Female ornaments revisited – are they correlated with offspring quality?
The evolution and signalling content of female ornamentation has remained an enduring challenge to evolutionary biologists, despite the fact that secondary sexual characters are widespread in females. While females usually invest significant amounts of their resources, including carotenoids, in offspring, all the resources allocated to elaborate ornamentation reduce resources available for other purposes. This may in turn constrain female fitness leading to dishonest female signalling.
We review the literature for empirical studies on mutually ornamented species with conventional sex roles, by focusing on the association between female ornaments and quality of their offspring.
We found 43 papers where 33 (77%) are bird-studies, nine (21%) are on fishes, and one (2%) is a lizard-study. Nine of these report negative, 14 non-existing, and 20 positive associations between female ornament and offspring quality. Eighteen of the bird studies (55%) show a positive association between the two traits investigated, whereas five (15%) of the studies report a negative association. The number of fish studies, although few, is skewed in the opposite direction with two (22%) and four (44%) studies supporting positive and negative association, respectively. A minority of studies on carotenoids-based ornaments reports a positive association (4 of 18 studies, or 22%) between the traits, which is low compared to studies on non-carotenoids-based ornaments (16 of 25 studies, or 64%).
The above-mentioned relative large number of studies with negative association, especially common in studies on fishes and in carotenoids-based-ornaments, challenges the generality of the direct selection hypothesis to account for female fineries. This is important because this hypothesis seems to have strong support in recent literature on the topic. In the present paper, we also propose possible explanations for the observed differences between taxa and suggest directions and ideas for future research on the evolution of female ornamentation.
Despite the fact that secondary sexual characters are widespread in females, the evolution and signalling content of female ornamentation remain enduring challenges to evolutionary biologists (Amundsen 2000; Kraaijeveld, Kraaijeweld-Smit & Komdeur 2007; Clutton-Brock 2007). It has been hypothesized that female ornaments signal the genetic quality of individuals (e.g. Zahavi 1975), or direct benefits such as parental care or other non-genetic maternal resources (e.g. Blount, Houston & Møller 2000; Massaro, Davis & Darby 2003; Gladbach et al. 2010). On the other hand, the resources allocated to ornamentation may reduce the resources available for offspring and thus constrain female fitness (Fitzpatrick, Berglund & Rosenqvist 1995; Price 1996; LeBas, Hockham & Ritchie 2003; Chenoweth, Doughty & Kokko 2006), thereby leading to dishonest female signalling (Funk & Tallamy 2000; Bonduriansky 2001). Thus, a male choosing to fertilize the eggs from an elaborately ornamented female might sire fewer or poorer quality offspring than his rivals choosing to mate with the drabber female, which has allocated relatively larger proportion of her resources into the eggs (Fitzpatrick, Berglund & Rosenqvist 1995). The outcome of these conflicting demands is suggested to depend on whether the value of information revealed by the signal outweighs the fecundity cost of the ornament (Fitzpatrick, Berglund & Rosenqvist 1995; Chenoweth, Doughty & Kokko 2006; LeBas 2006; Watson & Simmons 2010). This in turn relies on the ornament-fecundity relationship which may emerge as positive, negative or non-existent (e.g. Nordeide, Rudolfsen & Egeland 2006; Watson & Simmons 2010). In species with a negative association between ornaments and egg quality or fecundity, the potential information benefit of the female ornament (i.e. improved marketing value status) is not expected to outweigh the fitness cost of the ornament.
Two main hypotheses have been proposed to explain the evolution of female ornaments. The ‘direct selection hypothesis’ predicts that female ornaments are under direct sexual selection by males or selection via reproductive competition among females (Amundsen 2000; Kraaijeveld, Kraaijeveld-Smit & Komdeur 2007). Thus, according to this hypothesis, female ornaments are honest signals of some aspects of individual quality. The alternative hypothesis, the ‘genetic correlation hypothesis’, in turn suggests that female ornamentation is a genetically correlated response to selection for male ornamentation. This idea originally gained some support already from Darwin (1871). Later, Lande (1980) suggested that female ornamentation in mutually ornamented species may just be a temporal stage in the evolution of male ornaments. Lande (1980) hypothesized that in the first stage of the evolutionary process, females show mating preferences for ornamented males. In the second stage, mutual ornamentation evolves since both sexes share most of the genome (and thus also genes responsible for ornamentation). In the last stage, females evolve mechanisms to suppress the expression of their ornament, for example by down regulating the genes coding for the ornament, leaving only males ornamented. This final stage may be driven by sexual conflict and costs associated with the female ornaments (Lande 1980).
