade recursosen cada lugar vecinal no revel6 difercncias para la especieV. chalybeata.E1 mismotipo de anflisis mostr6diferencias significativas al comparara ambas especies.V. chalybeatamostraba tres repertoriosde canciones,en las mismasfireas en dondeV. purpurascensmostraba un solodialecto de canciones.La ausencia dediferencias en el hfbitat de avesvecinas con dialcctos particulares, sugiere que la diferencia en el cantode gruposvecinales en estasespecies de viudas, no puede explicarsea basede diferenciasen las propiedadesac6sticas de los hfbitats, ni en las diferenciasgen&icas que puedahaber entre poblacioneslocales, como adaptaci6n al hfbitat particular que ocupan. La evidenciaque sugiercpatrones independientes incluye: 1) la distribuci•nconsistente de dialectosparticulares en fireasvecinas donde han ocurridocambios en el hfbitat a travis de los aftos,2) el casoparticular de un grupo con su propio dialectoque sustitu¾•a otro sin que ocurrierancambios en el hfbitat, 3) "evoluci•ncultural" de pequefios&talles en los cantosque ocurrendc afioen afio,sin que mediencambios en el hfbitat, 4) movimientode avesentre las vecindades,¾ 5) otras especiesen el mismohfbitat que no mantiencnla armonia de dialectos-cancionespor fireasvecinales. Observaciones sobre el comportamiento socialde individuosparticulares sugiere que son adaptacionessociales competitivas el me- canismoprincipal para explicarel porqu• avesvccinas fienden a compartircanclones par- ticulares.

152 Vol.58, •o. 2 SongDialects and Neighborhood Habitats [ 153

Microgeographicvariation in song characterizesa number of avian species.The featuresof songthat are restrictedto local populationsare referredto as song"dialects." Local songneighborhoods in which birds sharefeatures of their songthat differ from other suchlocal areasoccur in arian specieswith a diversityof life styles--residentand migratory birds, monogamousand polygamousbirds, and specieswith and without parental care of their young(Krebs and Kroodsma1980, Kroodsmaand Baylis 1982, Payne 1981a, 1983). The ecologicalcauses and conse- quencesof local songdialects have been linked with local behavioral adaptationsof birds to changingsocial contexts, to dispersalhistories of local populations,to effectson geneflow and populationdifferentiation, and to adaptationto local environmentalconditions, and may involve more than one of these (Adret-Hausberger1982, 1986; Baptista 1975; Craig and Jenkins 1982; Handford 1981; King 1972; Nottebohm1975; Payne 1978a,b, 1981a,b, 1983, 1985a,b; Shields1982; Wiens 1982). The questionof ecologicaladaptation has been developed more prom- inently in interspecificthan in intraspecificstudies, but habitat differ- encesappear to be associatedwith acousticdifferences both within and among speciesof birds (Bowman 1979, 1983; Gish and Morton 1981; Handford 1981; Heuwinkel 1982; Hunter and Krebs 1979; Jilka and Leisler 1974; King 1972; Marten and Marler 1977; Morton 1975; Not- tebohm 1975; Sorjonen 1983; Wasserman 1979; Wiley and Richards 1982). The hypothesisof ecologicaladaptation of songis of generalin- terest insofar as other behavior (Ewald 1980, Fretwell 1972, Holmes et al. 1979, Holyoak 1973, Wolf et al. 1976) and morphology(Fretwell 1972, James 1982, Leisler and Winkler 1985, Lederer 1984) may be correlatedwith local habitat conditionsboth within and among species. However, the studiesof within-speciessong variation have usuallybeen carriedout with unmarkedbirds in populationsof unknownhistory, and without concurrenttests on other possibledeterminants of local song differencesincluding the social interactionsand dispersalbiology. In additionto consideringpopulation biology of a species,it may be useful to compareother speciesin the same area to test whether they show a comparablemicrogeographic differentiation in their songs,in order to determinethe generalityof any habitat associationof local behavior. Song dialectscharacterize local populationsof severalspecies of Af- rican brood-parasiticfinches, including the (Vidua chalybeata)and the (V. purpurascens).The indigobirds are species-specificbrood parasitesof the firefinches(Lagonosticta spp.) (Nicolai 1964, Payne 1973). Their fosterspecies are similar in many of their habitat requirementsand as many as five speciesmay coexistin the samearea (Fry 1966, Payne 1973). The similarity in life stylesof the indigobirdspecies suggests a test of the idea that songdialects are asso- ciated with differences in local habitats. This association,if it exists, might be due to commonacoustic design features for soundtransmission in each habitat, or to geneticadaptation of populationsfor local condi- tions. In the secondcase, song differences among populationsmight be- 154] R.B. Payne J.Field Ornithol. Spring 1987 haviorallylimit the dispersalof birds beyondtheir songneighborhood or they might lessdirectly indicate markers of local demesthat havediffer- entiatedboth in behaviorand in geneticcharacteristics (Payne 1981a). As the two indigobirdsare similar in ecologyand in song structure, differing mainly in their fosterspecies and in the mimicry of their foster species'songs, one speciesprovides a test of the ecologicalsignificance of variation in the other. Two main questionswere askedin the presentstudy: (1) Are the song neighborhoodsassociated with differenthabitats? (2) Do the two species show the same microgeographicpattern of responseto the habitat? In addition,the independenceof songdialects and local habitatswas tested with a comparisonof changein songand habitat overseveral years, with observation of whether movements of marked birds were restricted to one habitat or songneighborhood, and with a cross-speciescomparison of songvariation in other songbirdsin the samearea.

