Ornaments or offspring? Female sticklebacks (Gasterosteus aculeatus L.) trade off carotenoids between spines and eggs

Abstract

Hypotheses and models to explain female ornaments often assume that the elaborated traits are condition dependent; nevertheless, few empirical studies have addressed this topic. We studied a population of three-spined sticklebacks (Gasterosteus aculeatus) in which the females have conspicuous, carotenoid-based red coloration to their pelvic spines. The red coloration seems not to be condition dependent, as coloration is negatively associated with age and body length and not associated with condition. Furthermore, redder females did not have a lower density of leucocytes. We found a negative association between the females’ red carotenoid-based coloration in the spines and the amount of carotenoids in the female's gonads. Males choosing red-coloured females will fertilize eggs with small amounts of carotenoids and appear not to gain any benefit from their mates’ phenotypic quality that could result in offspring of improved quality. These results do not support the ‘direct selection hypothesis’ to explain the existence of the female ornaments.

Introduction

Numerous studies have shown that males of a variety of species have extravagant ornaments that evolved through sexual selection (Andersson, 1994). Only high quality males can afford to trade costly resources for showy ornaments at the expense of somatic maintenance, otherwise males of low quality could cheat by producing showy ornaments and the honesty of the signal content of the ornament would break down. Only condition-dependent ornaments are therefore regarded as honest indicators of quality (Zahavi, 1975; Andersson, 1994; Johnstone, 1997; Zahavi & Zahavi, 1997; Olson & Owens, 1998). Alternatively, ornaments may simply signal the individual's ability to attract mates (Fisher, 1930; Lande, 1981).

Females are generally less decorated than males and in species with conventional sex roles, females’ decorative or conspicuous traits – for simplicity termed ‘ornaments’ in this paper – have received less attention (Andersson, 1994; Amundsen, 2000). The inter-sexual difference in the elaboration of ornaments is often explained by skewed operational sex ratios with fewer females than males available for mating at a given time; this arises because sperm are small and cheap to produce in vast numbers compared with the time and energy burden of egg production, gestation, and parental care which typically falls on females (Emlen & Oring, 1977; Clutton-Brock & Vincent, 1991). An alternative model, proposed by Fitzpatrick et al. (1995), assumes condition-dependent ornaments in females, and suggests that males choosing excessively ornamented females receive fewer or poorer eggs than they could have obtained from a similar quality female which allocated less to ornamentation and more to reproduction. Such a fecundity cost will constrain the evolution of the male preference driving the sexual selection process, and, as a consequence, put a brake on further exaggeration of the ornament in the females (Fitzpatrick et al., 1995).

Two hypotheses have been proposed to explain the evolution of female ornaments. According to the ‘direct selection hypothesis’ females are ornamented as a result of sexual selection by males, whereas the ‘genetic correlation hypothesis’ suggests that females are ornamented due to a nonadaptive genetic correlation that arises as a consequence of sexual selection on males and the fact that most genes are common to both sexes (reviewed by Amundsen, 2000). In the direct selection hypothesis, males should prefer ornamented females and there should be a positive relationship between ornament elaboration and female quality.

