The Westermarck Effect and early childhood co-socialization

The aim of this study was to determine whether there are sex differences in regard to the effect of early childhood co- socialization on inbreeding-avoidance at sexual maturity. Multiple logistic regression methodology was used to model the relationship between cousin relationship and social proximity in childhood and its effect on mate choice beyond adolescence for a sample of Moroccan participants. Females showed a synergistic interaction effect for cousin marriage and childhood co-socialization and an additional interaction effect for co-socialization during the first seven years of childhood. Male participants showed no significant effect for early childhood co-socialization. The Westermarck Effect hypothesis that early childhood co-socialization leads to a lack of sexual interest at maturity is thus supported for female participants but not for male participants. These results are consistent with expectations from parental investment theory that females should show greater inbreeding-avoidance than males do because of the greater costs females suffer as a consequence of inbreeding depression.

Inbreeding-avoidance and the issue of sex differences

Darwinian psychology is in the process of usurping much territory that was once the province of Freudian psychology. Although attempts at rapprochement do exist, such as Nesse and Lloyd (1992), psychodynamic and evolutionary perspectives are typically arrayed in direct opposition to one another by evolutionary psychologists. The exemplar of incest or inbreeding-avoidance provides a paradigm case for drawing this distinction. Whereas the Freudian school proposes that the desire for incest is ubiquitous and must be met with taboo and repression, the Darwinian school holds that inbreeding- avoidance is the outcome of evolved adaptive mechanisms. The paradigmatic Darwinian inbreeding-avoidance mechanism is frequently referred to as the ‘Westermarck Effect’. The Westermarck Effect refers to the negative effect that early childhood social interaction and social proximity have on sexual interest in later life. The term was coined by Fox (1902) in honour of the Finnish sociologist Edward Westermarck (1891) who first proposed it. Shepher (1971) proposed that the effect be construed as a form of imprinting, or more specifically in this case, negative sexual imprinting. Shepher (1983) argued that there is sufficient evidence from genetic studies on inbreeding depression to warrant the conclusion that the reduced sexual interest of siblings raised in close proximity could be legitimately construed as an adaptive mechanism to prevent inbreeding.

The hypothesis we explore in this paper is that there are sex differences in sensitivity to the Westermarck Effect. Historically, the empirical study of inbreeding-avoidance in humans has been marked by a lack of consideration of the existence of sex differences in regard to the phenomenon. For example, the classic studies conducted by Wolf (1970) and Shepher (1971), as well as the theoretical model proposed by Fox (1962, 1980), all presume that the Westermarck Effect is symmetrical with regard to sex. Parental investment theory (Trivers, 1972) would predict that the cost of incestuous mating would be higher for females than for males, hence an expected prediction would be that the Westermarck Effect would be stronger for females than it is for males.

Pusey’s (1980) study on chimpanzee emigration patterns revealed that females, rather than males, emigrate from their natal groups, and that they do so in order to find mating opportunities with unrelated males. This observation led to further developments in the understanding of the relationship between sex-linked patterns of dispersal and inbreeding-avoidance in primates (Clutton-Brock, 1989; Moore, 1993; Moore & Ali, 1984). Although dispersal from the natal group at breeding maturity has been interpreted to be an event related to inbreeding-avoidance no matter which sex does the dispersing, Moore (1993) has argued that dispersal may be based on different causal factors for each sex. According to Moore, in species in which males do the dispersing, this is most likely the result of young males being forced out of their natal groups by dominant males who are protecting their breeding privileges. In this case, the non-occurrence of inbreeding by dispersing males is an unintended consequence of intrasexual mating competition by males that catalyses a search for alternative breeding venues (Moore, 1993). By contrast, Clutton-Brock (1989) has shown that in mammalian social species it is generally the case that female dispersal is found only where male breeding tenure exceeds the onset of female age of first conception. Thus, females emigrate in order to avoid inbreeding, but males emigrate in order to breed at all. The pattern depends on how long males tend to maintain their breeding privileges in their social systems.