METHODS Indigobirdswere observedand their songsrecorded in a studyarea of about 40 km 2 at Lochinvar National Park, Zambia, 15ø57'S, 27ø15'E, from 1972 through 1979. A sample of singingmale V. chalybeatawas selectedfor observationof socialbehavior and individually marked with coloredleg bands (Payne and Payne 1977). Males generallyshared all songtypes in their repertoirewith their neighborsand nonewith remote malesin other songneighborhoods. A "neighborhood"is definedas the set of males that share the same songtypes, or songrepertoire dialect. The songtypes and songneighborhoods were usually obviousupon hear- ing the singingbirds in the field. A sampleof more than 30,000 songs of more than 100 V. chalybeataand more than 1000 songsof about 30 V. purpurascens(not all were color-marked)were recordedwith a Uher 4000-seriestape recorderand later were audiospectrographedwith a Kay Elemetrics"Vibralyzer" 7030A or a Princetonreal-time spectrumana- lyzer (PAR-4512) and photographedon 35-ram film. The results con- firmed the distinctivenessof songdialects in the local V. chalybeataand the uniformity of vocal repertoiresin the local V. purpurascerzs(Payne 1985a,b). The three songneighborhoods of V. chalybeatawere named cowpie,junction, and diptera (Fig. 1). The songneighborhoods of V. chalybeatachanged in area and in the number of birds with the songs over the years, with one songpopulation (junction) disappearing and another(diptera) expanding into thejunction area. Analysis of the habitats was restrictedto the middle years of the study, 1975 and 1976, to avoid the confoundingeffects of changein vegetationthrough time. The habitat is a woodedgrassland south of the Kafue River flats, and has been describedby Clarke and Loe (1974), Douthwaite and van Lavieren (1977), Fanshawe (1971), Howard (1977), and Sheppe and Osborne(1971). The southernend of the park is wooded;the landssouth and east of the park have heavy human use. V. chalybeatais common Vol.•8, •o. 2 SongDialects andNeighborhood Habitats [ 155

3

FiOU}tE1. Locationsof 62 call-sitesof the indigobirdsV. chalybeataand V. purpurascens at LochinvarNational Park, 1975-1976.l, cowpiesong neighborhood; 2, junction song neighborhood;3, diptera song neighborhood, 1-3 all V.chalybeata; 4, V. purpurascens. 156] R.B. Payne j. VielaOrnithol. Spring1987 aroundthe villages,where its fosterspecies L. senegalanests in the thatch roofs.These two speciesas well as V. purpurascensand its fosterspecies L. rhodopareiaalso occur in the park. Each male indigobird focuseshis singing behavior in a single tree, termeda "call-site" (Payne 1973, Payne and Payne 1977). Males sing most of the day, femalesvisit males at the call-sites,and all matings observedwere at the call-sites.The focal sitesof activity were usedas the centersof samplingof the vegetationand other habitat features.Eight habitat characteristicswere measuredfor each major call-site usedbe- tween 1973 and 1976 in the 3 main dialectneighborhoods of V. chalybeata and for each call-site of V. purpurascensin the samearea. The habitat values were taken from a vegetationdescription (Douthwaite and van Lavieren 1977) supplementedwith an unpublishedmap by the authors and by our observations,from aerial photographs,from our photographic samplingof vegetationat eachcall-site, and from our quadrat sampling around the call-site betweenMarch and May in 1975 and 1976. Iden- tification of plants was basedon publishedregional floras and on com- parisonof material with the LochinvarPark referenceherbarium. The habitat charactersused for comparisonamong song neighborhoods were originallyamong a larger setthat was usedto comparethe mating success of individual malesin the populationwith their displayterritory (Payne and Payne 1977). They are the mostconspicuously measurable variables that seemedto be ecologicallysignificant to the indigobirds.They are associatedwith the densityof their fosterspecies and they includethe grassspecies that are the major foodsof the indigobirdsand their foster species.Some habitat variablesalso may be related to local differences in the acousticenvironment for transmissionof the songsof the indigo- birds (thickets,woodland, openland, and densityof the vegetationnear the call-site),but they were chosenmainly as overall ecologicallyappro- priate indicesof local habitat variation among indigobirdsin the study area. (1) rthicket. The shortest,most direct distance(m) from the call-site tree to the nearestriverine thicket was determinedfrom the vege- tation map and from aerial photographs.The rthicketdistance was thoughtto be significantto the indigobirdsinsofar as their Lagono- stictafoster specieswere abundant in this habitat. (2) wthicket. The distance (m) from the call-site tree to the nearest densethicket of othervegetation types. Lagonosticta firefinches were common in these thickets also. (3) woodland.The distance(m) to woodland denserin aerial photo- graphsthan the Acacianilotica habitat that was the most common habitat type of the call-sites.These includedBrachystegia wood- land with nearly closedtree canopy and the more open Albizia harveyiand Colophospermumtoopane woodlands and were a less favoredhabitat for the Lagonostictaspecies. These were thought to be a less-preferredhabitat. Vol.58, •o. 2 SongDialects and Neighborhood Habitats [ 157

