NOTES

Chapter One: Introduction

1. Posner (1994, 111) writes: “The ends [that sex serves] fall into three groups, which I shall call procreative, hedonistic, and sociable. The first is obvious. The second has two cells. One is relief from the urgency of sexual desire; the analogy is to scratching an itch, or to drinking water when one is thirsty. The other is ars erotica, the deliberate cultivation of the faculty of sexual pleasure; the analogy is to cultivating a taste for fine music or fine wine. The third group of sexual ends, the sociable, is the least obvious. It refers to the use of sex to construct or reinforce relationships with other people, such as spouses or friends.”

2. See the many studies in Sommer and Vasey 2006, which investigate behavior among bison, cats, dolphins, flamingos, and geese as well as many primates. Not all researchers agree about the reasons for homosexual behavior in nonhuman species, and a single explanation may not apply to all species. There are also interesting unresolved questions about whether humans show a particular propensity for exclusive homosexuality as opposed to bisexuality.

3. Many examples, including several cited in this paragraph, are described and discussed in Arnqvist and Rowe 2005 (e.g., 1–13). The various chapters in Muller and Wrangham 2009 offer a comprehensive summary of what is currently known about sexual coercion, including hypotheses about its evolutionary origins. Judson (2002, especially 9–20) offers an entertaining account of sexual conflict in various animal species.

4. See Barash and Lipton 2001; Birkhead 2000.

5. On lions, see Packer and Pusey 1983. The evidence for many primate species is summarized in van Schaik and Janson 2000, chapter 1, especially table 2.1 and pp. 40–41.

6. Arnqvist and Rowe 2005, 50–52.

7. See Whitchurch, Wilson, and Gilbert 2010. Kim Sterelny has suggested to me that sexual relations are unusual in this respect: “It does not seem to be true of other human co-operative alliances, where there is a lot to be gained and lost. We do not find potential co-authors more appealing, if we are not certain what they think of us. But writing a book together is a big and risky investment. So what makes uncertainty a bargaining lever in sex but only sex?” (personal communication, July 2011). On reflection I am not convinced that sexual relations are unusual in this respect. In labor markets, higher wages or bonuses are often paid to employees who can credibly threaten to leave; the threat of leaving itself conveys that the employee is more highly valued on the open market and therefore (perhaps) of greater value to the employer. Similarly, lovers whose affections may stray elsewhere effectively communicate that they are considered more attractive by others and may therefore seem a more valuable catch. In a different domain, one of the hypotheses of the attachment theory of parenting is that children need a “secure base” (Bowlby 1988) from which to explore the world. In this view, the more unpredictable the behavior of their caregivers, the more emotional energy the children need to invest in claiming their attention and the less they have available to explore their environment. On the application of attachment theory to adult relationships, see Crowell and Waters 2005 and the various contributions to Mikulincer and Goodman 2006.

8. Not all animals that reproduce sexually have two sexes, and in those that do, it is not always true that individuals are determinately of one sex or the other (some fish, for example, change sex according to the distribution of mating opportunities). These points, and the implication that sexual strategies are far more varied across nature than even Darwin realized, are well made in Roughgarden 2004 and 2009. Here I discuss species in which there are indeed two distinctive sexes and all or most individuals are determinately of one sex or the other.

9. A classic article by Trivers (1972) emphasized that it is the asymmetry of parental investment (including the investment in gestation) that creates these divergent incentives for males and females rather than the asymmetry of gamete size per se.

10. It can sometimes be illuminating to consider relations between the sexes as if they were a market in which males demand and females supply sexual services, a perspective developed, for example, in Baumeister and Vohs 2004. However, this view is a simplification for modeling purposes, and it should not be taken to imply that females do not value sex or the quality of the sexual encounters they engage in.

11. Bowles and Choi 2007.

12. Barash and Lipton 2001 has an extensive treatment of this evidence (e.g., p. 12). Birkhead 2000 focuses particularly on female infidelity and sperm competition (195–231). See Ledford 2008 for the surprise discovery of extra-pair copulations in a species of vole previously considered monogamous. Knight 2002 briefly and accessibly surveys studies of multiple sexuality among females of a wide range of species. Judson 2002, chapter 1, is another excellent source.

13. Baumeister (2010, 221–29) discusses at length the evidence that women on average desire sex substantially less often than men, based on how couples separately report their satisfaction with the frequency of sex within the relationship. Without distinguishing between the quality of sexual encounters, such evidence may be misleading. It's likely, for instance, that women may report sex with their current partners to be frequent enough given the quality of the experience, while still feeling that they might like sex more often if the experience were better—that is, if their partners were more generous and attentive. Men's preference for more sex than they currently get may reflect simply their greater preference than women for mediocre sex over no sex at all. Meston and Buss (2009) emphasize the wide range of motivations cited by women for desiring sex and the many kinds of cues that can trigger such a desire, suggesting that it may not be very informative to compare men and women with respect to an unconditional desire for sex, given how much the desire tends to be conditioned by circumstances.

14. Strictly speaking, all potential fathers might make some contribution to raising offspring, even if they are uncertain about their paternity without that uncertainty implying confusion (e.g., if several potential fathers have subjective paternity probabilities that add up to one). It is in the mother's interest to make each of them believe his probability of paternity is higher than it really is.

Chapter Two: Sex and Salesmanship

1. Centorrino et al. 2011.

2. On the signaling role of dance, see Brown et al. 2005.

3. Manfred Milinski (2003) has shown that scent plays an important role in facilitating mating between individuals whose immune systems have an intermediate degree of difference from each other.

4. Arnqvist and Rowe 2005, 74–77.

5. There is a substantial literature showing that perceived physical attractiveness is positively correlated with labor-market rewards (Hamermesh and Biddle 1994; Mobius and Rosenblat 2006) and electoral success (Berggren, Jordahl, and Poutvaara 2006). This finding is reviewed and its implications discussed in Hamermesh 2011 and Hakim 2011. The literature on signaling to friends and colleagues is vast: Miller 2009 provides an enjoyable, nontechnical overview. Bénabou and Tirole 2006 and Seabright 2009 develop models of the use of prosocial behavior for signaling motives. Ariely (2008, chapter 13) gives an amusing experimental demonstration of how ordering behavior in restaurants may be influenced by signaling motives, at a real cost in the perceived quality of the items ordered.

6. See Robin Hanson's blog, Overcoming Bias, September 15, 2010 (http://overcomingbias.com/2010/09/a-med-datum.html) for an excellent (anonymous) contribution from an ophthalmologist's assistant who describes the ostentatious alcohol swabbing of equipment in front of patients, while eye droppers are reused in ways that pose far more infection risk than would exist from unswabbed equipment.

7. There is evidence from both humans and other primate species that individuals perceived as sexually attractive are also treated with greater attention and generosity in nonsexual interactions, including those involving members of their own sex. In addition to the literature cited in n. 5 above, see Sapolsky 2005. Wallner and Dittami (1997) show that female Barbary macaques with larger than average anogenital swellings are not only more attractive to males but are also preferentially groomed by other females. Causality is not easy to establish here: do partners in nonsexual interactions tend to favor individuals that seem likely to enjoy greater sexual success, or do individuals attract more sexual partners because they appear to enjoy more social success? Both alternatives seem plausible.