Negative association between the intensity of female carotenoids-based ornamentation and female fitness, supporting the genetic correlation hypothesis, has a strong theoretical basis (e.g. Fitzpatrick, Berglund & Rosenqvist 1995; Chenoweth, Doughty & Kokko 2006). Despite this, reviews on the evolution of female ornaments so far have mainly promoted a positive association between ornament and offspring quality and thus given presumably stronger support for the direct selection hypothesis. For example, Amundsen (2000) argues that ‘…female ornaments have evolved quite independently of male showiness’ and that this independent evolution ‘…might be a widespread cause of female ornaments’. Kraaijeveld, Kraaijewelt-Smit and Komdeur (2007) conclude that there is ‘convincing evidence’ of the direct selection hypothesis, and ‘poor state of knowledge’ of the genetic correlation hypothesis. Clutton-Brock (2009) suggests that ‘…. the mechanisms responsible for the evolution of secondary sexual characters in females are similar to those operating in males….’ In a recent issue of this journal, an overview of positive and negative associations between carotenoids-based ornaments and carotenoids in eggs was published (Weiss et al. 2011). Here, we revisit this topic by focusing on mutually ornamented species and between taxa differences.
Our aim is to give an updated review of the literature on the potential associations between female ornaments and quality of their offspring in mutually ornamented species in all taxa. The specific focus in our study is to compare the published evidence for the direct selection and the genetic correlation hypotheses. In line with the original hypotheses, we interpret a positive correlation as support to the direct selection hypothesis and a negative correlation as support to the genetic correlation hypothesis. The topic is dominated by bird studies, but here we demonstrate how recently published articles mainly on fish ornaments can challenge the direct selection hypothesis as a general explanation for the evolution of female fineries.
Materials and methods
The literature was searched for empirical studies on the association between the elaboration of female ornaments and different qualitative and quantitative aspects of their eggs and offspring (see column 6 in Table 1). The quantitative and qualitative aspects of eggs and offspring are collectively and for simplicity termed ‘offspring quality’ in this study, although we realize that offspring number and egg size may not always be indicators of offspring quality (see 'Discussion'). The focus was on mutually ornamented species as they provide the clearest way to compare the relative importance of the abovementioned two alternative hypotheses. Female-specific ornamented species are less suited to test the genetic correlation hypothesis (Lande 1980), as outlined above, and such studies were therefore not included. Studies on sex-role reversed (male choice and female–female competition) were excluded as well. Moreover, we excluded experimental studies, in which females were supplemented with extra resources (e.g. carotenoids) before reproduction and where this treatment significantly increased either the ornament or the offspring quality or both (e.g. McGraw, Adkins-Regan & Parker 2005; Grether et al. 2008). Such studies were excluded because artificially enhanced access to resources may boost both mother's ornament and offspring quality, leading to artificial positive associations (Reznick, Nunney & Tessier 2000; Roff & Fairbairn 2007; Morales, Velando & Torres 2009). However, we did include associations between ornament and offspring quality for the control group (but not the manipulated groups) of such food-supplemented studies, when such control group data were presented separately. Our inclusion of the study was based on several search methods and not solely on an a priori set of search criteria in data bases. The possibility that some relevant articles escaped our attention cannot be excluded.
