William H James.  Journal of Biosocial Science. Volume 38, Issue 6, November 2006.

Introduction

In discussing the causes of male homosexuality and paedophilia, many recent workers have been sceptical in regard to the possibility of postnatal psychosocial determinants.

For instance, Blanchard (2004, p. 177) wrote in regard to male homosexuality: ‘Although this hypothesis may seem intuitively plausible, there are little (sic) empirical data to recommend it . On the contrary, the available data argue against such an explanation’. In a similar vein, Quinsey (2003, p. 110) wrote on the causes of paedophilia: ‘Childhood sexual experience has been an intuitively appealing and popular etiological candidate but can be rejected on several grounds’. Instead, both these authors espoused the ‘maternal immune hypothesis’. This proposes that some mothers are progressively immunized to male-specific antigens by their male fetuses and that there are increasing effects of anti-male antibodies on the brain of each succeeding male fetus, thus, in the words of Blanchard (2004, p. 179) ‘somehow divert[ing] the sexual differentiation of the fetal brain from the male-typical pathway so that the individual will later be attracted to men, rather than women’.

It is not my present purpose to argue that this hypothesis is false. Rather it is suggested that the evidence for it is not as strong as might be supposed from the presentations of its proponents. Further, it will be argued that the evidence for the postnatal experience hypothesis is not as weak as is suggested by the authors above. It should be emphasized that both hypotheses may, in principal, (to limited extents) be true.

It is generally agreed that the causes of male homosexuality are multifactorial: so, for instance, there can be little reasonable doubt that genetics (putative gene(s) on the X-chromosome) has some minor role to play in the aetiology of this condition (Hamer et al., 1993; Hu et al., 1995; Rahman & Wilson, 2003). This conclusion was strengthened by the findings of King et al. (2005) and Camperio-Ciani et al. (2004) that relatives of male homosexuals have more children than comparable relatives of male heterosexuals. However, as these latter authors note: ‘We emphasise that over 79% of the variance in male sexual orientation in our sample remains unaccounted for by an excess of maternal homosexual kin and number of older brothers’. This uncertainty concerning the aetiology of male homosexuality may be one reason why its potential determinants have inspired unusual degrees of partisanship in their proponents. Here an attempt is made to assess the evidence relating to the two hypotheses mentioned above (viz. of maternal immunization and of postnatal experience) remembering that logically, they are not mutually exclusive.

Evidence relating to the maternal immune hypothesis

This hypothesis was devised before the discovery of the ‘fraternal birth order’ effect in male homosexuality and gained plausibility from that discovery. This phenomenon is such that for each older brother, a man’s chance of being (or becoming) homosexual is raised by perhaps more than 30% (Bogaert, 2003). Blanchard (2004) and others likened the hypothesis to that of Gualtieri & Hicks (1985): the hypothesis of ‘selective male affliction’. In an earlier response to this suggestion, I noted that propositions on sex ratio offered by Gualtieri & Hicks (1985) in support of their hypothesis were almost certainly false (James, 2004a). However that, of itself, does not falsify their hypothesis, and since Blanchard (2004) continues to use their hypothesis as an illustrative model for his own, I shall first elaborate on the grounds for viewing the Gualtieri & Hicks (1985) hypothesis with reserve.

The nomenclature ‘selective male affliction’ is, I suggest, misleading. Males are selected against by some forms of affliction and not others. For instance, Arena & Smith (1978) found that 40 of 52 single localized defects of human morphogenesis showed a non-random statistically significant predilection for one sex or the other. These authors (and others e.g. Lubinsky, 1997) reported male excesses in cleft lip (with or without cleft palate), but a female excess in cleft palate only: there were male excesses in eight different cardiovascular malformations, and female excesses in three others; female excesses in four forms of central nervous system malformation (including anencephaly and spina bifida); and female excesses in several skeletal malformations including congenital dislocation of the hip. My point is that it seems not helpful to invoke an overall greater antigenicity of male fetuses when attempting to devise a general explanation for anomalies, many of which are female-biased. As far as I know, the hypothesis of Gualtieri & Hicks (1985) is not now generally accepted as the explanation for any neurodevelopmental disorder. So it seems quixotic to invoke it in support of a new hypothesis.