Recently Wolf (1995, 1998) re-evaluated his data on sim-pua marriage in China. ‘Sim-pua’ refers to a form of marriage formerly practiced in Southern China and Taiwan in which the female is adopted into the family of her future husband at a young age. Wolf (1970) and Wolf and Huang (1980) had argued that one consequence of the sibling-like relationship that ensued was an attenuation of sexual interest at sexual maturity. Wolf believed that the aversion was mutual and equal for both sexes. Responding to the hypothesis advanced by Walter (1990) that there might be sex differences in inbreeding-avoidance, Wolf re-examined his data and found evidence of a sex-linked difference (Wolf, 1995). However, he interpreted this outcome to be spurious, arguing that the sex difference is dependent on the age difference between the sim-pua husband and bride at the time the bride is adopted into the future groom’s family. Following Erickson (1989), Wolf (1995) suggested that the Westermarck Effect is an unintended consequence of attachment behaviour. He argued that sexual interest is interfered with by the attachment bonding that occurs when the older member of the dyad provides parent-like forms of protection and support to the younger member of the dyad. Thus, Wolf sees the mechanism that produces sibling non-incest to be similar to that proposed by Parker and Parker (1986) in explanation of father-daughter non-incest. Nevertheless, it remains unclear to what extent co-socialization leads to a significant reduction in sexual interest. The table Wolf provides shows differences in fertility rates in relation to age differences of husband and wife at the time of the wife’s adoption, but it is debatable whether these differences are significant. One has to question the strength of the Westermarck Effect when these marriages frequently produce between 4 and 8 children. Why would such marriages produce any children at all if the Westermarck Effect were a biological universal? Critics of the Westermarck hypothesis such as Kitcher (1985) and Leavitt (1990, 1992) have argued that such levels of reproductive success show the opposite of inbreeding- avoidance in the same way they argue that the existence of institutionalized brother-sister marriage in Roman Egypt refutes it (Hopkins, 1980).

Sociobiologists defend the inbreeding-avoidance paradigm in the face of such counter-evidence by developing theoretical models that show how inbreeding, in specific circumstances, can improve the inclusive fitness of those who engage in it (van den Berghe, 1983; Welham, 1990). These models tend to emphasize male reproductive strategies, although van den Berghe does offer an account of circumstances in which female inbreeding might be favoured despite the prevalent risk of inbreeding depression. It is important to note that the circumstances considered by van den Berghe have been obtained only rarely in a small number of societies, and only in recent history, and do not reflect what we might expect in typical ancestral environments of evolutionary adaptation (EEA) for our species. Those ancestral environments most likely contain parameters as described above (see Moore, 1993) Nevertheless, the underlying assumption is that males and females are alike with respect to decisions concerning inbreeding and outbreeding.

Hence, what both the critic and the supporter of the Westermarck hypothesis and the sociobiological explanation of inbreeding- avoidance frequently share is a failure to appreciate the potential role that sex differences play in the explanation of incest and inbreeding-avoidance. The issue of sex differences may be of key importance in the study of inbreeding-avoidance, since, if males control the mating system but do not show a disposition to inbreeding-avoidance, then any disposition to inbreeding-avoidance on the part of females may be compromised. Studies that focus exclusively on behaviour (e.g. Bevc & Silverman, 1993) fail to account for sex differences in the disposition to commit or avoid incest. It is interesting to note that in circumstances in which individuals are free to marry whom they want, such as was the case in Shepher’s (1971) study of mate choice on the Israeli Kibbutz, there were no marriages between individuals who were raised together in early childhood. As noted before, Shepher and others assume this ubiquitous avoidance to be mutual and equal on the part of both sexes. Preference data with respect to mate choice may be especially revealing in this regard as it would indicate what people would do if they were capable of exercising their preferences as distinct from what they are forced to do.

Cousin marriage and childhood social interaction in Morocco

It might seem paradoxical that evidence for inbreeding-avoidance should come from a culture that has a marriage custom that prescribes inbreeding between related individuals. Nevertheless, it is indeed the case that much of Westermarck’s evidence for the existence of innate inbreeding-avoidance was derived from cultures that practice some form of cousin marriage, especially Morocco where he did much of his fieldwork.

