1. Introduction

Skin colour has been used in the social construction of race. Therefore, the selective pressures involved in the evolution and maintenance of skin pigmentation have been obscured by ideological and political polemic. This work is an attempt to clarify the behavioural and physiological factors influencing the frequency of the genes that determine the pigmentation of the skin. We suggest that it is sexual selection (through male–male competition, which favours testosteronized men in polygynous societies and mate choice for light-skinned oestrogenized women in monogamous societies) which is the primary selection pressure that determines the skin's sensitivity to ultraviolet (UV) light and hence skin colour itself. Polygyny favours the evolution of high prenatal testosterone and this leads to a susceptibility to sunburn and skin infections. Monogamy favours high prenatal oestrogen that is protective against sunburn and skin infections. Very black skin evolves in populations with high UV and polygyny. Very light skin is associated with low intensities of UV and monogamy. Here we use the second to fourth digit ratio (2D:4D), a likely correlate of prenatal sex steroids, to disentangle the associations between sexual selection, skin colour and photo-protection.

Skin colour varies considerably within and between human groups. Much of this variation is dependent on the light-absorbing polymer melanin. Two types of melanin are common, the black-brown eumelanins and the yellow to reddish pheomelanins. The production of melanin occurs within melanocytes, which are found on the basal layer of the skin, between the epidermis and the dermis (Nordlund, Boissy, Hearing, King, & Ortonne, 1998). The melanocytes have many fine processes that extend to, and intertwine with, the surrounding cells called keratinocytes. Melanin is synthesised in the melanocytes, packaged into vesicles or melanosomes and exported to the surrounding cells. The majority of melanin within the skin is therefore to be found within the keratinocytes. Skin colour is dependent on the size and number of the melanosomes, in addition to the nature of their melanin content (Ortonne, 2002).

Much is now known about the biochemical pathways that lead to the production of melanin, and the genes which control these pathways (Sturm, Teasdale, & Box, 2001). However, the nature of the selection pressures which influence the frequencies of pigmentation genes remain obscure. There is evidence for at least four important selective pressures on human pigmentation. They include a widespread male preference for light-skinned women, an association between dark skin and polygyny which may arise from selection for high testosterone, the possibility that melanocytes, melanosomes, and melanin form a physical barrier to skin penetration by microorganisms, and a role for melanin as protection against UV. Each of these selective pressures have their advocates, and at present it is not at all clear which are the most important.

The evidence for a mate choice effect on skin colour has been summarised by Aoki (2002). In all human groups, males tend to be darker skinned than females (Wagner, Parra, Norton, Jovel, & Shriver, 2002). This sexual dimorphism is likely to arise from the differences in prenatal and adult oestrogen found in females and males. Women's skin lightens at puberty whereas men's skin becomes darker, and there is evidence from twin studies that this sexual differentiation is under genetic control (Omoto, 1965). Oestrogen may in fact increase the production of melanin, but the effect is not strong (Edwards & Duntley, 1949) and is only apparent at high concentrations Snell & Bischitz, 1963, Snell & Turner, 1966. The sex difference in skin colour may arise from early organisational effects of oestrogen, and indirectly from the increase in female subcutaneous fat seen at puberty because within this layer androgens are converted to oestrogens Mazess, 1967, Siiteri & MacDonald, 1973.

Of course skin pigmentation is partly dependent on exposure to light. The literature on skin colour makes a distinction between so-called “facultative pigmentation” which is measured from UV-exposed body parts, and “constitutive pigmentation” measured from the inner areas of the upper forearm and upper arm (Wagner et al., 2002). We accept this distinction, but it should be treated with caution. Men tan more readily than women (Harvey, 1985) and exposed parts of the body may show strong sex differences for pigment; areas such as the inner arm that are low in subcutaneous fat and measurements of pigment may underestimate the sex difference in colour. A male preference for women with lighter-than-average skin for their ethnic group has been shown by van den Berghe and Frost (1986) to be widespread. Their sample of 51 societies showed preferences for light-skinned women in 47, with no clear preference in the remainder. The sample included a number of traditional societies without strong class differences, so the findings argue against an environmental effect whereby skin colour is a proxy for status-related exposure to UV. Van den Berghe and Frost propose instead that a link between skin colour and fertility explains male preferences for light skin.