(4) openland.The distance(m) from the call-siteto opengrassland or abandonedonce-cultivated lands, or to maize fields. The indigo- birds fed in these habitats. (5) anilo. The proportionof the area mappedas Acacianilotica wood- land within 200 m of the call-site. Most call-sites were in this vegetationtype, which tendsto occuron poorly-drainedgray clay soils.Lagor•ostica firefinches were commonin this habitat. (6) vegpro.The densityof shrub and tree vegetationnear the call-site as sampled from standardizedphotographs. Photographs were taken from the four cardinal compassdirections at a height of 1.2 m abovethe groundand at 12 m from eachcall-site using a camera with a 55-mm Micro-Nikkor lens centered on the call-site tree (Figs. 2, 3). Each view was enlargedto 127 x 178 mm and the print was overlaidwith a transparent12.5 mm grid. The number of grid squaresthat coveredleafy shrub or tree vegetation(ex- ceptingthe call-sitetree) abovethe top of the grasswere counted. Scoreswere summed acrossthe four photographs.Topography was nearly flat and no hills were evidentin the photographs.The vegetationdensity was associatedwith the conspicuousnessof sing- ing malesand perhapsalso with acoustictransmission of song. (7) fechino.The frequencyof occurrenceof 1-m2 squarequadrats near the call-site that containedthe ephemeral seasonalgrass Echi- nochloacolonurn. The seedsof this grassare the mostcommon food of Vidua and Lagonostictafinches at Lochinvar Park. They were found in more than half of all samplesof indigobirdsand fire- finches(Payne and Payne 1977, Payne 1980). Quadrats were chosenat random(using a publishedrandom-numbers table) with regard to compassdirection and distancein 1975 (100 quadrats per call-site) or more rapidly at 5-m intervalsalong the NESW compassdirections from 1 to 50 m from the site in 1976 (24 quadratsper call-site).Data from 1975 were transformedto 1976 standardsby random selectionof data with distancesthat fell with- in 30 m of the call-site, the limit of most observedforaging trips of the singingmales. (8) fsetaria. The secondmost frequently observedfood items of the local fincheswere seedsof Setariaspp. grasses.These grasseswere sampledas for Echinochloa. The vegetationtype in which each call-site was situated was deter- mined with the vegetationmap, aerial photographs,and field descrip- tions. The frequencyof the different vegetationtypes associatedwith call-siteswas comparedbetween song neighborhoods and betweenspecies. Scoresof the eight habitat variableswere comparedbetween song neigh- borhoodsand betweenspecies with analysisof variance,t-tests, and the nonparametricKruskal-Wallis test and Mann-Whitney U-test (Siegel 1956). Multivariate comparisonsof call-site habitats were carried out using 158] R. B. Payne j. Field Ornithol. Spring 1987

FIGt•RE2. Moe call-site of a male V. chalybeatain the diptera songneighborhood, pho- tographedin Banakailavillage eastof LochinvarPark in May 1976. the computerprogram DISCRIMINANT available in the University of Michigan Interactive Data AnalysisSystems (MIDAS). Linear discrim- inant functions(Cooley and Lohnes 1971) were computedon standard- ized variablesusing each song neighborhood of V. chalybeataas a known group, using the two speciesas sample groups,and in a third test using the three songneighborhoods of V. chalybeataand the one of V. purpu- rascensas four groups.The canonicalrelations were comparedin graphic form to illustratethe similaritiesamong groups (Nie et al. 1975).

RESULTS Song populations of V. chalybeataoccurred in exclusivemicrogeo- graphicareas, with birds of the cowpiedialect mainly along a streamin a cultivatedand human-populatedarea south of the park, the diptera dialectmainly northeastof the park, and the junction dialect all within the park (Fig. 1). Habitats outsidethe park were generallyopen with mostlarge treeshaving been cleared for firewoodand cultivationof maize and the lands were intensivelygrazed by cattle. In all three areas the most frequently used habitat was open Acacia nilotica woodland (Table 1). The dialect neighborhoodswere not re- strictedto uniform habitat patchesand occurredin interveninghabitat types as well as in the acacia woodlands,which tended to be in low, VoL 58, No. 2 SongDialects and NeighborhoodHabitats [159

;

FIGURE3. Muplanka call-site of a male V. chalybeatain the diptera songneighborhood, photographednear Muplanka game guard camp in Lochinvar Park in May 1976.

poorly-drainedareas of gray soil with poolsof surfacewater well into the dry season.Vegetation classes were combinedon the basisof the densityof coverof the woodyplants in orderto createclasses large enough for statisticalcomparison: (1) Acacianilotica woodland, (2) open wood- land including combreturn,mopane, and acaciasother than A. nilotica and A. polyacantha(vegetation types W.2, W.6, W.7, W.8.1, W.8.2, W.8.3, W.10, W.12), and (3) closedvegetation (types W.1, W.3, W.4, W.5, W.9, W. 11) (Douthwaite and van Lavieren 1977 vegetationtypes, see Table 1). Frequency distributionsof the call-sitesin each dialect populationin Table 1 were computedfor thesevegetation types. There was no significantdifference in habitats amongthe three songpopula- tions of V. chalybeata(X 2 = 8.49, df = 4, P > 0.20). A comparisonof the two specieslikewise showedno significantdifference in habitat dis- tribution (x 2 = 6.88, df = 2, P > 0.05) though V. purpurascenstended to occurless frequently than V. chalybeatain the more open vegetation types. Habitat characterswere comparedto test whether male V. chalybeata with different songdialects differed in habitat. There was no consistent differencebetween sites used by birds of the three songneighborhoods (Table 2). Analysisof variance(ANOVA) indicateda significantdiffer- enceamong song populations for two of the eight variables,vegpro (F = 160] R.B. Payne j. FieldOrnithol. Spring1987

T^BI•E1. Vegetationtypes associated with call-sitesin songpopulations of indigobirds.

N call-sitesin each vege- tation type V.chalybeata V. pur- Junc- Dip- pura- Vegetationtype a Cowpie tion tera scens W.1. Brachystegiawoodland 0 0 0 0 W.2. Combretum-Pericopsis-Xeroderriswoodland 0 0 4 0 W.3. Combreturn-Acacia 1 0 1 1 W.4. Acaciapolyacantha 0 0 2 2 W.5. Albiziaharveyi 2 0 0 0 W.6. Colophospermumtoopane 2 0 0 0 W.7. thickets on termitaria 0 0 0 0 W.8. Acaciasavanna on alluvial clayb 9 14 12 4 W.9. riparian woodland 0 0 0 1 W.10. secondarywoodland of abandonedcultivation 1 I 3 0 W.11. thickets 0 0 0 1 W.12. hot springsvegetation 0 0 0 0

aNomenclature follows the vegetation types of Douthwaiteand van Lavieren (1977). The mapaccompanying that work was not published, but a simplifiedversion appears in Payne and Payne(1977). bAll call-sitesin thiscategory were in vegetationtype W.8.4, theAcacia nilotica woodlands of Douthwaiteand van Lavieren (1977), thoughthe acaciasA. gerrardiiand A. sieberana (whichelsewhere at Lochinvarform more extensive woodland types W.8.2 andW.8.3) also were commonin someareas in the cowpieand diptera neighborhoods.