Table 1. Overview of studies on the association between the expression of female ornaments and fitness-aspects of eggs and offspring in mutually ornamented species with conventional sex roles. ‘Comment’ is either a direct citation or, to save space, an edited version of text in the study
|Least auklet||Facial plume, red bill, bill knob size||–||Field experiment||No||Chick survival||No||None||Jones & Montgomerie (1992)||Ornament not related to reproductive success|
|Pied flycatcher||White patch||–||Field||No||Number of fledged chicks||No||None||Potti (1993)||The ornament seems to be neutral relative to female fitness|
|House finch||Plumage colour||Carotenoid||Field||No||Number of fledged chicks||No||None||Hill (1993)||No relationship between female plumage coloration and reproductive success|
|Barn swallow||Tail length||–||Field||No||Number of fledged chicks||Positive||Direct selection||Møller (1993)||Female tail length positively associated with total seasonal reproductive success|
|Barn swallow||Tail length||–||Field experiment||No||For example brood size at fledging||No||None||Cuervo, deLope & Møller (1996)||Original female tail length before manipulation was unrelated to reproductive performance.|
|Barn swallow||Ventral plum-age coloration, tail length||–||Field||No||Number of fledged chicks||Positive||Direct selection||Safran & McGraw (2004)||Plumage coloration, but not tail streamers, predicts annual reproductive success|
|Zebra finch||Bill colour||Carotenoid||Experiment||No||Number of chicks surviving ||Negative||Genetic correlation||Price & Burley (1994)||Red-billed females died sooner and produced fewer offspring than did less red-billed females|
|Lesser kestrel||Grey rump and tail||–||Field||No||Number of fledged chicks||No||None||Tella et al. (1997)||Females with grey plumage do not achieve fitness benefits.|
|Bluethroat||Blue, chestnut, white and black colours (combined)||–||Field||No||Fledging body mass||Negative||Genetic correlation||Rohde, Johnsen & Lifjeld (1999)||No indication that female plumage coloration was positively related to seasonal reproductive performance. Negative relationship between ornament and fledging body mass revealed in 1 of 2 years|
|Magellanic penguin||Flipper length||–||Field||No||Number of chicks raised||Positive||Direct selection||Forero et al. (2001)||Females with larger flippers had a higher probability of raising two chicks (relative to < two chicks)|
|Scissor-tailed flycatcher||Tail length||–||Field||No||Clutch size||Positive||Direct selection||Regosin & Pruett-Jones (2001)||Female tail length was correlated with early clutch initiation, and, in one year, larger clutches|
|Inca tern||White feather moustache, wattles length||–||Field||No||Mean chick mass||Positive||Direct selection||Velando, Lessells & Marquez (2001)||Female ornament (parent moustache) predicts reproductive performance and chick quality|
|Barn owl||Size black feather spots||Melanin||Field experiment||No||Date of egg-laying||Negative||Genetic correlation||Roulin et al. (2001)||Lighter coloured females reproduced earlier in the season. Results support genetic correlation hypothesis|
| || || || |
|Blue tit||Yellow coloration||Carotenoid||Field experiment||No||Chicks tarsus length at fledging||No||None||Senar, Figuerola & Pascual (2002)||Plumage coloration of genetic female had no effect on tarsus length of fledglings|
|Rock sparrow||Yellow crest patch||–||Field experiment||No||Number of broods||Positive||Direct selection||Pilastro, Griggio & Matessi (2003)||Double-brooding and primary females had larger patches and higher breeding success, than single-brooding and secondary females|
|Blue tit||Yellow plumage||Carotenoid||Expt||Yes||Carotenoids concentration in eggs||No||None||Biard, Surai & Møller (2005)||No relationship between yellow plumage colour of adult females and carotenoid deposition in eggs|
|Blue tit||UV-blue cap & yellow collar feathers||Carotenoid (collar feathers)||Experiment||No||Clutch size, chick fledgling success||Positive||Direct selection||Doutrelant et al. (2008)||Carotenoid-based female coloration was positively linked to key proxies of bird lifetime reproductive success|
|Yellow-eyedPenguin||Iris, post-ocular stripe||Carotenoid||Field||No||Number of fledged chicks ||Positive||Direct selection||Massaro, Davis & Darby (2003)||Males can use eye and plumage coloration as an indirect cue in assessing age and quality of females during mate choice|
