The Auk 111(1):80-86, 1994

TAIL ORNAMENTATION, SIZE DIMORPHISM AND WING LENGTH IN THE (PLOCEINAE)

STAFFAN ANDERSSON 1 AND MALTE ANDERSSON Departmentof Zoology,University of Gb'teborg,Medicinaregatan 18 S-41390 G•teborg,Sweden

AlaSTRACT.--Sexualwing dimorphism in relation to tail ornaments and body size was studiedin the strikingly sexuallydimorphic widowbirds and bishops(Euplectes) of the African tropics.Seven widowbirds grow long tails, varying from 7 cm in Fan-tailedWidowbird (E. axillaris)to 0.5 m in the Long-tailedWidowbird (E. progne).Aerodynamic drag increaseswith tail length, and adaptationsto compensatefor this cost might be expected(e.g. increasing wing length), a prediction that was supportedamong the widowbirds. After controlling for overall size dimorphism (estimated by tarsus length), 70% of the variation in residual wing dimorphism among the widowbirds was explained by tail dimorphism. Bishopswere more dimorphic in wing length than expected,which may be related to their slow display flight. The resultssuggest caution in using wing length alone for interspecificcomparisons of sexual size dimorphism. Basedon tarsuslength, the lekking E. jacksoniis more size dimorphic than the average of its congeners, in contrast to what has been concluded in previous studies basedon wing length. Received13 May 1993,accepted 2 July 1993.

SEXUALLY SELECTED characters, such as the which all rectricesare elongated) compared to elongated tails or the exaggerateddisplays of pin tails and fork tails (Balmford et al. 1993). some , are expected to evolve until their As predicted based on higher energy costs of mating advantagesare balancedby costs(Fisher flight, sunbirds with experimentally elongated 1930), like predation or physiological con- tails spent lesstime flying than controls (Evans straints (Harvey and Bradbury 1991, Andersson and Hatchwell 1992). Based on a correlation be- 1994). At this dynamic equilibrium, morpho- tween wing length and tail length in one of the logical or behavioral adaptations that reduce (Nectarinia johnstoni), Evans and Hatch- the costs (hence, improve survival and allow well (1992) suggestedthat the sunbirds com- further exaggeration of the trait) can be ex- pensatefor the energetic costsof a long tail by pected to evolve. increasing wing span. Two such indirect effects of sexual selection Likewise, intraspecificrelationships between on avian morphology have recently been re- ornamental tail length and wing length have ported. Hedenstr6m and M611er (1992) used been found in the Long-tailed Widowbird (Eu- aerodynamic theory to predict morphological plectesprogne; Craig 1989) and Jackson'sWidow- adaptationsto the demandsof songflight. Con- (E. jacksoni;Andersson 1992a), suggesting sistentwith their predictions,they found great- similar compensationsfor the cost of carrying er sexual dimorphism in wing span and wing long tails. In particular, the exceptionallylong area in eight passerineswith song flight than wings of the Long-tailedWidowbird have made in related specieswithout this (supposedlysex- it an outlier in comparative studies of sexual ually selected) behavior. Evans and Thomas size dimorphism based on wing length (Payne (1992) estimated the aerodynamiccosts of elon- 1984) gated tails in three speciesof sunbirds (Nectar- In this paper, we investigate the possibility inia spp). An increased tail area was shown to that sexualdimorphism in wing length in eight affectthe aerodynamicsin severalways, but most widowbird speciesis a sexually selected adap- importantly increasethe drag on the bird, which tation to compensate for the aerodynamic costs is a major component of flight cost (Norberg of tail ornamentation. We also discuss the re- 1990). This cost should be particularly strong lationship between sexual size dimorphism es- for birds with graduated tail ornaments(i.e. in timated from wing-length data and that esti- mated from tarsuslength, and the relationship • Presentaddress: Department of Biology 0116, Uni- between size dimorphism and body size. Sexual versityof Californiaat SanDiego, La Jolla,California selection and, hence, sexual dimorphism are be- 92093, USA. lieved to be particularly strong in lek mating