4.46, df -- 2, P < 0.05) andfsetaria(F -- 5.25, df -- 2, P < 0.01). The 3 songneighborhoods were compared further for these two variables with a Shefigmultiple comparison test. The resultsshowed significant differ- encesin mean scoresof vegetationdensity between the junctionand dipteraneighborhoods (F -- 8.71,P < 0.01) andin meanscores of the frequencyof Setariagrasses between the cowpieand diptera (F = 5.20, P < 0.05) and betweenthe junctionand dipteraneighborhoods (F = 9.17,P < 0.01). Nonparametriccomparisons were made with the Krus- kal-Wallis statisticand the resultingvariables were comparedin pairs with the Mann-Whitney U-test. The resultswere identicalwith the correspondingparametric tests; the samepairs of songneighborhoods differedsignificantly in theirhabitat scores (Table 2). Vegetationdensity washigher in thejunction dialect area than in the dipteraarea, due to the intensiveclearing of woodlandsoutside the park. Setariagrasses were lessfrequent around the call-sitesin dipteraarea but thesegrasses grew in all three indigobirdareas. Comparisonof the 52 call-sitesof V. chalybeatawith the 10 call-sites of V. purpurascensshowed significant species differences in two habitat variables(Table 2). V. purpurascenscall-sites had on the averagealmost 50% more tree and shrub cover(mean scoresfor V. chalybeata111, for V.purpurascens 152, t = 2.63,P < 0.05)and had 30% less Acacia nilotica Vol.58, No. 2 SoNgDialects and Neighborhood Habitats [ 161

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

ß .

+l+l+l+l+l+l+l+l 162] R.B. Payne J.Field Ornithol. Spring 1987 habitat within 200 m of the call-site (mean scoresfor •/. chalybeata66, for •/. purpurascens36, t = 2.37, P < 0.05). The valuesfor vegproand anilo in all 62 siteswere negativelyassociated, but not quite significantly so (r, = -0.22, r0.0s= -0.25). •/. purpurascenslike its fosterspecies L. rhodopareiatended to occuronly in areas with densethicket vegetation, whereas •/. chalybeataand its fosterspecies L. senegalawere somewhat more widespreadand extendedinto the more open vegetationtypes. The song neighborhoodswere defined as groups and the variables transformedto standardizeddistributions for multivariate analysis.Ma- halanobis distancesD 2 were computed between group centroidsin a seriesof simultaneouslinear discriminantfunction analysesusing the pooled within-stratum correlation matrix. Variation in habitat among song neighborhoodswas not significantlygreater than among birds of the samesong neighborhood (D2•,• = 0.79, D•,3 = 2.22, D2•,3= 2.22; pairwise group F-statistics0.63, 2.12, 2.12; P > 0.05). In this analysis, the discriminantfunction classified correctly the songdialect for only 30 of the 52 (58%) call-sitesof •/. chalybeatain the three songneighborhoods. A variety of habitatswere usedwithin a songneighborhood, and neigh- boring males often sang in the same habitat patch. The lack of a close associationof songneighborhoods and habitats indicatesthat the song neighborhoodsare not explainedby habitat differences. Habitats of the two specieswere comparedin the samemanner. The habitatsof V. chalybeatagenerally were distinguishablefrom thoseof V. purpurascens(D 2 = 2.32, F = 2.15, P < 0.05), and 53 of the 62 cases (85%) were correctlyclassified by the discriminantfunction. The two indigobird speciesnevertheless were similar in their habitats, and their call-sites overlapped in all habitat features that were measured. I ob- servedreplacement of males of one speciesby males of the other both within a seasonand betweenyears on a few call-sites. In the third test the canonicalcomponents of the discriminantscores of all four populationswere summarizedgraphically (Fig. 4). The vari- ablesanalyzed failed to provide evidenceof overall ecologicaldifference amongthe four groups.DF1 characterizesthe variation both within and amongthe three V. chalybeataneighborhoods and showsthe considerable overlap.DF2 characterizesthe differencesbetween V. chalybeataand V. purpurascens.The habitatsof the two speciesoverlap somewhat less than do the habitatsof the three songpopulations of V. chalybeata.The cor- relations of the habitat characters with these two discriminant functions show which characterscontribute to thesetrends (Table 3). The vege- tation density and the frequencyof Setaria grasseswere most highly correlatedwith DF1, and the vegetationdensity and the proportionof Acacianilotica habitat to DF2. The vegetationdensity was the character with the greatestdiscriminating power in a seriesof stepwise(both for- ward and backward)discriminant analyses that are not presentedhere in detail. Althoughthese stepwise discriminant analyses suggested some additionalcombinations of significantdifferences among song neighbor- hoods,the statisticalinference as in all multivariate testsreported here Vol.sS, No. 2 SongDialects and Neighborhood Habitats [ 163

.

1'" - 3"'3-•- 2-- -' __ U') .... ? 1223 33 •:.,., I-- '- 2 12 •322433 3 "3,, Z

FIGURE 4. Discriminant function analysis of eight habitat variables of 62 call-sites of indigobirds. 1, cowpie song neighborhood;2, junction song neighborhood;3, diptera song neighborhood,1-3 all V. chalybeata;4, V. purpurascens.Canonical relation 1 accounts for 51.1% of the total variance and canonical relation 2 accounts for an additional 37.7% (overall variance 88.85% of the total variance). Note the lack of separationby habitatamong the three songneighborhoods of V. chalybeataand of the 2 indigobirdspecies.

is questionableboth becausethe samplesizes were small and becausethe test (Box's M) for equality of covariancesshowed significant differences amonggroups (Nie et al. 1975). The main result of the comparisonsis that habitats of the local V. chalybeatasong populations do not differ very much and differ lessfrom each other than from the other indigobird species.