|Burrowing parrot||Red feather patch ||–||Field||No||Mass of first-hatched nestlings||Positive||Direct selection||Masello & Quillfeldt (2003)||Females with larger red abdominal patch fledged nestlings with higher pre-fledging masses and longer pre-fledging bills|
|European shag||Crest of feathers||–||Field||No||Reproducing vs. non-reproducing. Egg-laying date||Positive||Direct selection||Daunt et al. (2003)||Reproducing females have larger crest than non-reproducing. Positive relationship between ornament size and egg-laying date. Early breeding is generally associated with high breeding performance|
|Shelduck||Immaculateness of plumage||–||Field|| ||Chick survival||Positive||Direct selection||Ferns & Lang (2003)||Broods of parents (including females) with more immaculate plumage were more likely to survive|
|Red-tailed tropicbird||Streamer length||–||Field||No||Chick mass or fate, egg size||No||None||Veit & Jones (2003)||Streamer length was not correlated with chick performance|
|Northern cardinal||Red feathers & bill, head crest, face mask||Carotenoid (in red feathers & bill)||Field||No||Number of fledged nests||Positive||Direct selection||Jawor et al. (2004)||Redness of female feathers correlated positively with reproductive success, whereas crest and mask were not|
|European starling||Iridescent throat feather hue & length||–||Field||No||Size clutch & hatching success||Positive||Direct selection/ (none)||Komdeur et al. (2005)||Ornamented females fledged more young, but ornament (PC1 of hue and length) may be confounded with age|
|Rufous bush chat||Size of white tail patch||± melanin||Field||No||Number of fledged chicks||No||None||Alvarez (2004)||Only in males were ornament and reproductive outcome related|
|Black swan||Curled feathers on wings||–||Field||No||Offspring survival||Positive||Direct selection||Kraaijeveld et al. (2004)||Curled feathers (of male and female combined) appears to signal social dominance which is highly correlated with reproductive success |
|Great tit||Yellow central plumage||Carotenoid||Field||No||Number of fledged chicks||Negative||Genetic correlation||Mänd, Tilgar & Møller (2005)||Ornamented females had on average lower fledging success and fewer fledglings in the year with unfavourable breeding conditions|
|Great tit||Yellow, blue & white feathers, & UV chroma of yellow feathers||Carotenoid & melanin traits, white patch||Field||No||Total anti-oxidants in eggs||Positive to cheek patch||Direct selection||Remeš, Matysioková & Klejdus (2011)||Total antioxidants in eggs increased with female immaculateness of white cheek patch only|
|Eastern bluebird||Structural blue, chestnut||Melanin||Field & experiment||Yes||Number of fledged chicks||Positive||Direct selection||Siefferman & Hill (2005)||Structural coloration predicted annual reproductive success|
|Great black-backed gull||Bill, gape & eye-ring coloration||Carotenoid||Field||No||Larger eggs and clutches ||Positive||Direct selection||Kristiansen et al. (2006)||Females with high colour intensity had larger eggs and clutches|
|Fowl||Comb size||–||Experiment||No||Egg mass||Positive||Direct selection||Cornwallis & Birkhead (2007)||Females with large comb lay heavier eggs|
|Blue-footed booby||Foot colour||Carotenoid||Experiment||Yes||Egg mass and volume||Negative (control)||Genetic correlation||Morales, Velando & Torres (2009)||Carotenoid not added (controls): Ornamentation was negatively associated with the mass and volume of eggs, but the association was reversed in experiment groups (dietary carotenoids added to feed) |
|Black-legged kittiwake||Bill, gape and tongue coloration||Carotenoid||Field||No||Number of fledged chicks||No||None||Leclaire et al. (2011)||Coloration of integuments was correlated with fledgling success in males but not in females|
| Fishes |
|Stickleback||Skin colour (spines)||Carotenoid||Field||No||Carotenoids in eggs||Negative||Genetic correlation||Nordeide, Rudolfsen & Egeland (2006)||Negative association between red spines and amount of carotenoids in eggs |
|Arctic charr||Skin colour (Throat & belly)||Carotenoid||Field||No||Carotenoids in eggs||None/(negative) ||None/(genetic correlation)||Nordeide et al. (2008)||The ornament was not significantly associated with amounts of carotenoids in the eggs, although being negative and close to significance|
|Arctic charr||Skin colour||Carotenoid||Experiment||No||Number of egg surviving||Negative||Genetic correlation||Janhunen et al. (2011a)||Negative relationship between female colouration and her offspring survivorship|
