2.1. Modern social networks and fertility

The pattern of social interactions that occurs in modern social networks is an evolutionary novelty. Living in kin-based social groups is a human characteristic with evolutionary roots that probably extend back to the Miocene (Foley, 1996). Compared to other ape societies, all human societies are remarkable in the extent to which cooperation occurs between nonrelatives (Richerson & Boyd, 2005). Nevertheless, traditional societies are highly kin-oriented, compared to modern societies. Industrialization introduces cheap mechanized transport and allows family members to pursue economic opportunities far from their place of birth. As a result, the vast majority of interactions occurring in postindustrial societies are between individuals who are not genetically related, and interactions that take place via contemporary telecommunication media are often between people who have not even met. That modernization is the trigger for a dramatic change in social networks has been well documented (e.g., Notestein, 1945, Ogburn & Nimkoff, 1955, Thornton & Frick, 1987, Zelinsky, 1971).

Links between a widening of social networks and the adoption of reproductive behavior that is inconsistent with maximizing reproductive success are also well documented. Individuals with wider social networks are the first in a population to adopt family limitation (e.g., Axinn & Barber, 2001, Axinn & Yabiku, 2001, Barber et al., 2002, Behrman et al., 2002, Bongaarts & Watkins, 1996, Boulay & Valente, 1999, Kohler, 2001, Valente et al., 1997, Watkins, 1990, Watkins & Danzi, 1995, Weinstein et al., 1990, Godley, 2001). Reproductive decisions that limit family size might be adaptive in environments in which resources are restricted (e.g., Mace, 1998), but the family limitation that accompanies modernization occurs at a time of rapid increase in the availability of resources. The norms of modern societies encourage parents to believe that they must invest large amounts of time and resources in their children if they are to be successful, but there is no evidence that this investment pays off in terms of reproductive success. Kaplan, Lancaster, Johnson, and Bock (1995) found that men raised in small families did not achieve greater fitness than those raised in large families.

Turke (1989) has also suggested that the modern fertility decline may be the result of a reduction in contact between kin, suggesting that psychological mechanisms, which evolved to solve the problem of allocating life effort in a manner that maximizes reproductive success, monitor the availability of committed caregivers to provide help with raising children. When kin cease to be available, these mechanisms determine reproductive resources to be low, even if the couple is well supplied with physical resources. Therefore, instead of investing in a large number of offspring, couples concentrate their reproductive investment on producing small number of socially competitive children.

2.2. Culture and reproductive decisions

Newson et al. (2005) argue that, although the amount of practical support available from a kinship network may be one of the factors that influence reproductive decisions at the individual level, if viewed at the population level, kin altruism may have a more important effect.

Individuals operating in a social network provide each other with a vast amount of social information. A long tradition of research in social psychology has shown that the exchange of social information that occurs within a group creates and maintains the social norms or culture of the group (e.g., Turner, 1991, Postmes et al., 2005). Day-to-day discussions between group members develop and continually revise the canon of values and beliefs that provide the proximate explanations for many of the decisions, including reproductive decisions, that are made by group members. For example, the belief that each child needs his own bedroom will motivate a couple who can only afford a three-bedroom home to avoid having a third child. Reproductive decisions are influenced by many such cultural elements, and modern societies have developed a large number of elements that combine to make even very prosperous people believe they cannot or should not raise a large family.

The kin influence hypothesis does not suggest that kin obsessively encourage wise reproductive decisions or that people who are not related spitefully encourage behavior that reduces reproductive success. The bias is weak, so the content of conversations between both friends and relatives will largely reflect prevailing cultural norms. However, it has been seen that when social networks are dense with kin, cultural institutions encourage couples to produce as many children as they can successfully raise (Lorimer, 1954). For example, an explanation of why African societies may “offer greater resistance to fertility decline” (Caldwell & Caldwell, 1987, p. 409) summarizes cultural characteristics that acted (and, to some extent, still do act) to maintain high fertility. In interviews about childbearing conducted in Africa, a majority responded that it is “fearful” to die without children. Limiting of family size was considered extremely risky because of a keen awareness of the possibility that all one's children might die. The rare cases of an entire family perishing were widely reported and long remembered. According to Caldwell and Caldwell (1987), these attitudes to reproduction were underpinned by remnants of ancient religious beliefs in which gods or ancestors were concerned with continuance of the family. High fertility was seen to be not only a divine reward but also evidence of correct behavior. People who did not have a son to perform the correct burial rituals become unhappy wandering ghosts. Thus, even if raising many sons was extremely costly in ecological terms, in cultural terms, the cost of not having a surviving son was incalculable.

