Ornis Hungarica 2012. 20(1): 1–25. doi: 10.2478/orhu-2013-0001 Sexual selection, range size and population size

ANDERS PAPE MØLLER1 & LÁSZLÓ ZSOLT GARAMSZEGI2

Anders Pape Møller & László Zsolt Garamszegi 2012. Ivari szelekció, elterjedési terület és populációméret. – Ornis Hungarica 20(1): 1–25.

Abstract Sexual selection may impose fi tness costs on both males and females due to the costs of developing and maintaining exaggerated sexual signals, reducing average fi tness in strongly sexually selected species. Such reductions in average fi tness could affect local extinction risk and hence distribu- tion range. However, given that both sexually monochromatic and dichromatic species are common and widesp- read, benefi ts of sexual selection must be invoked to maintain equilibrium. We tested for differences in breeding range size and population size between monochromatic and dichromatic species of birds in a comparative analysis of species from the Western Palaearctic. In an analysis of standardized linear contrasts of the relationship between sexual dichromatism and range size and population size, respectively, that controlled for similarity among taxa due to common descent, we found no signifi cant relationship. However, when we analyzed carotenoid-based se- xual dichromatism sexually dichromatic species had larger distribution areas and higher northernmost distribution limits, but not southernmost distribution limits than sexually monochromatic species. In contrast, melanin-based sexual dichromatism was not signifi cantly associated with range size or population size. Therefore, population density of sexually dichromatic species with carotenoid-based coloration was lower than that of monochromatic species, because dichromatic species had similar population sizes but larger ranges than monochromatic species. These fi ndings suggest that the different physiological roles of pigments associated with sexual dichromatism have effects on total range size of birds.

Keywords: birds, carotenoids, melanin, sexual dichromatism, sexual selection

Összefoglalás Az ivari szelekció kihatással lehet az egyedi rátermettségre mind hímeknél, mind a tojóknál, mert a másodlagos nemi jellegek kifejlesztése és fenntartása bizonyos költségekkel terheltek, melyek visszahatnak a rátermettségre az erősen ivari szelekció alatt álló fajoknál. A rátermettségben megmutatkozó költségek további befolyással bírnak a helyi extinkciós rátára, és így az elterjedési területre. Ennek ellenére úgy tűnik, hogy az iva- rilag monokromatikus és dikromatikus fajok gyakoriak és elterjedtek, így az ivari szelekció előnyei egyensúlyt teremtenek a költségekkel. Jelen komparatív vizsgálatban azt teszteltük a nyugati Palearktikus régióban költő madaraknál, hogy az ivarilag monokromatikus és az ivarilag dikromatikus fajok elterjedési területe és populáció- mérete is különbözik-e. A lineáris standardizált kontrasztok módszerét használva, amikor a fajok közötti rokon- sági kapcsolatot is számításba vettük, nem találtunk összefüggést a vizsgált változók között. Amikor azonban csak a karotin alapú színezetre fókuszáltunk, kiderült, hogy a dikromatikus fajoknak nagyobb és északabbra nyúló elterjedési területe van, mint a monokromatikus fajoknak. Ezzel szemben, a melanin alapú színezetre nem találtunk ilyen összefüggést. Az eredményekből még arra is következtetünk, hogy a karotin alapon dikromatikus fajok populációs denzitása alacsonyabb, mint karotin alapon monokromatikus fajoké, mert a dikromatikus fajok hasonló populációmérettel bírnak, mint a monokromatikus fajok. Összegzéskeppen elmondhatjuk, hogy valószí- nűleg a különböző pigmentekhez kapcsolódó fi ziológiai mechanizmusok különböző szerepet játszanak az ivari dikromatizmus és az elterjedési területek kapcsolatának fenntartásában a madaraknál.

Kulcsszavak: madarak, karotinoidok, melanin, ivari dikromatizmus, ivari szelekció

1Laboratoire d’Ecologie, Systematique et Evolution, Université Paris-Sud, Orsay France, 2Department of Evolu- tionary Ecology, Estacion Biologica de Donana-CSIC, Seville, Spain, e-mail: [email protected] 2 ORNIS HUNGARICA 2012. 20(1)