80 January1994] SexualDimorphism in Widowbirds 81 systems, but the comparative evidence is in- through female choice, but in some casesmay conclusive,and exceptionsabound (Payne 1984, also function asagonistic signals. The long tails H6glund 1989, Trail 1990, Oakes 1992). For ex- are of the graduatedtype that should handicap ample, the only lek breeder in this genus,Jack- flight considerably(Balmford et al. 1993).In this son'sWidowbird, was reported as no more size paper we test the prediction that adaptationsto dimorphic than its resource-defensepolygy- reduce such costs have evolved in the widow- nous congeners (Payne 1984, HtSglund 1989, birds. Oakes 1992). These comparisonswere basedon wing lengths, and we discussthe validity of METHODS these conclusions and the new patterns that Table 1 gives the biometric data used for each spe- emerge from using tarsuslengths instead. cies. Data were obtained by S.A. primarily from the The widowbirds (formerly Coliuspasser)to- collection of bird skins in the National Museums of gether with the bishops form the genus Eu- , Nairobi, and (for E. hartlaubipsammocromius) plectes,a group of 16 small (13-45 g) granivorous in the British Museum of Natural History, Tring. weaverbirds (subfamily Ploceinae) of the Af- Specimensthat could not unequivocally be assigned rican tropics(Delacour and Edmond-Blanc1933, by sex or specieswere excluded. As some samples, Hall and Moreau 1970). All speciesare similar particularly of females,were very small or equivocal in general biology and behavior (Craig 1980) (uncertain sexing and/or speciesidentification), we included data from the literature (Maclean 1988, Craig and, obviously, closely related as hybridization 1989). For tarsus and tail length of albonotatusand occurs in captivity (Colahan and Craig 1981). macrourus,MacLean (1988) did not give sample sizes. All Euplectesare strikingly sexually dimorphic Measurementsof aferand fransiscanuswere provided in plumage. Breeding males have black body by A. J. F. K. Craig. The smallestsamples included plumage with patchesof bright red, yellow, or are for gierowiifemales (2), and macrourusmales (4) white. The male bishopsare the mostbrilliantly and females(4). All other samplesare larger than five. colored and have a short tail, while male wid- Due to lack of data, three species (Golden-backed owbirds replace the tail feathersduring the pre- Bishop [E. aureus],Fire-fronted Bishop [E. diadematus] breeding molt, and grow graduated ornamental and Red Bishop[E. orix])were excludedfrom the anal- tails of various lengths. Females are in most ysis.Most Euplectesare widespreadin Africa and some have geographically isolated populations and sub- casesnoticeably smaller, streakybuff or brown, species(Hall and Moreau 1970,Mackworth-Praed and and difficult to distinguish in the field. Non- Grant 1985). When applicable, the full names of the breeding malesresemble the femalesbut in some subspeciesincluded in this analysisare given in Table speciesretain part of the nuptial coloration on 1. For one species,the Long-tailed Widowbird, there rump or wings. were large sample sizes for more than one of the Habitat preferencesrange from reed bedsand geographicallyisolated populationsin Craig (1989), savanna with sparsebushes through to open of which we randomly selected one (E. prognedela- ,and probably all speciesare polyg- merei)by throwing a die. However, we alsomade sure ynous (Craig 1980).Except in the lek-breeding that inclusion of any of the other two did not signif- Jackson'sWidowbird, males defend nesting ter- icantly affect the results.The scientificand common names used follow Sibley and Monroe (1990). ritories and participate to some extent in nest On the museum skins, tarsuslength was measured building, but not (or very rarely) in incubation with calipers as the distance from the middle of the or feeding. intertarsal joint to the distal edge of the last complete The tail ornamentsvary greatly in size, from tarsal scale (i.e. before the toes begin). Wing (chord) the slightly elongated7-cm tail in the male Fan- length was measuredwith a ruler