1. Introduction

Evolutionary psychology suggests that human traits that are considered attractive function as reliable markers of the biologic condition and reproductive potential of individuals (Grammer et al., 2003, Symons, 1995) Women, at least in Western societies, are considered attractive when they have relatively narrow waists, large breasts, and symmetrical features (Henss, 2000, Singh, 1993). We have recently documented that waist-to-hip ratio and breast size are indeed indicators of reproductive potential in women because they are related to the levels of reproductive hormones (Jasienska, Ziomkiewicz, Ellison, Lipson, & Thune, 2004).

In humans, individuals of both sexes who are more symmetrical are considered more attractive (Gangestad et al., 1994, Grammer et al., 2003, Grammer & Thornhill, 1994, Thornhill & Gangestad, 1994, but see Kowner, 1996, Noor & Evans, 2003, Swaddle & Cuthill, 1995). The degree of asymmetry may be an important indicator of developmental stability and a biomarker of fitness (Jones et al., 2001). Random deviation from the perfect symmetry expected in bilateral structures of bilaterally symmetrical organisms is referred to as fluctuating asymmetry (FA) (Møller & Swaddle, 1997). FA may result from the influences of genetic and environmental stressors operating during fetal development (Møller & Swaddle, 1997). It has been linked to some aspects of human health (Manning, 1995, Manning et al., 1997, Shackelford & Larsen, 1997, Waynforth, 1998), although other studies failed to find significant associations (Hume & Montgomerie, 2001, Milne et al., 2003, Rhodes et al., 2001, Tomkinson & Olds, 2000).

Very few studies show that the degree of FA correlates with reproductive potential. In men, more symmetrical individuals have higher sperm number per ejaculate, greater sperm speed and migration ability (Manning, Scutt, & Lewis-Jones, 1998), and higher sperm motility (Soler et al., 2003). In women, studies have indicated a relationship between breast symmetry and fertility-related traits. Women with higher breast symmetry have a higher number of offspring (Møller, Soler, & Thornhill, 1995)—a finding confirmed by another study that, in addition, documented an earlier age at first birth for more symmetrical women (Manning, Scutt, Whitehouse, & Leinster, 1997). However, both the higher number of children and lower age at first birth may be caused indirectly by the increased attractiveness of symmetrical women. Attractive women may be able to attract more partners and choose those with better socioeconomic status, which, in turn, may lead to a higher number of children in such unions. Similarly, earlier age at first birth in more attractive women may result from an earlier age at marriage, which is a strong predictor of first reproductive event (Wood, 1994). Therefore, the above data do not provide conclusive evidence of a causal biologic link between FA and fertility. To prove that such a link really exists, one needs to show a relationship between asymmetry and a woman's physiological reproductive potential. Such potential can be assessed by measurements of ovarian steroid hormones, which are indicators of the probability of conception and successful pregnancy (Baird et al., 1997, Baird et al., 1999, Gorkemli et al., 2004, Lipson & Ellison, 1996, Lukaszuk et al., 2005, Sharara & McClamrock, 1999).

We hypothesize that in women, FA will be related to reproductive potential. More specifically, we expect that more symmetrical women would have higher levels of estradiol, which are predictors and determinants of successful pregnancy.

2. Materials and methods

2.4. Statistical analysis

We tested differences in anthropometric, lifestyle, and reproductive variables between symmetrical and asymmetrical women by two-sided t tests (Table 1). The mean lengths of the fourth fingers of the right and left hands were not significantly different [68.6 and 68.7 mm, respectively; paired t test, t(170)=1.0, p=.32]. In addition, the variance in the length of the right fourth finger was not significantly different from the variance in the length of the left fourth finger (p>.25). In contrast, average lengths of the second fingers of the right and left hands were significantly different [68.1 and 67.5 mm, respectively; paired t test, t(170)=4.3, p<.0001]; consequently, because this indicated a directional asymmetry, measurements of the length of the second finger were not used in analyses. FA of the fourth finger had a symmetrical distribution, with low skewness (0.354, p>.05), but a significant leptokurtosis (kurtosis=3.508, p<.001). Leptokurtosis seems to be a common feature of the distribution of FA (Gangestad & Thornhill, 1999) and is suggested to be a result of early loss of highly asymmetrical individuals (Martin, Manning, & Dowrick, 1999).

Table 1. Anthropometric, reproductive, and lifestyle characteristics of groups of symmetrical and asymmetrical women