Empirical studies are ambiguous in their support for the competing hypotheses. A review of bird studies gave them approximately equal support, with (i) males preferring ornamented females in six studies and showing no preference or a preference for the less ornamented of the females in five studies, and (ii) females’ ornaments and phenotypic quality traits being positively associated in five and not associated in six studies (Amundsen, 2000). In recent years, several studies of birds have demonstrated a positive association between female ornamentation and phenotypic quality (Roulin, 1999, 2004; Johnsen et al., 2003; Massaro et al., 2003; Pilastro et al., 2003; Pizzari et al., 2003; Griggio et al., 2005) whereas Wolf et al. (2004) revealed no male preference for ornamented females. Similar studies in other taxa are few. LeBas & Marshall (2000) reported male agamid lizards (Ctenophorus ornatus) to prefer females with higher throat chroma, but found no evidence that throat chroma was an indicator of female quality. In the dance fly Rhamphomyia tarsata, in which only the female is ornamented, females with large ornaments were preferred by males and were more fecund than less ornamented females (LeBas et al., 2003). Fish mate choice studies have demonstrated male preference for sword-ornamented females in Priiapella olmecae (Basolo & Delaney, 2001), carotenoid-ornamented females in sockeye salmon (Oncorhynchus nerka) (Craig & Foote, 2001; Foote et al., 2004), and for carotenoid-ornamented females in the two-spotted goby (Gobiusculus flavescens) (Amundsen & Forsgren, 2001, 2003). However, male convict cichlids (Cichlasoma nigrofasciatum) revealed no male courtship preference based on carotenoid-based ornaments (Beeching et al., 1998), and males of a different swordtail fish, Xiphophorus helleri and the three-spined stickleback (Gasterosteus aculeatus) discriminated against sword- and carotenoid-based ornamented females, respectively, during courtship (Basolo & Delaney, 2001; Nordeide, 2002) consistent with the genetic correlation hypothesis. The confusingly different conclusions from the carotenoid-based ornament studies of male mate choice in fishes might not be as contradictory as they seem at first glance. For example, sockeye salmon is anadromous and inhabit ‘…productive marine environments in which carotenoids are readily available in their krill-dominated diet’ (Foote et al., 2004). In such environments carotenoids might not be a limiting resource, hence males preferring red-ornamented females might not suffer by fertilizing eggs with low levels of antioxidant carotenoids. Moreover, female two-spotted gobies have carotenoid-based coloration at their bellies and the colour ‘…is mainly caused by the pigmented eggs being visible through the skin…’ (Amundsen & Forsgren, 2001). There is therefore most likely a direct relationship between the gobies’ red coloration and the amount of carotenoids in their eggs. Hence, males that choose to spawn with ornamented females will fertilize eggs with high amounts of antioxidant carotenoids. The carotenoid-based ornaments of the three-spined sticklebacks differ from the two-spotted goby in at least three important aspects. First, there is no obvious causal relationship between the ornament and the amount of carotenoids in the eggs of females, as carotenoid is seen solely in the skin at the inner part and bases of the pelvic spines and not because the eggs are visible through the skin as in gobies. Secondly, only female gobies have the reddish ornament on their bellies (Amundsen & Forsgren, 2001), whereas both female and male sticklebacks have similar more or less conspicuous red coloration on their pelvic spines, at least in northern and south-western Norway (Nordeide, 2002). Furthermore, the sticklebacks’ spines are apparently more red during the spawning season than during the rest of the year, and territorial males do in addition develop the nuptial red throat and blue eyes (J.T. Nordeide, unpublished), which is typical for this species (Rowland, 1994). Finally and different from other fish species, red serves as a strong signal eliciting territorial aggression (ter Pelkwijk & Tinbergen, 1937; Tinbergen, 1948) or a dual effect of aggression and fear in male three-spined sticklebacks (Rowland, 1994), in addition to being an important mate choice cue for females (reviewed by Rowland, 1994).

Fish studies testing the prediction about association between female ornaments and phenotypic quality are particularly rare. In the only study in fish that we are aware of, support for a condition-dependent association between a carotenoid-based ornament in female Arctic charr (Salvelinus alpinus) was equivocal (Skarstein & Folstad, 1996).

Carotenoid-based ornaments are common (Goodwin, 1984; Andersson, 1994; Møller et al., 2000). The colours of the different carotenoids vary from yellow, to orange and red (Goodwin, 1984), although for simplicity we term the colour of the pelvic spines ‘red’ in this study. Carotenoids are lipid-soluble hydrocarbons that are synthesized by plants, algae, some bacteria and some fungi, while animals must obtain them through their diet (Goodwin, 1984). Carotenoids are generally considered as rare and valuable because of their beneficial role in a variety of physiological processes including being powerful immunostimulants and antioxidants (reviewed by Olson & Owens, 1998; see also Hartley & Kennedy, 2004), and may improve egg quality and larval survival in fish (Verakunpiriya et al., 1997; Watanabe & Vassallo-Agius, 2003). Antioxidants remove harmful radicals produced as by-products of normal cellular activity, and from environmental stressors (reviewed by Chew, 1996). With no antioxidants present these highly unstable radicals pair with other electrons in the cell and damage molecules, in for example membranes, which can result in impaired immune function (Chew, 1996). Carotenoid-based ornaments are often regarded as honest, condition-dependent indicators of male quality (Andersson, 1994; Johnstone, 1997; Olson & Owens, 1998). Females may differ from males as they are hypothesized to trade-off carotenoids in their eggs and somatic maintenance (Blount et al., 2000), and possibly carotenoid-based signals. This is due to embryos being rapidly growing organisms with a high rate of metabolism that can produce more radicals and have more oxidative stress than slower growing organisms (Blount et al., 2002). Developing embryos, at least in birds, rely heavily on oxidation of fatty acids from the lipid-rich yolk supplied by the mother (Blount et al., 2002). However, it remains unclear whether carotenoid display in females in general advertises their reproductive potential (Blount, 2004).