Introduction consequences for the variance in individual contributions to populations. Second, popu- Sexual selection arises from the fi tness ad- lations consist of individuals that differ in vantages of certain individuals over others their degree of sexual ornamentation, and in competition for mates, resulting in the this should have consequences for the local evolution of exaggerated secondary sexual risk of extinction. Accordingly, Doherty et characters (Darwin 1871). While the vari- al. (2003) have shown for bird census data ance in individual mating success increases from North America that local extinction risk as a consequence of sexual selection, this and local turnover rate are greater for sexu- increase in variance may also have impor- ally dichromatic than for monochromatic tant implications for population processes. species. Therefore, it is not surprising that For example, an increase in the variance in Doherty et al. (2003) for North American reproductive success may increase demo- birds and Prinzing et al. (2002) for Europe- graphic stochasticity with consequences for an birds did not fi nd a relationship between extinction risk (Sæther et al. 2004). Fur- sexual dichromatism and population trends thermore, average fi tness of individuals of because local extinctions would be expected strongly sexually selected species may be to be balanced by a high local turnover rate. reduced compared to species subject to less Third, given that species differ inherently in intense sexual selection, and such load due the costs and benefi ts of sexual selection, to sexual selection will invariably suppress we should expect sexually dichromatic spe- population size relative to that expected cies to run greater risks of extinction than in the absence of sexual selection (Tanaka monochromatic species. Indeed, McLain 1996). However, the world is full of wide- et al. (1995, 1999) and Sorci et al. (1998) spread and common monochromatic and have shown for introduced birds to oceanic dichromatic species, implying that sexual islands that the risk of immediate extinc- selection may also be advantageous. Given tion is elevated for dichromatic compared to that the proportion of sexually dichromatic monochromatic species, even when control- species is variable among taxa, and that di- ling for potentially confounding variables chromatism has evolved numerous times such as inoculate size. (Price & Birch 1996), we can assume that While numerous studies have investi- benefi ts as well as costs are present, main- gated the effects of sexual selection on fi t- taining the frequency of sexual dichroma- ness components at the level of individuals, tism at an equilibrium level. relatively few studies have investigated the The effects of sexual selection on popu- effects of sexual selection at the population lation processes should be visible at, at or species level. Here we test the prediction least, three different levels. First, individu- that population size, distribution range and als should differ in their ability to cope with northernmost and southernmost distribu- the costs of sexual selection, with mating tion limits differ between species that vary success, fecundity and viability being re- in the intensity of sexual selection. We used lated to the expression of secondary sexual sexual dichromatism as a proxy for sexual characters as predicted by models of condi- selection, given that mating success (Gon- tion-dependent secondary sexual charac- tard-Danek & Møller 1999) and fertilization ters (Andersson 1994). This should have success are positively related to sexual di- Anders Pape Møller & László Zsolt Garamszegi 3 chromatism within species (Møller & Ninni for carotenoid and melanin based coloration 1998). Similar patterns also occur among to test explicitly if the population consequen- species (Andersson 1994, Møller & Birk- ces of sexual selection differed between head 1994, Petrie et al. 1998). If sexual these two pigment categories. selection imposed signifi cant average costs upon individuals of a species, we would expect population size more often to be Materials and methods suppressed in sexually dichromatic than in monochromatic species. Likewise, if such Study species costs of sexual signals were present at the population level, we would expect marginal We included all bird species with a main populations of sexually dichromatic spe- breeding distribution within the Western cies more often to go extinct (Doherty et al. Palaearctic (Cramp & Perrins 1977–1994) 2003), resulting in a reduction in range size that resulted in a sample of 526 birds. of sexually dichromatic species. We tested these predictions by analyzing range size Sexual dichromatism and population size of the breeding birds of the Western Palaearctic because reliable in- We scored the breeding plumage of all spe- formation is readily available for the entire cies as sexually monochromatic if males and fauna. We have chosen range size and popu- females did not differ in coloration accord- lation as proxies of the effects of sexual se- ing to information provided by the descrip- lection at the species level, because these tions in Cramp and Perrins (1977–1994), traits can be measured in a standard way in and otherwise as sexually dichromatic. a large number of species We assume that This procedure was repeated separately for interspecifi c differences in range size, dis- carotenoid- and melanin-based coloration. tribution limits and population size refl ect We distinguished carotenoid-based sexual the outcome of population processes that monochromatism and dichromatism rely- occur due to local extinction risk and demo- ing on colors that were yellow, orange and graphic stochasticity that may be affected red as caused by carotenoids (see Tella et by the interspecifi c variance in reproductive al. 2004, Olson & Owens 2005 for similar success. criteria). For melanin-based coloration we Sexual coloration can be based on pig- included all colors that were brown, black ments or structural color, and pigment-based or reddish brown as typical for coloration coloration can be due to carotenoids or me- based on phaeo- and eu-melanin (see also lanins. Previous studies have implicated Gray 1996, Olson & Owens 2005). such pigments in various physiological func- tions such as free radical scavenging and im- Population size mune function (e.g. McGinness et al. 1970, Krinsky 1989, 1998, Rózanowska et al. Population sizes were obtained from Burfi eld 1999, Møller et al. 2000, Moreno & Møller and van Bommel (BirdLife International 2006), suggesting a trade-off between plum- 2004), who reported the total number of bree- age coloration and physiological function. ding pairs in the Western Palaearctic west of Therefore, we assessed sexual dichromatism the Ural Mountains, estimated in a consistent 4 ORNIS HUNGARICA 2012. 20(1) way from national bird census programs in all range sizes based on the equation above countries. We used the mean of the minimum were strongly positively correlated with es- and maximum estimates in that source. timates based on image analysis of breeding distributions of birds in the Western Palae- Range size and distribution limits arctic as reported in the electronic version of Cramp and Perrins (1977–1994) (r = 0.52, We estimated total geographical breeding N = 60, P < 0.001, Møller et al. unpublished range size as the area of the shape bounded information). by the greatest span of latitude and longi- The entire data set is provided in the ap- tude of each species’ entire breeding range, pendix. as published in Cramp and Perrins (1977– 1994). We extracted the northernmost, Comparative analyses southernmost, easternmost and westernmost distribution limits for the entire breeding Analyses of comparative data based on spe- range to the nearest 0.1 degree from the dis- cies may provide misleading conclusions, tribution maps in Cramp and Perrins (1977– if sister taxa are more similar with respect 1994). To take into account the curvature of to the variables under investigation than the earth (which was assumed to be spheri- randomly chosen species, and if species cal), this area was estimated by the equation richness differs considerably between cate- 2 Area = R × (Longitude1 – Longitude2) × gories of species such as monochromatic