		Symmetrical women (n = 123a) [mean (S.D.)]	Asymmetrical women (n = 48a) [mean (S.D.)]	p
	Age (years)	29.5 (3.29)	29.5 (3.37)	.99
	Age at menarche (years)	13.4 (1.42)	13.3 (1.11)	.73
	Education (years)	15.6 (3.27)	15.7 (3.03)	.88
	Birth weight (g)	3303 (629.6)	3254 (590.0)	.65
	Birth length (cm)	53.6 (3.46)	53.6 (4.46)	.97
	Body weight (kg)	61.6 (10.57)	60.6 (8.31)	.52
	Body height (cm)	162.8 (6.53)	165.1 (5.41)	.02
	BMI (kg/m2)	23.3 (3.82)	22.2 (2.73)	.07
	Body fat (%)	27.8 (7.52)	25.8 (6.35)	.14
	Breast-to-underbreast ratio	1.17 (0.043)	1.17 (0.039)	.42
	Waist-to-hip ratio	0.72 (0.049)	0.72 (0.044)	.96
	Absolute difference between fourth fingers (mm)	0.58 (0.495)	2.67 (1.191)	.0001
	Length of menstrual cycle (days)	28.7 (3.6)	28.2 (3.66)	.32
	Age at first birth (years)	23.2 (3.01)	23.0 (3.33)	.71
	No. of children	0.9 (1.17)	0.8 (1.17)	.63
	Physical activity (MET-h/day)	8.8 (3.02)	8.6 (2.63)	.72
	Smoking (cigarettes/day)	1.95 (4.81)	1.90 (4.66)	.96
	Alcohol consumption (U/day)	1.7 (1.64)	1.7 (1.52)	.97

p values derived from two-sided t tests.

a Numbers in groups may vary due to missing values.

Differences between the groups of symmetrical and asymmetrical women in estradiol levels for aligned cycle Days −9 to +8 were analyzed by two-way ANOVA. Classification as symmetrical and asymmetrical was used as one factor and day of the menstrual cycle was used as the second factor (with 18 levels). Because the symmetry groups showed statistically significant differences in height and differences in body mass index (BMI) that were approaching statistical significance, we performed additional analyses to test the effects of these potential confounders: a three-way ANOVA with height (divided into tertiles) as the additional factor and a three-way ANOVA with BMI (divided into tertiles) as the third factor.

We also tested differences between symmetrical and asymmetrical women in mid-cycle estradiol levels (aligned cycle Days −3 to +1) by two-way ANOVA. The mid-cycle phase, which directly precedes ovulation, represents the part of the menstrual cycle with the highest estradiol levels.

In an attempt to eliminate possible confounding factors, all analyses were repeated with the sample restricted to include only urban women, as rural women have been shown to experience changes in reproductive hormones due to lifestyle influences, especially intense physical activity (Jasienska & Ellison, 1998, Jasienska & Ellison, 2004, Jasienska et al., in press).

3. Results

Symmetrical women had 13% higher mean estradiol levels than asymmetrical women [19.4 and 17.1 pmol/l, respectively; F(1, 2805)=17.24, p=.0001]. The difference was even more pronounced when this analysis was restricted to urban women. Symmetrical urban women had 21% higher mean estradiol levels than asymmetrical urban women [20.4 and 16.9 pmol/l, respectively; F(1, 2023)=27.91, p=.0001] (Fig. 1).

Fig. 1. Mean (with 95% confidence intervals) profiles of estradiol for symmetrical and asymmetrical urban women.

Symmetrical women had 21% higher mean mid-cycle estradiol levels than asymmetrical women [24.4 and 20.2 pmol/l, respectively; F(1, 807)=12.98, p=.0003]. In an analysis restricted to urban women, the difference in mean mid-cycle estradiol levels between symmetrical and asymmetrical women was even more pronounced, with levels in symmetrical women 28% higher than those in asymmetrical women [25.9 and 20.2 pmol/l, respectively; F(1, 585)=16.14, p=.0001].

Symmetrical and asymmetrical women differed in body height and BMI; the difference was significant for height and approached significance for BMI. There was no significant difference in other factors that could be potential confounders (Table 1). The differences in estradiol levels between symmetrical and asymmetrical women remained statistically significant [F(1, 2733)=17.76, p=.0001] in a three-way ANOVA with height tertiles as the third factor. Height tertiles did not differ significantly in estradiol levels [F(2, 2733)=1.98, p=.14]. Similarly, the results of a three-way ANOVA with BMI tertiles as the third factor showed that the differences in estradiol levels between symmetrical and asymmetrical women remained statistically significant after adjusting for the BMI [F(1, 2733)=16.29, p=.0001] (Table 2). BMI also showed a significant relationship with estradiol levels [F(2, 2733)=7.79, p=.0004]. The interaction between symmetry and tertiles of BMI was not statistically significant [F(1, 2733)=0.07, p=.93].

Table 2. Mean estradiol levels (from aligned cycle Days −9 to +8) in groups of symmetrical and asymmetrical women in tertiles of BMI

		Symmetrical women		Asymmetrical women		p
		Estradiol (pmol/l) [mean (S.D.)]	n	Mean estradiol (pmol/l) [mean (S.D.)]	n
	Low BMI	18.4 (13.02)	36	16.3 (9.4)	19	.022
	Moderate BMI	20.8 (15.72)	46	18.2 (10.73)	21	.003
	High BMI	18.7 (14.47)	41	16.1 (9.61)	8	.045