It may be acknowledged that in searching for possible prenatal causes of homosexuality, we are engaged in the quest for atypical influences on the fetal brain (rather than on some of the other structures considered above). Blanchard (2004, p. 174) wrote that one of the purposes of his paper was to ‘elaborate the maternal immune hypothesis beyond previous statements of it’. So I assume that that paper adduces all the evidence for the hypothesis that anti-male antibodies of maternal origin might influence sexual differentiation in the male fetal brain. Here I shall try to address each piece of that evidence.

Experimental evidence

Blanchard (2004) cited evidence that blood serum (a potential source of antibodies) from mothers of children with autism or dyslexia, when injected into pregnant female mice, produced behavioural deficits in the mouse offspring (Dalton et al., 2003). Serum from control mothers had no such effect. However, one has no assurance that the effect was mediated by anti-male antibodies. Blanchard (2004) also cited three papers on the effects of immunizing maternal rodents with paternal antigens (Saji et al., 1980; Lu & Dawson, 1986; Gentile et al., 1992). These reports conflict on the effects on placental and fetal size and on abortion rate: some found enhancement, others an inhibitory effect. Lastly, there has been one study (which Blanchard indicated was the only one) on maternal immunization and filial sexual behaviour (Singh & Verma, 1987). These authors found that male mice whose mothers were immunized to H-Y prior to pregnancy were much less likely to mate successfully with receptive females. (But we are not told whether these male mice tried to mount, or were receptive to, other males.) So, taken in toto, though these experimental studies may seem weakly supportive of Blanchard’s hypothesis, they cannot be construed as decisively demonstrating the operation of anti-male antibodies.

However, some female mice do reject male skin grafts and in doing so produce male-specific antibody that distinguishes male from female cells defining serologically detected male antigen (SDMA) (Goldberg et al., 1991). So the question arises: what is known about the properties of these antibodies?

In view of this uncertainty, it seemed worth consulting Medline and the World of Knowledge SCI and SSCI databases. The search term ‘anti-male antibody’ elicited three ‘hits’ from Medline (since 1951) and two from WOK since 1945. Of these five hits, four related to papers invoking the concept in explanation for male sexual orientation or behaviour. In short, anti-male antibodies are not a well-described phenomenon in immunology, so their use as explanatory tools in behavioural science seems premature.

In particular, since the hypothesized antigens have not been specified, mothers of homosexual men have not been tested for the presence of anti-male antibodies. Yet that would seem a reasonable minimal criterion for acceptance of the hypothesis. Blanchard’s hypothesis is more speculative than his presentation would seem to imply.

Epidemiological evidence

The sex ratios of sibs of probands with behavioural and brain anomalies. Blanchard (2004) cited evidence that boys with learning disability or mental retardation had excesses of older brothers (Ackerman et al., 1988; Flannery & Liederman, 1994). But he did not cite the subsequent study of Liederman & Flannery (1995) which failed to detect a difference between the sib sex ratios of 2080 neurodevelopmentally disordered children and of 11,213 controls. Nor did he mention the evidence for excesses of brothers among the sibs of male stutterers, and of children with convulsions and of children with benign focal sharp waves; and for excesses of sisters among the sibs of bulimics and female stutterers (for references, see James, 1998). In short, I can discern no persuasive evidence for the maternal immune hypothesis in the sex ratios of sibs of probands with various forms of neurodevelopmental and behavioural disorder. A worker who wished to invoke these male excesses to favour such a hypothesis has the burden of explaining the female excesses too.

