Parus Caeruleus)

PATTERNS AND SIGNIFICANCE OF GEOGRAPHICAL VARIATION IN THE BLUE TIT (PARUS CAERULEUS)

JEAN-LOUIS MARTIN CentreNational de la RechercheScientifique et Centre d'l•cologie Fontionnelle et Evolutive, (CEFE/CNRS), B.P. 5051, F-34033 MontpellierCedex, France, • and AmericanMuseum of Natural History,Department of Ornithology, Central Park West at 79th Street, New York, New York 10024-5192 USA

ABSTRACT.--Istudied geographicalvariation in mensural charactersand plumage pattern over the breeding range of the Blue Tit (Paruscaeruleus). In Eurasiabody size decreases clinally from north-northeastto south-southwest,and bill thinnessand relative tarsallength increaseclinally from the northeastto the southwest.This variationis consistentwith changes in environmental factorsknown to affect these characters.Geographical isolates in North Africavary in sizeand in bill andtarsus morphology in a way consistentwith habitatvariation among isolates.Similarities in mensural charactersamong samplesare more consistentwith similaritiesin ecologicalconditions than with taxonomy.Geographical variation in plumage pattern essentiallyis clinal in Eurasiabut discontinuousin North Africa. On a broad scale, plumagepattern variation is consistentwith the populationgroups defined by taxonomists (Eurasianand North African populationgroups). Within eachpopulation group consistency between variation in plumage pattern and the probable population phylogeny is poor. My resultssubstantiate the importanceof the unifying effectof gene flow in adjacentsamples even over a large geographicalarea and the necessityof geographicalisolation to fosterthe appearanceof gaps in clinal variation or of markedly different population characteristics. Received20 June1990, accepted 11 March 1991.

NUMEROUS authors have stressed the rela- though plumage color has been used exten- tionshipsbetween morphologicaltraits and en- sively to study geographicalvariation in birds vironmental characteristics (for reviews on eco- and to establishtheir taxonomy(Mayr 1969;for logicalinferences from morphologicaldata, see the Blue Tit, Hartert 1910, 1923, 1932-1938; Vau- Hespenheide 1973, Leisler 1980, Leisler and rie 1957). Climatic variation or variation in Winkler 1985, James1982; for relations between quality of food can influencevariation in color bill sizeor shapewith prey sizeand shape,see (Gloger 1933,Snow 1954b,Slagsvoid and Lifjeld Snow 1954b; Betts 1955; Schoener 1965; John- 1985, Vevers 1985). son 1966;Lack 1971;Grant 1972;Partridge 1976; Basedon this knowledgeI undertooka com- Herrera 1978, 1981; for relations between tarsus prehensive analysisof the variation in men- lengthand perching substrate, see Fretwell 1969, sural characters(size and shape of bill, wing, Lack 1971, Grant 1971; and for relations be- and leg) and plumage(extent of color patches) tweenwing shapeand maneuverabilityin flight, in the Blue Tit (Parus caeruleus)over its entire seeNachtigall and Kempf 1971, Hespenheide breeding range (Fig. 1). I attempt to define the 1973,Kokshaysky 1973). The heritabilityof these principal pattern of geographicalvariation in- morphologicaltraits has been demonstrated dependentof currently recognizedsubspecies. (Smithand Zach 1979,van Noordwijk et al. 1980, I compared patterns of differentiation in iso- Dhondt 1982,Boag 1983, Grant 1985,Schluter latedversus continuously distributed continen- and Smith 1986, Alatalo and Gustafsson 1988). tal populationssubject to greatergene flow un- However, geographicalvariation in morphol- der the assumptionthat geographicalisolation ogyis not alwaysgenetically based (James 1983, allows a better match between the species'mor- Zink 1989). phology and the localenvironment by reducing The relation of plumage pattern to environ- gene flow (Slatkin 1981, 1987; Rockwell and mental featuresis more poorly documented,al- Barrowclough1987). For this, I first identified patterns of geographicalvariation in the se- lected characters.I then searchedfor congru- Addressto which reprint requestsshould be sent. ence of these patterns with geographicalvari-