8. Goffman 1963; Yoshino 2006. As Yoshino acknowledges, even members of majority cultures do something similar to covering, because their identities are multiple and they are continually balancing the requirements of one context against those of another.

9. See Fox 1984.

10. Tungate 2007, 15.

11. The advertising campaign used the supermodel in a number of different contexts, one of which was to have her (apparently) crashing the car deliberately to show off its airbags and other safety features. Here the fact that the advertisers were prepared to take risks with such a famous model would itself have signaled their confidence in the reliability of the car's safety systems. See YouTube video, “Citroen Xsara Advert: Crash Test with Claudia Schiffer,” http://youtube.com/watch?v=G71qlOk-qoY, accessed May 18, 2011.

12. See Doniger 2005, especially chapter 1, which discusses Cary Grant along with many other cases.

13. Miller 2009, 72.

14. Veblen 1915, 43.

15. Veyne 1996.

16. Foreman 2009.

17. Sexton and Sexton 2011.

18. Foreman 2009.

19. Calculation based on a return fare of €13,400 (US$16,400) from Paris to Tokyo with Air France, quoted on June 30, 2010, with a total flying time of just under twenty-four hours.

20. See Searcy and Novicki 2008 for a survey and discussion of the problem of honest signaling in song sparrows.

21. Reid et al. 2005. This kind of research is more difficult than it seems, and the difficulty is not simply that of crawling around in the undergrowth trying not to frighten the birds. Strictly speaking, the fact that birds with larger repertoires have more offspring could be due solely to the fact that more fertile females choose birds with larger repertoires. The researchers present other evidence to suggest that the males' contribution is important, notably that males with larger repertoires also live longer on average, suggesting that they have better underlying health.

22. MacDonald et al. 2006.

23. I hope readers will feel that this claim does not need to be corroborated with a scientific study.

Chapter Three: Seduction and the Emotions

1. See Damasio 1996. This is not to say that specific regions of the brain embody specific emotions: the embodiment appears to be functional rather than structural (see Lindquist et al. 2011).

2. This work is surveyed in the papers collected in Ellison and Gray 2009.

3. Damasio 1994.

4. Frank 1988. Fessler and Haley (2003) survey and discuss the role of emotions in human cooperation. I discuss the evidence for the emotions as a positive force for human prosociality in Seabright 2010, chapters 4 and 5, laying particular emphasis on their role in the reduction of violence in modern human societies. Bowles and Gintis (2011) discuss in great detail the evolution of human prosocial behavior.

5. Stendhal 1962, 37 (my translation).

6. Gambetta 2009, 258–59.

7. See Miller 2000, 238–41.

8. Kim Sterelny has pointed out to me that it is an established finding in psychology that “variable reinforcement schedules lead to highly persistent behaviour, so the variability of female experience might equally be taken as an adaptation to ensure continuing sexual interest” (personal communication, July 2011). This is possible, but because the male's continuing interest is typically more important to the female than vice versa, this logic would suggest that we should expect the male orgasm to be the less reliable of the two.

9. I consider evidence that this was so in chapter 4. The argument also evidently requires that no individual sexual encounter have too high a probability of leading to pregnancy.

10. See Caldwell and Young 2006 for a survey of the effects of both vasopressin and oxytocin, Kosfeld et al. 2005 for the social impact of oxytocin on trust in human subjects, and Walum et al. 2008 for evidence that genetic variation in vasopressin receptors correlates with strength of pair-bonding behavior in humans. The physiology of orgasm is discussed in great detail in Komisaruk et al. 2006. Lloyd (2005, especially 107–48) argues in favor of Donald Symons's theory of female orgasm as a pure by-product of male orgasm and claims that rival theories indicate important biases in the largely male community of scientists studying the question. As I suggest here, a by-product view of the origins of female orgasm might be compatible with a view that natural selection might co-opt such a mechanism for certain adaptive purposes, such as screening.

11. See Buss 2003, 223–34. In particular, female orgasm may have had the advantage of also responding selectively to lovers with “good” genes, since orgasm is known to increase the rate of sperm uptake into the uterus.

12. Cunningham, Barbee, and Pike 1990.

13. See Owren and Bacharowski 2001, 156.

14. We report our findings in Centorrino et al. 2011.

15. To avoid spurious correlation, in case individuals who have decided to trust others try to justify their acts by rating smiles as genuine, we compare an individual's trusting behavior with the ratings of smiles by other participants.

16. The effect on trust of smiles in still photographs was first shown by Scharlemann et al. (2001).

17. This concept was introduced into biology by Zahavi (1975) and formalized by Grafen (1990). Similar theories were being developed in parallel by the economists Michael Spence (1974) and James Mirrlees (1997).

18. In the screening model of Michael Spence (1974), workers may be of high or low productivity, but this value is not observable by employers. Employers can, however, observe how much education the employees choose to acquire. If high-productivity employees have lower costs of acquiring education (because they can succeed in class with less work), they will choose higher levels of education than low-productivity employees even if education brings neither them nor their employers any benefit at all (and is a pure handicap). The value of education to high-productivity workers is that it signals not just that those workers can bear the handicap but also that they will have higher productivity when hired by employers.

19. Weatherhead and Robinson 1979.

20. Gibson and Höglund 1992. Galef (2008) summarizes studies of mate-choice copying in Japanese quail: not only is this finding confirmed for females, but the opposite is found for males, which is consistent with the greater scarcity of the female bird's gametes than those of the male. However, the evidence on whether women find attached men more sexually attractive is mixed: for instance, Uller and Johansson (2003) find no such effect.

21. A variant of this strategy is used (or so is my unscientific impression) by younger women in the company of older and ostentatiously wealthier men: by spending a lot of time telephoning or texting, they can signal, “I'm not really with this guy.” This behavior is consistent with the dual strategy discussed by Thornhill and Gangestad (2008), who argue that human females have evolved two distinct sexualities: an “extended sexuality,” when conception is impossible, whose purpose is to elicit material benefits from males, and an “estrous sexuality,” whose purpose is to obtain “good genes” when conception is a possibility. There may be visibly different tendencies in sexual behavior under these different conditions without the tendencies being entirely distinct: behavior that influences the possibility of better genes would also have a (probabilistic) impact on the behavior of contributing males, and natural selection might have made human female psychology very sensitive to the mingling of these two currents.

22. Darwin 1981.

23. Desmond and Moore 2009.

24. Carroll 2010, 122–27.

25. Coyne 2009, 92.

26. Ridley 2004, 186.

27. Gee, Howlett, and Campbell 2009, 15.

28. Darwin 1856.

29. Shakespeare, sonnet 116. Boyd 2012 offers a powerful analysis of Shakespeare's sonnets that uses the insights of signaling theory, and of evolutionary theory more generally, to convincing effect. This sonnet is one of those addressed to the Fair Youth, so the context is not that of the standard Petrarchan sonnet addressed by a man to a woman, but that does not make signaling any less central to its concerns.

