I am puzzled at the frequent invocation of sexual selection to solve awkward evolutionary puzzles, especially in human evolution. Hairless skin? No problem: sexual selection! Big brain? Sexual selection! Language abilities? Sexual selection! Musical and artistic skills? Sexual selection! An extreme example of this trend is Geoffrey Miller's The Mating Mind, which I commented on here.

This uncritical enthusiasm for sexual selection leads me to set out some reasons for caution...

I base my discussion mainly on Darwin's Descent of Man, Malte Andersson's Sexual Selection, and Patrick Bateson (ed) Mate Choice, along with Fisher, Hamilton, Trivers, and the other usual suspects. My comments on the marriage customs of primitive people are based on Westermarck's History of Human Marriage and a variety of more recent sources, including those cited here.

By Darwin's definition, a sexually selected trait is one which gives an individual 'an advantage over other individuals of the same sex and species, in exclusive relation to reproduction'. The advantage consists in an increase in the quantity or quality of sexual partners as compared with individuals who do not possess the trait. Darwin further distinguished between sexual selection based on a physical contest between rivals (usually males), and selection based on active choice by the opposite sex. Sexual selection by contest was widely accepted by Darwin's successors, but selection by choice was more controversial. It is now generally agreed that selection by choice does occur, but there is still disagreement about the basis for that choice. Hence the controversies about the Handicap Principle, selection for genetic quality, the role of protection against parasites, the feasibility of Fisher's 'runaway' process, and so on.

The following deals only with sexual selection by choice. The mere existence of choice does not automatically lead to sexual selection in Darwin's sense. The choice must lead to some reproductive advantage, measured by the number of surviving offspring. In the case of highly polygynous breeding systems, and where there is no paternal care of offspring, it is easy for a female preference to lead to a reproductive advantage for the preferred males: they simply mate with more females and have more offspring. In (approximately) monogamous breeding systems it is more difficult, as Darwin himself emphasised. If there is a large surplus of one sex over the other, selection can operate through the failure of unpreferred individuals to get mates, but it is rare for there to be a large surplus of one sex (exceptions are discussed in a famous paper by Hamilton). Where the sex ratio is approximately equal, and mating is approximately monogamous, sexual selection can only operate at the margins, through differences in mate quality, or through extra-pair copulations.

How can we recognise if a trait is due to sexual selection? I don't think there is any foolproof test, but here are some indicators noted by Darwin and others as prima facie marks of sexually selected traits:

1. The trait is found only in one sex, or is found much more highly developed in one sex than the other.

2. The selected sex is the one with the greater variance in reproductive success, which is nearly always the male. Males can father an indefinite number of offspring, and their reproductive success usually depends primarily on the number of matings they secure, whereas female reproduction is limited by nutrition and physiology.

3. Sexual selection is strongest among highly polygynous species, such as birds with a lek mating system, or mammals where males keep a harem. Conversely sexual selection in weakest among species with monogamy and biparental care, though males can still increase their reproductive success by extra-pair copulations, and both sexes can increase it by obtaining high quality mates.

4. It is likely that a trait is sexually selected if it first appears at sexual maturity.

5. It is likely that a trait is sexually selected if it appears only during a specific mating season.

6. It is likely that a trait is sexually selected if it used prominently in courtship.

7. A trait is plausibly due to sexual selection if it is costly and appears to have no direct survival value - the classic example being the peacock's tail. But note that there may also be a sexual preference for the average or optimum state of a trait already favoured by natural selection, in which case sexual selection will act as a stabilising or retarding factor in evolution. This factor does not seem to be sufficiently discussed in the literature.

8. An animal seldom has more than one type of trait strongly developed by sexual selection. For example, birds which are good singers seldom have spectacular plumage, and birds with spectacular plumage are seldom good singers.

9. Sexually selected traits often differ conspicuously between closely related species. This may be because sexual selection has a purely arbitrary component, or because sexually selected traits serve to distinguish one species from another and prevent infertile matings.

10. My last point is tentative, but I wanted to get to 10! I suggest that sexually selected traits are unlikely to be due to single genes, and especially recessive genes. The reason is that a new gene mutation will be so rare (and in the case of a recessive gene even more rarely expressed) that it will not provide good material for sexual preferences to work on. If the trait is very rare, any individual with a preference for that trait will be at a disadvantage (if only by the time wasted in searching), so the preference is unlikely to evolve. New traits such as blue eyes, which depend on recessive genes, are prima facie unlikely to spread by sexual selection. This would require the selecting sex to have a widespread predisposition in favour of such traits, for no apparent reason.