DISCUSSION Songneighborhoods of V. chalybeatawere largely independentof any habitat differencesbetween them. One songneighborhood occurred en- tirely within the park, but so did a few malesof the other songneigh- borhoods.The few significantdifferences in habitat betweensong neigh- borhoodscompared pairwise appear to be a consequenceof the fact that birds with a shared song repertoire generally are each other's closest neighborsso tend to occurin the samepatch of habitat where they are within visualand acousticcontact. No habitatvariables were significantly different amongall three songneighborhoods compared simultaneously and no multivariatefunction was foundto separatewith clarity the hab- itats of the songneighborhoods. On the other hand the habitat samplingand quantitativeanalysis were adequateto show habitat differencesbetween V. chalybeataand V. pur- purascensin the same area. V. purpurascensand its foster speciesL. 164] R.B. Payne J.Field Ornithol. Spring1987

T^BLE 3. Correlations of habitat characters with habitat discriminant functions of four songpopulations of indigobirds.

Correlation

Habitat character DF1 DF2

rthicket 0.108 0.232 wthicket -0.051 0.095 woodland - 0.108 0.388 openland 0.318 - 0.225 anilo 0.138 0.551 vegpro -0.535 -0.651 fechino 0.088 0.466 fsetaria -0.701 0.232

ro.o5= 0.250, ro.o•= 0.325. rhodopareiatend not to occurin the more clearedhabitats; both occurin the relativelyprotected areas of LochinvarPark. The resultsindicate less differencein habitat betweenthe songneighborhoods of V. chalybeata than betweenthe speciesV. chalybeataand V. purpurascens. The microgeographicscale of the songdialects indicates that birds with the same sharedrepertoire are generallyneighbors and live near eachother (Fig. 1). The songdialects are geographicallyseparated from each other in neighborhoodsrather than being intermixed in a fine- grainedmosaic as they might be if songswere adaptedon a more local, territory-by-territoryscale to match a local habitat, as has beenreported in certainother species(King 1972, Wasserman1979). The analysiswas carried out on a site-by-sitebasis according to the resident'ssong rep- ertoire rather than by groupingsites according to any microgeographic criteria. Within the area usedby birds with a commonrepertoire, hab- itats varied, and some sites were more similar to sites in other song neighborhoodsthan to certainsites in the samesong neighborhood. So the songrepertoire at a sitehad morein commonwith the songrepertoire of malesat neighboringsites than with the observedhabitat differences in the studyarea woodlands. The acousticproperties of local sitesand songswere not testedexper- imentally.All malessang on exposedbare branchesin the topsof trees at heightsof 3-8 m. Transmissionproperties of soundswith different songrepertoires near the call-siteswere probablysimilar in local con- ditions,but would be more subjectto degradationif birds were to sing in densefoliage or near the ground (Wiley and Richards 1982). The overall structureof audiospectrogramsof indigobirdsongs were similar acrosssong neighborhoods and species(Payne 1973, 1979, 1985a). The songsdiffered in fine detail of the notesrather than in overalltem- poral patterningand frequencyrange and envelopeof the song,though theselast charactersmay covarywith soundattenuation and degradation in different habitatsin other species(Bowman 1979, Heuwinkel 1982, Vol.58, •o. 2 SongDialects and Neighborhood Habitats [165