30. See Cooper 2008.

31. Trivers 2000, 115: “True and false information is simultaneously stored in an organism with a bias towards the true information's being stored in the unconscious mind, the false in the conscious. And, it is argued, this way of organizing knowledge is oriented towards an outside observer, who sees first the conscious mind and its productions and only later spots true information hidden in the other's unconscious.” He also suggests a different theory of self-deception, suggesting that it arises through the manipulation of components of our beliefs by others, such as parents. Trivers 2011 develops his views on self-deception at book length. The experimental evidence that tends to support Trivers's first theory is reported in Valdesolo and DeSteno 2008.

32. There is an abundant literature in psychology on motivated reasoning, where the desire to reach certain conclusions influences cognitive processes to an extent that is constrained by the need to provide plausible rationalizations for the motivated beliefs (see Kunda 1990). Haidt 2007 suggests that this plays a fundamental role in moral reasoning, since much explicit moral argument seeks to confirm the soundness of preexisting moral intuitions. Whether empirical or ethical, such forms of reasoning pose an obvious question: why did natural selection make us that way, instead of giving us reasoning faculties much more focused on seeking out truth? There is no consensus about this among researchers. Trivers's theory is one suggestion whose applicability is limited to truths about facts it would be strategically valuable to conceal from others. Haidt believes that motivated reasoning in morality is a by-product of recent evolution: “Language and the ability to engage in conscious moral reasoning came much later, perhaps only in the past 100 thousand years, so it is implausible that the neural mechanisms that control human judgment and behavior were suddenly rewired to hand control of the organism over to this new deliberative faculty” (2007, 998).

33. It now seems clear that the pain of rejection by a lover is a real, physical pain that activates brain regions similar to those activated by painful physical stimuli; see Kross et al. 2011.

34. For a wonderful tour of the confusions and paradoxes of sexual signaling, see the three books by Wendy Doniger (1999, 2000, 2005): the second is most explicitly about sex, while the first is about gender and the third about love.

Chapter Four: Social Primates

1. The original Latin phrase is “Bellum omnium contra omnes.” It appears in Hobbes 2008, chapter 1.

2. Darwin, 1979, chapter 14.

3. Darwin 1981, 162–63.

4. The literature on primate societies is vast; an excellent place to start is de Waal 2001.

5. Marmot 2004. Recent work by Armin Falk and coauthors (2011) has shown that in experimental settings, receiving pay that is perceived as unfairly low increases heart-rate variability, which is known to predict cardiovascular disease in the long run. They also show in data from firms that there is a strong association between cardiovascular health and perceived fairness of pay.

6. Muller and Wrangham 2004.

7. de Waal 1982, 138. Kim Sterelny has suggested to me that the relative instability of the coalitions in chimp societies, compared to the relatively norm-driven and institutionally structured social worlds of humans, can be compared to the contrast between market transactions without fixed prices, in which everything is continually negotiated, and fixed-price shopping (personal communication, July 2011).

8. Gesquière et al. 2011.

9. Harcourt et al. 1981. Most of the relevant research has been done in primates, but there's no reason that the relationship should be confined to primates. The very large testes of whales and dolphins (those of the right whale weigh over a ton) suggest likely polygyny: see Ridley 1993, 220. This relationship holds between species rather than within them: there's no evidence that men with larger testicles are able to have sex with larger numbers of women (despite the claims of the many websites devoted to techniques for testicle enlargement).

10. See Simmons 2001 for an account of sperm competition in insects.

11. Wrangham and Peterson, 1996, 225–27.

12. Stanford 1999, 42.

13. See Leduc 1992 for a discussion of marriage as gift exchange in classical Greece. Leduc makes the very interesting point that it was in the more socially conservative city-states, such as Sparta and Gortyn, that women could be citizens and own property in their own right; indeed, citizenship derived from membership in a community of landowners that was closed to outsiders. In Athens, which was socially more innovative and inclusive with respect to male foreigners, the locus of citizenship remained the (male-headed) household, and women passed from father's to husband's household, treated not so much like chattels as like children. As she memorably expresses it, “Women were the chief victims of the invention of democracy” (239).

14. Baron-Cohen 2003.

15. Croson and Gneezy 2008.

16. There is much controversy on this topic. That measured IQ scores show significant racial differences on average is not in serious dispute. What is controversial is the extent to which these differences are attributable to genetic variation, both because of selection effects and because there is strong evidence for large environmental variation of the relevant kind. Rushton and Jensen (2005) have argued that the proportion attributable to genetic variation is large; Fryer and Levitt (2006) use tests on very young children (where observable racial differences in scores are very small) to argue that this is unlikely. Hunt (2011, chapter 11) has a good, balanced overview that emphasizes how little we still know about this difficult question, and I discuss the deficiencies of IQ as a measure of intelligence in chapter 5. Here I want merely to make the point that the logic of sexual selection would lead us to expect very small differences in intelligence between human populations but potentially large differences between men and women. It is therefore both surprising and fascinating how hard the latter differences are to find.

17. Darwin 1998.

18. See Wilder, Mobasher, and Hammer 2004; Seabright 2010, 6.

19. Seabright 2010, 5–6. Cochrane and Harpending (2009) discuss ways in which relatively recent human evolution has been rapid, though these are consistent with the argument I advance here.

20. The Flynn effect (see chapter 5) shows how large are the likely environmental effects on test scores, and how small by comparison are the differences in scores between populations.

21. See Baron-Cohen 2003, for example.

22. Sterelny (2012) discusses these changes at length, emphasizing how tiny cultural modifications accumulated over a long time can add up to major qualitative change. Some innovations had an important effect on the balance of power between men, such as the development of projectile weapons, which made physical contests less dependent on simple strength. The gender balance of power may have been affected by other innovations, such as the making of fabric, which from the beginning appears to have been an essentially female occupation (see Barber 1994). Adovasio, Soffer, and Page (2007) have argued that this and other technologies manufactured by women (including ropes and nets) played a central role in human social evolution.

23. Ryan and Jetha 2010, 12.

24. See Mead 1973; Freeman 1983. In fairness to Mead, it should be said that social anthropology in the early twentieth century combined fascination with sexual practices of non-Western cultures with a free-and-easy approach to evidentiary rigor. A flavor of the latter can be found in Ernest Crawley's Studies of Savages and Sex (1929), with its disquisitions on such topics as “the sexual impulse of the savage,” in which Crawley makes statements such as “Even the negress is by no means very amorous” (9), on the authority of Havelock Ellis's reference to “a French army surgeon familiar with the black races in various French colonies.” Mead's evidence gathering was comparatively thorough.

25. Ryan and Jetha 2010, 217.

26. Hrdy 2009, 155.

27. See Thornhill and Gangestad 2008.

28. Cornwallis et al. 2010.

29. Dunbar 1992. The relationship, though present in the data, is not statistically very robust: the hypothesis that living in larger groups requires greater neural processing capacity, and might therefore lead to selection for larger brains, rests on a range of evidence, much of which is summarized in Sterelny 2003.

30. See Kaplan et al. 2000; Hooper 2011, especially figure 3.7.

31. See Hawkes et al. 2000. There is controversy about how central grandmothers were to human social evolution: see Sterelny (2012, especially chapter 4.3) for a view that makes them part of the story but a less crucial part than claimed by Hawkes et al.