There is evidence that marriage systems are associated with skin colour. Frost (1994) has pointed out that sub-Saharan peoples have dark skins and very high frequencies of “generalised polygyny” (85% of societies have polygyny in >20% of sexual unions), whereas circum-Mediterranean peoples have lighter skins and low frequencies (36%) of generalised polygyny. The reasons for this difference in the distribution of polygyny may relate to female food gathering. In traditional groups African women contribute more to food procurement than women in Europe, perhaps because the latter experience longer winters which restrict gathering. Polygyny in sub-Saharan Africa is therefore, from a male perspective, less “expensive” per wife. In comparison to sub-Saharan Africa polygyny is much less common in New World tropical societies, and this may explain why skin colour is also lighter in the New World. However, one needs to be cautious when considering the evolution of skin colour in the New World. Amerindians may have arrived in the Americas as little as 15,000 years ago, and one may question whether this is enough time for skin pigmentation to evolve in response to UV. However, the modern Scandinavians colonised Scandinavia only about 4000 to 5000 years ago, and that was apparently enough time to evolve the pale skins that appear to fit them to their low UV environment Diamond, 1991, Diamond, 1997, Frost, 2001. Returning to Old World peoples, comparison between societies from different continents risks the possibility that historical connection within continents may generate spurious between-continent relationships. However, Frost (1994) has compared mating systems and skin colour within sub-Saharan Africa. Thus, Khoisan peoples and pygmies are weakly polygynous and have lighter skin than the strongly polygynous Bantu group. The light skin colour of pygmies has been explained because they live in the shade of dense rain forest, but the Khoisan also have light skin and they are found in open desert. It seems that there is indeed some connection between polygyny and dark skin.

One strong correlate of polygyny is high pathogen load (Low, 2000). Therefore, a function of melanin as a barrier against skin pathogens may be an important aspect of the evolution of dark skin. The case for a role of melanin as an inhibitor of proliferation of bacterial and fungal infections in the dermis and epidermis has been convincingly made by Mackintosh (2001). He points out that the distribution of melanin in different tissues of the body does not strongly suggest a photo-protection function. Thus, melanocytes can be plentiful in areas, which are not often exposed to UV such as the skin of the genitalia, the throat, nasal and auditory passages, and internal membranes such as the peritoneum and brain tissues. Skin protected by dense fur may be highly melanised e.g., in dogs, seals and polar bears, and nocturnal animals such as bats and possums may also have melanised skin. From a phylogenetic perspective melanin is present throughout the Metazoa suggesting an early evolutionary origin. The production of melanin involves the stepwise modification of the amino acid tyrosine and its derived compounds. In invertebrates melanin production is triggered by the chemical signatures characteristic of microorganisms and parasites, the invading bacteria and eukaryotic parasites are then trapped within humoral capsules packed with melanin. In humans and other vertebrates melanocytes may process and present antigens to CD4+ T cells, they may also act as phagocytes against bacteria, and melanin can act as a physical barrier to the entry of microorganisms into the dermis.