820 The Auk 108: 820-832. October 1991 October1991] Variationin theBlue Tit 821

0 500 •000krn

•3 5 21 N3 NE4• 3• •0, •8 I t5 to • t9 t0 I 6 I 5 I t5

1- .... 't ...... t ..... • r ...... I 8 _3• 33 I I

44 23 i t2 20 9 : 5 i 33

• •28 t2 •7,% ...... SES•:_ '-.• zr• i'.-- ", sE4: .... .--- '-,,;',,,

15 •e

Fig. 1. Range of tSe Blue Tit (stippled line) and definition of tSe 54 sampling units. Letters represent 9 main geographicalareas of Eurasia(NE = northeast;E = east;5E = southeast;N = north; C = central; 5 = south;NW = northwest;W = west;and SW = southwest),6 Mediterraneanislands (MI1 = Majorca;MI2 = Corsica;MI3 = Sardinia, MI4 = 5icily; MI5: Crete; and MI6 = Greek islands),a samplefrom Cyrenaica, Libya(CY), samplesfrom tSe MagSreb(NA), and samplesfrom tSe Cana• Islands(CA). TSe numberattacSed to letters(e.g. CA2) identifiessampling unit witSin eac5 geograpSicalarea. NumberswitSin eac5 sampling unit refer to samplesizes (male value above female value). ation in the species'environment; for example, more local scale, consistencybetween sample climaticvariation or habitatvariation (vegeta- similarity in plumage pattern and recognized tion structure and composition,feeding re- subspecieswas weak. sources).Finally, I analyzed the relation between the presenceof geographicalbarriers to MATERIAL AND METHODS dispersalof individualsand the amountof phe- The Blue Tit occupiesmost of the western Palearctic notypicgeographical variation. I followedmany (Fig. 1). It inhabits deciduousbroad-leaved forests of the recommendations of Zink and Remsen (Snow 1954a, Lack 1971, Perrins 1979) except in the (1986)for improvingthe studyof geographical Mediterranean region where it inhabits a variety of variation. deciduousand evergreenhabitats distributed in a mo- I demonstratedthat geographicalvariation in saic pattern (Blondel 1985, Blondel et al. 1987, Isen- mensural characters is consistent with variation mann 1987). In contrastto the northern part of its in environmental factors known to affect the breeding range, the hot dry Mediterraneansummer charactersunder study. Variation is stronglyen- probablyis the seasonwhen it is most difficult for birds to obtain food and water (see Blondel et al. hanced in geographicalisolates. The type of 1987).As the high number of overwinteringbirds variation observedin isolateschanges with the from the north implies (Blondel 1969), the wet and habitat type used by species.The consistency mild winter is more benign than the centralEuropean of morphologicalsimilarity with population winter (see Blondel et al. 1987). On the Canary Is- systematicsis poor. Broadpatterns of plumage lands,where the Blue Tit is the only memberof the variation seembest explained by history. On a genus Parus,it breeds in Tamarixwoodlands on the 822 J•,N-LouIs MARTIN [Auk, Vol. 108 two dry easternmostislands (Bannerman 1963, Ba- larger (Fig. 1). I decidedto report only the resultsfor callado 1976), and is more or less restricted to pine the 1,276 male specimens,with the understanding forests on the five outer islands (Lack and Southern that they can be generalizedto the females. 1949,Cullen et al. 1952, Hemingsen 1963,Grant 1979). Standardized(or mean centered) principal com- Vaurie (1957)distinguished two subspeciesgroups. ponent analysis(PCA) was used to analyze the un- The first, caeruleus,includes all Eurasianpopulations. transformed data (SAS software; for details on the The second,teneriffae, includes all populationsfrom method and its usein biology, seePielou 1977,Orloci North Africa (Maghreb and Cyrenaica), the Cana,-y 1978, Gauch 1982). This method is well suited to dis- Islands, and Pantelleria (a small Italian island lying closecharacter sets related to size and shapefactors off the Tunisian coast) (Brichetti and Violani 1986). (Morrison 1967),and to elucidatethe phenetic group- Vaurie (1957) recognized8 subspeciesin the caeruleus ings of individuals. The calculationswere made with group and 6 subspeciesin the teneriffaegroup pri- the individual male specimensas active elements. marily on the basisof color. However, to get a graphic reduction of the data to Differencesin colorationwithin the caeruleusgroup manageableproportions, I figured the averagecoor- are slight and clinal. The southern and westernmost dinates of the individuals collected in a given geo- forms are darker and more brightly colored than the graphicalunit in the principal componentspace (see northern and easternmost forms (see Snow 1954b, Fig. 1 for samplesizes in the 54 samplingunits). The 1955).In the teneriffaegroup, the groundcolor of the analysisof the characters'correlation circles in the backis slategray insteadof olive green,as in caeruleus, PCA plane providesa descriptionof the characters' and the head cap is of dark ultramarine blue instead "position"and "representativehess"on the plane (best of light blue; the light gray neckring is absent(Etch- for charactersplaced on the circle'scircumference). ecopar and Hiie 1964). In contrastto the caeruleus This allowsinterpretation of the axesin termsof phe- group, the six subspeciesof the teneriffaegroup are notypic characters. well defined by unambiguouscolor characters.The Clusteranalysis was used to get further insightinto geographicalrange of eachof them is clearly iden- between-samplesimilarities on a hierarchical basis. tified, correspondingto one or severaladjacent geo- First, I constructedclasses of sampling units by "Dy- graphicalisolates (Vaurie 1957,Etchecopar and Hiie namic Clouds" (Diday and Simon 1976), a method 1964 for the illustration). that is an extension of the k-means (McQueen 1967). Characters studied.--I measured 13 mensural and 9 For this I usedthe mean PCA coordinatesof the spec- plumage-patterncharacters of museumspecimens (see imens in the 54 sampling units on components1-4. Appendix). Most characterswere measuredwith dial The hierarchicalorganization of the classesof sam- calipersaccurate to 0.02min. The length of the tarsus pling units obtainedby "Dynamic Clouds"was built was measured with dividers and a ruler, the length by a hierarchicalcluster-analysis procedure using the of wing was measuredwith a ruler, and three plum- averagePC scoresof the sampling-unitclasses (CAH age-patterncharacters were quantified accordingto procedurein ADDAD software)(Roux 1985;also see charts.I estimated(a) the proportion of the green Sheath and Sokal 1973). wash on the back (1 = less than 1%, 2 = 2-5%, 3 = 6- 20%, 4 = 21-50%, 5 = 50-100%), (b) the importance RESULTS of the blackpatch on the belly (1 = absent,2 = small, 3 = large, 4 = very large) and (c) the proportion of VARIATION OF MENSURAL CHARACTERS white (versusyellow) on the belly and chest(1 = less than 1%, 2 = 2-10%, 3 = 11-40%, 4 = 41-80%, 5 = 81- The first two principal components summa- 95%, 6 = 96-100%). rize, respectively,29.2% and 25.6%of the vari- The sex of each specimen,its age (1 = juvenile, 2 ance (Fig. 2a). Principal components3 and 4 = one-year-oldadult, 3 = older than one-year-old eachsummarize < 10%of the variance. This per- adult), and the extent of plumage wear (3 classes) centagedrops to <6% for the remaining prin- were recorded.The collectiondate and the geograph- cipal components.Interpretation will concen- ical coordinatesof the locality were noted. I examined trate on the first two axes. skins of 2,184 Blue Tits. Correlationcircle analysis.--The correlation cir- Dataanalysis.--I divided the breedingrange of the Blue Tit into 54 units on the basis of a latitude-lon- cle (Fig. 3) defines three main groups of mensural characters. The first one has four charac- gitude grid (Fig. 1). Geographicalisolates (in general ters and is correlated with PC1. The characters islands)were always consideredas distinct sampling units. are wing length (2 characters:WINL and GCWT, Becauseof sexualdimorphism (Geys 1968a,b; Mar- see Appendix), which is the best estimate of tin 1988),analyses were performed separately for males individual size for this species(Snow 1954b), and females.I excludedjuveniles and birds with the tail length (TAIL), and bill depth (BILH). Spec- heaviest feather wear (class 3). The results were sim- imenswith high valuesfor thesecharacters have ilar for both sexes,but the samplesize of maleswas negative scoreson PC1. October1991] Variationin theBlue Tit 823