32. Wrangham et al. 1999; Wrangham 2009.

33. See Hawkes et al. 1991; Kaplan et al. 2001.

34. See Hawkes 2004, for example. This point does not invalidate the example, since males still exert substantially more control over females than in other ape species, even if the control remains incomplete and variable across cultures and circumstances.

35. Hrdy 2009, 151.

36. Copeland et al. 2011; Alvarez 2004.

37. Hawkes 1991.

38. Boehm 1999.

39. Bowles 2009. Seabright 2010, especially chapters 3–5, discusses the evidence for violence in forager societies and explanations for the much lower modern levels of violence. This is also the theme of Pinker 2011, which documents the downward modern trend in violence in compelling detail.

40. Steckel and Wallis 2009, table 2.

41. Seabright 2010, 265 and n. 2.

42. See Stearns 2000, chapter 1.

43. Diamond 1987.

Part Two: Today

1. World Health Organization 2011, file DTH6 2004.xls.

2. World Health Organization 2011, file vid.680.xls.

3. See Hill, Hurtado, and Walker (2007, figure 1), on the Hiwi of Venezuela. Gurven, Kaplan, and Supa (2007, table 5) report similar evidence for the Tsimane of Bolivia, as do Hill and Hurtado (1996) for the Ache of Paraguay and Marlowe (2010, 137) for the Hadza of Tanzania. I am grateful to Paul Hooper for directing my attention to this information.

4. Stevenson and Wolfers (2009) present impressive cross-country and time-series questionnaire data indicating that self-reported levels of women's happiness have declined over thirty-five years relative to those of men in 125 of 147 countries examined, including the United States and the countries of the European Union. Self-reported happiness has also declined in absolute terms in the United States, though not in the European Union. For most of the paper, Stevenson and Wolfers treat these reports as equivalent to actual happiness, though at one point they acknowledge that “one might regard our rather striking observation as an opportunity to better understand the determinants of subjective well-being, and the mapping between responses to survey questions about happiness and notions of welfare” (194). This is a very much more important issue than they acknowledge. It seems likely that in the 1950s and 1960s, prior to the beginning of their US data set, there was a widespread expectation that women, especially married women, would be upbeat and positive about their condition, stifling private doubts about their own happiness in order to project a good image for the sake of their families. (Koontz [1992] is eloquent on various aspects of what she calls the “nostalgia trap” of American families.) One of the consequences of the feminist movement is that women may feel freer to express dissatisfaction; it has also showed them that there were viable alternatives to their current lives against which those lives could be assessed. Opinions may differ about whether that was a good thing (and it's certainly conceivable that expressing dissatisfaction may increase the real levels of dissatisfaction subjects feel), but it means that we cannot interpret answers to questions about how satisfied women have been with their lives in the same way across many decades. This does not mean, of course, that the authors' conclusions are incorrect: indeed, it's possible that there is increasing female dissatisfaction fueled by the growing disparity between women's and men's levels of education. I discuss this possibility further in chapter 8.

5. Wilder, Mobasher, and Hammer 2004.

6. See, for example Tiger 1999 (The Decline of Males); Garcia 2008 (The Decline of Men); Parker 2008 (Save the Males); and Rosin 2010 (“The End of Men”).

7. See Baumeister 2010, 17, on imprisonment and homelessness and Rosin 2010 on unemployment and educational attainment.

8. OECD 2011a.

9. See Mather and Adams 2007, especially figure 1. In addition, US Census data from 2010 corroborate the evidence from enrollment rates that increased female educational attainment, not any regression in male attainment, is the main reason for the growing gap between women's and men's attainment. See US Census Bureau 2011.

Chapter Five: Testing for Talent

1. Gowin 1915.

2. Case and Paxson 2008, 515.

3. Case and Paxson 2008, 503.

4. Case and Paxson 2008, table 4.

5. This finding is corroborated by the study reported in Abbot et al. 1998.

6. There has also been an independent finding of a correlation between height and a willingness to take risks: see Dohmen et al. 2011.

7. They do this with multiple regression analysis, which makes it possible to estimate the effect of height on earnings simultaneously with estimating the effect of talent on earnings, so that the former effect is measured taking the latter effect into account and vice versa.

8. Cinirella, Piopiunik, and Winter (2009) provide impressive evidence from Germany (where school tracking plays an important role in student educational achievement) that pupil height affects (independently of cognitive skills) the likelihood of being recommended by a primary-school teacher for enrollment in the most academic secondary-school track. They suggest this finding is likely due to teachers' rewarding higher social skills, and they note that taller children have better social skills as early as age 3. This latter effect could be due to an innate correlation of factors leading to height and to social skills or to a feedback effect whereby adults interact more with taller children, thereby reinforcing more strongly their social skills.

9. Ogilvie 2011, 56, 60, 113, 114.

10. See Ogilvie 2003, especially introduction.

11. Greenwood et al. (2005) argue that the scale and timing of the diffusion of a large range of household labor-saving devices, from refrigerators to washing machines and vacuum cleaners, are the most plausible explanation for the increase in female labor-force participation in the US economy.

12. Goldin and Katz (2002) show that by enabling single women to delay marriage, the pill increased returns to investment in education and professional training. They suggest that it indirectly increased the attractiveness to women of delaying marriage by increasing the probability that women who married later would make better choices in the marriage market (a woman marrying later would be a relatively more attractive partner when a shift in general practices meant she was not as unusual in doing so and therefore was less likely to be stereotyped for doing so).

13. Scandinavia is an exception, because the most radical changes, such as Sweden's Marriage Law of 1915, had been in preparation before the First World War. See Therborn 2004, 74.

14. Important changes in the law included not just those relating to voting and removing formal constraints on labor market participation but also those governing the status of the spouses within a marriage, divorce, and property ownership. See Therborn 2004, especially chapter 2.

15. See Brandt 2007, 84–85.

16. Bureau of Labor Statistics 2011a.

17. Bureau of Labor Statistics 2009.

18. Bureau of Labor Statistics 2011b.

19. Bureau of Labor Statistics 2011a.

20. On board members, see Soares et al. 2010; on CEOs, see CNN Money 2011.

21. Ryan and Haslam 2005.

22. Bureau of Labor Statistics 2011b.

23. See Astur et al. 1998 on spatial navigation. Hines (2011) summarizes evidence on the influence of testosterone on a number of cognitive skills, speculating plausibly that the strong preferences of young boys for playing with mechanical toys may be due not to any particularities of shape or color of such toys but to the fact that they are designed to be pushed through space. Buss (2004, 86–87) summarizes evidence that these differences in skill may reflect the specialization of men for hunting and women for gathering, a hypothesis that has given rise to a number of predictions about gender variations across skill types that are well supported by the evidence.

24. On verbal ability, see Hyde and Linn 1988. Hoffmann, Gneezy, and List (2011) find a gender difference in ability to solve a physical puzzle to be present in a patrilineal society but not in an adjoining matrilineal one.

25. See Fine 2011, especially chapter 3.

26. Shih, Pittinsky, and Ambady 1999. Stereotype threat has also been shown more recently to be important for responses to competition: see Günther et al. 2010.