Patrilateral parallel cousin marriage (alternatively, Father’s Brother’s Daughter or FBD marriage) is one of several marriage options that are available in Moroccan society. Such marriages may often bring together two cousins who have been co-socialized in early childhood. Explanations of the custom of FBD marriage in the Middle East frequently detail the various political and economic motives that make such marriage attractive to the families that resort to them. Cultural anthropologists such as Geertz (1979), Bourdieu (1977) and Holy (1989) have argued that cousin marriage is also a symbol system that facilitates the expression of solidarity and honour between these families. Khuri (1970) identifies three primary strategies behind FBD marriage in the Middle East. First, it is a means by which property can be kept in the family or, more properly speaking, the corporate agnatic kin group. It does this by preventing the alienation of property through inheritance or dowry that occurs when a lineage female marries into another lineage. Second, it is often used to consolidate political alliances between agnatic kin. Third, it is used as a way to facilitate segmentation by turning the lineage in on itself and drawing a boundary between itself and others. Nevertheless, FBD marriage is comparatively rare in Morocco and throughout the Middle East despite its theoretical popularity amongst anthropologists. Although only FBD marriages are ideologically prescribed, Davis and Davis (1989) indicated that marriages between nonFBD cousins were much more frequent than FBD marriages. Geertz (1979) found the number to be 9% for a community studied near Sefrou (cf. Maher, 1974). Similarly, Davis (1983) found the percentage of marriages between related persons to be about 12% for the village she studied. Rosen (1984) argues that kinship ties in Morocco are open to extensive negotiation and that corporate identities and loyalties can never be presumed. Hence, kinship ties are only one element, albeit an important one, in the nexus of concerns, obligations and strategies that are appealed to in political and economic alliance formation.

Biological co-factors may be equally complex. Bittles, Mason, Greene, and Rao (1991) have noted the prevalence of consanguineous marriage throughout the world and have demonstrated that reproductive benefits as well as decrements may be obtained. In particular, they found that fertility levels are higher in consanguineous marriages but that morbidity and mortality are also higher leading to approximately equal levels of long-term reproductive fitness. Much of the variation in mortality and morbidity is accounted for by differences in population structure, nutrition and social class. It would seem that even if the Westermarck Effect is a bona fide suppressor of sexual interest, its relationship to marriage patterns is complex. For example, Khuri (1970) recognizes the existence of attenuated sexual interest in cousin marriages, but he reveals that the lowering of sexual interest is desired by many parents because it focuses the marriage on factors they believe to be more important. Needless to say, this criterion may appeal to parents for the reasons Khuri gives, but it is less clear that it would appeal to the betrothed themselves. Beyond genetic relatedness, a key issue is the effect of childhood social interaction on sexual interest in adulthood. Hence, the issue of childhood social proximity and its effect on sexual desire remains of interest with respect to the psychology of mate selection, and Morocco is good venue for such a study given the residence patterns and marriage prescriptions that prevail there.

In Morocco, particularly in rural areas, it is not unusual for two adult sons to live with their wives and children under a single roof in the extended household of their parents (Rosen, 1984). It is thus also not uncommon for opposite sex cousins to grow up in close social proximity. Davis and Davis (1989) report that 18% of the households in the community they studied were of the three- generation extended family sort. They found that opposite sex children commonly sleep in the same room in their semi-rural village of Zawiya and that sometimes this includes adolescents as well. (Often, however, if there is only one other bedroom and post- pubertal children of both sexes need to be sheltered, the living room will be used for one of the sexes.) Maher (1974) notes that relatives living in town often are asked to accommodate the children of rural relatives so that the latter can attend primary or secondary school during the school year (see also Davis & Davis, 1989). This is sometimes facilitated by the process of fostering, which is an informal form of adoption commonly practiced in many regions of Morocco, and which may involve uterine kin as well as agnatic. Maher discusses an example in which one foster father’s brother’s son married the daughter of the foster father. Uterine kin of the wife may also end up in the household, and, although this is most common when the couple is childless, it also occurs in families that have their own children as well (e.g. when children are orphaned through the death of their parents). Thus, in situations such as these, cousins will be reared together much like siblings. Alternatively, many individuals who work abroad during the year return for the summer and reside with relatives, thus bringing cousins together for one to three months during the summer. In this latter case, children who are of primary school age or younger may be brought together under conditions that provide for close and intimate social interaction. Although such visitations may be more likely between brothers, it is also possible for non-FBD/S cousins to be brought together. Thus, even though co-socialization may be more likely to occur between patrilateral relatives, it is not unknown to have children related through their mothers living in close residence and hence playing together on a daily basis from early childhood on.