as the distance be- tailed Widowbird (E. axillaris) to the 50-cm tween the bend of the wing and the tip of the longest plumes trailing behind the male Long-tailed primary. The dry wing chords of the museum skins Widowbird--more than four times his body could not be flattened against the ruler to the extent length and one of the most extreme avian that is common practice when measuring live birds adornments. For the Long-tailed Widowbird (Svensson1975), producing a slight underestimate(1- 2 mm) of absolutewing length, but not affecting the (Andersson 1982) and the lekking Jackson's sexualdimorphism as the sexeswere treated equally. Widowbird (20-cm tail; Andersson 1989, 1992b) For tail length a thin ruler was positionedunder the there is evidence that tail length is a cue used upper tail covertssuch that it stoppedat the baseof by femalesin mate choice.It seemsreasonable the feathers, thus providing a measurement of dis- to assumethat the long tailsof male widowbirds tance from there to the tip of the longest rectrix. generally are sexually selected, probably We calculated relative dimorphism as the natural 82 ANDERSSONAND ANDERSSON [Auk, Vol. 111

the genus, jacksoni,is interesting with respect 0.4 ß pr to whether lek mating promotes strong sexual dimorphism. In three comparative studies (Payne 1984, H6glund 1989, Oakes 1992) with different approaches but all based on wing- length data, jacksoniwas included as being no •c•O.Z- ar•a •l•ca more dimorphic than its congeners.For exam- ple, Oakes (1992) used continuous data and compared dimorphism of lekkers with the av- 0.0 erage dimorphism of their congeners.As in his results,we found no difference in wing dimor- Body*size(tarsus-length) dimorphism phism betweenjacksoni and its congeners(tn = Fig. 1. Sexual dimorphism (In[male trait/female 0.03, P = 0.97), but tarsusdimorphism did differ trait]) in wing length regressedon sexualdimorphism (tn = -2.2, P = 0.046). Even so, it is doubtful in tarsuslength (indicating body size) for 13 Euplectes whether this should be considered as support- species.Bishops denoted by open and widowbirds by ing a relationship between lekking and sexual filled circles. First letters of speciesnames given (see Table 1). Test of regression:Fl,•2 = 7.1, P = 0.022, R2 sizedimorphism in this genus(jacksoni still only = 0.39. ranks 5); however, it clearly shows that the in- clusion of tarsuslength may yield rather dif- ferent conclusionsin interspecific comparisons log of [male trait/female trait] (e.g. Gaulin and Sailer of size dimorphism.Tarsus length may be more 1984, BjSrklund 1990). As an estimateof speciessize suitable than wing length in this respect, as it we used female size (i.e. natural log of female tarsus is a fixed skeletal trait and, generally, a better length). The bivariate relationships were explored index of avian body size (Freeman and Jackson with linear regressions,and tested with analyses of 1990, Bj•rklund 1991). However, like any other variance and Spearman rank correlations. Unfortu- nately, there is no phylogeny available for Euplectes, univariate measure of external morphology, which is why we treated the speciesas independent tarsuslength should also be used with caution. data points. Depending on the degree of evolutionary A composite measure based on a number of inertia of the investigated traits and the details of the skeletal characterswould be preferable (Free- true phylogeny, this might have unduly inflated our man and Jackson 1990), but it was not available sample size (e.g. see Harvey and Pagel 1991), but we in the present study. Unfortunately, skeletons still find the trends worth reporting. are generally rarely preserved in the museum collections. RESULTS AND DISCUSSION Sexual size dimorphism in many in- creaseswith body size (Andersson 1994). How- Due to an aerodynamically functional rela- ever, there was no such trend in Euplectes,or in tionship between wing span and body mass, bishops or widows treated separately (tarsusdi- wing-length tends to increase relative to body morphism vs. In[female tarsuslength]: bishops, size within and among bird species(Norberg r, = 0.50, n = 5, P = 