Fecundity Compensation via Kin Selection

In the early 1970s, Trivers and Wilson proposed the idea of fecundity compensation through kin selection [22], [23], suggesting that while subjects might not reproduce directly, increased fecundity in close relatives is promoted through direct support and assistance to close relatives who share genes, thus compensating for the lower reproductive rate observed among homosexuals. There is sufficient evidence from behavioral genetics that gene selection through direct support to kin is a powerful mechanism to enhance the transmission of genes from one generation to the next [24]. However, this genetic mechanism could not be confirmed for homosexuals in U.S. populations. Bobrow and Bailey observed that male homosexuals do not lend behavioral, emotional or financial support to close kin that would significantly enhance inclusive fitness, and social contact with both parents in some patterns even appears to be contrary [25]. However, subsequent evidence from a Samoan population has provided some support for kin selection. [26], [27], [28].

Unfortunately, further studies in both eastern and western industrialized societies have not provided evidence for kin selection [29], [30], [31] The cross-cultural variability in the kin-directed altruism of homosexual males is not a universal characteristic that could account for the evolution of the genetic influence of homosexuality within the human population and does not exclude the possibility that kin selection might coexist with other mechanisms; thus, this hypothesis per se cannot be considered to be a self-sufficient solution for the homosexual conundrum. Therefore, two potential mechanisms for balancing selection in male homosexuality have been explored: sexually antagonistic selection [11], [21] and overdominance [19], [32], [33].

Sexually Antagonistic Selection

In 2004, Camperio Ciani et al. [11] suggested that male homosexuality was a by-product of a genetic factor that increases fecundity in females. This notion was based on data showing an increased frequency of homosexuality in the maternal line of homosexual probands [11],[34] and that both mothers and maternal aunts were more fecund in homosexual probands compared with heterosexual probands [11], [21], [34], [35], [36]. Additional evidence was found in support of the sexually antagonistic selection and female fecundity hypothesis independent of the fraternal birth order (see the further fraternal birth order section for an introduction to this effect). The responses from a large Internet survey also indicated that homosexual mothers exhibit higher fecundity compared with heterosexual mothers, independent of the fraternal birth order effect [37]. These results were later confirmed for maternal aunts in a wide Caucasian sample population [36] and were replicated in mothers of Samoan androphilic males, which have significantly more children than gynephilic males [38], [39]. However, these same authors recently showed that not only the maternal but also the paternal grandmothers of Samoan androphilic males (fa'afafine) produce more offspring than gynephilic males, which does not support the predictions of the X-chromosome linkage hypothesis [40].

Balancing selection via sexually antagonistic selection is a genetic mechanism based on a multi-locus genetic factor partially located on the X-chromosome, which could potentially promote androphilic behavior (attraction to males) in carriers. In males, this genetic factor increases the probability of homosexuality through androphilia, whereas in females, it increases fecundity through androphilia [11], [21], [34], [35], or better through a complex pattern of behavior, personality and enhanced fertility that increases attraction from males and fecundity [35]. According to the sexually antagonistic hypothesis, the localization of such a factor on the X-chromosome should increase fecundity only in maternal-line females sharing the X-chromosome with the homosexual subject and not in paternal-line females or any other related male. A mathematical analysis of the population dynamics has shown that this sexually antagonistic selection mechanism respects the empirical assumptions of both stability and low frequency [21].

Overdominance

In 2005, King and colleagues [32] proposed different conclusions while investigating a large British clinical sample. These authors performed a different analysis on probands based on family size, rather than fecundity and concluded that all members of a homosexual family compensate with increased fecundity [32]. Contrary to the predictions of the X-chromosome linkage hypothesis, King et al. (2005) showed that, compared with heterosexual males, homosexual males had significantly more aunts, uncles and cousins in the paternal line, but not the maternal line [32]. Schwartz et al. (2010) showed that, compared with heterosexual males, the paternal grandmothers of homosexual males have higher offspring production, but not the maternal kin [19]. The observation of an enlarged family size in homosexual probands suggests a balancing selection mechanism through overdominance, which has since been suggested in other studies [19], [41], [42].