|Arctic charr||Skin colour||Carotenoid||Experiment||No||Fecundity, and percentage embryo surviving||Negative||Genetic correlation||Janhunen et al. (2011b)||Intensely coloured females were less fecund, and offspring had lower viability|
|Chinook salmon||Skin colour||Carotenoid||Experiment||Yes||Egg-carotenoids concentration ||No||None||Garner, Neff & Bernards (2010)||Egg carotenoid concentration were not correlated with either skin carotenoid concentration or colouration|
|Chinook salmon||Brighter coloration & lateral integument||Conceivable linked to carotenoid ||Experiment||No||Reproductive success.||No||None||Neff et al. (2010)||Sexual selection favours increased body size and aspects of integument coloration in males, but not in females|
|Whitefish||Breeding tubercles||Keratin||Experiment||No||Swimming, predator–avoidance, yolk-volume||Positive||Direct selection||Kekäläinen et al. (2010)||Offspring of ornamented females had better swimming performance, predator-avoidance and larger yolk-volume than less ornamented females|
|Whitefish||Breeding tubercles||Keratin||Experiment||No||Mortality of eggs||Positive||Direct selection||Huuskonen et al. (2011)||Increasing number of female breeding tubercles was associated with low embryonic mortality|
|Sockeye salmon||Body colour||Carotenoid||Field||No||Colour of eggs||Negative||Genetic correlation||Ramstad, Woody & Allendorf (2010)||Our data are consistent with a trade-off between body and egg colour|
|Agamid lizard||Throat & chest coloration||–||Experiment||No||Egg number & weight||No||None||LeBas & Marshall (2000)||There was no evidence that female throat or chest coloration was an indicator of female quality|
Forty-three articles which fulfil the abovementioned search criteria for this review were found (Table 1). Thirty-three (77%) were bird studies, nine (21%) were on fishes, and one (2%) was a lizard study. No studies on other taxa which fulfil the criteria were found. Nine studies showed negative, 14 non-existing and 20 positive associations between female ornament and offspring quality (Table 1).
A majority of bird studies revealed a positive association between the two traits investigated with 18 (55%) and five (15%) of the studies showing a positive and negative association, respectively (Table 2). Fish studies, although few, were skewed in the opposite direction with two (22%) and four (44%) studies supporting positive and negative association, respectively. It is also worth noting that about 1/3 of the studies (both in birds and fishes, in addition to the lizard study) revealed neither positive nor negative association between the traits (Table 2).
Table 2. Overview of number of studies with negative, non-existing and positive association between ornament (carotenoids-based, and other ornaments than carotenoids-based pooled) and egg- or offspring quality
After splitting the studies into two groups (all taxa pooled) based on the nature of the ornament (carotenoids-based vs. non-carotenoids-based), the minority of studies on carotenoids-based ornaments reported a positive association (four of 18 studies, or 22%) between the traits (Table 3). This proportion was relatively low compared with the studies on ornaments based on other than carotenoids (16 of 25 studies or 64%, Table 3).
Table 3. Overview of number of studies with negative, non-existing and positive associations between ornament (carotenoids-based, and non-carotenoids-based) and egg- or offspring quality
The review reveals 9 (21%), 14 (33%) and 20 (47%) studies with negative, non-existing and positive associations between female ornament and her offspring quality, respectively. The direction of the relationship between female ornamentation and fitness differs between taxa (birds vs. fishes), between species within the same taxa and between different traits within the same species (see Great tit Parus major, Table 1). In our opinion, it is necessary and justifiable to speculate about these differences to better understand the evolution of female sexual ornamentation, although it is way too early to conclude comprehensively due to the limited number of studies published (Table 1).
Studies on ornaments based on carotenoids- vs. non-carotenoids differ considerably in their support for the direct selection hypothesis and the alternative genetic correlation hypothesis (Table 3). Given the high proportion of studies on carotenoids-based ornaments in our study (42%), we discuss below possible reasons why they are more commonly in disagreement with the direct selection hypothesis.