0.32; widows, r, = -0.04, 1990). As expected,sexual dimorphism in wing n = 8, P = 0.92; all, r, = 0.37, n = 13, P = 0.20). length among the widowbirds and bishopswas All else being equal, the costof flying should positively related to dimorphism in tarsuslength increasewith the length (and hencearea) of the (Fig. 1). However, the fit was not very close(R 2 tail, and males might through adaptationhave = 0.39) and, although some of the remaining countered this cost by increased wing span variation certainly is due to measurement er- (Evans and Thomas 1992). A simple test of this rors, biased sampling, etc., other factors also prediction is to regress the residual wing di- seem to be involved. morphism from Figure 1 on the sexual dimor- This suggeststhat wing dimorphism might phism in tail length (Fig. 2). The regressionfor be a misleading index of sexual size dimor- all species(line a in Fig. 2) is significant but phism. For example,the three top-ranking spe- doesnot explain more than one-half of the vari- ciesas regards to wing dimorphism (progne,hart- ation (R 2 = 0.51), mainly because the bishops laubi,and axillaris,respectively) rank 3, 7, and showedstronger wing dimorphism than would 1, respectively,in size dimorphism asestimated be expected from their short tails. The heavy from tarsuslength. The only lekking speciesin dependence of the regression on two species, January1994] SexualDimorphism in Widowbirds 83 84 ANDERSSONAND ANDERSSON [Auk, Vol. 111

O.Z ß pr•, tures in most species of Euplecteshas not been / o documented in enough detail for even a ten- E ha• //// (a) tative ranking of the aerodynamic demands on morphology, but a few intriguing observations are worth noting. First, all bishops (including the presumed widowbird capensis,also called the ) seem to perform some ver- Oho /// ar sion of a "bumble flight" (Craig 1980);the col- ored feathers of the back are "erected into a

0 af // ax great brilliant puff as the birds drone slowly acrosstheir territories on rapidly and noisily vibrating wings" (Emlen 1957).The costsof for- Tail dirnorphism ward flight increasedramatically when speed Fig. 2. Residualsexual dimorphism in wing length decreasesfrom normal transport towards slow (from regressionon body size dimorphism; Fig. 1) flight and hovering (Norberg 1990). Decreased plotted against sexual dimorphism in (nuptial) tai! wing loading through longer and broaderwings length. Separateregression lines refer to: (a) a!l spe- is a common morphological adaptation to slow cies;(b) the "subgenus"of widowbirds (filled circ!es); flight (Norberg 1990),for examplein many har- and (c) widowbirdswith flight display(i.e. excluding riers (Circusspp.) with slow searchflights. The jacksoni).First letters of speciesnames given (see Ta- slow bumble flight, thus, may be one reason for ble 1). Tests of regressions:(a) F..,• = 11.3, P = 0.006, R2 = 0.51; (b) F,.7 = 15.1, P = 0.008, R the cluster of bishops at a residual wing di- = 34.2, P = 0.002, R 2 = 0.87. morphism similar to the medium-tailed widow- bird species (Fig. 2). It would be particularly suggestiveif the flight display of nigroventrisis hartlaubiand progne,is evident in Figure 2 and either more laborious or performed more fre- from the nonsignificant rank correlation (r, = quently than those of the other bishops and, 0.29, P = 0.31). conversely,if the oppositewere true for afer. Restricting the relationship to the widow- The shorter than expectedwings of jacksoni birds (among which males of all species but may be related to the lack of a flight display in capensishave elongated nuptial tails), the rela- this species. As earlier suggested, the longer tionship is stronger (line b in Fig. 2) with 72% wings in nuptial compared to subadult males of residual wing dimorphism being explained could be a morphological adaptation to the cost by tail dimorphism. The rank correlation bor- of the jump display (Andersson 1993) and the dered on significancein spite of the small sam- costof carrying the tail in transportflight. How- ple (r• = 0.71, P = 0.059). This is consistentwith ever, the costsof the slow display flight in, for the prediction that widowbird