Unfortunately, there is not a great deal of literature on the quality and intensity of childhood play in Morocco. Most descriptions come from ethnographers who have studied family life in Morocco. Typically, boys and girls play in sex-segregated groups from early on and they do not have many opportunities for cross-sex social contact (Davis, 1983; Maher, 1974). Davis and Davis (1989), who conducted an in-depth study on adolescence in the village of Zawiya in north central Morocco, offer some insight into the matter of childhood social interaction. They claim that mixed play in childhood is rare after the age of 6 or 7 and that by the age of 12 years, males and females form virtually segregated groups in which males play games like soccer, whereas females cook, do needlework and help out around the house. Nevertheless, it is the custom in Morocco, as in other Muslim countries, for male and female children to attend the public bath (hamam) together prior to the age of seven. It is thought that no harm is possible at this age and so parents are rather permissive in connection with the social interaction of opposite sex children. After the age of seven, however, public bath attendance is no longer permitted and children begin to be segregated in their play. This segregation becomes increasingly stringent as children approach puberty, at which time strong sanctions against heterosexual social interaction are put into effect (see also Mernissi, 1987). Despite these obstacles, Davis and Davis (1989) relate how young men and women conspire to meet, and they discuss at length the prevalence of illicit sexual encounters that adolescents can manage to arrange.

To date, McCabe (1983) studied patterns of fertility and divorce in reference to cousin marriages in Lebanon in a manner similar to that employed by Wolf (1970) and found a pattern of lowered fertility and higher divorce. Unfortunately, her study is limited by the same considerations that apply to Wolf’s work in China. Walter (1997) employing regression analysis on marital preference data drawn from a Moroccan sample found strong evidence for a negative effect of childhood co-socialization and cousin relationship for female participants (p < .02) but not for male participants p = .13) in a model that contained other predictors of marital desirability as covariates. Although, the unreliable outcome for male participants is suggestive of a significant sex difference, Walter (1997) also found that males and females both rejected individuals whom they slept in the same room as when children. A problem with the latter result, however, was that the number of individuals who could remember such intimate childhood sleeping arrangements was quite small (n = 17).

Study objectives

The models presented here were designed with several objectives in mind. Because Walter (1997) examined only the effects of the first seven years of co-socialization on mate choice, the question remains as to what effect childhood co-socialization has on mate choice across the entire period leading up to sexual maturity. Co- socialization data for the second seven years of childhood were included in the model-building procedure in an attempt to determine whether co-socialization in the later phase of childhood is also an effective predictor of inbreeding-avoidance and, if so, whether there are significant sex differences. Interaction effects between early and late childhood co-socialization were also modelled. The present study obtained data specifically from cousins in order to examine the interaction between childhood co-socialization and cousin relationship including a second-level interaction for sex. Cousins were isolated from the data set for specific analysis because the quality of social interaction between cousins is likely to be more intimate than is typical in the wider population given the widespread pattern of co-residence between cousins that prevails in Moroccan society. In addition, Wolf’s (1995) hypothesis that age difference between male and female children under conditions of intimate social proximity explains sexual aversion was also tested.

Method

Participants

The data set used for the development of the models presented here is the same as that employed in Walter (1997). Participants consisted of 147 female and 107 male native Moroccans between the ages of 15 and 30 years.

Procedure

Each participant provided a list of potential marriage candidates whom they rank-ordered in a pairwise comparison procedure and then categorized as either high or low on the outcome variable and on each predictor. For example, a score of 1 on the outcome variable, marriage, means the candidate was categorized as a desirable marriage candidate, whereas a score of 0 means the candidate was rejected. The predictors are listed in Table 1. Each predictor in the model was demonstrated to be a significant predictor of marriage desirability in Walter (1997). They are included here in order to control covariation with the co-socialization and relatedness predictors.