In modern societies, in which social networks are largely made up of non-kin, the cultural institutions that encourage individuals to marry and have a family tend to be less strong and are becoming weaker (e.g., Alstott, 2004, Mack, 1997), and the cultural rewards for successfully pursuing alternative goals, such as a career, are increasing.

2.3. The cooperative nature of human reproduction

Why do cultural institutions have such a strong influence on reproductive decisions? How could natural selection have allowed human decisions that are so vital to individual fitness be subject to cultural influence? As Krebs & Dawkins (1984) point out, if an animal's behavior can be influenced by competing conspecifics, it is vulnerable to manipulation. Theorists such as Lumsden and Wilson (1981) and Tooby and Cosmides (1992) have argued that evolved mechanisms must place constraints on the elements of culture that individuals adopt because selection would have favored individuals who resisted adopting cultural norms when doing so would inhibit the propagation of their genes.

The strength of cultural influence might be explained by the cooperative nature of human reproduction, which has several unusual features associated with mothers being able to co-opt help with the raising of their young (Hill, 1993, Hrdy, 1999, Mace, 2000). Like other great apes, humans give birth to offspring that are helpless, slow to mature, and require a great deal of parental care. But, while human infants are most extreme in this respect, human mothers are more prolific breeders than other apes. There is variation between cultures, but humans typically space births between 2.5 and 3.5 years apart, compared to 4 to 5 years for chimps and nearly 8 for orangutans. This results in human mothers having several children at different stages of dependency, a situation that is only possible if the mother receives help with caring for and provisioning her young. Human life history has evolved so that a source of help is available within human groups. Compared to other mammals, humans, especially human females, are reproductively active for a small portion of their lives. Members of the group who are in the pre- and postreproductive phase of their lives can and do help care for and socialize the children of women who are reproductively active (Hawkes et al., 1997, Hrdy, 1999, Mace, 2000). This allows them to enhance their inclusive fitness and form useful reciprocal altruism-based alliances.

In humans, therefore, the social network is a resource on which mothers depend to successfully raise their offspring, and this means that each reproductively active woman's decisions about the timing of her births affects the reproductive opportunities of her peers. A woman whose reproductive behavior resulted in the group devoting an unfair portion of their efforts to her offspring could be punished by social exclusion, or group members could simply provide less care for the offspring and, thus, reduce their chance of survival. Such policing activities inevitably damage groups, however. The greatest reproductive success would have been achieved by members of groups that developed ways of establishing a consensus on the reproductive behavior that was appropriate, given the resource levels available (for an analysis on how such prosocial behaviors could evolve when culture is a factor, see Boyd & Richerson, 1985, Boyd et al., 2003, Richerson & Boyd, 2001, Richerson et al., 2003).

It may be that as long as an individual's social group contains a substantial proportion of people who are interested in promoting each other's reproductive success, the consensus can be relied upon to prescribe behavior consistent with the maximization of inclusive fitness. Indeed, observations of traditional societies by a number of human behavioral ecologists revealed that the reproductive norms of these societies encourage behavior that is indistinguishable from that which individuals would choose if striving to maximize their inclusive fitness (e.g., Chagnon, 1988, Hill & Hurtado, 1996, Borgerhoff Mulder, 1988, Cronk, 1989, Irons, 1979, Wang et al., 1995; and reviews by Cronk, 1991, Voland, 1998, and Low, 1993, Low, 1999, Low, 2000).