Our aim is to examine if the red carotenoid-based coloration at the females’ pelvic spines is condition dependent, because such studies are few in birds and almost nonexistent in fishes. Of particular interest is the association between the intensity of the carotenoid-based pigment in the pelvic spines and the amount of carotenoids in the gonads of the female, as it remains unclear whether carotenoid ornamentation in females generally advertises their reproductive potential (see above).

Material and methods

Sticklebacks were caught from a landlocked population in Lake Nedre Vollvatn at an altitude of 125 m at 67°17′N, 14°25′E in Bodø, northern Norway. The lake is oligotrophic and most likely offers poor access to carotenoids. This population is perennial and both sexes start spawning at an age of 2 years (J.T. Nordeide, unpublished). The fish were captured a few days before the start of the spawning season for this population (J.T. Nordeide, unpublished) on the 24th and 25th of May 2003. Visual inspection of the gonads indicated that a batch of eggs would be fully developed within few days. We consider this sampling time to be a good compromise between sampling (i) too early and before the eggs were fully developed, and (ii) too late, after carotenoids were expelled from the gonads through batches of eggs spawned earlier this year. The fish were caught in traps made from 1.5-L lemonade bottles of transparent polyester. The females were first anaesthetized by an overdose of benzocain (ethyl 4-aminobenzoate) and were dead after 15–25 s. A blood sample was collected from the fish body anterior to the caudal fin before the fish were frozen in liquid nitrogen. We spent on average 61 s (range 44–106) from first handling the females until they were frozen. After 3 days the females were transferred to a freezer at −40 °C. Blood smears were prepared and preserved with methanol in the field. In the laboratory we measured each fish for total length to the nearest mm, total body mass to the nearest mg, gonad mass to the nearest 0.1 mg, gender determined by inspecting the gonads, and age was determined from the otoliths. No fish was infected by the cestode Schistocephalus solidus or any other parasite as determined by visually inspecting the body cavity, gills and external body of each fish. The condition (K) of each female stickleback was estimated as:

• 

where the exponent was estimated from a linear regression of ln-transformed somatic body mass (total body mass − gonad mass) over length of all 69 females.

The blood smears were stained by the May–Grunwald–Giemsa staining method and two areas of the stained blood smear of each fish were photographed at 100× (Axioskop 2 mot plus, Zeiss AxioCam HRc, Carl Zeisswerk, Göttingen, Germany). Erythrocytes and leucocytes (lymphocytes and granulocytes) in each photo were identified according to Amin et al. (1991) and counted from printouts of the photos. Mean number (±SD) of erythrocytes and leucocytes counted in each photo was 675 (±231) and 7.8 (±7.0), respectively, and there was a general agreement between the two counts (two photos) of each fish (General Linear Method, GLM: F_1,67 = 64.7, P < 0.001, r² = 0.491). The relative density of leucocytes (L) was calculated as:

• 

based on the average values of the two counts.