Table 1. Outcome and predictor variables and values

Childhood social proximity was measured for each case belonging to a participant by means of three 6-point scales. Participants were asked to pick the level of contact that best applied to each candidate on their list. The three scales referred to the level of social contact during the first seven years of childhood, the second seven years of childhood, and to how frequently the participant and the ratee slept in the same room during the first seven years of childhood. The scales appeared on the interview schedule as follows:

every day summers only weekly monthly yearly never

A dichotomous predictor was devised by dividing the scale such that cousins who saw each other every day for at least the summer months or longer were distinguished from those who saw each other once a week or less. Cousin relationship was recorded as a dichotomous dimension that reflected the presence or absence of cousin relationship between the participant and the marriage candidate. Other covariates included participant’s age, region of origin (north vs. central Morocco), urban or rural residence, and educational level. These variables were included in order to control for differences due to demographic variation.

As in Walter (1997), statistical analysis was conducted using the technique of logistic regression (see Hosmer & Lemeshow, 1989) to model the outcome, desirable for marriage, as a function of the predictor variables. As mentioned in the section above, each participant rank-ordered and then categorized a number of candidates on each predictor and the outcome variable. Female participants had 3-11 candidates, with a mean of 5.6, whereas males had 3-12 candidates, with a mean of 6.0. In total, there were 830 candidates provided by the female participants and 638 candidates provided by the male candidates. Records were clustered by participant using generalized estimating equations. This methodology, an implementation of mixed effects models, adapts to the heterogeneity in response due to differences from rater to rater. It thus controls for a rater who, say, especially prizes one attribute and also provided many candidates (see Diggle, Liang, & Zeger, 1994).

Each main effect predictor included in the model is there because it provides a significant contribution to the outcome variable of marriage desirability (Walter, 1997). In the models presented here, they serve two purposes. First, as covariates to ensure that the relationships between co-socialization and marriage desirability are not confounded with other known effective predictors. Second, interaction effects may exist between these predictors and co- socialization. The development of these models differs from Walter (1997) in the number of interaction effects included in the modelling procedure. The models shown are the best using the criterion of predictive mean squared error, assessed via bootstrapping (Efron & Tibshirani, 1997). That is, the models do the best job of successful prediction of the outcome.

An odds ratio of 1 corresponds to a regression coefficient of 0, increasingly smaller fractions of 1 for odds ratios correspond to increasingly negative regression coefficients, and odds ratios increasingly larger than 1 correspond to increasingly large positive regression coefficients. In particular, an odds ratio of 0.20 for a given predictor means that the participant classifies the stimulus object as being five times less marriage-desirable on that predictor, whereas an odds ratio of 5 means that the participant classifies the stimulus object as being five times more desirable with reference to that predictor.

Results

Table 2(F&M) presents the full logistic regression models. Several interaction effects were found that improve the model presented in Walter (1997). Table 2(F) reveals that for female raters, cousin relationship combined with high childhood co- socialization is a significant negative predictor of adult marital preference (p = .01). Table 2(M) reveals that for male raters, this relationship is unreliable (p = .11). The sex difference for males and females on the relationship between early childhood co- socialization and marriage desirability was highly significant at p < .01. Females show a negative response compared with males on this dimension.

Tables 3 and 4 consist of two-by-two arrays of odds ratios for the different levels of the specific predictors under consideration. The reference case is arrayed in the upper left quadrant. The odds ratio for this group is set at 1 and the odds ratios for the other groups are calculated in comparison with it. The numbers in parentheses under the odds ratio present the 95% confidence intervals for the odds ratios. The odds ratio itself represents the best estimate. Table 3 presents the odds ratios that show the interaction of co-socialization and cousin relationship for female participants. Male cousins with high co-socialization levels were rated by females to be five times more likely to be rejected for marriage compared to low co-socialized non-cousins. Compared with this, both high co-socialized non-cousins and low co-socialized male cousins both show ‘negative’ but non-significant effects.

Table 2. Logistic regression coefficients for best model for (F) female raters and (M) male raters on marriage desirability. Odds ratios > 1 indicate that an increase in the variable is associated in an increase in candidate’s desirability for marriage.