a. MORPHOLOGY b. PLUMAGE PATTERN

30' 30-

20-

10- •1 O- PC2 o

-0.4 I 3 5 7 9 11 13 I 3 5 7 9

AXES OF PRINCIPAL COMPONENT ANALYSIS Fig. 2. Percentof variancesummarized by the suc- cessiveprincipal componentaxes of the standardized -1 principalcomponent analysis performed on 13 men- 1 -0.4 0 0.4 1 sural characters(a) and nine plumage pattern cl•ar- PC1 acters (b). Fig. 3. Coordinatesof charactersin the PC1-PC2 correlation circle obtained in the standardized principal component analysis on mensural characters. The secondgroup of charactersis correlated Symbolsare defined in the Appendix. with PC2. The characters are PR1L and GCP1 (seeAppendix), both related to the length of the first primary,bill length (BILL), and tarsus cludes the specimens from Libya and the Ca- length (TARS).Specimens with high valuesfor nary Islands. thesecharacters have negativescores on PC2. Three morphologicaltypes are definedin the Becausethe length of the first primary is related plane between PC1 and PC2. The first type to the bird's maneuverabilityin the vegetation (negative scoreson PC1, samplesE4 and El) (Nachtigall and Kempf 1971, James1982), all includes specimensfrom the easternmostpart the charactersbelonging to this secondgroup of the species'breeding range (Figs. 1, 4). They can be related to foraging behavior. The third are associatedwith the charactersof group 1 in group of charactershas an intermediate posi- Fig. 3. They have a large size and a stout bill. tion relative to the two previousgroups. It in- The second morphologicaltype includes the cludes three characters related to toe size specimens from the large Mediterranean is- (MTON, HTON, HTOE; seeAppendix). Finally lands (Corsica [MI2], Sardinia [MI3], and Crete an isolatedcharacter, the differencein length [MI15]) and those from the southern Iberian between the first and secondprimary (P1P2), peninsula(SW2) (classB of clusteranalysis, Fig. showshigh correlationsin the plane between 4). These specimensare smaller for all the char- PC1 and PC2. actersmeasured (Fig. 3). Finally, the third mor- Between-specimensegregation by PCA.--Prin- phologicaltype is representedby the specimens cipal componentone (Fig. 4) segregatesinsular from the westernmostCanary Islands(La Palma sampling units (geographicalor ecologicalis- [CA1], Hiefro [CA2]; Fig. 4). They are slightly lands and mostpeninsular samples) from con- smallerthan average(wing and tail length) but tinental sampling units. All continental sam- have a long, rather thin bill, long tarsus,and pling units belong to the Eurasianpopulation long first primaries (absolute and relative complex(negative scores).PC2 segregatesthe length). Eurasiangeographical isolates (positive scores) The Eurasiansamples are distributedalong a from the North African geographicalisolates northeastto southwestgeographical cline. They (negative scores). range from large specimenswith a stoutbill in Seven classes(A to G) are defined by cluster the northeast(cluster D on Fig. 4) to small spec- analysison PC scores(Fig. 4). The hierarchical imenswith a relativelythinner bill in the south- tree segregatesthe specimensinto two main western (south of the Iberian peninsula and groups(Fig. 5). The first group (clustersA to D) large isolatedMediterranean islands; cluster B, includes the specimensfrom Eurasiaand from Fig. 4). Along that cline averagedistances be- the Maghreb, the second(clusters F to G) in- tween meansof adjacentsampling units in the 824 JEAN-LOUISMARTIN [Auk,Vol. 108

•-•.1- D O ,', ,I- • W3 W.•

/ N2C2 W2 S4 'El •E•3N4.Wl SE2 N301c4S2/ NE4 03 /

-2- E

! I -2 -1 LARGE SIZE, STOUT BILL PC1 SMALLSIZE, THIN BILL Fig. 4. Mean coordinatesof the specimensfrom the 54 sampling units in the PC1-PC2 plane of the standardizedprincipal component analysis performed on the 13 mensuralcharacters. Symbols are sampling units (envelopes)defined in Fig. 1, lettersA-G refer to the sampleclasses defined by clusteranalysis, and the brokenline separatesthe two main populationgroups defined by Vaurie (1957)(CAER = caeruleusgroup, Eurasia;TENE = teneriffaegroup, North Africa).

PC1-PC2 plane (Fig. 4) are smallestin the cen- of averagesize. The North African specimens tral part of the species'range (central, eastern from Libya (sampling unit CY) and from the and northern Europe;cluster D, Fig. 4). Overall two easternCanary Islands (CA6, CA7; Fig. 6) variationis smalldespite the huge geographical (clustersF and G on Fig. 4) have a small size, a area. Average distancebetween samplingunit thin but ratherlong bill, and ratherlong tarsus means tends to increase towards both ends of and first primary. Finally, the North African the cline, especiallyat the southernand south- specimensfrom the 5 western Canary Islands western ends of the species'breeding range (CA1 to CA5, Fig. 6; clusterE, Fig. 4) are slightly (peninsulasand islands,Figs. 1, 4). smallerthan average,like thoseof the Maghreb, The North African samplingunits, in marked but are characterizedby a long, rather thin bill contrast,show larger averagedistance between and long tarsusand first primary. sampling units and gapsbetween distinct mor- phologicalgroups (Fig. 4). Specimensfrom the VARIATION IN PLUMAGE PATTERN Maghreb (Algeria [NA1], Morocco [NA2], Tu- nisia [NA3]; part of cluster A, Fig. 4) are similar The first principal component summarizes to the specimensfrom the middle part of the 32.9%of the variance(Fig. 2b). Thereis a marked Eurasian cline (western pericontinental and gap between PC1 and PC2 (15.9% of the vari- south continental specimens,all included in ance).The decreasein variancesummarized by cluster A). Their body size is smaller than av- PC2 to PC7 showsno marked gap. erage,whereas bill, tarsus,and first primary are Correlationcircle analysis.--Characters GBAK October1991] Variationin theBlue Tit 825

CANARY ISLANDS Lapalma Tenerife •CA7 A B CD E F G •CAI•CA4,• Lanzarote •1• Gomera• Fuerteventu• Hierro CA3 Gran Canaria ,,,•l'•'r-• CA,?. CA5 L,] Fig. 6. Map of the Canaryarchipelago with name and symbol for each island.