27. Heritability refers to the degree of population variation in a trait that is explained by population genetic variation, and this is not the same thing as the total contribution of genes to the possession of a trait. For example, the number of fingers we have is almost completely determined by our genes, but its heritability is close to zero, because almost everyone has ten fingers, and those who do not have almost always lost a finger through an accident, so that the small amount of population variability is almost all due to environmental factors. Visscher et al. (2008) provide a valuable review and explanation of the notion of heritability. Estimates of the degree of heritability of IQ are sensitive to whether truly genetic contributions to IQ scores are separated from contributions from the maternal environment (even identical twins reared apart will have shared the same womb). Devlin et al. (1997) show that taking the maternal environment into account can substantially reduce estimates of the genetic heritability of IQ, from the 60–80 percent claimed by Herrnstein and Murray (1994) to somewhere between 34 percent and 48 percent by their own estimates.

28. Flynn 2007.

29. However, Blair et al. (2005) make a plausible case that it is to do with a combination of increased population access to formal schooling and (in more recent decades) to the increasing cognitive demands of mathematics education for young children, at the age when the prefrontal cortex exhibits high neural plasticity in response to experience. Ramsden et al. (2011) report neuroimaging data suggesting that changes in IQ measures in a sample of teenagers are not due to noise in the measurement of unchanging underlying skills but are correlated with modifications over time in local brain structure, specifically with gray matter in areas activated by speech (for verbal measures of IQ) and finger movements (for nonverbal measures).

30. Lynn and Irwing (2004), in a meta-analysis of fifty-seven studies using the Raven Progressive Matrices Test, find no advantage in favor of boys age 6–14 but a five-point advantage in favor of adult men, which they interpret in support of Lynn's (1999) theory that differences emerge after puberty because males mature more slowly than females. Jorm et al. (2004) found that differences on tests that favored men tended to disappear when various sociodemographic and health variables were controlled for, but on tests that favored women they tended to be accentuated. Deary et al. (2007), in a single survey comparing pairs of opposite-sex siblings, find a very small (less than 1-point) advantage in favor of men, though they report men as having higher variance (see n. 47 below). Colom et al. (2002), on a single Spanish implementation of the Wechsler Adult Intelligence Scale, found that there were no gender differences in g but differences in the total IQ scores, of around 3.6 points, in favor of men, indicating that the method of aggregating component scores gives more weight to men than is warranted by calculations of g. Thus the fact that a test is highly g-loaded does not imply that differences in mean scores imply differences in the underlying g scores. However, Jackson and Rushton (2006), analyzing the performance of 100,000 17–18-year-olds on the Scholastic Assessment Test, find a difference in both average IQ scores and mean g, equivalent to 3.63 points on an IQ scale.

31. This point is also illustrated by the results reported in Geary et al. 2000, who show that a group of men performed better than a group of women on a test of arithmetical reasoning, and that individual differences in arithmetical reasoning were related to individual differences in IQ, but that there was no sex difference on the IQ test.

32. Hunt 2011, 407.

33. Lynn (1999) simply asserts that intelligence should be defined as the sum of the verbal comprehension, reasoning, and spatial abilities, which is not an argument at all. This is like claiming that a place's attractiveness to live in should be defined as the sum of its nightclubs plus the sum of its fields.

34. See Astur et al. 1998.

35. See Milner et al. 1968.

36. See Almlund et al. 2011.

37. See Miller 2009 for a general, nontechnical introduction to personality psychology, especially chapters 9 and 12–14. Schmitt et al. (2008) suggest some reasons for caution in interpreting results as stable across cultures, notably because more prosperous societies display larger differences in personality, especially between men and women.

38. Bowles and Gintis (1976) were pioneers in this area, and Bowles, Gintis, and Osborne 2001a and 2001b summarize more recent research.

39. Schmitt et al. (2008) find higher scores for women on conscientiousness in most nations in a fifty-five-nation study. However, Müller and Plug (2006) find no difference between women and men on this trait. Nyhus and Pons (2005) also find no effect of conscientiousness on earnings (and no significant difference between men and women in average conscientiousness) in a careful study based on Dutch panel data.

40. Almlund et al. 2011, figure 16, summarizes correlations with job performance, and their section 7.B discusses such findings in detail. The positive association between emotional stability and productivity was previously found in meta-analyses by Barrick and Mount (1991) and Salgado (1997).

41. Müller and Plug 2006.

42. Almlund et al. 2011.

43. See Duckworth and Seligman 2005, for instance.

44. Müller and Plug 2006.

45. See Johnson, Carothers, and Deary 2008, 2009.

46. See Zechner et al. 2001. Miller's hypothesis is set out in Miller 2000.

47. See discussion in Hunt 2011, 382–86. Hedges and Nowell (1995, table 2) report differences in gender means and in gender variances for 37 different tests, with males having higher means in 23 of the 37 but higher variance in 35 of the 37 tests. A different selection of tests would doubtless have produced different proportions, though it seems safe to bet that the variance discrepancy would be likely to remain higher than the discrepancy in means. Deary et al. (2007) report substantially greater male than female variance in measures of g, sufficient to mean that roughly twice as many men and women are represented among the top 2 percent of scores.

48. See Halpern et al. 2007. Such differences are also negatively correlated across countries with other measures of gender empowerment, which supports the idea that socialization is an important influence; see Guiso et al. 2008.

49. See the various contributions to Gallagher and Kaufman 2005 for an idea of the controversy. Ellison and Swanson (2010) document a large gender gap at the highest percentiles in the United States. Although the gap does not vary much across schools, which might suggest a limited role for environmental influences, they also document that the highest-achieving girls are concentrated in a very small subset of schools. This suggests that the small degree of observed environmental variation gives little insight into the potential distribution of ability: as the authors write, “There is limited value in trying to put a lot of effort into estimating ‘ability’ distributions when almost all girls who would be capable of achieving extremely high scores do not do so” (29). Hyde and Mertz (2009) find that a gap does exist in the United States, though it has narrowed over time. It does not exist in some ethnic groups and in some other nations.

50. Zechner et al. 2001.

51. See Kostyniuk et al. 1996, Mayhew et al. 2003. According to a report by the Associated Press (2007), a study in 2007 by Carnegie-Mellon University for the American Automobile Association reported that there are 1.35 male deaths per 100 million miles driven, compared to 0.77 female deaths; however, I have not been able to track down the original study.

52. Gouchie and Kimura 1991; Hunt 2011, 406. See Hines 2011 for a review of the developmental effects of testosterone and resulting gender differences in behavior. However, a very recent study by Kocoska-Maras et al. (2011) finds no effect of testosterone on spatial ability in a large, double-blind, randomized study.

53. Even where physical strength matters, the discrepancies are not always what you might expect. Two of the industries in which women earn most on average, relative to men, are construction and agriculture: see Bureau of Labor Statistics 2011c.

Chapter Six: What Do Women Want?

1. An extended development of this view is set out in Browne 2002.

2. I am indebted to Bertrand 2011, which provides an excellent survey on gender differences in psychological attributes, with far more detailed discussion and references than I can include here.

3. Carey 2011.

4. These are summarized in Croson and Gneezy 2008.

5. Beyer 1990; Barber and Odean 2001; Niederle and Westerlund 2007. These gender differences are immune to the general criticisms leveled by Harris and Hahn (2011) against studies of overconfidence in general; the biases diagnosed in these studies by Harris and Hahn should not be expected to operate differently for women and men.