Overdominance suggests that a co-dominant genetic factor in autosomes promotes both fecundity and homosexuality. If this factor were present as a single copy (heterozygous) in an individual genome, an increase in fecundity would be observed, whereas if two copies (homozygous) of this factor were present, the probability of male homosexuality would increase and should not affect female fecundity. It is possible that a multi-locus genetic trait in males results in behavior that is more conducive to reproducing and supporting offspring among heterozygotes, and that this trait in the homozygotic condition might produce male homosexuality. The greater reproductive fitness or fecundity the biological relatives of homosexuals men would thus offset the selection pressure against homosexuality [33]. A mathematical analysis of population dynamics suggests that this mechanism, if based on at least two autosomal loci, respects both of the empirical assumptions of stability and low frequency [21]. Thus, if confirmed, such a mechanism could resolve the conundrum.

Fraternal Birth Order Effect

Numerous studies have shown that the existence of older brothers increases the probability of homosexuality in later-born human males [43]. The probability of a male homosexuality increases with increasing numbers of older biological brothers, a phenomenon referred to as the fraternal birth order effect. The fraternal birth order effect has been reviewed several times [44], [45], [46]. Biological brothers increase the probability of homosexuality in later-born males, even if these siblings are reared in different households, in contrast to stepbrothers or adoptive brothers, which have no effect on sexual orientation [47].

The most likely explanation for the fraternal birth order effect is the progressive immunization of certain mothers to male-specific antigens through subsequent male fetuses and the increasing effects of anti-male antibodies on the sexual differentiation of the brain in successive male fetuses. These explanations do not necessarily suggest a direct genetic influence on sexual orientation but rather a potential maternal immune response to Y-linked histocompatibility antigens [46], [48]. Alternatively, it has been suggested that potential sexual contact with older brothers [49] or less rigid parenting might also potentially enable younger sons to more freely admit their sexuality [32]. However, Bogaert (2006) finding that the fraternal birth order effect only exists with biological brothers, contradict these alternative interpretations [47]. The maternal effect, defined as the fraternal birth order effect, has generally been examined in relation to the number of older brothers [43]. Blanchard [48] estimated that the population “risk” for homosexual orientation due to the fraternal birth order effect is approximately 33% for each older brother. Further support for the effect of fraternal birth order includes the much lower ratio of first-born to second- or third-born status among homosexual probands compared with the corresponding ratio in heterosexuals, particularly in low fecundity populations [11]. Few studies have reported that homosexual men also have greater numbers of older sisters than expected [32], [37], [38], [39], [50].

The fraternal birth order effect has now been observed in a large number of studies. Theoretically, however, the maternal effect cannot account for the conundrum. A systematic mathematical analysis of the propagation and equilibrium of putative genetic factors associated with male homosexuality in a population based on the use of the selection equation for one or two biallelic loci and Bayesian statistics for pedigree investigation showed that maternal effects violate the empirical requisite of both stability and low frequency [20], [21]. Thus, if the compensatory effect of homosexuality only results from a maternal effect (fraternal birth order) mechanism, then a rapid extinction or total diffusion of homosexuality in most populations would be observed. Both effects are inconsistent with historical and empirical observations. The persistence of male homosexuality in the human population has been confirmed through archaeological evidence, indicating that homosexuality occurred prehistorically and is therefore not a recent phenomenon [51], [52], [53], [54].

This finding does not suggest that a fraternal birth order effect does not occur, but rather that the fraternal birth order effect alone cannot explain the persistence of genetic factors that partially influence homosexuality associated with the low frequency of reduced fecundity in all populations. The same conclusion applies to other maternal effects, such as maternal genomic imprinting [21].