Principal componenttwo segregatesthe Cur- asianspecimens along a geographicalcline go- ing from eastern Europe (negative scores;El, E2), to southwestern Europe (positive scores; MI2 to MI5, SW2). There is a progressive de- creaseof the proportion of white on the belly, the size of the white forehead patch and, to a Fig. 5. Hierarchicaltree linking the classesof sam- lesserextent, the black markings on the belly, ples defined by cluster analysis.Sample classesare the wing-bar width, and the amount of white defined in Figure 4. on the secondaries from east-northeast towards south-southwest.In the easternmost samples, (amount of green on the back, GNEC (width of the tail featuresare narrowly tippedwith white. the gray neck ring) (see Appendix), WBAR This clinal distributionhas a gapbetween spec- (wing-barwidth) and SWPA (width of the white imens from continental southern Europe (S3, patch on the first secondaries) are correlated S5, SW1) and specimensfrom the large Medi- with PC1 (Fig. 7). terranean islands (MI2, MI3, MI5) and southern A second set of characters is correlated with Iberian peninsula (SW2). The four exceptions PC2. They are WBEL (the amount of white rel- to this pattern are the specimensfrom the Bal- ative to yellow on the belly), FHPT (the width earic islands (MI1; Fig. 1), the Elburz region of the white patch on the forehead), BBEL(the size of the black patch in the middle of the belly) and the size of the white patch on the tail tip (TWPA). One character,BNEC (width of black neck ring), shows low correlationswith both PC1 and PC2. Betweensample segregation by PCA.--Principal component one (Fig. 8) segregatesthe 54 sampling units into two distinct subgroups.The PC2 o first group, with mostly negative scores,in- ~ cludesall Eurasianspecimens, characterized by the presenceof a gray neck ring, higher values -0.4 for the proportion of green on the back (close to 100% for all these specimens),and broader wing bars and more white on secondaries.The secondgroup (positive scoreson PC1) includes -1 the specimensof all North Africa. They are 1 -0.4 0 0.4 1 characterizedby the absenceof the gray neck PC1 ring, little or no green on the back (which is Fig. 7. Coordinatesof the plumage-patternchar- essentiallyslate gray except for the specimens acters on the PC1-PC2 correlation circle obtained in of Hiefro [Fig. 5] for which the slategray feath- the principalcomponent analysis on plumage-pattern ers are tipped with green), narrower wing bars, characters.For the definition of the symbolssee Ap- and less white on secondaries. pendix. 826 J•N-LouIs MARTIN [Auk, Vol. 108

TENE

0u. o.•-2 z • o ,•

o ,,0-3

I I I I I I -2 -1 0 1 2 3 4 HIGH PROPORTION OF GREEN ON BACK, BROADWING BARS PC 1 Fig. 8. Mean coordinatesof the specimensfor eachof the 54 samplingunits in the PC1-PC2plane of the principal componentanalysis performed on the nine plumage-patterncharacters. Symbols refer to the sampling unitsdefined in Figure1; letters A-H referto the sampleclasses (envelopes) defined by clusteranalysis. The broken line separatesthe two main populationgroups defined by Vaurie (1957)(CAER = caeruleusgroup, Eurasia;TENE = teneriffaegroup, North Africa).

(SE4;Fig. 1), the Greek islands (MI6; Fig. 1) and Maghreb (NA1 to NA3) and from the two east- western Turkey (S1; Fig. 1). ernmost Canary Islands (CA6, CA7). In contrast to the Eurasian condition, North The hierarchicaltree (Fig. 9) linking the eight African sampling units (positive scoreson PC 1) clustersdefined by clusteranalysis (Fig. 8) seg- were separated into three clear-cut groups by regates the Eurasian (classesD to H) from all the cluster analysis(Fig. 8). The sample from exceptone North African sample (classesB and La Palma (CA1; cluster A) in the Canary Islands C; Fig. 9). The exception(class A) is the birds represented a phenotypical extreme of exten- of Palma,which havecharacters typical of North sive white on the belly and white patches on African populationsand characterspresent in the tail tip. The other phenotypical extreme, the easternmostsamples from the Eurasian clusterB, includesspecimens from Hierro (CA2, group (classH on Fig. 9) one of the islands geographicallyclosest to La

Palma [CA1]), from Cyrenaica (CY) in Libya (the DISCUSSION most distant north African population from La Palma), and from the three central Canary Is- GEOGRAPHICAL VARIATION IN lands(CA3 to CA5). Thesespecimens have deep MENSURAL CHARACTERS yellow bellies with little or no white, and tail tips with very little or no white. The third North Variation in body size.--Clinal size variation African group (cluster C) has an intermediate among Eurasiansamples on PC1-PC2 (Fig. 4) is position and includesthe specimensfrom the consistentwith Bergmann's(1847) ecoclimatic October1991] Variationin theBlue Tit 827 rule, which states that homeotherms have an increased body size with increasing latitude. A B C D E F G H This may be related to heat economy(see maps of temperatureisotherms in January,Times At- I I las of the World 1967). The ecologicalsignificance of Bergmann'srule is widely debated (James1970, McNab 1971, Murphy 1985, Zink and Remsen 1986). James(1970) and Murphy (1985) suggestthat intraspecificsize variation in homeotherms can be related to a combination of climatic factorsthat include temperatureand moisture. In this case small size is associated with hot humid conditions, larger size with cooler or drier conditions. Data that fit Bergmann'srule (Snow 1954b, for parids) might also be explained by Geist's hypothesisthat relateslatitudinal size increase Fig. 9. Hierarchicaltree linking the classesof sam- in homeotherms to a latitudinal increase in the ples obtained by cluster analysis. ClassesA-H are defined in Figure 8. seasonalresource peak in phasewith the grow- ing period of the offspring (Geist 1987). Ac- cording to Geist, the seasonalpeak increases ture of the perching substrate(Grant 1971). A from 0 ø to 60 ø latitude and decreases thereafter. longer tarsusis associatedwith stableperches, The smaller size of Blue Tits in the Mediterra- a shortertarsus with unstableperches. nean is consistentwith Geist'shypothesis. In- The thinner bill in Mediterranean Blue Tit deed Cramm (1982) and Clamens (1987) have populationscould be relatedto the progressive shown that the resourcepeak used by Blue Tits changefrom habitatsdominated by deciduous to raisetheir young is lower in evergreenMed- trees(smooth foraging substrate)in the north- iterranean broad-leaved forests than in decid- easternpart of the species'breeding range to uous broad-leaved forests. Together with re- habitatsdominated by evergreen trees (harsh suits of long-term field studies (Blondel et al. foragingsubstrate) in the southwestof the range 1987, Clamens and Isenmann 1989), this sug- (seemap of naturalvegetation types in The Times geststhat there is a relationship between the Atlas of the World 1967). Differencesin wing lower spring resourcepeak in evergreen Med- shape, bill dimensions, or tarsus length be- iterranean habitats and the smaller adult size tween North African populationsmay similarly of Blue Tits in these habitats. Adult size de- be linked to differencesin the generalstructure creaseis strongestin populationsin evergreen of the habitat (e.g. broad-leavedversus conif- habitatsgeographically isolated from popula- erouswoodlands). In this context the segrega- tions living predominantlyin deciduoushabi- tion observedbetween the different sampling tats (Martin and Bellot 1990). The smallest Blue units from the Canary Islands is informative. Tits are found in the mostarid regionsof North The specimensfrom the two driestislands (Lan- Africa, which could result from even lower lev- zarote and Fuerteventura, where the tits nest in els of available resources in these habitats. The semi-aridhabitats [Bacallado 1976] and especial- contribution of environment and geneticsto ly in Tamarix canariensiswoods [Bannerman size variation remains an open question. 1963])group with thosefrom Libya (Figs.4 and Variationin•bill and leg dimensions.--Vegetation 5). In Libya, Blue Tits also live in semi-arid hab- structureand prey typesare essentialelements itats dominatedby short-leavedtrees or bushes in the processof morphologicaladaptation to (Juniperuswoods [Bundy 1976] but also histor- foraging. In addition to the relationship be- ically cypressand pine [Hartert 1923]). In con- tween bill size/shape and prey size/shape or trast, the Blue Tits from the middle Canary betweenwing shapeand flight, paridsthat for- Islandsbreed mainly in Pinuscanariensis (a long- age in broad-leavedtrees have stouterbills than leaved pine) and in laurel forestsdominated by those in coniferous habitats (Snow 1954a, 1955; Visnea mocanera. In the outer islands the birds Lack 1971). Further, the length of the tarsus are restrictedalmost completely to pine woods relative to body weight is related to the struc- (Lack and Southern 1949, Cullen et al. 1952, 828 JEAN-LOuISMARTIN [Auk, Vol. 108