Different animal groups, such as birds and fishes and potentially even species within each taxa, may differ in their ability to transport resources and nutrients for ornaments internally between different body parts or, alternatively, vary in their timing of the resource allocation between the eggs and sexual ornamentation. This may cause differences between taxa with carotenoids-based ornaments and non-carotenoids-based ornaments, if the allocation routes work differently. For example, many salmonid fish are known to store carotenoids in their muscle tissue during the non-reproductive season and redistribute large amounts of these pigments from muscles to the skin and gonads during the sexual maturation process (Steven 1949; Crozier 1970; Ando & Hatano 1987; Torrissen & Naevdal 1988; Bjerkeng, Storebakken & Liaaen-Jensen 1992; Hatlen et al. 1995; Hatlen, Arnesen & Jobling 1996; Garner, Neff & Bernards 2010). Birds are also known to use various endogenous tissues as temporary reservoirs for carotenoids pigments, but to what extent these stores can be re-directed to pigmentation of feathers, skin and beak remains ambiguous and seems to differ between bird species (reviewed by McGraw 2006; see also Mänd, Tilgar & Møller 2005; Griggio et al. 2009; Biard, Surai & Møller 2005). Unfortunately, to our knowledge, these patterns are largely unknown for non-carotenoids-based ornaments.
In some species, the incidents of acquiring carotenoids in different tissues, like ornament and other body parts, are seasonally separated. Females of such species may allocate carotenoids to their ornaments (feathers, legs or beak) mainly during the non-reproductive season expressing current condition at moulting (Fitzpatrick, Berglund & Rosenqvist 1995; Johnsen et al. 1996; Hegyi et al. 2008), and later in their eggs during the period before egg development, like for example rock sparrows Petronia petronia do (Griggio et al. 2009). If future research shows that females of some species (i.e. birds) are less able to re-allocate carotenoids between different tissues or have a time-lag from moulting to reproduction compared with other species (like fishes), this might explain some of the differences between the taxa as indicated in this review. This would help us to understand why some studies clearly support the genetic correlation and others the direct selection hypothesis (Tables 1 and 2). The direct selection hypothesis would fit species with no or less reallocation of resources between body parts and to species with a time-lag between moulting and reproduction (i.e. species where carotenoid allocation to different body parts are seasonally separated). The genetic correlation hypothesis may predominantly fit species in which the development of ornamentation and eggs occurs relatively simultaneously, or to species that are able to reallocate carotenoids between different body tissues and the ornament. Similar reasoning can potentially also be applied for non-carotenoids-based ornaments that are produced by costly resources or nutrients.
Carotenoids are pigments synthesized only by plants, algae, some bacteria and some fungi, and thus animals must obtain them through their diets (Goodwin 1984; but see Moran & Jarvik 2010). Carotenoids are generally considered valuable for individuals because of their beneficial role in various physiological processes. They act as powerful immunostimulants and antioxidants, for example, by removing harmful radicals produced as by-products of normal cellular activity and environmental stressors. Carotenoids are especially important in reducing damage caused by radicals in the rapidly developing eggs and embryos (reviewed by Olson & Owens 1998; Blount, Houston & Møller 2000), and thus, carotenoids may improve egg and larval quality and survival in fish (e.g. Verakunpiriya et al. 1997; Watanabe & Vassallo-Agius 2003). However, not all red, orange or yellow ornaments are carotenoids based (e.g. Grether, Hudon & Endler 2001; Weiss et al. 2011; Weiss, Foerster & Hudon 2012). Moreover, non-carotenoids-based ornaments, made of for example melanin and keratin, or feathers (see Table 1), may be less essential for the development of eggs and offspring than are carotenoids. Some non-carotenoids-based ornaments (see Table 1) might therefore be less costly and less condition dependent. It can be argued that this difference between carotenoids-based and non-carotenoids-based ornaments might explain part of the higher percentage of studies with a positive association between the traits in the non-carotenoids-based ornament group (Table 3). On the other hand, no support for carotenoids being more condition dependent than non-carotenoids-based melanin ornaments was revealed in a meta-analysis by Griffith, Parker & Olson (2006), although the authors stress the very limited data available.