Balancing Selection: Evidence from First-born Homosexuals and Younger Siblings

The identification of a balancing selective mechanism is complicated by the fact that fraternal birth order predicts an increase of older brothers due to a maternal effect in a family of homosexual probands. Thus, alternative balancing selection hypotheses should also show increased fecundity outside of the predictions of fraternal birth order. Increases in fecundity as an artifact (sampling bias) of the fraternal birth order effect are possible because randomly selected homosexual groups tend to include more older brothers than well-matched heterosexual comparison groups; consequently, homosexual groups exhibit larger sibships [42]. The sampling bias of fraternal birth order is observed when homosexuals are used as probands to investigate the fecundity of their mother. To clarify whether increases in fecundity resulting from a maternal effect are distinct from increases in fecundity that result from balancing selection, it is necessary to examine fecundity when the fraternal birth order effect is absent. The absence of the fraternal birth order effect can be observed through a comparison of the fecundity of mothers with homosexual firstborn sons, verifying the existence of a larger number of total offspring [11], [34], [41], [55] or comparing the number of younger siblings between similarly matched homosexual and heterosexual probands [37]. These data support balancing selection as a factor independent of maternal effects. An independent role of fecundity from the fraternal birth order effect has been suggested for high fertility populations, such as Samoa. [38] This question could also be addressed through a comparison of the fecundity of non-parental female relatives, particularly, individuals not selected for having a homosexual son and, thus, whose fecundity cannot be ascribed to the fraternal birth order effect, but notwithstanding should be more fecund according to the balancing selection hypothesis. Specifically, the fecundity of aunts and grandmothers, independent from having a homosexual male offspring is informative, as these individuals would not be selected from their own offspring and thus the fraternal birth order effect would not bias the sample toward an artificially larger family size and would not interfere with fecundity predictions. In this case, any differences in sibship size (averaging out religious, socioeconomic, ethnic, or other confounding cultural or demographic effects trough an adequate sampling design) could be reasonably attributed to differences in specific parental fertility.

Research Question

The aims of this study are to determine the balancing selection mechanisms, if any, which allow the persistence of genetic factors that influence male homosexuality in the population. The fecundity of aunts and grandmothers of homosexual vs. heterosexual probands facilitates an analysis of three competing hypotheses with mutually exclusive predictions.

1. The lack of differences in either maternal or paternal female fecundity will provide strong empirical evidence that all fecundity variations in homosexuals are exclusively due to a side effect of fraternal birth order, as indirectly suggested [55].
2. Increased fecundity in both maternal females and paternal females will provide empirical evidence for overdominance as type of the balancing selection, as indirectly suggested [19], [32].
3. Increased fecundity in only the maternal line of female relatives of homosexual probands will provide evidence for sexually antagonistic selection as the type of balancing selection, as predicted [21].