A relationship between melanin, mate choice, polygyny and resistance to microorganisms does not preclude a role for melanin in photoprotection. In some parts of the world, particularly Old World populations, there are strong associations among skin colour, latitude, and UV in indigenous peoples Jablonski & Chaplin, 2000, Relethford, 1997. The implication of these correlations is that skin pigment is adaptive in areas of high UV because it is protective against sunburn and the breakdown of folate, and maladaptive in areas of low UV because light promotes the formation of vitamin D in a skin layer below the melanised region. The support for this influential theory is mixed. A very low level of skin pigment, such as is found in albinos, is related to severe sunburn, elevated rates of skin cancer and early mortality in the Tropics (Robins, 1991), but in more normally pigmented individuals, males show greater burn responses to UV than do females despite having darker skin (Wagner et al., 2002). There are also inconsistencies in the associations between latitude and pigmentation suggesting that “black” skin may not provide a significant advantage, in terms of UV protection, over “brown” skin. For example, the Khoisan peoples have yellowish-brown skin but are found in the “high-UV” environment of the Kalahari and the tropical Amerindians also have an intermediate skin colour. In the case of the former this does not appear to be the result of lack of time for an adaptive response. The Khoisan occupied much of central, eastern and southern Africa before being pushed into their Kalahari “refuge” by “black” peoples from West Africa. Many of the world's darkest peoples are found at low latitudes in sub-Saharan Africa, and pigmentation decreases as one goes north through the circum-Mediterranean region (Jablonski & Chaplin, 2000). However, similarly strong clinal effects in skin pigmentation are not found in the New World where very pigmented indigenous peoples are not found at low latitudes (Diamond, 1991). It is true that published correlations between UV and skin pigmentation are strong, but most up-to-date samples of skin reflectance scores and UV rely heavily on Old World samples, and caution must be exercised in their interpretation. In addition the requirement for a light skin at high latitudes in order that Vitamin D can be synthesised has been challenged. Robins (1991) has pointed out that vitamin D produced during the long summer in high latitudes may be stored in muscle and fat for use during the low light intensities of the winter. Therefore, light skin at high latitudes is not essential if one is to avoid vitamin D deficiency and rickets. Robins's arguments are controversial (e.g., see Jablonski & Chaplin, 2000), but there are sufficient doubts about the melanin-and-photoprotection and light skin-and-vitamin D theory to warrant a reinterpretation of the data. Here we suggest that sexual selection may predispose some populations to sunburn and a tendency to skin problems. The relationship between forms of sexual selection and latitude may then result in correlations between skin colour and latitude. The key to understanding these effects may lie in identifying the prenatal influences of oestrogen and testosterone on the skin. We use 2D:4D as a tool to investigate the associations between skin pigmentation and prenatal sex steroids.

The relationships between “constitutive” pigmentation and oestrogen are likely to be independent of short-term effects of light and may arise as early as the prenatal period. If this is so we might expect that 2D:4D ratio would be predictive of skin colour because there is accumulating evidence that 2D:4D is a positive correlate of in utero levels of oestrogen. Thus, mean values of 2D:4D are higher in females than in males (Manning, Scutt, Wilson, & Lewis-Jones, 1998), the dimorphism is found in children as young as 2 years, and there appears to be little change in the sex difference at puberty (Manning et al., 1998). The sex difference in 2D:4D is robust across populations but there are also strong differences among ethnicities (Manning et al., 2000), for reasons unknown. Perhaps selective pressures related to marriage systems affect levels of prenatal sex steroids, which in turn influence immune status, pathogen loads, and digit ratios. Manning (2002) has reported a curvilinear relationship between latitude and 2D:4D in a sample of nine populations: Black populations, including Jamaicans and South Africans, had low mean 2D:4D in both sexes, while Caucasian populations showed considerable variation with some having high mean 2D:4D ratios and a suggestion of low ratios at very high latitudes. Immune mechanisms and pathogen loads both show considerable sexual dimorphism, which is dependent on prenatal and adult sex steroids, and pathogen loads may be related to 2D:4D (Manning, 2002). Overall, the evidence therefore indicates that 2D:4D and skin colour are both related to sex, pathogen load and geographical differences, and hence that 2D:4D may help us to understand why skin colour is associated with sex, pathogens, and ethnicity.

Here we report data from two studies conducted on Caucasian samples. The first concerns the relationship between skin colour and 2D:4D in men and women, the second the association between 2D:4D and susceptibility to sunburn and diseases of the skin, namely eczema, dandruff, and athlete's foot.