Sorjonen1983, Wiley and Richards1982). Most songsin the individual/ neighborhoodrepertoire of indigobirdsare complexin structure.A song has 5-9 kinds of notes,some notesare repeatedwithin the song,and nearly all notesare broad-bandin nature and cover a wide range of frequencies.The temporalstructure is complexinsofar as differentnotes in a typical songare of different duration,the rate of frequencymodu- lation often variesbetween beginning and end of a note;many notesare short (50-200 msec)and noteshave complexpatterns of inflection (il- lustrationsin Payne 1979, 1985a). The harsh, broadband,temporally complexsongs may be designedfor long-rangesignalling in opencountry where atmosphericturbulence and attenuation by atmosphericabsorp- tion over distanceare more critical than scatteringby vegetation.Indi- gobird songsemphasize frequency modulation (Payne 1985a). The ac- cumulationof random fluctuationsand reverberationsfrom atmospheric degradationwould likely maskany distinctiveamplitude modulation over long distancesin thesesongs (Morton 1975, Wiley and Richards 1978, 1982). On the other hand someindigobird songs are mimetic,matching the localsongs of their fosterspecies of firefinch,and thesesongs are often slurred tonal whistleswith lessvarying temporal complexityof the notes. It may be significantthat these mimetic songsare given at closerange when a male interactswith a visiting female on the ground, as well as being the less-frequentlydelivered songs from the high song perches (Payne 1979) in advertisement. The broadbandcomplex songs are derivedthrough a processof cul- tural evolutionwithin a songtype from the foster-speciesmimetic songs, with the tonal elementschanging over a few yearsto noteswith a broad frequencyrange and rapid inflectionas well as an introductionof am- plitude modulation(Payne 1985a), so it seemsunlikely that the design of the foster-speciessong sets long-term constraints on the designfeatures of the advertisingsong of their broodparasites. Although there is ample evidencethat the songsof indigobirdsare designedin part for transmis- sion in their particular physicalas well as socialenvironment, the point emphasizedhere is that the differencesin songsamong the repertoire neighborhoodsdo not appearto be designedfor localhabitat differences, becausethere are no obvioushabitat differencesamong the songneigh- borhoods. The songneighborhoods were not isolatedby distanceor unsuitable habitat. Habitats betweensong neighborhoods were slightly drier and were lessbushy and often had tall perennial grasseswhere firefinches were uncommon.Color-marked individual V. chalybeatawere observed to movefrom one songneighborhood to anotherand to flock togetherin the non-breedingseason (Payne 1985b). It is unknown whether most youngindigobirds born into a songneighborhood area remain there to breed. Nearly all marked young (38 of 40) disappearedby the next breedingseason. If the songdialects of V. chalybeataat LochinvarPark were associated with habitat differences,one might expectother speciesof small birds 166] R.B. Payne j. FieldOrnithol. Spring 1987 that use the samehabitats in a similar way to have correspondingsong neighborhoods.V. purpurascenshad local song dialectsin other areas, but within the studyarea the birdscomprised a singlesong neighborhood. The fosterspecies firefinches Lagonosticta senegala and L. rhodopareiado not live in local populationsthat correspondto the songneighborhood areasof V. chalybeata.Songs of 20 L. senegalathat I recordedwithin the three songneighborhoods of V. chalybeatain 1972 and 1975 showedno sign of local dialectsin and around the park. Individually marked L. sengalawere observedto move between indigobird song neighborhood areas (Payne 1980). No local dialectswithin the area were found in recordingsof songsand calls of L. rhodopareia,and movementsof these firefinchesacross the songneighborhood areas of V. chalybeatawere ob- served.Flappet Larks, Miralta rufocinnamomea,live in the samehabitat as the indigobirdsand have signal dialects(wingflap patterns) in local neighborhoods(Payne 1978b, 1981b), but thesedo not coincidewith the indigobirdsong neighborhoods. Other speciesthat I have tape recorded at Lochinvar and elsewhere in southern and central Africa and found to vary locally or regionally in songsdid not show a commonmicrogeo- graphicpattern with the songneighborhoods of V. chalybeataat Lochin- var Park: these included Blue Waxbill, Uraeginthusangolensis, Melba Finch, Pytilia melba,and African Boubou,Laniarius ferrugineus. All six speciesof finchesmentioned live in the same area in Lochinvar Park. Crop samplesof all six showedthe most commonfood in the breeding seasonto be the seedof the grassesEchinochloa colonurn and Setariaspp. (samplesizes ranged from 6 to 220 birds for eachspecies of finch). The lack of correspondencein the songneighborhood areas of V. chalybeata and the other birds suggeststhat whateverfeatures of the local ecology are associatedwith the songneighborhoods in the indigobirds,they do not affect the other speciesin the same manner. V. chalybeataand V. purpurascenssing on elevatedsites in rather open habitat. In the species that do clearly show an associationof habitat and songstructure, the birds occupya denserwoodland habitat (Gish and Morton 1981, Hunter and Krebs 1979, Sorjonen1983), and the acousticdesign features of song vary more amongthose habitats. The comparisonwith other local species suggeststhat differencesin habitat observedat Lochinvar Park are not predictorsof local songneighborhoods. During the yearsof the field study,the distributionof the songneigh- borhoodschanged within the park. Junction neighborhooddeclined from 15 singingmales in 1973 to one in 1976. The area occupiedby junction neighborhoodin 1973 and 1974 was taken over in part in 1975 and almost completelyin 1976 by previouslyunmarked males of diptera neighborhood(Payne 1985a). The overlap of call-site distributionsin Figure • resultsin part from this shiftingboundary between song neigh- borhoods.No abrupt habitat changeaccompanied the shift in songneigh- borhooddistributions from 1974 to 1976 though there was continuing habitat degradation.The numbersof finchesusing bushythickets and tall grassappeared to decreasewith the lossof their habitat throughthe Vol.