After less than 48 h of capture the sticklebacks were removed from the liquid nitrogen and during the next few minutes the ventral part of each fish was photographed together with a standard piece of red cardboard against a standard black background. We used a Canon EOS D60 digital camera (Canon, Tokyo, Japan) with a Canon EF 100 mm f/2.8 Macro lens and Canon Macrolite MR-14 EX ring flash to produce close up pictures of the ventral part of the fish with raised spines. The digital photos were analysed in RGB mode by Adobe Photoshop version 7.0. (San Jose, CA, USA) and adjusted for differences in brightness between different photos according to the red cardboard (Image – Adjustments – Brightness). We first drew a line in Adobe Photoshop enclosing an area around the entire right skin fold, which is located between the inner half of the pelvic spine and part of the fish's body close to the pelvic spine. The average density values for all three primary colours, red, green, and blue (R, G and B) were quantified from all the pixels enclosed by this area (Villafuerte & Negro, 1998). This was then repeated for the left skin fold and for the cardboard standard for each fish. The mean R, G and B values of the two skin folds were used in further calculations. A red skin fold of an ornamented fish differs from a drab skin fold mainly by having red coloration in a larger part of the skin fold, and to a smaller degree by the colour variation (yellow–red) itself. The method used to quantify red coloration in this study (see above) will yield a small average red value for the entire skin fold of a fish with (i) a small area at their skin fold covered with red (or yellow) coloration, and (ii) a yellow colour. The alternative way of quantifying red coloration, as the percentage of the skin fold which has red coloration, is less good as there is no well-defined edge between the red coloured part and the drab part of the skin fold. Red coloration in fishes may also be caused by pteridines, which have similar spectral properties as carotenoids but, unlike carotenoids, are not extracted in acetone (Grether et al., 2001). We conclude that no pteridines was present in the stickleback skin folds in our study, as they became colourless after extracting carotenoids by acetone (E.S. Egeland, unpublished). The degree of red coloration of the skin folds was estimated both as hue and as intensity. Hue is the wavelength of a colour measured in degrees with red coloration having low values, and hue of the pelvic spines was estimated directly by Adobe Photoshop from the density values after adjusting for fine differences in hue of each cardboard standard in each photo relative to the average hue in all photos. Repeatability of estimates of hue in pelvic spines is high (Nordeide, 2002). The intensity (I_R) of the red colour of both the red cardboard and the skin folds of the pelvic spines was calculated as:

• 

and we adjusted the final I_R-value of the skin fold of the pelvic spines of each fish according to the I_R-value of the cardboard in each photo relative to the average I_R-value of all photos. Repeatability of measurements of I_R of the skin fold was high (GLM; F_1,13 = 333.6, P < 0.001, r² = 0.96).

The chemical composition of the carotenoids of gonads was analysed by high-pressure liquid chromatography (HPLC) separately from each fish. The gonads of each female (except one individual due to a mistake) were added to 0.500 mL of freeze-cold 30% methanol in acetone (−20 °C) and stored in a freezer (−20 °C) for at least 72 h. All extracts were stored under nitrogen to prevent carotenoid degradation. Before HPLC analysis, the extract was decanted and filtered through an Anotope 0.2-μm syringe filter (Whatman, Maidstone, England). The HPLC analyses were performed on a Agilent 1100 instrument with quaternary pump, thermostatted autosampler, thermostatted column compartment and diode array UV-visible detector (Palo Alto, CA, USA). The column was a Spheri 5 RP-18 5-μm column with precolumn (Brownlee labs 0711–0017 from Perkin-Elmer, Wellesley, MA, USA), and a gradient system was used as eluant (0 min, 1 : 4 1 m ammonium acetate (aq):methanol; 30 min, methanol:acetone 7 : 3, 50 min, methanol:acetone:hexane 3 : 5 : 2), flow 1.25 mL min⁻¹. Injection volumes were 100 μL for spines extracts and 50 μL for gonads. All analyses were performed with column temperature 25 °C and detection wavelength 420, 450 and 480 nm (HPLC system 4 in Egeland et al., 1995). The retention times was compared with standards kindly provided by Prof. Synnøve Liaaen-Jensen, NTNU, Trondheim. Due to the small amount of material, neither mass spectrum nor co-chromatography in two different chromatographic systems (Schiedt & Liaaen-Jensen, 1995) could be obtained for the single carotenoids. Qualification of different carotenoids is based on retention times (HPLC) and visible light absorption spectrum, whereas quantification is based on chromatographic area only.

Statistical analyses were carried out by GLMs using spss version 11.5 (SPSS Inc., Chicago, IL, USA). None of the variables needed to be transformed to meet the assumptions of independence, heterogeneity of variance, normality of error, and linearity (Grafen & Hails, 2002). In the models we included all likely biologically relevant explanatory variables that could potentially explain the dependent variable (length, gutted mass, age, carotenoids in gonads, number of leucocytes, gonad mass) and their interactions. We then dropped terms sequentially until the model included only terms whose elimination would decrease the explanatory power of the model.

The study is in accordance with ethical guidelines stated by the Norwegian Ministry of Agriculture through the Animal Welfare Act. Ethical concerns were given the highest priority.