I-IerrLingsen1963, Grant 1979).Accordingly, the fully adapting to Mediterranean biotopes.This Blue Tits of the middle and outer Canary Islands interpretation is strengthened by the observed have the longest tarsus and the longest and increasein biotopesused by Blue Tits on Corsica thinnest bills within the species. Indeed, the where they are common breedersin matoffals more a population is restrictedto pine the lon- (Blondel 1981, 1985; Martin 1982; Blondel et al. ger the bill and tarsus (PC1, Fig. 4; see also 1988). Grant's (1979) analysis of ecologicaland mor- Consistencywith taxonomyand phylogeny.-- phologicalvariation of CanaryIsland BlueTits). There is little consistencybetween the classi- Lack and Southern (1949) suggestedthat the fication of the samplesbased on similarity in longer and thinner bill of the Blue Tits from mensural characters(Fig. 5) and the classifica- Tenerife was the result of convergencetowards tion proposedon the basisof plumage charac- the morphologyof Coal Tits (Parusater). How- ters (e.g. Vaurie 1957).The dichotomybetween ever, Partridge and Pring-Mill (1977) rejected Eurasianand North African population groups the convergence hypothesis on experimental basedon plumage color does not appear in my grounds.Together with Snow (1954a),they at- analysisof mensural characters(Figs. 4 and 5). tributed the increase in bill and tarsus length The specimensfrom the Maghreb group with in the Canarian Blue Tits to Allen's rule. This these from Eurasia(Fig. 4, cluster A; Fig. 5). rule claims there is a decrease in the extremities Further, mensural charactersimilarity can be of homeotherms as latitude increases, an ob- greater between geographically distant popu- servation also explained by body temperature lations that face similar ecological conditions regulation. Allen's rule seemsan unlikely ex- than between geographicallyclose populations planation. Indeed, the increasein bill and tarsus subjectto different ecologicalconditions (Cher- length from mainland North Africa to the outer noff 1982, Dillon 1984). The resemblance be- Canary Islands (Fig. 4) can hardly be related to tween the Maghrebbirds and populationsfrom latitudinal effects because the difference in lat- southwestern Eurasia, despite their probable itude between the Canary Islandsand the near- different phylogenetichistory, is an example. est African populations is less than 2 degrees, The similarity of the birds from Cyrenaicato and among the Canarian populations this dif- the geographicallydistant specimens from the ference is lessthan 0 degrees. inner CanaryIslands is anotherexample of ap- The hierarchical tree that links the clusters parent morphological convergence.The hier- of samplingunits (Fig. 5) displaysstrong con- archicalclassification of the samplingunits (Fig. sistencybetween overall samplesimilarity and 5) splits the sample into two main groups of similarity in habitat. Group one (clustersA to clustersaccording to environmental similarities D) containsthe groups that forage on broad- and independentof the probablephylogeny. leaved trees (deciduousor evergreen). All samples from populationsthat forage in evergreen GEOGRAPHICAL VARIATION IN broad-leaved trees fall in either cluster A or B. PLUMAGE PATFERN Group two (clusr•'•sE to G) comprisesbirds that forage either on narrow-leaved trees such as In Eurasia geographical plumage-pattern Tamarix,Juniperus, and cypress(clusters F and variation (Fig. 8) is consistentwith the known G), or more or lessexclusively on long-needled variation in plumage color (Snow 1954b, 1955). trees (Pinuscanariensis) (cluster E). Snow (1954b) evoked climatic variation to ex- Therole of geographicalisolation.--In both Eur- plain geographical variation in plumage col- asian and North African population group,,, oration in the BlueTit (and in paridsin general). there is a relation between sudden changesin In arid climateswith cold winters birds are pal- body-partdimensions (gaps on Fig. 4) and geo- er and grayer on the back and have lessyellow graphicalisolation. The importanceof isolation pigmentation. They are darker and of brighter in the emergenceof geographicalvariation in color in warm wet climates. In accord with this parids was pointed out by Snow (1954a).He explanation the extent and purity of white suggestedthat the range of biotypesused by markings increasesfrom wet warm climatesto the Maghrebpopulations is larger than that of arid climates with cold winters. The peculiari- populationsof southernEurope because the lack ties of the Elburz specimens(SE4) could be re- of geographical isolation has prevented the lated to the wetter climate of this area. Variation populationsfrom Mediterranean Europe from in plumage pattern in Eurasiais essentiallyclin- October1991] Variationin theBlue Tit 829 al and no gapsappear (Fig. 8) that relate to the responsesof similar genotypesto different en- subspeciesdefined by Vaurie (1957). vironments. Such variation may be enhanced The questionwhether geographicalvariation in its genetic component by geographicaliso- is adaptive or the consequencesof nutritional lation (on the large Mediterranean islandsfor factorson the processof pigmentationremains instance). Finally, variation within the North open. For example, Mediterranean Blue Tits African group is mainly linked to genetic drift reared in captivity have reduced yellow pig- within each of the isolatesor group of adjacent mentation(pers. obs.). Nongenetic variation in isolates. Similarities in coloration on the basis plumage color in relation to habitat and diet of phylogeneticrelationships alone seem un- has also been documented for the Great Tit (Par- likely becauseof the lack of consistencybe- usmajor) (Slagsvoid and Lifjeld 1985).Changes tween geographicalproximity and similarity in in coloration have also been reported due to plumage pattern (Figs. 8 and 9). increasedbleaching of the carotenoid pigment in the tropical magpie Cissaliving in open hab- RELEVANCE TO BIOSYSTEMATICS itatswhen comparedwith populationsof dense forests (Vevers 1985). In theory, two biosystematicalinterests can In the North African populations(teneriffae be served by this kind of study. One is to di- group;Vaurie t957), the basiccolor pattern fol- agnose phylogenetic relationships between lows the general trend of darker birds in the populations.The secondis to use the analysis southwest of the speciesrange (Gloger 1933). of the presentvariation as a pattern to illumi- However, the changein plumage-colorpattern nate the geography of speciation. Principal across the Strait of Gibraltar is dramatic and is components analysis of mensural characters not paralleledby any equivalentchange in eco- showsthat the datapresented better reflectecol- climatic conditions. ogy than population phylogeny. Principal com- Among North African populations various ponentsanalyses of plumage pattern could do isolatesdiffer in suchcharacters as the presence no more than substantiatephylogenetic segre- or absenceof white wing bars, the presenceor gationsthat were alreadyclear from the general absenceof white on the belly, or the presence differences in coloration between North Afri- or absenceof a green wash on the back. Each canand Eurasianpopulations. More satisfactory geographicalisolate or group of adjacent iso- answersin phylogenywill necessitatebiochem- lates is characterizedby certain features used ical analyses(reviews of Zink and Reinsent 986, by taxonomiststo identify subspecies.These Hillis 1987). featuresappear independent of thoseobserved More positively, my results substantiatethe in neighboring populationsand unrelated to importanceof the unifying effectof gene flow climatic or environmental determinants. For in- betweenadjacent nonisolated samples even over stance,cluster analysis groups specimens of the a large geographicalarea. Its corollarythat geo- driest Canary Islands with those from the Ma- graphicalisolation is necessaryto makepossible ghreb. Alternatively the specimensfrom very the appearanceof gapsin clinal variation or of dry Cyrenaica are more similar to those from markedly different population characteristics the wetter middle and outer CanaryIslands (ex- was strongly supported.Such studies,especial- cept for Palma). Clusters A, B, and C (Fig. 8) ly if supplementedby knowledgeof thebiology includesampling units that belongrespectively of the species,can lead to a better understand- to one subspecies(CAt), two subspecies(NAt, ing of the evolutionary implicationsof a spe- NA2, NA3 and CA6, CA7) and three subspecies cies' population structure and of its geograph- (CY; CA2; and CA3, CA4, CA5). ical, environmental, and historical determinants. I believe that plumage-patternvariation shows that the general differences observed between North African and Eurasian samples directly ACKNOWLEDGMENTS reflect phylogeny (i.e. they are due to the di- This study would not have been possiblewithout vergenceof two geneticallyisolated population the help of the curatorsand assistantsof the M•zseu groups). Secondly,variation within the Eur- de Zoologiaof Barcelona;Humboldt Museum f•ir Na- asian group is essentially linked with clinal turkunde,Berlin; Alexander Koenig Museum, Bonn; changein local conditions.Differences may rest Institut Royal des SciencesNaturelies, Brussells;Hun- on genetic differencesor result from different garianNatural History Museum,Budapest; Zoologisk 830 Jr•.N-LOUISMARTIN [Auk,Vol. 108