2. Study I

2.2. Results

Descriptive statistics (mean±S.D.) of the sample were as follows: age 24.64±7.37 years; forearm pigment score males 667.16±31.03, females 656.84±32.43; hand pigment score males 710.91±32.06, females 693.36±39.41; right 2D:4D males 0.96±0.03, females 0.97±0.03; left 2D:4D males 0.95±0.03, females 0.96±0.03.

The intraclass correlation coefficient for the pigment scores was high (forearm r1=.98, hand r1=.99). A repeated-measures ANOVA test showed between-individual differences in pigment score was significantly higher than within-individual differences or measurement error (forearm F=108.43, P=.0001; hand F=269.93, P=.0001). We concluded that our pigment measurements represented real between-individual differences. The reliability of the colour scores was further supported by significant correlations between right and left forearm pigment scores (r=.89, P=.0001) and right and left hand pigment scores (r=.67, P=.0001). The intraclass coefficients for right and left 2D:4D were high (right 2D:4D r1=.94, left 2D:4D r1=.96) and between-individual differences in 2D:4D were significantly higher than within-individual differences or measurement error (repeated-measures ANOVA right 2D:4D F=31.61, P=.0001, left 2D:4D F=43.09, P=.0001). We concluded that our measurements of 2D:4D represented real differences between participants.

As expected, we found significant sex differences in the pigment scores with females having lighter skin than males for both forearm and hand scores (forearm, t=2.47, P=.01; hand, t=3.71, P=.0003). There were also significant sex differences in right hand 2D:4D ratio with males having lower 2D:4D than females (t=3.32, P=.001), while the sex difference in left 2D:4D was in the same direction and very close to significance (t=1.97, P=.05). Age was not related to forearm pigment scores (r=.004, P=.96), but there was a significant positive association between age and pigment score of the hands (r=.24, P=.0003).

With regard to digit ratio and skin colour we are most interested in the relationships between 2D:4D and our measure of “constitutive” pigmentation of the forearm. Left and right 2D:4D were negatively correlated with forearm pigment scores in women but not in men (Table 1 and Fig. 1). The female associations survived Bonferroni correction. There were no significant associations between 2D:4D and “facultative” pigment scores of the hand.

Table 1. Relationships between 2D:4D and “constitutive” skin colour of the upper forearm and “facultative” colour of the hand

		2D:4D right		2D:4D left
		r	P	r	P
	Colour score—forearm (“constitutive” pigmentation)
	Males	−.02	.83	−.04	.65
	Females	−.25	.009*	−.33	.0003**
	Colour score—hand (“facultative” pigmentation)
	Males	.02	.81	.01	.90
	Females	−.12	.22	−.16	.09

* Bonferroni correction at P=.03. ** Bonferroni correction at P=.001.

Fig. 1. The relationship between 2D:4D of the left hand of 115 white female Caucasian participants and skin pigment scores from the inner arm. The formula for the line is y=−345.229x+989.563; r2=.11.

3. Study II

3.2. Results

Descriptive statistics of the sample (mean±S.D.) were as follows: age 21.05±2.52 years; right 2D:4D males 0.97±0.02, females 0.98±0.03; left 2D:4D males 0.97±0.03, females 0.99±0.03. As expected, males tended to have lower values of 2D:4D than females and this was significant for the left hand but not the right (right t=1.06, P=.27; left t=2.07, P=.04).

The intraclass correlation coefficients for first and second measurements of 2D:4D were high for both right and left hands (right r1=.90, left r1=.94). Repeated measures ANOVA analyses showed that between-subject differences in 2D:4D were significantly greater than within-subject differences (measurement error) for both right and left hands (right F=18.34, P=.0001; left F=33.13, P=.0001). We concluded that our measurements of 2D:4D reflected real between-subject differences in digit ratio.