But when groups are largely made up of unrelated individuals, with no interest in each other's fitness, the consensus reached may reflect other shared or individual goals.

2.4. An interweaving of social instincts

Richerson and Boyd (2001) suggest that human behavioral decisions are influenced by two sets of “social instincts.” One is a set which humans share with other apes. These incline them to compete with other members of their social group. The other set of social instincts is more recently evolved and unique to humans. These incline individuals to cooperate with their fellow group members. The way competition and cooperation are interwoven can be observed in every aspect of human behavior. For example, when a modern father heavily invests in his daughter to ensure she gets a good education, he is partly motivated by competition: he desires that she will be successful relative to her peers. But he is also behaving cooperatively in the sense that he has adopted the modern system of social norms that highly values economic success. The man's fitness would be better served by having more children and encouraging them to begin childbearing early (Kaplan et al., 1995).

The kin influence hypothesis assumes that the interweaving of cooperation and competition influences people as they transmit the social communications that maintain cultural norms and values. It also assumes that genetic self-interest is a factor in the competitive strand. Parents would like their children to be healthy, happy and successful in the terms of their shared culture—but they would also like some grandchildren. Thus, when a modern woman talks to her daughter about future plans, cultural norms will incline her to be concerned for her daughter's happiness and career success, but in most cases, her daughter's reproductive plans will also be salient, more salient than they would be during a similar conversation with younger friend, work colleague, or student.

3.1. Method

We asked female participants to read one of 16 scenarios of 600–800 words, which depicted a childless woman of reproductive age (called “Nicola”) asking an older woman (called “Barbara”) if, given her situation, she should have a baby. We then asked the participants (1) to write down what they think Barbara will say to Nicola and (2) what they themselves think Nicola should do.

In 8 of the 16 scenarios, Barbara and Nicola are depicted as being mother and daughter, and in the other eight, they are depicted as being close friends who are a generation apart in age. In all the scenarios, “Barbara” is depicted as being a happily married, financially secure homeowner with an adult child. Nicola's situation varies. In eight of the scenarios, she is depicted as being in her early 20s, and in the other eight, she is depicted as being in her early 30s. Her personal circumstances are depicted in four different ways:

Texts of the scenarios and questions asked of participants can be viewed on http://psychology.ex.ac.uk/archive/In202/newson2007e_hb_scenarios.pdf.

3.2. Results

The tone and content of the responses participants gave when writing what they thought Barbara would say to Nicola provide ample evidence that the participants understood the scenario and wanted to express themselves clearly. Most of the answers were written in the first person and mentioned details from the scenario. Some replies also included accounts of personal experiences of the participant and evidence of emotional engagement. However, only a third of participants included an explicit answer to Nicola's question when writing what they thought Barbara would reply. Many responses explicitly stated that it had to be Nicola's (or Nicola and her husband's) decision. However, over 90% did express a preference in the advice they wrote, in the statement of their own belief, or in both. The frequencies of the responses in the 16 conditions are displayed in Fig. 1.

Fig. 1. Frequency of responses to scenarios. Responses of participants evaluating whether character should try to have a baby. Participants were assigned to mother or friend condition. The character's circumstances (one of four scenarios) and age (in early 20s or early 30s) were also manipulated.

Not surprisingly, there were correlations between the three dependent variables representing the response of the participants. What the respondent said was her own belief correlated more strongly with the answer in her written reply (Kendall's tau-b=.455, p<.001) than the advice included (Kendall's tau-b=.362, p<.001) in the written reply. Answer and advice were also correlated (Kendall's tau-b=.474, p<.001). Whether or not the participant was a mother herself was found to be correlated with some of her responses and, so, was included in the model as a covariate. The age of the participants was not found to be correlated with any of the responses and, so, was not included. The advice participants wrote in their replies was not significantly correlated with any of the independent variables, but the answer included in their replies and their own beliefs stated afterwards were.