6. This was first shown by Bateman (1948); see also Trivers 1972, especially 37–39.

7. For example, Sukumar and Gadgil (1988) show that there are differences in risk-taking behavior between male and female Asian elephants, with males tending to take greater risks.

8. Bertrand and Mullainathan (2010) provide strong evidence that CEOs are rewarded as much for the results of luck as for anything that might represent the influence of their own efforts on firm performance, and that this tendency is stronger in firms that, by other criteria, appear to be less well governed (less in the interests of shareholders, that is). They interpret this finding, plausibly, as reflecting the fact that CEOs try to set their own pay and have as much interest in rewarding themselves for the results of luck as for the results of their efforts. This leaves a number of questions unanswered, notably why even poorly governed firms do not resist payment for luck more strongly than payment for effort; they clearly must offer some resistance, as CEOs cannot choose to pay themselves any amount they like.

9. Scotchmer (2008) develops a model of promotion in hierarchies that formalizes this insight.

10. Andreoni and Vesterlund 2001. Engel (2011) finds in a meta-analysis of dictator games that women systematically behave more altruistically than men. However, this is far from being an agreed conclusion. Baumeister (2010, especially chapter 5) argues that men and women tend to express altruism in different settings, with women expressing it preferentially in small groups and within intimate relationships, and men expressing it in larger groups and more public settings.

11. See Chaudhuri 2011. Pinker (2011, 684–89) speculates that overall levels of violence in society might be influenced by the extent to which women hold positions of power.

12. Niederle and Westerlund 2007. Interestingly, even the women in the highest quartile of performance were less likely to choose the tournament than the men in the lowest quartile.

13. Kuhn and Villeval 2011.

14. Gneezy, Niederle, and Rustichini 2003; Niederle and Westerlund 2008. Booth and Nolen (2009) find that girls from single-sex schools compete as enthusiastically as boys, suggesting social learning and not genetics as the reason behind observed gender differences in competitive behavior. These studies are limited to school-age children, and the findings may not generalize to other age groups.

15. Kuhn and Villeval 2011.

16. See Gupta, Poulsen, and Villeval 2005 for an example of a study where it does not make a difference.

17. Gneezy and Rustichini (2004) found that girls competed more strongly against boys than against other girls in a test of running.

18. See Dreber, Von Essen, and Ranehill 2011 for a study on Swedish children that fails to find the differences previously found for Israeli children by Gneezy and Rustichini 2004. Cárdenas et al. (2011) find no gender differences in competitiveness in Colombia, whereas in Sweden the results are mixed depending on the task studied.

19. Günther et al. 2010.

20. See Chen, Katusczak, and Ozdenoren 2009; Apicella et al. 2008; Dreber et al. 2009; Dreber and Hoffman 2010. However Apicella et al. (2011) find no hormonal correlates of choosing whether to compete in solving mazes.

21. Manning and Swaffield 2008; Manning and Saidi 2010. Booth 2009 gives an excellent overview of the issues.

22. Babcock and Laschever 2003, 7.

23. Bowles, Babcock, and McGinn 2005; Small et al. 2007.

24. Bowles, Babcock, and Lai 2007.

25. This is a point made long ago about discrimination by Gary Becker (1971).

26. See Goldacre 2009, chapter 4.

27. This point is made very effectively by Dougherty (2005).

28. Bertrand, Goldin, and Katz 2010.

29. This doesn't prove that there's no difference in talent: there might be certain characteristics of women who choose to become mothers that make them less likely to be successful some years later. But any such difference would have had to be invisible to employers before the women's decision to take career breaks, so in the absence of more concrete evidence, the career breaks themselves look like a more plausible explanation for the divergence in salaries.

30. Bertrand, Goldin, and Katz 2010, 240.

31. Yoshino 2006, ix.

Chapter Seven: Coalitions of the Willing

1. de Waal 1989, 48.

2. de Waal 1989, 51.

3. Goodall 1986; Nishida 1996.

4. de Waal 1989, 122.

5. Low 2000, 181.

6. On primates, see Henazi and Barrett 1999; Silk, Alberts, and Altmann 2004; Pandit and van Schaik 2003; van Schaik, Pandit, and Vogel 2004, 2005. On other species, see Low 2000, 181–82.

7. Granovetter 1973.

8. It's quite possible that common membership in a group may by itself determine willingness to treat one another favorably. If men treat other men more favorably than they treat women just because they are men, this would be an example of classic discrimination, which may indeed be part of the reason why women appear disadvantaged in the labor market. However, such discrimination is illegal in many countries (except when recruiting for a few occupations, such as military ones). I am interested here in a more indirect and subtle process whereby more favorable treatment is determined by membership of a coalition in which the members have invested some effort in building bilateral ties.

9. Granovetter 1973.

10. Moore 1990.

11. A study by Chow and Ng (2007) finds that among coworkers, women are less likely than men to socialize with those whom they subsequently need to approach for instrumental favors, though the sample was small and far from random (seventy-two executive-education participants, of whom two-thirds were men). A study by Forret and Dougherty (2004), reporting a tendency for women's networking behavior to be less effective at promoting their professional interests than that of men, indicates a small negative correlation between female gender and three main measures of networking activity, namely “maintaining external contacts,” “socializing,” and “engaging in professional activities,” though it does not indicate the significance of this correlation. Campbell (1988) reports that in a sample of employed people changing jobs, women had contacts in a smaller range of occupations than did men.

12. Ibarra 1997.

13. Bu and Roy 2005.

14. Burt 1998.

15. Green and Singleton 2009; Igarashi, Takai, and Yoshida 2005; Lemish and Cohen 2005.

16. Two studies that have done so are Smoreda and Licoppe, 2000 and Wajman, Bittman, and Brown 2009.

17. Friebel and Seabright 2011.

18. Tannen 1990 and 1994 are studies of gender differences in conversational strategies (the latter focusing on the workplace), though based on examples rather than statistical evidence.

19. Forret and Dougherty (2004) find that self-reports of networking activities along several dimensions are sometimes insignificantly related with measures of professional success for women even when they are positively related for men, though this finding is not consistent across different measures. Aguilera (2008) reports that women who find employment through personal contacts do significantly better than those who find employment through ordinary labor market procedures, while this effect is not present for men; however, this result is compatible with the possibility that men's employment is already more remunerative because of male networking, whether or not the job is found through personal contacts.

20. We also have information on educational links and on connections via membership of not-for-profit associations.

21. Lalanne and Seabright 2011.

22. The basic difficulty is that we cannot measure talent directly, so we need to find a proxy for talent. If we use as explanatory variables in our regression analyses some other variable that is imperfectly correlated with talent, it will not take out of our measured effect enough of the hidden effect of talent, so our estimates will still be biased in the direction of finding more of an effect than really exists. Alternatively, if we use past values of salaries as proxies for talent, this has the opposite bias: namely that we risk overlooking an effect where one really exists, because any effect of networks on past salaries will be excluded from our analysis by definition. We use two variants of this second approach to tackle the problem, and fortunately both methods give comparable results. One is to use salaries several years in the past as control variables so as to allow several years for the effects of networks on salaries to manifest themselves. This is particularly important because the contacts people make in one year may not benefit their careers until several years later. The other variant is to look just at what happens when people's networks change over time: do increases in a given person's network lead to increases in their salaries? For further details, see Lalanne and Seabright 2011.