males have example, the similarly sized progne,might be a evolved longer wings (thereby increasing wing stronger selection pressure on wing span. In area and decreasingwing loading) to compen- any case, when we restricted the analysis to sate for the cost of the long tail. Even if this include only display-flying widowbirds (line c interpretation turns out to be correct, however, in Fig. 2), the proportion of variance explained it doesnot explain the relatively long wings of was 87%. most male bishops, and the shorter than ex- If the true phylogeny of Euplecteshad been pected wings of jacksoni. known and controlled for in the above regres- Apart from elongated tails, another likely se- sions, the sample sizes (i.e. independent evo- lection pressure on nuptial wing morphology lutionary events) might have been smaller, per- is the type and extent of flight display. Heden- haps making the relationships statistically str6m and M611er(1992) gave a detailed account nonsignificant. No phylogeny for this genus is of the influence of various morphological traits, available. Hall and Moreau (1970) suggesteda including wing span, on different aspects of number of "superspecies" (e.g. hartlaubi and passerinesong flight. Among other things they progneas one, and macrourusand albonotatusas predicted that if flapping-flight endurance is another). However, they based this on similar- important, selection should favor increased ities in male nuptial plumage suchas tail length, wing span.The nature of the flight display and and controlling for this phylogeny would by the strength of sexual selection for display fea- definition deflate our data. For example, progne January1994] SexualDimorphism in Widowbirds 85 and hartlaubipsammocromius might then have ANDERSSON, S. 1989. Sexual selection and cues for long wings and tails because their recent an- female choice in leks of Jackson'sWidowbird Eu- cestor evolved these traits (i.e making it one plectesjacksoni. Behav. Ecol. Sociobiol.25:403-410. and not two independent associationsof the ANDERSSON,S. 1992a. Lek mating and sexual selec- two traits). On the other hand, within both the tion in Jackson'sWidowbird (Euplectesjacksoni). superspeciespairs--hartlaubi/progne and mac- Ph.D. dissertation, Univ. G6teborg, G6teborg, Sweden. rourus/albonotatus--the positive relationship ANDERSSON,S. 1992b. Female preference for long between wing and tail dimorphism is support- tails in lekking Jackson'sWidowbirds: Experi- ed. mental evidence. Anim. Behav. 43:379-388. Another alternative interpretation of the re- ANDERSSON,S. 1993. Sexual dimorphism and modes lationship between tail and wing dimorphism of sexual selectionin lekking Jackson'sWidow- is that it results from a genetic correlation be- birds Euplectesjacksoni. Biol. J. Linn. Soc.49:1-17. tween different aspectsof plumage growth (i.e. BALMFORD,A., A. L. R. THOMAS,AND I. L. JONES. 1993. that selection for longer male tails produces a Aerodynamicsand the evolution of long tails in [nonadaptive] increase in male wing length). birds. Nature 361:628-631. Taking it one step further, the entire variation BJ•RKLUND,M. 1990. A phylogeneticinterpretation in tail and wing length might be attributed to of sexualdimorphism in body sizeand ornament genetic drift. This possibility could not be re- in relation to mating systems in birds. J. Evol. Biol. 3:171-183. futed by Savalli (1993), who applied the test by BJ•RKLUND,M. 1991. Patternsof morphologicalvari- Turelli et al. (1988) to Euplectestail lengths. It ation among cardueline finches (Fringillidae: seems, however, unlikely that drift alone can Carduelinae). Biol. J. Linn. Soc. 43:239-248. explain the tenfold range in Euplectestail COLAHAN,B.D., AND A. J.F. K. CRAIG.1981. Euplectes lengths. hybrids. Ostrich 52:58-59. In sum, much of the interspecificvariation in CRAIG,A. J. F. K. 1980. Behaviour and evolution in widowbird wing length seemsto be the result the genusEuplectes. J. Ornithol. 