The model testing the association between responses and the situation for raising a child revealed a large effect of this variable (Table 1). Participants were more positive about pregnancy when Nicola was depicted as being in the “difficult” situation of wanting her dying husband's child than when she is in the relatively “easy” situation of the career couple or the stepmother. They were 12 times as likely to include yes in their reply and 25 times as likely to say that they themselves thought Nicola should get pregnant. The age of the woman contemplating pregnancy also had an effect. Participants who received scenarios depicting Nicola as being in her 30s were over four times as likely to say yes to pregnancy and nearly eight times as likely to believe she should get pregnant than those who received scenarios depicting her to be in her early twenties. The relationship depicted did not have a significant general effect on the participants' responses. Whether or not the participant is a mother herself is associated with her own belief about Nicola's choices. For any given combination of situation, e.g., age and relationship, those who do not have children themselves are more likely than mothers to think Nicola should get pregnant. The response pattern for the combination of age and situation revealed that participants were less likely to take age into account when the situation was difficult.

Table 1. Parameter estimates (S.E.) for the ordinal regression models describing associations between participants' responses and Nicola's situation along with other covariates

		Answer	Advice	Own belief
	Situation=difficult	2.45⁎⁎⁎ (.56)	.59 (.45)	3.25⁎⁎⁎ (.58)
	Age=older	1.49⁎⁎ (.49)	.57 (.40)	2.04⁎⁎⁎ (.53)
	Relationship=mother	.70 (.44)	.54 (.83)	.80 (.49)
	Situation×relationship	−.24 (.71)	−.92 (.63)	−2.06⁎⁎ (.76)
	Age×situation	−1.83⁎ (.80)	−.68 (.66)	−2.16⁎⁎ (.79)
	Age×relationship	.97 (.65)	.58 (.57)	−.42 (.67)
	Situation×relationship×age	−1.28 (1.07)	−.32 (.92)	.73 (1.06)
	Covariate: whether participant has children	.01 (.27)	.33 (.24)	.92⁎⁎ (.27)

⁎ p<.001. ⁎⁎ p<.01. ⁎⁎⁎ p<.05.

The combination of situation and relationship in the own belief variable reveals that when the situation for raising a child was difficult, participants who had played the role of mother were less likely to be positive about pregnancy than those who had played the role of “friend.” Under the proportional odds model, the natural logarithm of the odds of a neutral or negative response for combinations of independent variables is obtained by adding the appropriate parameter estimates and changing the sign (McCullagh & Nelder, 1989). In this case, 0.80±2.06 yields an odds ratio estimate of e1.24 or 3.56. The odds of a positive response then would be the reciprocal or 0.29. So, when the situation for raising a child was depicted as difficult, participants who had played the role of mother where about a quarter as likely as those who had played the role of friends to think the young women should get pregnant. By contrast, when the situation for raising a child was depicted as easy, those completing the task in the mother role were about twice as likely as friends to believe she should get pregnant.

The response pattern for the combination of age and situation revealed that participants were less likely to take age into account when the situation was difficult. When responding to the career couple and stepmother stories, women were more inclined to discourage a young Nicola from becoming pregnant than Nicola depicted as being in her 30s. This difference was not seen in those responding to the widow scenario.

The model testing the effect of norms on responses revealed that whether or not becoming pregnant would be considered deviant had a very large effect (Table 2). When they had received a scenario depicting a woman wanting to have a child without the involvement of a socially acknowledged father, participants were a tenth as likely to say yes to pregnancy than when they had received the story depicting a woman contemplating widowhood. Recipients of the single woman scenario were 0.13 as likely to say they believed the woman should consider having a baby in her situation. No other independent variables or combinations were found to have a significant effect.

Table 2. Parameter estimates (S.E.) for the ordinal regression models describing associations between participants' responses and whether or not becoming pregnant would be considered deviant

		Answer	Advice	Own belief
	Norm=deviant	−2.35⁎⁎ (.63)	−.82 (.52)	−2.01⁎ (.58)
	Age=older	−.32 (.62)	−.12 (.53)	−.21 (.58)
	Relationship=mother	.37 (.56)	−.35 (.50)	−.46 (.55)
	Norm×relationship	−.73 (.82)	.01 (.76)	1.07 (.80)
	Age×norm	.77 (.88)	−.1.35 (.86)	.71 (.83)
	Age×relationship	−.23 (.84)	.23 (.73)	.52 (.79)
	Norm×relationship×age	1.35 (1.22)	1.42 (1.15)	−.75 (1.1)
	Covariate: whether participant has children	.57 (.33)	.31 (.30)	.34 (.31)

⁎ p<.01. ⁎⁎ p<.001.