23. This part of the analysis is based on joint work with Nicoletta Berardi of the Banque de France, reported in Berardi and Seabright 2011.

24. Other researchers have found that gender differences in deferred remuneration are even greater than those in salaries: see Albanesi and Olivetti 2008; Kulich et al. 2009; Bebchuk and Fried 2004; and Geletkanycz, Boyd, and Finkelstein 2001.

25. This is not a new concern: Daily, Certo, and Dalton (1999) made much the same point when reviewing evidence of female boardroom representation in the 1990s. A recent report in the Observer (Bawden 2011) suggested that firms in the United Kingdom were focusing only on nonexecutive and not on executive board appointments, in response to the Davies report that recommended a voluntary target of 25 percent female boardroom representation. A related concern is expressed by Gregoric et al. (2010), who present evidence suggesting that boards with higher representation of women tend to be less diverse in other dimensions.

26. See Hunt 2010, 97.

27. There is indeed a literature discussing which of these motivations is the more important, with Geletkanycz, Boyd, and Finkelstein (2001) favoring (a particular version of) the prudence theory and Bebchuk and Fried (2004) favoring a version of the preference theory.

28. Low 2000, chapter 10.

29. Scott (1983, 319) reports that “respondents reporting to someone of the same gender had significantly higher trust in their superior than did men or women reporting to a superior of the opposite sex.” Unfortunately, this finding does not show that their trust was warranted. And it is directly contradicted by a study by Jeanquart-Barone (1993), which found that “the highest level of trust was found between female subordinates reporting to male supervisors” (1).

30. Bonein and Serra (2009) find evidence in favor of greater trust between players of the same gender, while Sutter et al. (2009) find evidence against it.

Chapter Eight: The Scarcity of Charm

1. Giridharadas 2007.

2. Giraudoux 1997, 32.

3. Mayhew 1861, 1: 311–12.

4. Nasscom 2009.

5. On Africa, see GSMA/A. T. Kearney 2011a, which reports 620 million mobile subscriptions in the top twenty-five African countries (p. 4); on India and China, see GSMA/A. T. Kearney 2011b, which reports 752 million subscriptions in India and 842 million in China (p. 1).

6. GSMA/A. T. Kearney, 2011b, 4.

7. International Telecommunication Union, 2011a for Internet subscriptions, 2011b for Internet users.

8. There have been more than 52 million visits logged for Khan Academy videos, though how many visitors have fully used the courses is a matter of conjecture. See www.khanacademy.org.

9. This is a potentially vast subject that in principle would require another book. Interesting theoretical papers from an economic point of view include Falkinger 2007 and 2008, and Anderson and De Palma 2009. Klingberg 2009 gives a very accessible account of the neuroscientific constraints on human attention and their consequences for our everyday lives.

10. Bureau of Labor Statistics 2010.

11. The data on occupational earnings discussed in the next few paragraphs are all available at Bureau of Labor Statistics, 2011d; I have drawn on the series “May 2009 National Occupational Employment and Wage Estimates” and “2000 National Occupational Employment and Wage Estimates.”

12. More precisely, an ordinary least-square regression analysis (over the 686 occupations for which there are data for the two years) of the ratio of the earnings spread in 2009 to the earnings spread in 2000, with the percentage change in annual median earnings and the change in total employment as regressors, yields coefficients of 8.9 percent on the first and 2.4 percent on the second; each coefficient is statistically significant at a little under 2 percent.

13. Farrell 1993.

14. OECD 2011b.

15. See US Census Bureau 2011.

16. For marriage rates among whites and African Americans, see “Down or Out” 2011. For differences in educational attainment, see US Census Bureau 2011. Among the population aged 30–34, 27.4 percent of black women have at least a bachelor's degree, while only 16.9 percent of black men do. Banks (2011) argues that this disparity is creating a real crisis in relationships for both black men and black women.

17. Stevenson and Wolfers 2009.

18. The authors also report strong positive trends for the reported happiness of African American women. There are in any case some difficulties in interpreting trends in reported happiness over time. See the introduction to part two, n. 4, for a discussion of the issues involved.

Chapter Nine: The Tender War

1. The ape anthropologists would of course acknowledge that some elements of their story had been anticipated by the work of human anthropologists before them. Hrdy 2009 and Kaplan et al. 2009 summarize the literature on the first of these points. Hrdy's book begins with an intriguing thought experiment imagining what would happen if her fellow passengers on a plane journey were other species of ape. “What if I were traveling with a planeload of chimpanzees? Any one of us would be lucky to disembark with all ten fingers and toes still attached, with the baby still breathing and unmaimed. Bloody earlobes and other appendages would litter the aisles” (3). Bowles and Gintis (2011) discuss how the evolution of prosocial behavior for elaborate cooperative purposes has driven the development of most of the other characteristics that distinguish humans from the other great apes. Burkart, Hrdy, and van Schaik (2009) propose a theory of the evolution of the motivational components of the distinctively human psychological complex based on cooperative breeding; chimpanzees, they argue, have many of the purely cognitive components already in place.

2. The literature on this topic is surveyed in Seabright 2010, especially chapters 3–5. Pinker 2011 has a comprehensive overview of the literature on the downward trend of violence in recorded history.

3. This point is made most forcefully in Boehm 1999.

4. See, for instance, de Waal 2001, especially chapter 2.

5. This is a central theme, perhaps the principal theme, of Foucault 1990.

6. The public interest in the affair allegedly centered on whether Clinton had committed perjury in his statements about Lewinsky in relation to the lawsuit brought against him by Paula Jones for sexual harassment. But it would be hard for an objective observer to deny that the press and public fascination with presidential adultery became overwhelming. Otherwise it is hard to see why the allegations of a consensual relationship with Lewinsky should have been considered relevant to allegations of harassment of Jones or why Clinton should have felt such a strong incentive to lie under oath about the former.

7. Educated and reflective people are by no means immune from such reactions. For example, in a discussion of sexual morality, the philosopher Elizabeth Anscombe wrote of “the rewardless trouble of the spirit associated with the sort of sexual activity which from its type is guaranteed sterile: the solitary or again the homosexual sort” (1976). This led her commentators in the same volume, Michael Tanner and Bernard Williams (1976), to inquire, with reference to homosexual acts, “How does Professor Anscombe claim to know?”

8. See, for example, Marlowe 2010, 175.

9. Several chapters in Muller and Wrangham 2009 discuss sexual coercion in humans and the insights that emerge from comparing it with coercion in nonhuman primates; the final chapter, in particular, draws general lessons from the many case studies in the volume.

10. See Muller and Wrangham 2009, chapter 18.

11. This has been an important issue in the development of hostility between radical Islam and modern industrialized societies. Sayyid Qutb, the Egyptian founder of the Muslim Brotherhood and an important influence on Osama bin Laden and the 9/11 hijackers, was significantly influenced by a visit to the United States between 1948 and 1950. In his best-selling work Milestones, published two years before his execution by Gamal Abdel Nasser's regime, Qutb railed against what he called “this behavior, like animals, which you call ‘free mixing of the sexes.’ This vulgarity which you call emancipation of women” (Qutb 1964, 139). He also described churches as “sexual playgrounds” (cited in Irwin 2001) and wrote that “the American girl is well acquainted with her body's seductive capacity. She knows seductiveness lies in the round breasts, the full buttocks, and in the shapely thighs, sleek legs and she shows all this and does not hide it” (cited in von Drehle 2006).