The offspring in different animals face different selective pressures determined by the environment. At the individual level trade-offs are expected, but in a good, nutrient-rich environment such trade-offs may not necessarily become apparent (in contrast to low-quality environments). For example, the great tit chicks of brightly ornamented mothers had lower fledging success compared with chicks of drabber mothers only during environmental unfavourable conditions (Mänd, Tilgar & Møller 2005). The anadromous sockeye salmon Oncorhynchus nerka inhabiting carotenoids-rich marine environments is an example of a species where females have carotenoids-based ornaments but are not limited by access to carotenoids (Foote, Brown & Hawryshyn 2004). Moreover, in low-pathogen-risk environments the trade-off between ornaments and offspring may not be particularly important either, since females may not have to allocate so much carotenoids into their self-maintenance functions (immunity). Then more carotenoids could be available for both ornament and eggs, reducing the strength of possible compromise between these two targets. On the other hand, in such circumstances the need of protective carotenoids may also be reduced in the females' eggs, allowing the potentially concomitant selection for more showy ornaments. Similar reasoning would apply also for non-carotenoids-based ornaments, if these ornaments are produced by costly material or resources.
The balancing patterns of natural and sexual selection can also be expected to vary between species with divergent life-history strategies. As an example, many iteroparous salmonid species, including Arctic charr Salvelinus alpinus can remain reproductively active for a period of several years, and hence, they must trade-off the resource investment during a given breeding season against future possibilities for reproduction. Semelparous salmonids (genus Oncorhynchus), in contrast, put all their efforts into a single reproductive bout, and thus, they neither have to store carotenoids or other essential resources for later use. This may partially explain why the trade-off between female ornamentation and investment in offspring may be absent or less obvious in the latter group (e.g. Foote, Brown & Hawryshyn 2004; Garner, Neff & Bernards 2010). Correspondingly, the trade-off in carotenoids allocation should be stronger, on average, in species with larger female fecundity- or egg-biomass. In many fishes, females may well produce thousands or even hundreds of thousands of eggs and are thus likely to face a higher natural selection pressure to invest considerable amounts of carotenoids in their egg production, relative to species which produce only a few eggs per breeding season (most birds). However, this difference between taxa may be partly compensated by the larger size of bird eggs.
In general, males are expected to base their mate choice principally on phenotypic indicators that are associated with female fecundity either directly (e.g. body size; Sargent, Gross & van den Berghe 1986; Kraak & Bakker 1998; Bonduriansky 2001; Byrne & Rice 2006) or indirectly (ornaments; Amundsen & Forsgren 2001; Massironi, Rasotto & Mazzoldi 2005). Even in species where female ornaments are negatively associated with offspring fitness, elaborately ornamented females may potentially have higher fitness than non-ornamented females. This may be very unlikely, but would occur if higher sexual attractiveness of daughters of the bright mother more than compensates her reduced number of surviving offspring (because of the trade-off), as revealed in studies over several generations. In this scenario, direct selection hypothesis would work despite the fact that we may observe a trade-off between offspring survival and female ornamentation. However, in species where ornament-offspring trade-offs have been demonstrated so far (see Table 1), males preferring ornamented relative to drab females have been shown in the Bluethroat Luscinia svecica, (Amundsen, Forsgren & Hansen 1997) only, as far as we know. Males seem to prefer red females in sockeye salmon (Foote, Brown & Hawryshyn 2004), but in this population carotenoids may not be a limited resource (see above). Zebra finch Taeniopygia guttata males prefer to associate with females with bill colours in the middle of the phenotypic range (Burley & Coopersmith 1987), and male sticklebacks Gasterosteus aculeatus seem to prefer females with drab spines (Nordeide 2002). We are not aware of studies demonstrating males to prefer females based on their ornament in the remaining five species with negative correlation between ornament and offspring quality in Table 1 (Barn owl Tyto alba, Blue tit Cyanistes caeruleus, Great tit, Blue-footed booby Sula nebouxii and Arctic charr). It is also worth to note that even when we observe positive correlation between individual fitness components and female ornamentation, it does not mean that the total fitness of bright females is necessarily higher. So far there are no studies on the total fitness of the offspring generations of brightly coloured females with costly ornamentation, the final answer has yet to be said on this topic.