4. Discussion

We have shown that a measurement of symmetry in women is related to levels of estradiol during the menstrual cycle. Women characterized as more symmetrical had higher levels of estradiol than less symmetrical women (Fig. 1). Estradiol produced during the menstrual cycle is crucial for successful conception, and levels of estradiol are important indicators of a woman's ability to conceive (Baird et al., 1997, Baird et al., 1999, Gorkemli et al., 2004, Lipson & Ellison, 1996, Lukaszuk et al., 2005, Sharara & McClamrock, 1999). In women from the United States, the average estradiol levels in a menstrual cycle were associated with the probability of conception of 12%, whereas a 37% rise in estradiol levels led to an increase of probability of conception to approximately 35% (Lipson & Ellison, 1996). In our study, mean mid-cycle estradiol levels in symmetrical urban women were almost 30% higher than those in asymmetrical women from the same population. Such difference in hormone levels suggests a substantial increase in the probability of conception for symmetrical women. We have recently shown that levels of estradiol in reproductive age women are related to their size at birth (Jasienska, Ziomkiewicz, Lipson, Thune, & Ellison, 2006), indicating that developmental conditions influence future fecundity. This finding may suggest a link between FA, which is also believed to arise due to developmental stressors, and estradiol levels in women.

Although we used a simple univariate indicator of symmetry (and this study would benefit from the use of more comprehensive multivariate measure of symmetry) in our study, we succeeded in measuring estradiol in samples collected daily for one entire menstrual cycle. Due to large intracycle variations in daily estradiol levels, such frequent sampling is necessary for precise assessment of individual hormonal levels (Jasienska, 2002). In addition, levels of estradiol measured in saliva, in contrast to serum measurements, represent the unbound—and therefore biologically active—fraction of that hormone (Ellison, 1988). We were also able to control for a wide array of factors that may potentially influence levels of ovarian hormones in women. Results of this study indicate that FA is related to levels of estradiol, even after controlling for potential confounders.

In this study, more symmetrical women differed from less symmetrical women in the mean levels of estradiol, which are related to the probability of conception (i.e., in their reproductive potential) but not in reproductive history traits. Age at menarche, age at first birth, and number of children were similar for both groups of women (Table 1). However, in modern family planning societies, a relationship between physiological reproductive potential and reproductive success may no longer exist. Age at menarche depends on nutritional status both in utero (Adair, 2001, Romundstad et al., 2003) and during childhood development (Ellison, 1982, Ellison, 1990, Vermeulen, 1993). In modern societies, age at first birth and number of children are mainly consequences of family planning. Although the women in our study did not use hormone-based contraception for at least 6 months prior to their participation, they might have used hormone-based contraception and other methods of family planning during the years before our study began. In addition, our study was not designed to investigate differences in reproductive success in relation to FA because the women in our study were relatively young and most likely their family size was not yet complete. In general, in modern societies, age at first birth and completed family size may depend more on behavioral than on physiological determinants, rendering a relationship between FA and these reproductive characteristics no longer detectable (but see Manning et al., 1997, Møller et al., 1995).

Because FA is hypothesized to be a biomarker of an individual's fitness, several investigations have searched for a relationship between symmetry and health, with mixed results (Al-Eisa et al., 2004, Henderson & Anglin, 2003, Manning et al., 1997, Martin et al., 1999, Milne et al., 2003; see also recent reviews: Grammer et al., 2005, Weeden & Sabini, 2005). Our study shows that more symmetrical women of reproductive age have higher levels of estrogen. Based on these findings, one may hypothesize that the relationship between symmetry and health may be different for premenopausal and postmenopausal women.

In women, many aspects of health are estrogen dependent, such as cardiovascular health, osteoporosis, depression, and reproductive cancers (Barrett-Conor & Bush, 1991, Jasienska, 2002, Jasienska & Thune, 2001a, Jasienska & Thune, 2001b, Jasienska et al., 2000, Key & Pike, 1988, Nguyen et al., 1995, Pike et al., 1993). Therefore, in women of reproductive age (premenopausal), those with higher levels of estrogens may also be in better health, because estrogen at normal physiological levels acts as an immunostimulant (Jacobson & Ansari, 2004, Kovacs et al., 2002). However, women who had high levels of estrogens during their reproductive years may suffer detrimental health effects in their postmenopausal years, as high lifetime levels of reproductive hormones are related to an increased risk of breast cancer and other hormone-dependent cancers (Jasienska & Thune, 2001a, Jasienska & Thune, 2001b, Jasienska et al., 2000, Key & Pike, 1988, Pike et al., 1993). In contrast, due to lower estrogen levels in menstrual cycles, asymmetrical women may have more health problems and reduced fecundity prior to menopause than more symmetrical women but may benefit from a lower risk of hormone-dependent cancers in the postmenopausal period. Results of our research suggest that in studies on the relationship between symmetry and health in women, predictions should be made contingent upon a woman's menopausal status.

Other authors have suggested that the human preference for more symmetrical individuals evolved to facilitate the choice of partners with greater developmental stability. The results of this study suggest that, in women, symmetry is related to higher levels of estradiol and, thus, higher potential fertility. As a consequence, compared with men who do not use symmetry as a signal of women's potential fertility, men attracted to more symmetrical women may achieve higher reproductive success.