If we consider human evolution with these points in mind, there are only a few traits which seem good candidates for explanation by sexual selection. The facial hair and deep voice of the human male, which appear only at sexual maturity, are good candidates. (But do women actually find men's facial hair attractive?) The greater average size of the male is also a plausible candidate, though this might alternatively be explained by 'contest' selection or by ecological differences between males and females. The pubic hair of both sexes is consistent with point 4, but not point 1, and it is possible that it functions merely as an indicator of sexual maturity, and not as a basis for selection between mature individuals.
[Added: There are also difficulties for the idea that humans' hairless skin is a result of sexual selection. Children have no visible body hair, then it appears, on chest, arms, legs, etc, at sexual maturity. Men have more than women, though undepilated women have more than children. This is the reverse of the pattern one would expect if hairlessness were due to sexual selection: we would expect children to be shaggy, then to lose their body hair at puberty. The pattern we actually see suggests that hairlessness serves some other purpose, but that the body hair of adults plays a sexual role of some kind - though possibly more as an indicator of maturity than as a sexual attractor. I note that male chest hair only appears about 5 years after puberty, when the man is fully grown and able to compete with adults.]

As is well-known, Darwin believed that many of the differences between human races, such as skin colour, are due to sexual selection, but he probably underrated the importance of natural selection in this area, and he had no knowledge of genetic drift (though he did suggest that selectively neutral characters might occur as 'fluctuating elements'). Most of the human racial differences are found in both sexes, and in children as well as adults, contrary to points 1 and 4 above. Moreover, blond hair, which has sometimes been cited as a sexually selected racial trait, often darkens before sexual maturity, which strongly contradicts point 4.

There are general reasons for doubting that sexual selection has been a major factor in human evolution. Sexual selection is strongest when there is wide variance in reproductive success. But if we disregard freak examples like Genghis Khan, differences in human reproductive success are relatively modest. Humans are slow, small-scale breeders and are only moderately polygynous. And in primitive societies the choice of mates is tightly constrained both by the number of available partners and by tribal customs. Choice for the individual is very limited, and often made mainly by relatives of the individuals concerned. Most females are married at puberty to older males. If women are subsequently widowed, they are usually remarried either to relatives of their dead husband or to a younger male who is not yet successful enough to have a desirable young wife. We do not know how far back in human history such customs go, but the number of available partners for marriage has probably always been small until recent times. In small, scattered populations the challenge is to find any suitable partner, not to choose among a plethora of alternatives! In short, it would be difficult to find a worse candidate species for strong sexual selection than homo sapiens. I'm sure I am not the first person to make this observation - I think it is somewhere in Maynard Smith - but it is worth making again.

Of course, even weak sexual selection might have important effects if sustained over long periods and not opposed by natural selection. But this is no use to those who want to explain the rapid evolution of costly traits such as human intelligence.

I don't know why explanations by sexual selection are so popular, but there may be a misconception that sexual selection is faster or stronger than natural selection. This is not likely to be true except in the case of a highly polygynous species. Fisher's phrase of 'runaway' sexual selection may also be responsible for the misconception, but again this process is probably the exception rather than the rule. Runaway selection is among the most difficult and controversial areas of evolutionary biology (see the essays by O'Donald and Arnold in Bateson (ed), and chapter 2 in Andersson). This has attracted some very clever theoretical biologists to study the subject, and there is a large literature, but theory far outruns empirical evidence. I also note that runaway selection is particularly unlikely to apply to selection of female traits. The reason is that in this case the process would require there to be a 'sexy daughter' effect, analogous to the 'sexy son' effect in runaway selection for male traits. But the 'sexy daughter' effect would be weak, because a daughter who attracts numerous sexual partners will not greatly increase her number of surviving offspring, unlike a 'sexy son'.

Runaway sexual selection does offer the hope of explaining seemingly arbitrary traits, because it does not require any strong initial reason for the female sexual preference. Once the process gets going, initially arbitrary preferences are self-reinforcing. But again, this probably requires special circumstances to take off, and in general it cannot be assumed that sexual preferences are arbitrary. Fisher himself proposed that the starting point of runaway selection was a preference for a trait mildly favoured by natural selection. Preferences for anything other than the existing optimum traits would normally be opposed by natural selection, so any change in sexual preferences - for example, a trend towards preferring lighter or darker skin - would itself need to be accounted for. In the most successful models of runaway selection, it cannot get started when the female preference is very rare (Andersson, chapter 2). These models also assume polygyny, though it has been shown that a moderate form of the Fisher process may work in a monogamous system (Andersson p. 42).

Overall, sexual selection is a more complicated and difficult process than natural selection, and its explanatory advantages are largely illusory. Like group selection, I think it should be an explanatory last and not first resort. Of course, this doesn't mean it should be ruled out a priori, or that it is not legitimate to propose bold hypotheses, just that the difficulties need to be recognised and squarely faced.