•8, No. 2 SongDialects and Neighborhood Habitats [ 167 years of the study. A 1-km transectthat was used to determinebird densitiesin 1972 and 1973 was run again three times in March and April 1979. The transectwas in an area little usedby cattleand was not representativeof the widespreadhabitat destructiondue to cattle in the park woodlands.The firefinchL. senegalawas seenonly twice (1 and 2 birds) and L. rhodopareiawas seenon all three censuses(1, 1, 2 birds). These valueswere slightly lower than the correspondingmonthly means in 1973 (Payne 1980). The vegetationalso changed progressively due to differencein burning. The park was burned early eachyear in the early 1970s,but in 1976 and later a few areaswere not burned at all, including the junction site. In 1979 the grasseswere thicker in this area than in earlier years when the site was a centerof indigobirdactivity. I found no Echinochloagrass around the sitein 1979. Chlorisguyana and Digitaria spp.were abundant,but we saw no indigobirdsfeeding on thesegrasses. An herb layer aroundjunction in 1979 madeup 20% of the cover,more than in the early 1970s. Although vegetationwas not sampled quanti- tatively in 1979, the changesdid not appear to involve those habitat variables that differed significantlybetween the songneighborhoods in earlier years (Table 2). Thus the change in the distribution of song neighborhoodsin the park over the yearscoincided in time with change in habitat quality but not in closeassociation and no direct causeand effectof habitat quality and the distributionof the different songneigh- borhoods is indicated. The distinguishingcharacteristics of the local songdialects at Lochin- var Park involvedetails in the duration, inflection,and frequencymod- ulation of certain noteswithin the songs(Payne 1985a). If thesefine details that define the neighborhoodsong differenceswere related to acousticdesign for transmissionin different habitats,then the differences shouldpersist across years, at least in areaswhere the habitat remained the same. However, all 22-24 songsin the song repertoire of an indi- gobirdsong neighborhood change from year to year, everyyear, and all birds make the samechanges both within and betweenyears in a con- tinuing processof cultural evolution(Payne 1985a). This ongoingchange, observedin all songtypes regardlessof their different functions(Payne 1979, 1985a), indicatesthat the detailswhich distinguishthe differences in the songdialects are not stableadaptations to persistenthabitat dif- ferencesin their neighborhoods.To comparethe details of songdiffer- encesamong the three populationsin a quantitative manner, I chose19 descriptivecharacters associated with songcomplexity, frequency range and envelope,and temporal patterning and comparedthese in univariate analysesand in a multivariateclustering scheme (Payne 1983). This was alsodone for 6 consecutiveyears of songrecording for all songtypes in all three dialects(Payne, unpublished data). The resultsshowed no dif- ferencein the rate of changein acousticstructure of songsamong the three dialectneighborhoods. The songsdid not changeany fasterin the park, where the habitat was changing,than in the village areas outside the park, where the habitat appearednot to changeas much over the 168] R.B. Payne j. FieldOrnithol. Spring1987 yearsof the study.Thus the two predictionsof a hypothesisof differential adaptationof song among the neighborhooddialects (stability of song acrossyears and greater changein songsin a more drasticallychanging habitat) were not realized,and the resultsagain point to processesother than differential acousticadaptation of songto local habitatsin explain- ing the continuityand changein the indigobirdsong neighborhoods. Although the songdialect neighborhoodsof V. chalybeatado not cor- respondto distincthabitats, the birdsthat sharetheir songscorrespond to the set of birds that visit each other and competefor the same call-sites and mates. Femalesvisit males within a songneighborhood, which cor- respondsto an explodedmating arena (Payne and Payne 1977). Indi- gobirdssometimes disperse between song neighborhoods and then switch their songrepertoire to match their current socialneighborhood (Payne 1985a,b). Males copy each other within the year and betweenyears, as shownby the consistentchange among all malesin a songneighborhood. Part of the behavioralhomogeneity that characterizesa songneighbor- hoodin thesebirds is associatedwith differentialbreeding success: song variants of males with large numbers of matings tend to be copied in detailby their neighbors(Payne 1985a). The declineof thejunction song neighborhoodaccompanied the disappearanceof one particularly suc- cessfulmale, and the social changemay have been associatedwith the replacementof onesong neighborhood by another(Payne 1985a,b). Song dialectsin the indigobird speciesappear to be the result of socialadap- tationsrather than habitat adaptations,though of coursein generalthe two factorsneed not be mutually exclusive. Habitats and songdialects may be more closelyassociated in certain other speciesof birds. However, in any study of songdialects and local ecology,it is necessaryto look also at the social behavior of individual birds and to test simultaneouslymore than one hypothesisto avoid spu- rious associations(Payne 1981a), and it is possiblethat acousticand socialadaptations may interact (E. S. Morton, pers. comm.).The multi- variate tests used for the habitats of the indigobirdsmay be useful in other studies of association of local behavior traits and habitat differences within and betweenspecies of birds.