Museum,Copenhagen; Staatlisches Museum f/JrTier- ßD. CHESSEL,& B. FROCHOT.1988. Bird species kunde, Dresden; ForschungsinstitutSenckenberg, impoverishment,niche expansion,and density Frankfurt; Martin Luther Universit/it, Halle; E1/iintie- inflation in Mediterranean island habitats. Ecol- den Museo,Helsinki; Zoological Institute, Academy ogy 69: 1899-1917. of Sciences,Leningrad; Rijksmuseum van Natuurlijke ß A. CLAMEl,IS, P. CRAMM, H. GAUBERT,& P. Historie, Leiden; Civico Museo di Storia Naturale, ISENMANN. 1987. Population studies on tits in Milano; Zoologische Staatssammlung, Munich; the Mediterranean region. Ardea 75: 21-34. American Museum of Natural History, New York; BOAG,P.T. 1983. The heritability of external mor- Museum d'Histoire Naturelie, Paris; National Natural phology in Darwin's ground finches (Geospiza) History MuseumßSofia; Naturhistoriska Riksmuseet, on Isla Daphne MajorßGalapagos. Evolution 37: Stockholm; Staatlisches Museum f/Jr Naturkunde, 877-894. Stuttgart;Museo Civico di Storia Naturale, Terrassini; BRICHETTI,P., & C. VIOLANI. 1986. Une population British Museum (Natural History), Tring; Naturhis- nicheusede (Paruscaeruleus ultramarinus Bp.) sur torisches Museum, Vienna; National Museum of Nat- l'ile de Pantelleria (Canal de Sicile). L'Oiseau et ural History, SmithsonianInstitutionß Washington; RFO 56: 77-81. PeabodyMuseum of Natural History, Yale; Museo BUND',',G. 1976. The birds of Libya. BOU checklist Ornithologico F. Foschi,Forli; and Collection Priolo, No. 1. London, Br. Ornithol. Union. Catania.B. Massa and C. Violanihelped in organizing CLAMENS, A. 1987. RSle de la nourriture dans la the search for some specimens.I benefited from a d•termination des param•tresd•mographiques FrankM. ChapmanMemorial Fund grantand fellow- des M•sanges (Aves, Paridae) en ch•naie verte. ship (AMNH, New York). I thank J. Blondel, A. Cap- Alauda 33: 254-266. parella, M. Debussche,A. Dhont, F. B. Gill, P. Isen- ß& P. ISENMANN.1989. Effectof supplemental mann, J. Woodbury, M. Lecroy, D. Snow, and R. M. food on breeding of Blue and Great tits (Parus Zink for criticismsand suggestionsat variousstages caeruleusand Parusmajor) in Mediterranean hab- of the manuscript. itats. Ornis Scandinavica 20: 36-42. CRAMM,P. 1982. La reproductiondes m•sanges dans LITERATURE CITED une ch•naie verte du Languedoc.L'Oiseau et RFO 52: 347-360. ALATALO, R. V., & L. GUSTAFSSON. 1988. Genetic com- CHERNOFF,B. 1982. Charactervariation among pop- ponent of morphological differentiation in Coal ulations and the analysisof bitgeography. Am. Tits under competitiverelease. Evolution 42: 200- Zool. 22: 425-439. 203. CULLEN, J. M., P.E. GUITON, G. A. HORRIDGE,& S. BACALLADO,J.J. 1976. Notas sobre la distribution y PIERSON. 1952. Birds on Palma and Goreera (Ca- evolucion de la avifauna canaria.Pp. 413-431 in nary Islands). Ibis 24: 282-323. Bitgeographyand ecologyin the Canary Islands DHOND?,A. A. 1982. Heritability of Blue Tit tarsus (G. Kunkel, Ed.). The Hague, Junk. length from normal and crossfostered broods. BANNERMAN, D. A. 1963. Birds of the Atlantic Is- Evolution 36: 418-419. lands,vol. I. A History of the birds of the Canary DIDAY,E., & J. C. SIMON. 1976. Clustering Analysis. Islandsand of the Salvages.Edinburgh and Lon- Pp. 47-94 in Communicationand cybernetics10, don,Oliver & Boy..d. digital pattern recognition(K. S. Fu, Ed.). Berlin, BERGMANN, C. 1847. Uber die Verh/iltnisse des W/ir- Springer Verlag. mefkonomie der Tiere zu ihrer Grfsse. Gfttinger DILLON,R. T., JR. 1984. Geographicdistance, envi- Studien 1: 595-708. ronmental difference and divergence between BETrS, M.M. 1955. The food of titmice in oak wood- isolatedpopulations. Syst. Zool. 33: 69-82. lands. J. Animal Ecol. 24: 282-323. ETCHECOPAR,R. D., & F. HOE. 1964. Les oiseaux du BLONDEL,J. 1969. Syn6cologiedes passereauxr6si- nord de l'Afrique. Paris,Boub•e. dents et migrateurs dans un 6chantilion de la FRETWELL,S. D. 1969. Ecotypicvariation in the non- r6gion M6diterran•enne frangaise.Th•se •s Sci- breedingseason in migratorypopulations: a study ences.Marseille, Centre R•gional de Documen- of tarsallength in someFringillidae. Evolution 23: tation P•dagogique. 406-420. 1981. Structureand dynamicsof bird com- GAUCH,H.G.,JR. 1982. Multivariate analysisin com- munities in Mediterranean habitats.Pp. 361-385 munity ecology. Cambridge studies in ecology. in Maquis and chaparrals (F. Di Castri, D. W. Cambridge, UK, Cambridge Univ. Press. Goodall, and M. L. Specht, Eds.). Collection Eco- GEIST,V. 1987. Bergmann'srule is invalid. Can. J. systemsof the World. Amsterdam, Elsevier. Zool. 65: 1035-1038. --. 1985. Habitat selection in island vesus main- GEes,J.F. 1968a. Vleugeln staartlengte bij de Pim- land birds. Pp. 477-516 in Habitat selection in pelmees(Parus caeruleus L.). Natuurwet. Tijdschr. birds (M. L. Cody, Ed.). New York, Academic 50: 174-185. Press. --. 1968b. Beak-measurements in the Blue Tit October1991] Variationin theBlue Tit 831