Comparisons of mean 2D:4D in subjects who reported susceptibility to sunburn, eczema, dandruff, and athlete's foot with the mean 2D:4D and those not reporting susceptibility are given in Table 2. Where there were significant differences, these were consistently in the direction of an association between lower ratios and susceptibility. For sunburn, males who reported susceptibility had significantly lower left 2D:4D than those who did not (P=.008). For eczema, susceptible men had lower right 2D:4D ratios than nonsusceptible men (P=.003). Susceptibility to dandruff was related to lower left 2D:4D in females (P=.009). Finally, for athlete's foot, mean left 2D:4D was lower in susceptible participants than in the nonsusceptible, with both sexes combined (P=.04), but on correction for multiple tests this contrast was no longer significant (P=.16).

Table 2. Associations between 2D:4D and sunburn, eczema, dandruff, and athlete's foot

		2D:4D right				2D:4D left
		yes	no	t	P	yes	no	t	P
	Sunburn
	Males	.98±.02	.97±.03	0.51	.66	.96±.04	.98±.03	2.69	.008***
	Females	.99±.02	.98±.03	1.36	.18	.99±.03	.99±.03	0.10	.92
	23 males and 26 females reported a susceptibility to sunburn.
	Eczema
	Males	.95±.02	.98±.02	3.07	.003****	.97±.03	.98±.03	0.55	.58
	Females	.98±.02	.98±.03	0.81	.42	.98±.03	.99±.03	0.79	.43
	7 males and 35 females reported a diagnosis of eczema.
	Dandruff
	Males	.97±.02	.98±.02	1.94	.06	.98±.03	.98±.03	0.51	.61
	Females	.97±.04	.98±.02	1.67	.10	.97±.04	.99±.03	2.70	.009**
	40 males and 26 females reported a susceptibility to dandruff.
	Athlete's foot
	Males	.97±.02	.98±.03	0.17	.86	.97±.04	.98±.02	1.69	.10
	Females	.98±.02	.98±.03	0.51	.61	.98±.04	.99±.04	2.11	.04*
	44 males and 31 females reported a susceptibility to athlete's foot.

*** Bonferroni correction at P=.02. **** Bonferroni correction at P=.01. ** Bonferroni correction at P=.04. * Bonferroni correction at P=.16.

4. Discussion

With regard to our data on 2D:4D ratio we have two associations which bear on the question of the evolution of skin colour.

Firstly, we have found that women with light skin tend to have high female-type finger ratios. This is the first report of an association between skin colour and 2D:4D ratio, and since there is evidence that 2D:4D is positively related to in utero oestrogen, skin colour in women is therefore likely to be negatively associated with prenatal oestrogen, making it a possible marker of the organisational effects of oestrogen on “constitutive” skin pigmentation. However, this does not exclude the possibility that female skin colour also explains some of the variance in adult oestrogen concentrations. A high 2D:4D has been found to correlate with large family size in women (Manning et al., 2000). Therefore, the widespread male preference for women with light skin may be adaptive in that the payoff from such a preference may be an increase in fertility. As male choice for lighter-than-average skin is widespread, this aspect of sexual selection will tend to reduce skin pigment universally. A consequence of selection for light skin will be an increase in prenatal oestrogen. The increase is unlikely to be restricted to female foetuses as correlated effects are expected to be found in male foetuses (Manning, 2002), but an increase in oestrogen in males will have deleterious consequences on such things as the formation of the heart and blood vessels (Manning, 2002), so male-related foetal effects are likely to slow the spread of genes for lighter skin and increased oestrogen. In addition, the variance of female reproductive success is quite small, being approximately equal to that of males in strictly monogamous groups and lower than that of males in polygynous societies. This is another factor that is likely to diminish the strength of selection for lighter skin. In polygynous societies sexual selection for high prenatal testosterone in male foetuses is likely to strongly oppose the genetic effects of mate choice preferences. Also male mating preferences are less important in strongly polygynous societies in which intense male–male competition means that all women quickly find mates and do not lose reproductive time. In weakly polygynous populations such as the !Kung, 10% to 20% of women may lose 5 to 15 years of reproductive time while waiting between partners (Howell, 1979). Therefore, male choice for light-skinned females may be a stronger force for the reduction of skin pigment in monogamous peoples than in polygynous groups.