121:144-161. of sexually selectedadaptations to compensate CRAIG,A. J. F. K. 1989. Tail length and sexualse- for costsincurred by tail ornamentation. As a lectionin the polygynousLongtailed Widow (Eu- consequence,wing lengths may give mislead- plectesprogne): A cautionarytale. S. Afr. J. Sci.85: ing estimatesof sexual dimorphism in overall 523-524. body size among widowbirds and bishops.This DELACOUR.J., AND F. EDMOND-BLANC. 1933. Mono- might apply also to other avian groups with graphiedes veuves (Revision des genres Euplectes et Vidua). Oiseau Rev. Fr. Ornithol. 3:519-562, aerodynamically handicapping displays or or- 687-726. naments. EMLEN,J. T., JR. 1957. Display and mateselection in ACKNOWLEDGMENTS the whydahs and bishop birds. Ostrich 28:202- 213. We thank the staff at the Nairobi Museum's Orni- EVANS,M. R., AND B. J. HATCHWELL. 1992. An ex- thology Department, and the British Museum of Nat- perimental study of male adornment in the Scar- ural History (Tring) for accessto the collections of let-tufted Malachite Sunbird: II The role of the bird skins, Mats-Ola Larssonfor practicalhelp, and elongatedtail in mate choiceand experimental A. Balmfordand two anonymousreviewers for valu- evidencefor a handicap.Behav. Ecol. Sociobiol. able commentson the manuscript.A. J. F. K. Craig 29:421-427. commentedon an earlier version of this manuscript, EVANS, M. R., AND A. L. R. THOMAS. 1992. The aero- and alsoprovided measurements for E. aferand fran- dynamicand mechanical effects of elongatedtails siscanus.Financial support to S.A. camefrom the Ernst in the Scarlet-tufted Malachite Sunbird: Measur- Collianders,Hierta-Retzius, Kungl. och Hvitfeldtska, ing the costof a handicap.Anim. Behav.43:337- and J. A. Wahlbergsfoundations, Wenner-Gren Cen- 347. ter, and the Royal Society of Arts and Sciencesin Gothenburg. M.A. was supported by the Swedish FISHER,R.g. 1930. The geneticaltheory of natural selection. Clarendon Press, Oxford. Natural Sciences Research Council. FREEMAN,S., AND W. M. JACKSON. 1990. Univariate LITERATURE CITED metricsare not adequateto measureavian body size. Auk 107:69-74. ANDERSSON,M. 1982. Female choice selects for ex- GAULIN, S. J. C., AND L. D. SAILER. 1984. Sexual di- treme tail length in a widowbird. Nature 299: morphism in weight among the primates: The 818-820. relative impactsof allometry and sexual selec- ANDERSSON, M. 1994. Sexual selection. Princeton tion. Int. J. Primatol. 5:515-535. Univ. Press,Princeton, New Jersey.In press. HALL, B. P., AND R. E. MOREAU. 1970. An atlas of 86 ANVœRSSON^Nr)ANr)ERSSON [Auk,Vol.

speciationin African passefinebirds. British Mu- OAKES,E.J. 1992. Lekking and the evolution of sex- seum of Natural History, London. ual dimorphismin birds:Comparative approach- HARVEY, iv. H., AND J. W. BRADBURY. 1991. Sexual es. Am. Nat. 140:665-684. selection.Pages 199-233 in Behaviouralecology: PAYNE, R. B. 1984. Sexual selection, lek and arena An evolutionary approach (J. R. Krebs and N. B. behavior, and sexual size dimorphism in birds. Davies, Eds.). Blackwell, Oxford. Ornithol. Monogr. 33. HARVEY,IV. H., AND M.D. IVAGEL1991. The compar- S^V^LLI,U. M. 1993. An application of the neutral ative method in evolutionary biology. Oxford model to the evolution of tail length in the genus Univ. Press, Oxford. Euplectes(Aves, ). Evolution 47:696-699. HEDENSTR•M,A., AND A. IV.MtSLLER. 1992. Morpho- SIBLEY,C. G., AND B. L. MONROE, JR. 1990. Distri- logical adaptationsto song flight in passefine bution and of the birds of the world. birds: A comparative study. Proc. R. Soc. Lond. Yale Univ. Press, New Haven, Connecticut. B. Biol. Sci. 247:183-187. SVENSSON,L. 1975. Identification guide to European H'6GI•IJNI:),J. 1989. Size and plumage dimorphism passetines. Naturhistoriska Riksmuseet, Stock- in lek-breeding birds: A comparative analysis. holm. Am. Nat. 134:72-87. TRAIL, IV. W. 1990. Why should lek-breeders be M^CKWORTH-PRAED,C. W., AND C. H. B. GRANT. 1985. monomorphic?Evolution 44:1837-1852. Birds of easternand north-easternAfrica. Long- TURELLI,M., J. H. GILLESPIE,AND R. LANDE 1988. Rate roans, London. testsfor selectionon quantitativecharacters dur- M^C•,F•N, G.L. 1988. Roberts' birds of southern Af- ing macroevolution and microevolution. Evolu- rica. New Holland Publishers, London. tion 42:1085-1089. NORBERG,U. M. 1990. Vertebrate flight. Springer- Verlag, Berlin.