Results

Red coloration of the pelvic spines

The two visual parameters of the pelvic spines, intensity (I_R) and hue, are significant although weakly associated (F_1,67 = 6.120, P < 0.05, r² = 0.084). Females with high I_R, have red hue value (low hue value) and vice versa. We therefore present I_R and hue as the dependent variable in two separate models.

I_R as the dependent variable is significantly negatively associated with the total amount of carotenoids in the gonads as covariate whether age is accounted for (Table 1, Fig. 1a) or not (F_1,67 = 10.497, P < 0.01, r² = 0.124). The model with age accounted for explained 23.8% of the variance. I_R and total amount of carotenoids in gonads are still significantly associated when 2- and 3-year-old females are tested separately (F_1,43 = 5.881, P < 0.05, r² = 0.102, and F_1,21 = 5.761, P < 0.05, r² = 0.185, respectively), and the estimated slopes (coef. ± SE coef.) are both negative (2 years old: −0.00018 ± 0.00008; 3 years old: −0.00019 ± 0.00008) and significantly different from zero (2 years old: P < 0.05; 3 years old: P = 0.03). That is, females with high levels of carotenoids in their gonads have drab pelvic spines measured as I_R, and this applies whether or not age is taken into account. Moreover, old females have less red pelvic spines than young females (Table 1, Fig. 1a). Length, relative density of leucocytes, condition, and gonad mass, are entered as covariates but do not significantly correlate with I_R and do not improve the fit of the model. No significant interaction was found.

Table 1.  Test statistics from a GLM type III (adjusted) sums of squares (SS) with colour of the pelvic spines measured as intensity of red (I_R) as the dependent variable, and effect of the covariates age and total amount of carotenoids in the gonads. Adjusted R² = 0.238.

Source	SS	d.f.	F	P-value
Analysis of variance
Age	0.039	1	10.867	0.002
Carotenoids gonads	0.045	1	12.586	0.001
Error	0.223	65
Total	20.445	68

Term	Coeff.	SE coef.	t	P-value
Coefficients
Constant	0.69	0.04	17.25	<0.001
Age	−0.044	0.013	3.39	0.002
Carotenoids gonads	−0.00019	0.00005	3.80	<0.001

Figure 1. Association between measurements of red coloration (intensity and hue) on the y-axis and measurements of quality on the x-axis among female sticklebacks, and their linear regression lines. Filled circles 2-year-old, open circles 3-year-old, and triangles 4-year-old females. (a) Total amount of carotenoids in the gonads plotted against intensity of the red colour of their pelvic spines, with regression lines drawn separately for 2-year-old (upper line) and for 3-year-old females. (b) Length plotted against hue of pelvic spines in females. The y-axis is inverted as red spines have low and drab spines high hue value.

Hue of the pelvic spines as the dependent variable was significantly associated with length as covariate in a type III model and the coefficient was significantly different from zero (Fig. 1b, Table 2). In other words, females with red hue (low hue value) were significantly smaller than drab females. The coefficient of the two covariates (i) total amount of carotenoids in the gonads and (ii) relative density of leucocytes, were marginally nonsignificantly different from zero (Table 2). That is, females with red hue (low hue value) have marginally nonsignificantly less carotenoids in the gonads and higher relative density of leucocytes than drabber females. The model with the three covariates explains 19.8% of the variance in hue. Age, gonad mass, or condition, are not significantly associated with hue when added as covariates. No interaction is found that is significant or improves the fit of the model.

Table 2.  Test statistics from a GLM type III (adjusted) sums of squares (SS) with colour of pelvic spines measured as hue as the dependent variable, and effects of the covariates length, total amount of carotenoids in gonads, and relative density of leucocytes. Adjusted R² = 0.198.

Source	SS	d.f.	F	P-value
Analysis of variance
Length	170.058	1	10.527	0.002
Carotenoids gonads	63.485	1	3.930	0.052
Leucocytes	63.889	1	3.955	0.051
Error	1033.86	64
Total	47084.3	68

Term	Coeff.*	SE coef.	t	P-value
*Note that red spines have low and drab spines have high hue value.
Coefficients
Constant	11.93	4.60	2.59	0.012
Length	0.274	0.084	3.26	0.002
Carotenoids gonads	0.071	0.036	1.98	0.052
Leucocytes	−1.117	0.562	1.99	0.051

Carotenoids in gonads and pelvic spines

Five of 68 gonads did not contain traceable amount of carotenoids. For each of the specimens with carotenoids, two to twelve different carotenoids could be detected. All these gonads contain from two to six short-chained carbonyl-conjugated carotenoids, absorbing light only in the UV region (387–416 nm), probably degraded astaxanthin from the fish's diet, although no intact astaxanthin is observed in any of the gonads.