(Parus caeruleusL.). Biol. Jaarboek Dodonaea 36: ß, & H. N. SOUTHERNß 1949. Birds on Tenerife. 150-158. Ibis 91: 607-626. GLOGER,C.L. 1933ßDas Ab•ndern der V6gel durch LEISLER,B. 1980. Morphologicalaspects of ecological Einfluss des Klimas. Breslau. specializationinbird genera. •kol. Vfgel.2: 199- GRANT, B. R. 1985. Selection on bill characters in a 220. population of Darwin's finches:Geospiza coniros- ß, & H. WINKLER.1985. Ecomorphology.Curr. tris on Isla Genovesa,Galapagos. Evolution 39: Ornithol. 2: 155-186. 523-532. MARTIN, J. L. 1982. L'infiltration des oiseaux fores- GRANT,P. R. 1971. Variation in the tarsuslength of tiers dans les milieux buissonnants de Corse. Rev. birdsin islandand mainlandregions. Evolution Ecol. (Terre Vie) 36: 397-419. 25: 599-614. ß 1988. Variation g•ographique,adaptation et --.. 1972. Bill dimensionsin three speciesof sp•ciation: l'exemple de Paruscaeruleus (Aves). Zosteropson Norfolk Island. Syst.Zool. 21: 289- Th6se,Universit• des Scienceset Techniquesdu 291. Languedoc,Montpellier, France. 1979. Ecologicaland morphologicalvaria- ß & g.D. BELLOT. 1990. Variation in mor- tion of Canary Island Blue Tits, Paruscaeruleus phology,laying dateand clutchsize between non- (Aves: Paridae). Biol. J. Linnean Soc. 11: 103-129. Mediterranean and Mediterranean Blue Tits. Pp. HARTERT,E. 1910. Die Vbgel der pal•arktischenFau- 157-164 in Populationbiology of passerinebirds: na, I. Berlin, Friedl'•nder. an integratedapproach (J. Blondel,A. G. Gosler, 1923. On the birds of Cyrena'ica.Novit. Zool. J. D. Lebreton, & R. H. McCleery, Eds.). Heidel- 30: 1-32. berg-Berlin,Springer Verlag. 1932-1938. Die Vbgel der pal•iarktischen MAYR,E. 1969. Principlesof systematiczoology. New Fauna. Erg'fi•nzungsband.Berlin, Friedl•inder. York, McGraw Hill. HEMINGSEN, A.g. 1963. Birds on Hierro and the MCNAB,B. K. 1971. On the ecologicalsignificance relation of number of speciesand of abundances of Bergmann'srule. Ecology52: 845-854. and body weights, to island area. Vidensk. Medd. MCQUEEN,J. B. 1967. Some methods for classifica- Dansk Naturhist. Fbren. 125: 207-236. tion analysisof multivariate observations.Proc. HERRERA,C.g. 1978. Niche-shiftin the genusParus 5th BerkeleySymp. 1: 233-248. MORRISONßD.F. 1967. Multivariate statistical meth- in southern Spain. Ibis 120: 236-240. ods. New York, McGraw-Hill. 1981. Combinationrules among western Eu- ropean Parusspecies. Ornis Scandinavica12: 140- MURPHY,E. C. 1985. Bergmann'srule, seasonalityß 147. and geographic variation in body size of House Sparrows.Evolution 39: 1327-1334. HESPENHEIDE,H.A. 1973. Ecologicalinferencesfrom NACHTIGALL,W., & B. KEMPF. 1971. Vergleichende morphologicaldata. Annu. Rev.Ecol. Syst. 4: 213- 229. Untersuchungenzur flugbiologischenFunktion des Daumenflttichs(Alula spuria)bei V/Sgeln.Z. HIL•IS, D.M. 1987. Molecular versusmorphological Vergl. Physiologie71: 326-341. approachesto biosystematics.Annu. Rev. Ecol. ORLOCI,L. 1978. Multivariate analysisin vegetation Syst. 18: 23-42. research,2rid ed. The Hague, Junk. ISENMANN,P. 1987. Geographicalvariation in clutch- PARTRIDGE,L. 1976. Someaspects of the morphology size: the exampleof the Blue Tit (Paruscaeruleus) of Blue Tits (Paruscaeruleus) and Coal Tits (Parus in the Mediterraneanarea. Vogelwarte 34: 93- ater) in relation to their behaviour. J. Zool., Lon- 99. don 179: 125-133. JAMES,F.C. 1970. Geographicvariation in birdsand ß& F. PRING-MILL. 1977. Canary Island Blue its relation to climate.Ecology 51: 365-390. Tits and EnglishCoal Tits: convergentevolution? 1982. The ecologicalmorphology of birds: Evolution 31: 657-665. a review. Ann. Zool. Fennici 19: 265-275. PERRINS,C. 1979. British tits. London, Collins. 1983. Environmental component of mor- PIELOU,E. C. 1977. Mathematicalecology, 2rid ed. phological differentiation in birds. Science221: New YorkßWiley. 184-186. ROCKWELL,R. F., & G. F. BARROWCLOUGHß1987. Gene JOHNSON,N. K. 1966. Bill size and the question of flow and the genetic structure of natural popu- competition in allopatric and sympatricpopula- lations. Pp. 224-255 in Avian genetics.Londonß tions of Dusky and Gray flycatchers.Syst. Zool. Academic Press. 15: 70-87. Roux,M. 1985. Alogorithmesde classification.Paris, KOKSHAYSKY,N.V. 1973. Functionalaspects of some Masson. detailsof bird wing configuration.Syst. Zool. 22: SCHLUTER,D., & J. •N•.M. SMITHß 1986. Natural se- 442-450. lection on beak and body size in the Song Spar- LACK,D. 1971. Ecologicalisolation in birds. Oxford, row. Evolution 40: 221-231. Blackwell Sci. Publ. SCHOENER,T.W. 1965. The evolution of bill size 832 JFAI•-LouIsMARTIN [Auk, Vol. 108