ACKNOWLEDGMENTS Observationsand tape recordingsat LochinvarPark were permittedby the Department of Wildlife and National Parks, Chilanga, Zambia. K. Klitz and D. M. Lewis assistedin the field. For arrangementsand accomodationsin Zambia I thank S. I. Bug•ie, H. Black, P. and D. Cartier van Dissell, R. J. Dowsett,W. K. Sinkamba,and G. C. N. Zyambo.L. P. van Lavierenprovided an unpublishedvegetation map. For commentson the manuscript I thank L. F. Baptista,E. H. Burtt Jr., s. M. Doehlert,L. R. Heaney, P. D. Heideman, E. S. Morton, P. Myers, and R. E. Strauss.The studywas supportedby grantsfrom the National ScienceFoundation, the National GeographicSociety, and The University of Michigan Faculty ResearchFund.

LITERATURE CITED

ADRET-HAUSBERGER,M. 1982. Socialinfluences on the whistledsongs of Starlings.Be- hav. Ecol. Sociobiol. 11:241-246. Vol.5a, r•o. 2 SongDialects and Neighborhood Habitats [ 169

1986. Temporal dynamicsof dialectsin the whistledsongs of Starlings.Ethology 71:]40-152. BAPTISTA,L. F. 1975. Songdialects and demesin sedentarypopulations of the White- crownedSparrow (Zonotrichialeucophrys nuttalli). Univ. Calif. Publ. Zool. 105. BOWMAN,R. I. 1979. Adaptive morphologyof song dialectsin Darwin's finches.J. Ornithol. 120:353-389. ß 1983. The evolutionof songin Darwin's finches.Pp. 237-537, in R. I. Bowman, M. Berson, and A. E. Leviton, eds. Patterns of evolution in Galapagosorganisms. American Assn. Advancement of Science, Pacific Division, San Francisco. CLARKE,J. E., ANOI.D. LOE. 1974. A guide to the national parks of Zambia. Anglo- American Corp. (Central Africa), Lusaka, Zambia. COOLEY,W. W., ANOP. R. LOHNES. 1971. Multivariate data analysis.Wiley, New York. CRAIG,J. L., ANDP. F. JENa:INS.1982. The evolutionof complexityin broadcastsong of .J. Theor. Biol. 95:415-422. DOUTHWAITE,R. J., ANDL. P. VAN LAVIEREN. 1977. A descriptionof the vegetationof Lochinvar National Park, Zambia. National Council for Scientific Research Tech. Rep. 34, Lusaka, Zambia. EWALD,P. W. 1980. Energeticsof resourcedefense: an experimentalapproach. Acta XVII Congr. Internat. Ornithol., Berlin:1093-1099. FANSHAWE,D. B. 1971. The vegetationof Zambia. Forest Research Bull. 7, Lusaka, Zambia. FRETWELL,S. D. 1972. Populationsin a seasonalenvironment. Princeton Univ. Press, Princeton,New Jersey. FRy, C. H. 1966. The ecologicaldistribution of birds in northern Guinea savanna,Ni- geria. Ostrich Suppl. 6:335-356. GISH, S. L., AND E. S. MORTON. 1981. Structural adaptationsto local habitat acoustics in Carolina wren songs.Z. Tierps¾chol.56:74-84. HANDFORD,P. 1981. Vegetational correlatesof variation in the song of Zonotrichiaca- pensis.Behav. Ecol. Sociobiol.8:203-206. HEUWINKEL,H. 1982. Schalldruckpegelund Frequenzspektrender GesSngevon Acro- cephalusarundinaceus, A. .s.cirpaceus, A. schoenobaenusund A. palustrisund ihre Bezie- hung zur Biotopakustik.Okol. V/Sgel4:85-174. HOLMES, R. T., R. E. BONNEY,AND S. W. PACALA. 1979. Guild structure of the Hubbard Brookbird community:a multivariateapproach. Ecology 48:31-35. HOLYOAK,D.T. 1973. Significanceof colourdimorphism in Polynesianpopulations of Egrettasacra. Ibis 115:419-420. HOWARD,G.W. 1977. National parks of the Kafue Basin. Pp. 47-56, in G. J. Williams and G. W. Howard, eds.Development and ecologyin the lower Kafue Basin in the nineteenseventies. Kafue Basin ResearchCommittee, University of Zambia, Lusaka, Zambia. HUNTER,M. L., ANOJ. R. KREBS. 1979. Geographicalvariation in the songof the Great Tit (Parusmajor) in relation to ecologicalfactors. J. Anim. Ecol. 48:759-785. JAMES,F. C. 1982. The ecologicalmorphology of birds:a review. Ann. Zool. Fenn. 19: 265-275. JILKA,A., ANOB. LE•SLER. 1974. Die Einpassungdreier RohrsSngerarten(Acrocephalus schoenobaenus,A. scirpaceus, A. arundinaceus)in ihre LebenstrSumein bezug auf das Frequenzspektrumihrer ReviergesSnge.J. Ornithol. 115:192-212. K•NG, J. R. 1972. Variation in the song of the Rufous-collaredSparrow, Zonotrichia capensis,in northwesternArgentina. Z. Tierpsychol.25:155-169. KREB$,J. R., AND D. E. KROODSMA.1980. Repertoiresand geographicalvariation in bird song.Adv. Study Behav. 11:143-177. KROODSMA,D. E., ANDJ. R. BAYLI$. 1982. Appendix: a world surveyof evidencefor vocal learning in birds. Pp. 311-337, in D. E. Kroodsma and E. H. Miller, eds. Acousticcommunication in birds, vol. 2. Academic Press, New York. LEDERER,R.J. 1984. A viewof avianecomorphological hypotheses. •)kol. Vfgel 6:119- 126. 170] R.B.Payne J.Field Ornithol. Spring 1987 LEISLER,B., AND H. WINKLER. 1985. Ecomorphology.Pp. 155-186, in R. F. Johnston, ed. Current ornithology,vol. 2. Plenum, New York. MARTEN, m., AND P. MARLER. 1977. Sound transmissionand its significancefor vocalization.I. Temperate habitatsßBehav. Ecol. Sociobiol.2:271-290. MORTON,E.S. 1975. Ecologicalsources of selectionon avian soundsßAmer. Natur. 108: 17-34. NICOLAI,J. 1964. Der Brutparasitismusder Viduinae als ethologischesProblem. Z. Tierpsychol.21:129-204ß NIE, N.H., C. H. HULL, J. G. JENKINS,K. STEINBRENNER,AND D. H. BRENT. 1975. SPSS statisticalpackage for the socialsciencesß Second ed. McGraw-Hill, New York. NOTTEBOHM,F. 1975. Continental patterns of song variability. in Zonotrichiacapensis: somepossible ecological correlatesß Amer. Natur. 109:605-624. PAYNE,R. B. 1973. Behavior,mimetic songsand songdialects, and relationshipsof the parasiticindigobirds (Vidua) of AfricaßOrnithol. Monogr. 11. ß 1978a. Microgeographicvariation in songsof SplendidSunbirds Nectarinia coc- cinigaster:population phenetics, habitats, and songdialectsß Behaviour 65:282-308ß 1978b. Local dialectsin the wingflaps of Flappet Larks Mirafra rufocinnarnornea. Ibi} 120:204-207ß ß 1979. Songstructure, behaviour, and sequenceof songtypes in a populationof Village Indigobirds,Vidua chalybeata. Anim. Behav. 27:997-1013ß ß 1980. Seasonalincidence of breeding,moult, and local dispersalof Red-billed FirefinchesLagonosticta senegala in Zambia. Ibis 122:43-56. ß 1981a. Populationstructure and socialbehavior: models for testingthe ecological significanceof songdialects in birds. Pp. 108-120, in R. D. Alexander and D. W. Tinkle, eds. Natural selectionand social behavior. Chiron Press, New York. 1981b. Persistenceof local wingflap dialectsin Flappet Larks Mirafra rufocin- narnornea. Ibis 123:507-511. ß 1983. Bird song,sexual selection, and femalemating strategiesßPp. 55-90, in S. K. Wasser, ed. Social behavior of female vertebratesßAcademic Press, New York. ß 1985a. Behavioralcontinuity and changein local songpopulations of Village IndigobirdsVidua chalybeata. Z. Tierpsychol.70:1-44ß 1985b. Songpopulations and dispersalin Steelblueand Purple Widowfinches. Oslrich56:135-146. • ANDm. PAYNEß1977. Socialorganization and mating successin songpopulations of Village Indigobirds,Vidua chalybeata.Z. Tierpsychol.45:113-173. SHEPPE,W., ANDT. OSBORNE.1971. Patternsof use of a flood plain by Zambian mam- mals. Ecol. Monogr. 41:179-205. SHIELDS,W. M. 1982. Philopatry, inbreeding,and the evolutionof sex. State University of New York Press,Albany, New York. SIEGEL,S. 1956. Nonparametricstatistics for the behavioralsciencesß McGraw Hill, New York. SORJONEN,J. 1983. Transmissionof the two most characteristicphrases of the songof the Thrush Nightingale Luscinialuscinia in different environmentalconditions. Ornis Scand. 14:278-288. WASSERMAN,F. E. 1979. The relationshipbetween habitat and song in the White- throated SparrowßCondor 81:424-426. WreNS,J.A. 1982. Songpattern variationin the SageSparrow (ArnphispizabellO: dialects or epiphenomena?Auk 99:208-229ß WILEY, R. H., ANDD. G. RICHARDS.1978. Physicalconstraints on acousticcommunication in the atmosphere:implications for the evolutionof animal vocalizations.Behav. Ecol. Sociobiol. 3:69-94ß • AND•. 1982. Adaptationsfor acousticcommunication in birds: soundtrans- missionand signal detection.Pp. 131-181, in D. E. Kroodsma,and E. H. Miller, eds. Acoustic communication in birds, vol. 1. Academic Press, New York. WOLF, L. L., F. G. STILES,AND F. R. HAINSWORTH. 1976. Ecologicalorganization of a tropicalhighland hummingbird community. J. Anim. Ecol. 45:349-379. Received12 June 1986; accepted15 Nov. 1986.