Fourteen (33%) of the 43 reviewed studies indicate non-existing association between ornament and offspring quality (Table 2). Non-significant association may be a result of wrong traits being examined, traits or offspring quality being inappropriately measured, publication bias against non-significant associations, or Type II-errors. However, the most likely and especially interesting explanation is naturally that there is simply no association between the ornament and offspring quality. It might be argued that studies showing non-existing associations due to the latter explanation are actually studies in support of the genetic correlation hypothesis, since only positive association between ornament and trait can falsify this hypothesis. This is in contrast to the direct selection hypothesis, which may more easily be falsified by both non-existing and negative associations. As in the present review, we have not considered non-existing associations as supporting evidence for the genetic correlation hypothesis, and it is possible that the number of such articles has been underestimated.
Acknowledging parasites and pathogens complicate the reasoning in the present study. According to the good-genes (handicap) hypotheses, sexual ornaments may signal heritable differences in the genes coding for immune defence against parasites (Hamilton & Zuk 1982). Empirical tests of this hypothesis in females are scarce in species known to trade-off resources between ornament and eggs. In the two studies we are aware of, parasitized female Arctic charr were in fact also the most ornamented individuals (Skarstein & Folstad 1996; see also Skarstein, Folstad & Rønning 2005), whereas no such association was found in another study of the same species (Nordeide et al. 2008). However, in males of several other species, parasitized males are, in general, less showy than unparasitized males. If the similar pattern also proves to be generally true for females, males of such species might be faced with problems to gain accurate information from female ornamentation. Females may be dull either due to having invested their resources into eggs at the expense of ornaments or due to low quality of the genes coding for immune system. This would mean that males should avoid both the drabbest, that is, most parasitized females, as well as the brightest ornamented females with poor eggs. As a result, this should lead to male preference for intermediately ornamented females, which indeed is in accordance with some earlier studies (Burley & Coopersmith 1987; Chenoweth & Blows 2005; Chenoweth, Doughty & Kokko 2006).
Future perspectives and conclusion
Predictions from ideas highlighted in the present study are possible to test. Especially, suitable models for this work are closely related animal groups like fish species or genera that live in different environmental conditions (e.g. carotenoids-rich marine vs. poor freshwater habitats) or exhibit divergent life-history patterns. In addition, recent advances in genomics and next generation sequencing have opened new and fascinating possibilities also in the study of sexual selection (Ellegren & Sheldon 2008; Mardis 2008). Quantitative traits loci (QTL) analysis has, for example, been used to reveal a major QTL for bisexual expression of the ornamental comb mass and several female-specific QTLs in a mutually ornamented domestic fowl Gallus gallus (Wright et al. 2008). Strong genetic correlation in beak redness between sexes and four genomic regions suggesting linkage with beak redness made in a QTL linkage mapping of Zebra finches (Schielzeth et al. 2012) indicate that largely the same genes influence beak colour in both sexes in this species. Fully sequenced species like the three-spined stickleback (Peichel et al. 2001) are particularly suited for QTL-analysis. A QTL-analysis revealing different location of loci coding for ornaments in males and females, provide support for the direct selection hypothesis. Such a QTL-analysis combined with studies of expressed genes (cDNA) has the potential to become a powerful tool to test the relative importance of the direct selection hypothesis and the genetic correlation hypothesis as an explanation to the evolution of mutual ornaments.
To conclude, a growing number of studies demonstrate a trade-off between resources in ornaments on one hand and eggs or offspring on the other hand. This illustrates the inadequacy of the direct selection hypothesis as a general explanation to account for evolution of ornaments in females. There is accumulating evidence supporting the tenability of the alternative genetic correlation hypothesis in animal species across different taxonomic groups, in accordance with the ideas of Lande (1980).
We thank the referees for the valuable comments on the earlier versions of the manuscript. This research has been supported by the Academy of Finland (project 127398 to RK) and the Kone Foundation (to JK).