(sin(Latitude1) − sin(Latitude2)) where R is and dichromatic species. We used statisti- the radius of the earth (6366.2 km) and lati- cally independent standardized linear con- tude and longitude are expressed in radians. trasts (Felsenstein 1985), which controls for We used the northernmost and the southern- similarity among species due to common most distribution limits as estimates of dis- descent to test the predictions. Contrasts tribution limits. were calculated using the software of Purvis In widespread species Old and New World and Rambaut (1995), implemented in the ranges were calculated separately and sub- computer program CAIC. Standardization sequently summed in order to obtain more of contrast values was checked by examina- precise estimate on range sizes at the global tion of absolute values of standardized con- level. The method over-estimates the real trasts versus their standard deviations (Gar- geographical range, but the error should be land 1992, Garland et al. 1992). Plotting the random with respect to the variables under resulting contrasts against the variances of test. Estimates of area were strongly posi- the corresponding nodes revealed that these tively correlated with geographical range transformations made the variables suitable size as calculated by counting one-degree for regression analyses. grid cells overlain on published distribu- We log10-transformed distribution area tion maps for a sample of 20 Palaearctic and and population size before analyses. Nearctic bird species (r = 0.87, P < 0.001), Given that sexual dichromatism was a and with range size as reported for a sam- dichotomous variable we used the Brunch ple of 11 threatened species (Stattersfi eld & procedure in CAIC to identify all indepen- Capper 2000) (r = 0.98, P < 0.001, based on dent contrasts for nodes where transitions log-transformed data). Likewise, estimated occurred in sexual dichromatism. At these Anders Pape Møller & László Zsolt Garamszegi 5 nodes positive contrasts in breeding distri- Nakagawa (2004), who emphasized the im- bution or population size imply that they portance of unbiased reports of effect sizes. vary in the same direction as sexual dichro- We used the software Comprehensive Meta matism. Using a t-test, we tested whether Analysis (BioStat, 2000, http://www.meta- the mean of these contrasts differed from analysis.com/) to calculate effect sizes and zero, as expected for correlated evolution of corresponding confi dence intervals (CI). traits. To address problems of outlier con- Comparative analyses rely on a phyloge- trasts, we analyzed the distribution of con- netic hypothesis for identifying independent trasts with a non-parametric test (Signed- contrasts due to a transition from one kind rank test), which provided equivalent results of sexual coloration to another. We used a to those obtained from parametric tests. As composite phylogeny created by using in- more breeding pairs may be present if the formation from Sibley and Ahlquist (1990). distribution range is larger, we also calcu- This phylogeny for higher taxa was supple- lated population size while controlling for mented with information from other sources breeding distribution. We controlled statisti- to resolve relationships between species cally for this problem in a phylogenetically (Randi et al. 1991a, b, Sheldon et al. 1992, adjusted regression model of population size Seibold et al. 1993, Sheldon & Winkler on breeding density. For this phylogeneti- 1993, Suhonen et al. 1994, Wittmann et al. cally adjusted regression, using the Crunch 1995, Blondel et al. 1996, Badyaev 1997, procedure in CAIC, we regressed contrasts Leisler et al. 1997, Slikas 1997, Cibois & for the two continuous variables through the Pasquet 1999, Kimball et al. 1999, Sven- origin. Then, we fi tted the slope to the raw sson & Hedenström 1999, Voelker 1999, species data, and calculated residuals from Johnson & Clayton 2000, Kennedy et al. this regression line. These residuals, repre- 2000, Sheldon et al. 2000, Geffen & Yom- senting breeding density, were later ana- Tov 2001, Johnson et al. 2001, Møller et al. lyzed by using the Brunch procedure to test 2001, Barker et al. 2002, Dimcheff et al. for any effect of sexual dichromatism. 