Balancing Selection and Fraternal Birth Order Effect

Camperio Ciani, Corna and Capiluppi [11] showed that mothers of homosexual men were significantly more fecund than the control-matched mothers of heterosexual men. Blanchard [55] noted that a research design based only on maternal fecundity was somewhat problematic because of the fraternal birth order, which could inflate the mean size of the sibships of the homosexual group. However, Camperio Ciani, Corna, and Capiluppi [11] maintained that a fecundity advantage was also present in the mothers of first-born homosexuals, for whom the fraternal birth order cannot be implicated. However, due to the small sample size of the initial study, their results were not highly significant. Later, Iemmola and Camperio Ciani [34] performed a study in a larger sample, which provided higher statistical significance, even in mothers of first-born homosexuals, suggesting that the balancing effect was accurate. Blanchard [55] confirmed the increased fecundity of mothers of first-born homosexuals through the analysis of four large sample sets of mothers of first-born homosexuals, showing that first-born homosexual mothers did not show a fecundity increase and suggesting that the fecundity increase could be an artifact of the fraternal birth order [55]. Riger et al. [41] promptly replied to Blanchard [55], arguing that in their large sample of homosexuals, the mothers of a first-born homosexual were indeed significantly more fecund than the mothers of heterosexuals, thus confirming the conclusions obtained in previous studies [11], [18], [21], [34]. The database assembled by Blanchard [55] was impressive and cannot be easily dismissed. However, several observations can be made. First, the age difference between homosexuals and heterosexuals in the previous analysis was large [55], showing a range of greater than 10 years in some samples, and included heterosexuals who were much younger than homosexuals. Thus, correcting for age in this case could result in particular distortions [63], [64]. Notably, the correction for age is necessary to control for the cohort effect, as mothers born earlier belong to a more fecund cohort if all other assumptions, such as the linearity of the variable regressed, are respected. However, in our species, the age distribution of female fecundity is different from that of males, which is linear from age 20 to 50. In contrast, among modern Western females, the fecundity distribution is highly skewed. The fertility peaks in women in their early 20s and considerably declines after the age of 35, becoming null before 50 [65]. Thus, a linear regression of age, as shown through ANCOVA, could artificially overestimate fecundity in younger mothers, with some residual fecundity that might be expressed, thereby distorting the ANCOVA and resulting in an overestimation of younger female fecundity; thus, an increase in the fecundity of the heterosexual sample would be observed in the Blanchard data. Because of this well-established and persistent problem with ANCOVA [63], [64], [66], [67], the raw data should be included to estimate the actual effect of age co-variation [68], as performed in the present study. An advantage of the present study is that it focuses only on the older female relatives of the probands, whose fecundity was definitively complete at the time of investigation.

A second observation refers to the inclusion [55] of a sample from a previous study [37] in the analysis of the fecundity of mothers with a first-born homosexual son [55]. However, the inclusion of the Blanchard and Lippa BBC Internet database [37] was problematic, as the authors observed that the fraternal birth order effect was weaker and more inconsistent than previous studies. Blanchard and Lippa suspected that the participants responded less conscientiously through the Internet than conventionally examined participants. Consequently, a series of filters and statistical operations was employed to remove inaccurate respondents and confounding effects. Furthermore, the probands in this Internet population study belonged to a population adopting a so-called “stopping rule”, which comprises individuals that continue having additional offspring until acquiring children of both sexes [69], [70]. The adherence to this common, but confounding rule obliged the authors to limit their results only to right-handed and non-last-born homosexuals to determine fraternal birth order effects. Approximately 20% of the sample was excluded due to unreliable answers based on incongruence between the reported family size and sibling composition. However, this exclusion could not be realized for the sub-sample size in the Internet study of the mothers of first-born homosexuals [55]. We hypothesize that a similar percentage of unreliable respondents might also be present among the first-born respondents. Unfortunately, this aspect cannot be controlled because cross checking was not possible for the first-born probands based on the limited questions included in the original study [37], [71]. Notably, the Blanchard and Lippa showed that whereas the fraternal birth order effect was highly complex and difficult to observe, the significantly higher number of overall offspring of the mothers of homosexuals, in accordance with the balancing selection hypotheses, was indeed evident [37].

X-chromosome-linked Sexually Antagonistic Selection Versus Autosomal Overdominance