12. Fischer 1994 is a detailed social history of the telephone in America up to and including the 1930s.

13. Besides the poster reproduced in the figure, others can be viewed at the Sodahead blog, http://sodahead.com/living/when-smoking-was-good-for-your-health-socially-acceptable/blog-63461, accessed June 12, 2011.

14. Stephanie Coontz's book Marriage: A History (2005), which documents this development in great detail, has the subtitle From Obedience to Intimacy, or How Love Conquered Marriage.

15. There is no doubt that the physical manifestations of erotic passion can be understood through the adaptive logic of natural selection (see Fisher 2004 for a detailed account and Young 2009 for a summary of some of the scientific contributions to this literature; Frazzetto 2010 emphasizes that even online dating, with its entirely artificial setting, can be usefully illuminated using the same explanatory tools). But this does not mean that a romantic relationship can be best understood by analogy with an organ of the body.

16. DeSteno, Vadesolo, and Bartlett (2006), for instance, report experimental evidence suggesting that jealousy reactions are prompted by threats to self-esteem, which implies (though it does not prove) that contextual variation in the extent of threats to self-esteem is likely to affect the extent to which jealousy is felt in otherwise similar relationship dynamics. Hupka and Ryan (1990) report that manifestations of aggressive male responses in situations eliciting jealousy are substantially variable across a sample of ninety-two preindustrial societies in response to the presence or absence of a number of stable cultural norms, but that female responses do not vary significantly with such norms.

17. Wasson 2010, 127–29. Rather less subtly, an advertisement for a fashion brand in the free Parisian commuter magazine 20 Minutes had the slogan “Elle change de look comme elle change de boyfriend” (March 2, 2011).

18. Though space precludes discussion of this point in detail here, the same applies to relationships that differ in other ways from the model. It is common, for instance, for critics to decry commercial sex as sordid and tawdry (see Jeffreys 1997, for instance) and to deny that it represents a real choice for the seller. What matters is not how it compares with an ideal consensual encounter based on mutual desire but rather how it compares with the actual alternatives available to the parties (though child prostitution raises issues that go far beyond those relevant to commercial sex between adults). Unfortunately here again, legal or social constraints on consensual encounters make it more difficult to take action against coerced encounters: a study of street-level prostitution in Chicago in 2007 showed that a prostitute was more likely to have given free sexual services to a police officer than to have been arrested by one (Levitt and Venkatesh 2007). It is also likely that criminalizing commercial encounters between consenting adults discourages cooperation with the police by some of the most likely witnesses of sexual crimes against nonconsenting adults, or against children.

Many ethnographies have emphasized the variety of motivations and circumstances surrounding prostitution, both on the side of clients and on the side of the prostitutes themselves. See Clouet 2008 for a recent example; Meston and Buss 2009, chapter 8, for evidence based on an online survey; and Zelizer 2006 for many examples and a comprehensive discussion. Ringdal 1997 is a detailed historical account of the way the institutions of prostitution have varied across many different societies. Edlund and Korn (2002) have shown that many aspects of commercial sex are consistent with the idea that prostitutes undertake it as an economic choice: it is not one that many readers would envy, perhaps, but the alternatives available to many prostitutes are not enviable either.

Edlund and Korn emphasize, though, that the short-t erm attractions of female prostitution typically come with a long-term cost, namely the destruction of the marriage prospects of the women concerned. This cost is evidently much higher for street prostitutes than for others: indeed, many of the part-time student prostitutes described in Clouet 2008 are unlikely ever to reveal their former occupation to a future husband, who is unlikely ever to find out unless they are careless. And it hardly needs repeating that in cultures all over the world, courtship may involve substantial in-kind investments by one party (usually though not always male) as inducements for the sexual favors of the other, even when no money changes hands, so that the boundaries between commercial and noncommercial sex can be very blurred. This does not mean, as Zelizer (2006) emphasizes, that courtship is “nothing but” commercial sex: on the contrary, each type of exchange comes freighted with complex attitudes, norms, and expectations that distinguish it from others. But nor are these wholly distinct realms with nothing in common: they have a family resemblance, one might say. What distinguishes commercial sex may be the short-term nature and explicit character of the transaction rather than the fact of the transaction per se. In a different market in which unease has often been expressed for similar reasons about commercial transactions, namely the market for human gametes, a study by Almeling (2011), based on detailed interviews, emphasizes that differences in norms and conventions mean that “sperm donation is considered a job and egg donation a gift” in spite of the fact that both are remunerated in the United States (168).

19. There is evidence that job satisfaction is more highly correlated with reported life satisfaction for men than for women (Della Giusta, Jewell, and Kambhampati 2011), but evidently this finding does not tell us to what extent the correlation arises because of inheritance rather than social reinforcement. It's possible that there are genetic foundations for a degree of variability in the single-mindedness with which individuals focus on their jobs to the exclusion of other aspects of their lives, consistent, for instance, with the observation that autism is more common among men than among women (see Baron-Cohen 2003). But it is also my (unscientific) impression that very single-minded men are much more likely to be profiled admiringly in the press than are equivalently single-minded women. The most admiring press profiles are reserved for women who manage to “have it all,” whereas profiles of those who pursue single goals (at least professional goals) with great concentration often carry a hint of monstrosity, with the partial exception of sportswomen whose careers finish early enough for them to switch their focus to other goals, including a family or a second career. If my impression is accurate (and I would be interested to see a careful study that tested it), it would imply that there is an important contribution of social conditioning to the tendency of men to pursue less diverse goals.

20. Apter 1995.

21. Fernández and Cheng Wong (2011) report a large difference in education and labor-force participation between women born in 1935 and those born twenty years later.

22. US Census Bureau 2011. Women in their late forties and early fifties are slightly more likely to be educated to bachelor's degree level and above than men in the same cohort. In Lalanne and Seabright 2011, we report that women in our sample are slightly more educated than men. However, the composition of degrees may also matter for remuneration, and men are somewhat more likely to have degrees in finance and in engineering.

23. Systematic evidence on how this has worked in the few countries that have tried it is scarce. See Bennhold 2010. Existing parental-leave provisions are highly asymmetric between parents: for regularly updated international comparisons, see International Labor Organization 2011.

24. See Cools, Fiva, and Kirkeboen (2011), who study the effect of a reform in Norway in 1993 that substantially increased the incentives for fathers to take paternity leave. They find an essentially negligible impact of the provision on a range of variables, including average school performance, family fertility, and divorce rates (though these findings may reflect the rather minor nature of the reform). However, children's school performance became more responsive than before the reform to the education levels of the father, suggesting that fathers are becoming more involved in care of children. They find that maternal labor supply is complementary to paternal labor supply, so in families where fathers take time off work to care for children, mothers are more rather than less likely to reduce working hours and earnings.