In order to test Shepher’s (1971) hypothesis that the first six years constitute a critical period in which sexual aversion develops in response to social proximity, we compared cousins under varying levels of early and late childhood social proximity. (Our scales differ from Shepher’s in that ours divides childhood into two seven- year intervals.) Restricting to cousins, the study included 134 female participants rating 311 candidates, and 98 male participants rating 209 candidates. Individuals who were seen every day during the first seven years constitute the high early childhood group. Individuals who were seen every day during the second seven years form the high late childhood group. If the case being rated was seen once a week or less, these were categorized as being low early or low late, respectively. Thus we obtain two-by-two tables for each sex (see Table 4F & M) comparing the following conditions: low early and low late childhood co-socialization, low early and high late childhood co-socialization, high early and low late childhood co- socialization, and high early and high late co-socialization. Table 4 F & M shows the results of a model that contains an interaction term for early and late childhood co-socialization.

Table 3. Odds ratios for childhood co-socialization x cousin relationship in female raters. Odds ratios < 1 indicate a reduction in the candidate's desirability for marriage. Confidence intervals for the odds ratios are given in parentheses.

Table 4. Odds ratios for early co-socialization x late co- socialization, when restricted to cousins. Odds ratios < 1 indicate a reduction in the candidate's desirability for marriage; odds ratios > 1 increase in the candidate’s desirability. Confidence intervals for the odds ratios are given in parentheses.

For male raters, both the high early/low late childhood co- socialization group and the low early/high late co-socialization group reveal ‘positive’ odds ratios (1.93 for the former and 3.16 for the latter), whereas the high early/high late co-socialization group shows a slightly ‘negative’ odds ratio of 0.85. However, the differences between these groups as represented in these results are not sufficiently different to achieve statistical significance. These results demonstrate that for male raters, cousin relationship, combined with either early, late or continuous childhood co- socialization is an unreliable predictor of adult mate preference. For female raters, however, the term for low late/high early childhood co-socialization is statistically highly significant (p<.0001). Female participants who experienced high early childhood co-socialization and low late childhood co-socialization with an individual cousin, unanimously rejected the individual for marriage later in life. The odds ratio of 0.0003 indicates that early childhood co-socialization makes a male over 3000 times more likely to be rejected as a marriage partner later in life. One cannot expect a stronger effect. This supports Shepher's hypothesis that the first six or seven years of childhood constitute a critical period for negative sexual imprinting, but it supports this claim only for females and not males.

We also tested Wolf’s (1995) hypothesis that the age difference between the participant and the person rated makes a difference in the relationship between childhood co-socialization and adult sexual attraction. This was tested in the following manner. Age of rater was removed from the list of covariates and a new term was added to the model that represented the age difference between raters and ratees. We categorized the age difference into three categories: less than 3 years, 3-6 years, and greater than 6 years. We found no statistically significant effect for females (p = .23) or males (p = .10).

Discussion

These results demonstrate that females show a greater aversion than do males to marriage with individuals with whom they shared daily social contact in early childhood, as would be predicted by Trivers’ parental investment hypothesis (1972). Although males do not show the same effect for early childhood co-socialization, they do show an aversion to marriage with cousins (Walter, 1997). Nevertheless, rejection of cousin marriage per se cannot be adduced to support a claim for inbreeding-avoidance on the part of males as the level of genetic identity encountered in cousin marriages falls outside the parameters associated with inbreeding depression as discussed by Thornhill (1991).

What relationship, if any, exists between the various predictors of mate attraction (i.e. intelligence, physical attractiveness, kindness and responsibility) that are presented in Walter (1997) and the inbreeding-avoidance variables presented in this study? Because there were no interaction effects between the inbreeding-avoidance variables (i.e. cousin relationship, childhood co-socialization) and the other primary predictors (i.e. responsibility, intelligence, kindness, physical attractiveness), these appear to be additive or subtractive in nature. This situation forces us to re-examine the nature or purported strength of inbreeding-avoidance on mate attraction more generally. Previous theorists such as Shepher (1971), Fox (1962, 1980) and Wolf (1970, 1995, 1998) argue as though the existence of an active inbreeding-avoidance response nullified the effects of any and all other potential predictors of marital attraction. The results presented in Walter (1997) suggested no overwhelming disqualifying effect of sleeping in the same room as children for both sexes, merely a subtractive one in which sleeping in the same room as children made the candidate about five times less likely to be categorized as marriage desirable. It is possible that other factors in combination might potentially outweigh such a negative bias. The results presented here suggest, on the contrary, that, although the relationships between the mate choice and the co- socialization predictors are still additive or subtractive, daily social contact in the first seven years ubiquitously disqualifies individuals as marriage candidates for female participants. The results also suggest that the sleeping in the same room question may have been misunderstood by at least one female participant in Walter (1997), as sleeping in the same room everyday necessarily implies daily social contact. Odds ratios for male raters in Walter (1997) showed that females with whom they shared the same bedroom on a daily basis in early childhood were four times more likely to be classified as undesirable for marriage. Thus, although males show a significant Westermarck Effect under conditions of closest childhood social proximity, females show a significant Westermarck Effect at lower levels of childhood social contact. This finding thus supports the prediction of a sex difference in inbreeding-avoidance as predicted by Trivers’ parental investment theory.