Secondly, we have evidence that low male-type 2D:4D is associated with susceptibility to sunburn, eczema and dandruff. Low 2D:4D is likely to be a correlate of high prenatal testosterone, and may be common in those groups where male–male competition for females is strong (Manning, 2002). Polygyny is expected to result in high variance in male fitness and strong selective pressure for high prenatal testosterone in male foetuses. This will be countered to some extent by correlated effects of testosterone on female foetuses but if the reproductive payoff to successful males is sufficiently large then genes for high prenatal testosterone will spread. The association between low 2D:4D and sunburn, eczema, and dandruff may be evidence of a deleterious organising effect of testosterone on the developing immune system and in particular that part associated with the skin. Interestingly polygyny is associated with high pathogen loads (Low, 2000), which may result in part from the negative effects of prenatal testosterone on immunocompetence. In addition, the negative associations between 2D:4D and sunburn, eczema, and dandruff may also indicate that prenatal testosterone can have a disruptive effect on the overall integrity of the skin, reducing the water-holding capacity of the dermis, increasing water loss from its outer layers and causing disorder to the normal process of producing new cells and sloughing-off old cells. With regard to sunburn, UV suppresses cutaneous cell-mediated immunity and the immune system responds to UV by initiating the inflammatory response. Oestrogen is thought to afford protection against UV by increasing the water-holding capacity of the dermis, and testosterone may have opposing effects (Wagner et al., 2002). With eczema or atopic dermatitis, there are structural problems in the outer layer of the epidermis, the stratum corneum, which tends to dry and crack, resulting in water loss and an increased colonisation by microorganisms such as Staphylococcus aureus(Heaton, Mallon, Venaille, & Holt, 2003). As for dandruff, here we see a disruption in the normal orderly process of producing new cells in the scalp while losing dead cells: dandruff is the shedding of dead skin at an excessive rate, resulting in the premature loss of cells, many of which are not dead. As with eczema dandruff may be related to colonisation by microorganisms, e.g., the fungus Pityrosporum ovale, but this may be secondary to the original condition of disordered cell production and loss (McGinley, Leyden, Marples, & Kligman, 1975). Melanin may block the penetration of the skin by microorganisms, thus helping to prevent disruption of the skin's integrity (Mackintosh, 2001). Therefore, patterns of 2D:4D support an association between melanised skin and polygyny.

Comparative evidence regarding marriage systems indicates considerable geographical differences in the distribution of polygyny in relation to latitude (Murdock, 1967). In the Old World polygynous societies tend to be found in the Tropics, but in the New World there is no such association between polygyny and latitude. Our 2D:4D data suggest that polygynous groups may be most at risk for damage by UV and deleterious skin conditions which will lead to penetration of the skin by microorganisms. Highly melanised skin should therefore be associated with polygyny, particularly in areas where UV intensity is high. Indigenous groups in sub-Saharan Africa are very highly melanised and most are polygynous, whereas at similar latitudes in the New World polygyny is uncommon and pigmentation is not very high (Frost, 1994). Polygyny at high latitudes may result in the evolution of high prenatal testosterone and dark skin despite moderate levels of UV. For example the now-extinct natives of Tasmania lived at the temperate latitude of 42° S, but had dark skin. Murdock (1967) gives the Tasmanian marriage system as polygynous with female capture. Such a system would be expected to result in strong male–male competition for wives, high prenatal testosterone and a susceptibility to sunburn. Therefore, dark skin is not unexpected despite the temperate latitude.

In conclusion, we think our findings with regard to 2D:4D, skin colour, and sunburn are supportive of an important role for sexual selection in the evolution of skin pigmentation. We suggest that protection against UV and against microorganisms, which colonise the skin is important in polygynous societies because of a link with high prenatal testosterone. It is in these populations where a role for melanin is most important. A widespread preference for light-skinned women may assume more selective importance in monogamous groups than in polygynous societies.