A further question that naturally arises is why other studies have not observed the increased female fecundity of the maternal line. King et al. (2005) and Schwartz et al. (2010) observed a generalized increase in fecundity [19], [32]. However King et al. provided some questionable results because the study did not analyze the specific fecundity of each gender class, but rather, described family size; thus, it is difficult to interpret the individual fecundity contribution to the increased family size [32]. Schwartz and colleagues [19] conducted a study that replicated the study of King and colleagues [32] with a larger sample collected from festivals and gay pride events in the U.S. However, because the results substantially replicated those of the previous study [32], the same questions emerged regarding their pedigree. As Riger et al. noted [41], the heterosexual sample that Schwartz et al. used showed a surprisingly low fecundity compared with similar samples, which might indicate sampling problems. Moreover, the collection method of Schwartz et al. was unclear with respect to consideration for the relative age of the relatives and the systematic age differences between the paternal and maternal relatives, as this information is relevant with respect to the considerably large 6-year difference between the homosexual and heterosexual samples in the Schwartz et al. study, which is unexplained through a demographic comparison specifically designed to assess the differential fecundity between probands and their relatives. The application of ANCOVA for this sample would reveal the previously discussed weaknesses of the covariance analyses controlling for age [63], [64], [66], [67]. Notwithstanding, the results of this study were enlightening. Although these authors argued against a prevalence of male homosexuality in the maternal line, the reanalyzed results from their published data (Table 2 [19]) showed a 5.07% enrichment of homosexuality in the maternal line versus a 2.61% enrichment in the paternal line (X² = 72, df 1, p<0.0001), consistent with the sexually antagonistic hypothesis. This outcome can be derived by calculating the ratio of homosexuals among the maternal relatives sharing the X-chromosome with the proband (i.e., brothers, maternal uncles, and maternal cousins from aunts) and the ratio of homosexuals in the proband classes that did not share the X-chromosome (i.e., fathers, grandfathers, paternal uncles, paternal cousins of both aunts and uncles and maternal cousins from maternal uncles) from their published table.

Indeed, the data presented in Tables 2 and 3 by Schwartz et al. suggests a maternal increase in fecundity [19]. In Table 3, the presented fecundities are aggregated between males and females making it difficult to determine an X-chromosome selective fecundity effect. However, we were able to calculate the number of maternal aunts in the Schwartz et al. sample using the number of maternal uncles published in Table 2. By subtracting maternal uncles from the aggregated number of maternal uncles and aunts, the number of maternal aunts was determined as 1080. Using the same procedure, the number of paternal aunts was determined as 1065. Subsequently, from the published number of male cousins from maternal (n = 852) and paternal aunts (n = 696) published in Table 2 and considering that in human populations, the ratio of male live births to female live births is close to 106:100 [72], [73], we estimated the maternal aunt fecundity was calculated as 1.532 (1655/1080) offspring produced on average. In contrast, the fecundity of paternal aunts was 1.26 (1352/1065) offspring produced. Thus, these differences were highly significant (X² = 56.2, df 1, p<0.0001). The same comparative analysis of the fecundity of the aunts of the heterosexual probands showed no significant difference (X² = 2.6, df 1, p NS). Based on these calculations, the data of Schwartz et al. supports the hypothesis of sexually antagonistic selection based on empirical evidence [19]. Thus, we suggest that these data are not compatible with the overdominance hypothesis. Unfortunately, it was not possible to further test these hypotheses using the data presented in the tables from this study.

Sexually Antagonistic Selection as a General Genetic Mechanism to Balance Male Homosexuality in our Species

The maternal line fecundity effects observed in the present study have been previously demonstrated in European (Italian, Spanish and French) samples [11], [18], [21], [35] in a large BBC Internet sampling of primarily Anglo-Saxons [37] and Samoan populations [38], [39], [74] in the United States (if our re-analysis of the previous data [19] is correct) and in an English Caucasian sample [36]. However, the Rahman study [36] is also widely cited because of its non-white sub-sample, where fecundity was shown to be higher in heterosexual probands. Based on these results, many authors suggested that fecundity asymmetries involving increased maternal line female fecundity are a local variation present only in Western or Caucasian homosexuals [19], [34], [35], [75]. The different white versus non-white patterns of fecundity suggested that it might be important to consider ancestry [19]. However, in this study a sample of non-white homosexuals is described, with non-trivial confusion, as comprising 20 probands, with mothers producing 88 offspring, exhibiting a mean fecundity of 4.4 offspring, which was much lower than that of non-white heterosexuals (53 proband, with mothers producing 731 offspring), exhibiting a surprisingly high average fecundity of 13.79 offspring (previously unrecorded and incomparable, even with maternal aunts of the same sample; Table 4 in Rahman et al. [36]). However, in Table 6, the same authors presented data from a sibling sex composition of non-white probands, and we could re-calculate maternal fecundity, summing the frequency of all sibling classes, including the respective number of probands, to obtain an average fecundity of 5.2 for mothers of non-white heterosexuals and a fecundity of 3 for mothers of non-white homosexuals, which is not consistent with the average published Table 4 and establishes further confusion. We conclude that until the data on additional non-Western populations are properly examined, the maternal line female fecundity increase should be accepted as a general occurrence irrespective of ethnicity.