The one puzzling result that was obtained in this set of analyses, perhaps, is the finding that continued high levels of daily social contact into the second seven years appears to moderate the Westermarck Effect. Why should individuals who experienced high levels of social contact everyday for the first fourteen years he only 25% more likely to be deemed marriage-undesirable when high levels of daily social contact in early childhood only resulted in ubiquitous rejection? This odd result is most likely due to the presence in the sample of high late daily social contact group of many individuals who do not reside in the same home. (They are most likely close neighbours.) That is, genuine co-residence may be missing from many of these cases, thus compromising the measure. The strength of the effect of daily social contact solely in early childhood is probably due to the fact that daily social contact for this group does entail co-residence. Thus, we conclude that early childhood social interaction produces negative sexual imprinting in females but not in males.

Finally, we should note that the type, frequency and duration of social contact employed in this study are conceived somewhat broadly and could benefit from being made more precise. In future research, it may be feasible to introduce greater specificity into the scaling of these, and possibly other, potentially relevant dimensions of childhood social interaction. For example, different categories of play activity could be delineated and partitioned into finer slices of duration, frequency and intensity levels for different age grades. Logit-probit models may prove especially effective for teasing out significant aspects of such data. A key difficulty, of course, resides in the time-based trade-off between getting less accurate data about childhood social interaction many years after the fact or accurate data at the time of childhood but having to wait years until the onset of sexual maturity. Longitudinal studies no doubt would afford the greatest and most accurate long-term pay- off in this regard.

British Journal of Developmental Psychology (2003), 21, 353-365

(C) 2003 The British Psychological Society

References

Bevc, I., & Silverman, I. (1993). Early proximity and intimacy between siblings and incestuous behavior: A test of the Westermarck theory. Ethology and Sociobiology, 14, 171-181.

Bittles, A. H., Mason, W. M., Greene, J., & Rao, N. A. (1991). Reproductive behavior and health in consanguineous marriages. Science, 252, 789-794.

Bourdieu, P. (1977). Outline of a theory of practice. Cambridge: Cambridge University Press.

Clutton-Brock, P. (1989). Female transfer and inbreeding- avoidance in social mammals. Nature, 337, 70-72.

Davis, S. S. (1983). Patience and power: Women’s lives in a Moroccan village. Cambridge, MA: Schenkman.

Davis, S. S., & Davis, D. A. (1989). Adolescence in a Moroccan town. New Brunswick, NJ: Rutgers University Press.

Diggle, P., Liang, K.-Y., & Zeger, S. L. (1994). Analysis of longitudinal data. New York: Oxford University Press.

Efron, B., & Tibshirani, R. (1997). Improvements on cross- validation: The .632+ bootstrap method. Journal of the American Statistical Association, 92, 548-560.

Erickson, M. (1989). Incest avoidance and familial bonding. Journal of Anthropological Research 45, (3), 267-291.

Fox, J. R. (1962). Sibling incest. British Journal of Sociology, 13, 128-150.

Fox, R. (1980). The red lamp of incest. New York: Dutton.

Geertz, H. (1979). The meaning of family ties. In C. Geertz, H. Geertz, & L. Rosen (Eds.), Meaning and order in Moroccan society (pp. 315-391). Cambridge: Cambridge University Press.

Holy, L. (1989). Kinship, honour, and solidarity: Cousin marriage in the Middle East. Manchester: Manchester University Press.

Hopkins, K. (1980). Brother-sister marriage in Roman Egypt. Comparative Studies in Society and History, 22, 303-354.