Sixty gonads also contained orange-yellow carotenoid diols. The most common of these carotenoids is compatible with zeaxanthin (absorption maxima 450 and 477 nm). Lutein, a carotenoid present in stickleback skin (Wedekind et al., 1998) co-elutes with zeaxanthin in this HPLC system, so presence of traces of lutein cannot be excluded, but the absorption spectra indicates that zeaxanthin is dominant over lutein. Up to six other orange-yellow carotenoids are present in the gonads. An unknown, slightly more polar carotenoid than zeaxanthin (absorption maxima 452 and 482) is present in 38 of the gonads. This carotenoid has characteristics as for alloxanthin, reported in, e.g. the goldfish Carassius auratus (Ohkubo et al., 1999), but the identity cannot be confirmed. A carotenoid compatible with tunaxanthin, based on retention time only, is present in 14 of the gonads. Seventeen gonads also contained a carotenoid somewhat more polar than carotenes. Both the unknown carotenoids have similar absorption spectra as zeaxanthin, but with maxima at slightly higher wavelengths. Five gonads contain carotenoids with both retention times and visible light absorption spectrum comparable with bacterioruberin (absorption maxima 468, 496 and 530 nm), compared with earlier analyses of bacterioruberin using identical HPLC system. The identity cannot be confirmed, but these carotenoids must have a bacterial origin (Britton et al., 2004). Also, four gonads contain traces of a degraded chlorophyll, probably phaeophytin a based on retention time and visible light absorption spectrum (410, 508, 536, 608 and 667 nm) (Jeffrey et al., 1997). Three of these four gonads also contain the red bacterially synthesized carotenoids. All gonads containing carotenoids with bacterial or chlorophyll-derived origins, also contained the carotenoids common for all gonads.

Descriptive statistics of parameters measured are given in Appendix 1.

Discussion

The association between the amount of carotenoids in the females’ gonads and the measurement of the red coloration in the pelvic spines was significantly negative where red coloration was measured as intensity (I_R). Moreover, no positive relationship was detected between the red ornament and measurements of quality for the females, as the covariates were either significantly negatively associated (length, age), or not associated (condition) with red coloration, and as red-coloured females do not have lower relative density of leucocytes than drab females. In other words, males choosing red-coloured females as mates will fertilize eggs with small amounts of carotenoids and seem not to gain anything regarding their mates’ phenotypic quality that could result in offspring of improved quality. This suggests that males choosing to mate with drab females have a selective advantage. This result is consistent with a male mate choice experiment from the same populations, where males courted drab females more than ornamented ones when illuminated with white light but not with green light which prevented the use of red cues by sticklebacks (Nordeide, 2002). Males of other taxa may also avoid brightly coloured female traits or other masculine traits in females perhaps because they perceive ornamentation as male-like (Burley & Coopersmith, 1987; Price & Burley, 1993; Basolo & Delaney, 2001; Drickamer et al., 2001; Wolf et al., 2004).

In the eggs of our study females, derivates of what is likely to be degraded red astaxanthin and, in a few females, red bacteria-originated carotenoids were detected in addition to zeaxanthins and other orange-yellow to yellow carotenoids. Although no intact astaxanthin could be detected, astaxanthin and its derivates in eggs are not unexpected as astaxanthin seems to be especially important for egg and larval quality in fish species (Verakunpiriya et al., 1997; Watanabe & Vassallo-Agius, 2003).

We found no positive association between female sticklebacks’ phenotypic quality and their ornaments. This is one of very few studies that have addressed this relationship in female fishes. A nonsignificant association was found between red hue (low hue value) and white blood cells among female Arctic charr, whereas a significant negative association was found among males (Skarstein & Folstad, 1996). Redder Arctic charr females also had higher condition than drab females (Skarstein & Folstad, 1996). Several studies exist on the relationship between quality and carotenoid-based ornaments in male sticklebacks; condition has been shown to be both nonsignificantly (Baube, 1997; Rush et al., 2003) and significantly positively associated (Milinski & Bakker, 1992; Frischknecht, 1993; Barber et al., 2000). In yet two more studies the same association was either curvilinear (Candolin, 1999) or found to differ between subpopulations (Bakker & Mundwiler, 1994). The carotenoid-based ornament of male sticklebacks was nonsignificantly associated with leucocytes in a study by Barber et al. (2000).