Limitations

This paper attempts to explain the persistence of genetic factors influencing male homosexuality through a Darwinian paradox, thereby giving the impression that sexual orientation is exclusively determined through genetic factors, which is not true. In a previous study using pedigrees and questionnaires, it has been suggested that genetic and biological influences on sexual orientation could explain approximately 20% of the variance in sexual orientation [11], suggesting that genetic factors do not determine sexual orientation in males; thus, important research has been conducted to understand the social and developmental factors that influence the remaining variation. [16].

Sampling homosexual or heterosexual probands and deriving population demographic information could generate some biases, such as recalling ages of specific subjects, which should be considered. Accessing a ‘hidden’ population is difficult because no sampling framework exists, and the public acknowledgement of membership could potentially be prejudicial to the subjects or a fraction thereof. Thus, standard probabilistic sampling methods produce low response rates and unreliable responses [76]. Targeted sampling is a widely employed method for accessing hidden populations, which is preferred to snowball sampling [77] and other examples of chain referral. These procedures introduce well-documented biases [78]. More generally, estimates referring to population units different from the survey units (i.e., the population of the grandmothers or aunts of the interviewed subjects) were not corrected through weighing the units with the inverse of their probability of selection in the sample (homosexuals), as previously suggested [78], [79], reflecting the fact that these estimates were compared with similarly collected samples (heterosexuals). The same methodology was used in the same locations to reproduce potential biases and increase the internal validity of the comparisons. We considered these biases to have limited relevance in this study, and our procedures were unlikely associated with such biases, as only demographic variables (such as the number and age of grandparents, cousins or aunts and uncles) were assessed.

This study did not collect data on male fecundity. It is believed that male fecundity it is not relevant in this analysis for two reasons. First, according to general Darwinian theory, fecundity-related decisions are biologically more inherent to females than males, and the variables influencing the variance of reproductive success are different in males than in females. Second, paternity is much harder to assess than maternity. However, we cannot exclude the possibility that the inclusion of accurate male reproductive success data in a future study might lead to different conclusions.

Additional data, particularly data obtained from non-industrialized high fecundity societies, are needed to further confirm that sexually antagonistic selection is a widespread phenomenon explaining the male homosexuality conundrum. These data should focus on fecundity patterns in which no confusion with the fraternal birth order effect can occur, such as among grandmothers and aunts of probands, and to a lesser extent, the number of younger siblings of homosexual probands. The aim of this population genetics study was to understand the Darwinian paradox of the genetic component of male homosexuality, and no effort has been made to identify these genetic factors. Thus, we acknowledge that currently no genes influencing homosexuality have been identified, neither on the X chromosome nor on autosomes [15]; however, in light of the present results, it would be more promising to determine the genetic factors influencing fecundity in females.

Conclusions

The results of this study suggest that in homosexual males, the significant increase in the fecundity of the maternal line is not a by-product or an artifact of the fraternal birth order effect, but rather, is a coexisting factor potentially influencing homosexuality in males. In the families of homosexual individuals, the fecundity of maternal grandmothers and aunts significantly increases compared with the families of heterosexual individuals. This selective increase in the maternal line fecundity is sufficient to increase the total female fecundity of a family with homosexual individuals, even if the fecundity of the paternal females does not increase.

Regarding the genetic contribution to fecundity, we suggest that sexually antagonistic selection, rather than overdominance, is the balancing selection that potentially functions through X-linked genetic factors to compensate for the reduced fecundity of male homosexuals via increased maternal line female fecundity. The confirmation of these findings through future studies would resolve the long-standing Darwinian paradox associated with male homosexuality.