A more detailed report of the results, including the text of the responses, can be viewed on http://psychology.ex.ac.uk/archive/In202/newson2007e_hb_responses.pdf.

3.3. Discussion

Our aim was to investigate how specific changes in the framing of reproductive situations might affect a woman's evaluation of the options. The stories were designed to observe the effect that practical considerations and cultural norms have on a modern woman's belief about a reproductive decision and to detect if this is influenced by a preference for promoting the reproductive success of close kin.

Data supported one prediction of the kin influence hypothesis. Women primed by playing the mother role were more inclined to believe the Nicola character should get pregnant when she was in a good situation to raise a child and less inclined to believe she should get pregnant when raising a child would be difficult and, thus, likely to jeopardize her reproductive success in the longer term. The practicalities of child rearing were found to be of less concern when women were primed by playing the role of a friend. When Nicola was depicted as a woman wanting her dying husband's child, all the participants were largely sympathetic to her desire, but those in the friend condition were more likely to say they believed she should get pregnant.

This effect was seen in the own belief value, which was chosen by the participant herself after she had been primed by writing the response. This is likely to be the best indicator of the participant's opinion. In the answer or advice variables, the effect may have been diluted because many participants playing the role of mother invested a lot in expressing overt sympathy and support for Nicola's desire to have a child when writing their response. But then, in response to the question of what they themselves believed, many said they did not think she should get pregnant. Of course, the nature of the design of this role-play setting limits the generalizability of these results to actual reproductive advice, but the design does allow us to conclude that specific changes in the framing of reproductive situations affect a woman's evaluation of the options in line with predictions of the kin influence hypothesis.

As expected, the norms and beliefs of these culturally modern women had a larger effect than practical concerns. The belief that a woman should not choose to have a baby without the involvement of an acknowledged father had a powerful effect on participants' opinions about what Nicola should do, and this was not detectably mitigated by the age or relationship depicted.

Another prediction generated by the hypothesis was neither supported nor refuted by the data—that the age depicted would influence the evaluations of those playing the role of mother so that they would be more positive about pregnancy in a bad context if “a daughter” was nearing the end of her reproductive life. This might be because contemporary beliefs about the age at which women can and should have children do not reflect biological realities. We had not anticipated that many participants would believe that a woman in her early 30s has plenty of time left in which to decide to have a baby. Yet, this appears to have been the case. The content of several responses included statements like “ women give birth right into their 40s nowadays.” Nicola's age was found to have an effect in the model comparing easy and difficult situations for raising a child. Participants were more likely to recommend pregnancy when the Nicola character was depicted as being in her 30s. However, judging from the content of the replies, this was as much to do with a belief that a woman in her early 20s should postpone child-bearing until she has experienced more of life than a belief that a woman aged 33 years should not delay for fear of losing the chance of having children. When the situation for raising a child was depicted as being difficult because Nicola's husband did not have long to live, participants were more likely to encourage a 23-year-old Nicola to get pregnant immediately, undoubtedly because waiting would not be an option if she wanted her dying husband's child.

If humans are inclined to promote their inclusive fitness (Hamilton, 1964), there should be an innate bias to encourage kin to reproduce more when circumstances are favorable. This study provides convincing evidence for the existence of such a bias, but the effect of the preference is not large. People would prefer their kin to behave in fitness-enhancing ways, but as the kin influence hypothesis predicts, people's evaluation of a reproductive option largely reflect the beliefs, values, and norms extant at time the judgment is made. In the next section, however, we will show that, viewed on the population level and over time, the effect of even a small preference can be large.