differencesamong sympatric congeneric species ZI•qK,R.M. 1989. The study of geographicvariation. of birds. Evolution 19: 189-213. Auk 106: 157-160. $LAGSVOLD,T., & J. T. LIFJELD. 1985. Variation in --, & J. V. REMSEIqJR. 1986. Evolutionary pro- plumage color of the Great Tit Parusmajor in cessesand patterns of geographic variation in relation to habitat, season and food. J. Zool. 206: birds. Curr. Ornithol. 4: 1-69. 321-328. SLATKIN,M.K. 1981. Estimatinglevels of gene flow in natural populations.Genetics 99: 323-335. APPENDIX. Definition of the mensural and plumage 1987. Gene flow and the geographic struc- pattern characters measured. Mensural characters: ture of natural populations.Science 236: 787-792. I = bill characters,II = wing characters,III= tail SMITH, J. N.M., & R. ZACH. 1979. Heritability of length, IV = hind limb characters;plumage pat- some morphological charactersin a Song Spar- tern: I = white markings,II = belly and breastchar- row population. Evolution 33: 460-467. acters, III= neck characters. SI•EATH, P. H. A., & R. R. SOKAL. 1973. Numerical taxonomy. San Francisco,Freeman and Co. Symbol Character definition SNow, D. W. 1954a. The habitat of Eurasian tits (Pa- Mensural characters rus spp.). Ibis 96: 565-585. I BILL Bill length 1954b. Trends in geographicalvariation in BILH Bill depth II WINL Wing length Palearcticmembers of the genusParus. Evolution GCWT Greater covert tip to wing tip 8: 19-28. GCP1 Greatercovert tip to 1stprim. tip 1955. Geographicalvariation of the Coal Tit PR1L First (outermost)primary length Parus ater L. Ardea 43: 195-226. PIP2 First primary tip to 2nd primary tip THE TIMESATLAS OF THE WORLD. 1967. Edinburgh, P2WT Tip of 2nd primary to wing tip III TAIL Tail length The Times of London and J. Bartholomew and IV TARS Tarsus length Son; Boston,Houghton Mifflin Company. MTON Length of middle toe nail vAN NOORDWIJK,A. J., J. H. vAN BALEN,& W. SCHARLOO. HTOE Length of hind toe HTON Length of hind toe nail 1980. Heritability of ecologicallyimportant traits in the Great Tit. Ardea 68: 193-203. Plumage pattern VAURIE,C. 1957. Systematicnotes on the Palearctic I FHPT Width of foreheadwhite patch birds, number 26, Paridae. The Parus caeruleus WBAR Width of wing bar SWPA Width of the white patch on tail tip complex. Am. Museum Novit. 1833: 1-15. II BBEL Size of the black patch on belly VEVERS,H. G. 1985. Colour. Pp. 99-100 in A dictio- WBEL Proportion of white breast& belly nary of birds (B. Campbell and E. Lack, Eds.). III GNEC Width of gray neck ring BNEC Width of black neck ring Calton, T. & A.D. Poyser.