Hosmer, D. W., & Lemeshow, S. (1989). Introduction to applied logistic regression. New York: Wiley.

Khuri, F. (1970). Parallel cousin marriage reconsidered: A Middle Eastern practice that nullifies the effects of marriage on the intensity of family relationships: Man, 5(4), 587-618.

Kitcher, P. (1985). Vaulting ambition. Cambridge, MA: MIT Press.

Leavitt, G. C. (1990). Sociobiological explanations of incest avoidance: A critical review of evidential claims. American Anthropologist, 92, 971-993.

Leavitt, G. C. (1992). Inbreeding fitness: A reply to Uhlmann. American Anthropologist, 94, 448-450.

Maher, V. (1974). Women and property in Morocco. Cambridge: Cambridge University Press.

McCabe, J. (1983). FBD marriage: Further support for the Westermarck hypothesis of the incest taboo? American Anthropologist, 85, 50-69.

Mernissi, F. (1987). Beyond the veil. Bloomington: University of Indiana Press.

Moore, J. (1993). Inbreeding and outbreeding in primates: What’s wrong with ‘the dispersing sex’? In N. Thornhill (Ed.), The natural history of inbreeding and outbreeding (pp. 329-426). Chicago: University of Chicago Press.

Moore, J., & Ali, R. (1984). Are dispersal and inbreeding- avoidance related? Animal Behaviour, 32, 94-112.

Nesse, R., & Lloyd, A. (1992). The evolution of psychodynamic mechanisms. In J. Barkow, L. Cosmides, & J. Tooby (Eds.), The adapted mind (pp. 601-626). New York: Oxford University Press.

Parker, H., & Parker, S. (1986). Father-daughter sexual abuse: An emerging perspective. American Journal of Orthopsychiaty, 56, 531- 549.

Pusey, A. (1980). Inbreeding-avoidance in chimpanzees. Animal Behaviour, 28, 543-552.

Rosen, L. (1984). Bargaining for reality. Chicago: Unive
sity of Chicago Press.

Shepher, J. (1971). Mate-selection among second-generation kibbutz adolescents and adults: Incest-avoidance and negative imprinting. Archives of Sexual Behavior, 1, 293-307.

Shepher, J. (1983). Incest: A biosocial view. New York: Academic Press.

Thornhill, N. (1991). An evolutionary analysis of rules regulating human inbreeding and marriage. Behavioral and Brain Sciences, 14, 205-246.

Trivers, R. L. (1972). Parental investment and sexual selection. In B. Campbell (Ed.), Sexual selection and the descent of man 1871- 1971 (pp. 136-179). Chicago: Aldine.

Van den Berghe, P. (1983). Human inbreeding avoidance: Culture in nature. Behavioral and Brain Sciences, 6, 91-123.

Walter, A. (1990). Putting Freud and Westermarck in their places. Ethos, 18, 439-446.

Walter, A. (1997). The evolutionary psychology of mate selection in Morocco: A multivariate analysis. Human Nature, 8(2), 113-137.

Welham, C. V. J. (1990). Incest: An evolutionary model. Ethology and Sociobiology, 11, 97-111.

Westermarck, E. (1891). A history of human marriage. New York: Macmillan.

Wolf, A. (1970). Childhood association and sexual attraction: A further look at the Westermarck hypothesis. American Anthropologist, 72, 503-508.

Wolf, A. (1995). Sexual attraction and childhood association: A Chinese brief for Westermarck. Stanford, CA: Stanford University Press.

Wolf, A. (1998). Childhood association, sexual attraction, and the sexes or was Dr. Ellis right? Paper presented at the International Westermarck Symposium, Marriage, Morality, and Emotions: Updating the Work of Edward Westermarck, Helsinki, Finland, November 19-22.

Wolf, A., & Huang, C. (1980). Marriage and adoption in China: 1845-1945. Stanford, CA: Stanford University Press.

Received 4 February 2002; revised version received 21 October 2002

Alex Walter and Steven Buyske*

Rutgers University, USA

* Requests for reprints should be addressed to Steven Buyske, Department of Statistics, Rutgers University, 110 Frelinghuysen Road, Piscataway, NJ 08854, USA (e-mail: buyske@stat.rutgers.edu).

Copyright British Psychological Society Sep 2003