2002, Donne-Goussé et al. 2002, Broders et To determine the strength and direc- al. 2003, Riesing et al. 2003, Cibois & Crac- tion of the relationship between dichroma- raft 2004, Kruckenhauser et al. 2004, Tho- tism and distribution and population size, mas et al. 2004, Voelker & Spellman 2004, we estima ted effect sizes (such as Cohen’s Lerner & Mindell 2005, Webb & Moore sensu Cohen 1988), and the associated 95% 2005). We applied branch lengths from the confi dence intervals (CI) for each particular phylogeny of Sibley and Ahlquist (1990) for phylogenetic relationship. We preferred re- higher taxonomic levels. Within families the porting and focusing on effect sizes, instead distance between different genera was set to of using Bonferroni correction and signifi - 3.4 ∆T50H units, and between species within cance levels, because the latter approach has genera to 1.1 ∆T50H units (Sibley & Ahlquist been criticized in the fi eld of ecology and be- 1990, Bennett & Owens 2002). The phylo- havioural ecology due to mathematical and genetic hypothesis for the 526 species used logical reasons (Perneger 1998, Moran 2003, in the comparative analyses is available in Nakagawa 2004, Garamszegi 2006). There- “nexus” format as supplementary material. fore, to balance between Type I and II errors, Availability of information for different we followed the recent recommendations of variables varied, and hence sample sizes 6 ORNIS HUNGARICA 2012. 20(1) differed for the statistical tests. Although we detected strong effects for a phylogenetic had a huge sample size in terms of number pattern, with range size being larger for taxa of species, when using the Brunch proce- with sexually dichromatic carotenoid-based dure, contrasts could be calculated for nodes coloration compared to taxa with mono- only, where transition (from 0 to 1 or from chromatic coloration (Fig. 1, Table 1). 1 to 0) occurred in sexual dichromatism. The association between carotenoid-based These transition events are the focus of the sexual dichromatism and range size was due current study. to an increase in the northernmost distribution limit, whereas the southernmost distribution limit did not show strong differences between Results monochromatic and dichromatic species (Fig. 1, Table 1). In contrast, there was no such ef- Analyzing transitions from monochromatic fect at a robust magnitude for melanin-based to dischromatic coloration, we failed to fi nd sexual dichromatism or for overall sexual di- strong effects for the relationship between chromatism (Fig. 1, Table 1). range size and sexual dichromatism (Table Effect sizes corresponding to overall 1). That was also the case when the analysis population size and density were generally was restricted to melanin-based sexual di- small, indicating that these traits did not ap- chromatism (Fig. 1, Table 1). In contrast, we pear to differ considerably between mono-

Fig. 1. Mean (SE) contrasts associated with carotenoid-based and melanin-based sexual dichroma- tism and range size, northernmost and southernmost distribution limit, population size, and density. Numbers are sample sizes. 1. ábra Az egyes változókra (elterjedési terület, legészakabb és legdélebb elterjedési határvonal, populációméret és denzitás) számolt átlagos (SE) stan