Why are there female sticklebacks with red spines if males find them less sexually attractive than the drab ones? Red-coloured females must have one or more advantages that compensate for the costs of being less attractive to males. First, ornamented females might have advantages in intra-specific interactions as suggested for the hummingbird Heliangelus amethysticollis (Bleiweiss, 1992) and revealed for the convict cichlid fish C. nigrofasciatum (Beeching et al., 1998). Secondly, red female sticklebacks may potentially compensate for low own attractiveness by deriving higher fitness from having redder brothers that are potentially very attractive, as red-throated males are preferred among female sticklebacks (reviewed by Rowland, 1994). This hypothesis requires that pelvic spine coloration in males has a genetic component, with red fathers having red sons, as demonstrated for throat coloration in sticklebacks (Bakker, 1993). Furthermore, red spines need to be attractive in males and red fathers should sire red-spined daughters.

Our results are to some degree consistent with Fitzpatrick et al. (1995), who argued that females may be less ornamented because males choosing excessively ornamented females receive fewer or poorer eggs than they could have obtained from a similar quality female which allocated less to ornamentation and more to reproduction. However, the models by Fitzpatrick et al. (1995) assume that the ornament is condition dependent, i.e. fecundity and trait expression covary strongly with condition. The lack of positive association between the female sticklebacks’ phenotypic quality and their pelvic spine coloration in our study suggests that the red coloration is not a condition dependent trait and therefore does not fit into the model proposed by Fitzpatrick et al. (1995).

The negative association between the females’ red-ornamented spines and the amount of carotenoids in their eggs revealed here, might have implications for a general rule with regard to the evolution of female ornaments. Female ornaments differ between species in two important aspects. First, ornaments may be basically the same in both sexes within a species, or only females may have the specific ornament in question. Secondly, this study suggests that we must distinguish between carotenoid-based female ornaments where a trade-off exists between ornaments and eggs and female ornaments where no such trade-off occurs. The ‘genetic correlation hypothesis’ is likely to explain females’ ornaments which are similar between sexes and where there is a trade-off between ornaments and gonads. The ‘direct selection hypothesis’ is expected to explain the evolution of female-specific ornaments where no such trade-off exists and where only females are ornamented. According to a recent paper by McGraw (2005), not only carotenoids, but melanins and several other endogenously manufactured chemical colorants in animals also act as antioxidants. Hence, the reasoning in this study may not only apply to carotenoid-based ornaments, but to any ornament made of chemical colorants which are also antioxidants and limited in some way.

In conclusion, we found nothing to indicate that the females’ carotenoid-based ornament was condition dependent, indeed females traded off carotenoids in their ornaments and gonads. Males seem to have nothing to gain by choosing the more ornamented females in this species. The results do not support the ‘direct selection hypothesis’ to explain the evolution of the ornament in female three-spined sticklebacks.

Acknowledgments

We would like to thank Anil B. Amin for teaching JTN how to identify erythrocytes and different leucocytes, Anders Berglund, Ivar Folstad, Natasha R. LeBas, Leigh W. Simmons, Joseph L. Tomkins and an anonymous reviewer for suggestions on previous drafts of the manuscript, School of Animal Biology, University of Western Australia for kind hospitality during JTN's sabbatical stay during which this paper was prepared, and Bodø University College for financing this study.

Appendix

Table Appendix1..  Descriptive statistics of parameters measured from the fish

	n	Min.	Max.	Mean	SD
Total body mass (g)	69	0.45	2.65	1.12	0.433
Somatic mass (g)	69	0.39	2.20	0.92	0.361
Age (years)	69	2	4	2.41	0.577
Length (mm)	69	40	65	50.55	5.878
Intensity	69	0.40	0.73	0.54	0.068
Hue (degrees)	69	17	35	25.93	4.456
Leucocytes (relative density)	69	0.00	4.30	1.24	0.887
Carotenoids gonads (mAU·s)	68	0.0	564.2	223.3	138.39
Gonad mass (g)	69	0.034	0.605	0.2011	0.1064