Birth weight and sexual orientation of males by the sexes of their previous sibs. Blanchard & Ellis (2001) reported three findings: (1) heterosexual males with older brothers weighed less at birth than heterosexual males with older sisters; (2) homosexual males with older brothers weighed less than heterosexual males with older brothers; and (3) homosexual and heterosexual males with no older siblings, or older sisters only, did not differ in birth weight.

The first of these findings had already been reported twice (Trotnow et al., 1976; Magnus et al., 1985) and has subsequently been confirmed (Cote et al., 2003). Blanchard (2004) also cited confirmatory evidence for the second and third findings. So let us provisionally accept all three and consider the possible explanations. Blanchard (2004) interpreted the comparatively low weight of male infants with older brothers as evidence for his hypothesized maternal immune process. However, there is an alternative explanation. It has been shown that the administration of testosterone to pregnant sheep lowers the birth weight of their offspring of both sexes (Manikkam et al., 2004). (Sheep are frequently used in experimental reproductive biology as models for the human reproductive system.) Moreover, there is very good evidence that high maternal testosterone levels are associated with the production of male offspring (James, 1996, 2004b). So a woman with two or more sons would be expected to have higher testosterone levels than a woman ascertained by a similar-sized family comprised of children of both sexes. Thus the first and third finding above are susceptible to a hormonal explanation. Indeed, preceding male births (like testosterone administration) are reportedly associated with lower birth weights in subsequent infants of both sexes (Trotnow et al., 1976; Magnus et al., 1985; though see Blanchard & Ellis, 2001). So the hormonal explanation seems preferable; and indeed, the experiment on sheep was interpreted by its authors as a model for intrauterine growth retardation.

The second finding reported above is that homosexuals with older brothers weigh less at birth than do control heterosexuals. This finding may be explained by the suggestion that these homosexuals (or some of them) were chosen for sexual purposes (by their older brothers or parents or paedophiles) because they were petite. The ground for supposing that such choice occurs is as follows. Boys who show early effeminate behaviour (‘sissy-boys’) are highly likely to become homosexual (Bailey & Zucker, 1995). But it is reported that ‘sissy-boys’ are ‘prettier’ than controls (Zucker et al., 1993). Presumably this is because they were ‘chosen’ (consciously or otherwise) by other people. (It is noteworthy that the same word ‘pretty’ was used by Fry (1998) to describe boys who were chosen for sexual purposes by older boys at the English Public School, Uppingham, in the 1970s (James, 2004c).) So unless ‘prettiness’ is associated with an inborn tendency to homosexuality, one may infer that male homosexuality has a learned or acquired component.

The sexual orientation of boys with older homosexual brothers. Green (2000) examined fourteen homosexual male-to-female transsexuals who had at least one older brother and at least one younger brother. He found that 21 of the 22 younger brothers were heterosexual. Green argued that the maternal immune hypothesis implies that the younger brothers of a homosexual male should have a higher probability of homosexuality, and that therefore the hypothesis was impugned. Blanchard (2004, pp. 179-181) countered this argument by claiming, in effect, that his hypothesized process is stochastic. That being so, it seems reasonable to ask what Blanchard would admit as negative evidence. What conceivable facts would count as falsifying the hypothesis?

Thus Blanchard (2004) seems to have cited no compelling experimental or epidemiological evidence that anti-male antibodies have an effect of the sort hypothesized. There is evidence that they—or other agents—may affect birth weight and viability, and some tenuous evidence that they adversely affect the brains of experimental animals. But the only reason for invoking anti-male antibodies in this context is the powerful evidence for a fraternal birth order effect. However, external explanations for this birth order effect already exist (as will later be described). That being so, there seems no immediate need to invoke maternal immune processes.

The range of phenomena associated with fraternal birth order and implications for the maternal immune hypothesis

There can be no reasonable doubt that a fraternal birth order effect exists in regard to male homosexuality (Blanchard, 2004). However, recent research has also uncovered such an effect in homosexual male-to-female transsexuals (Green, 2000; Poasa et al., 2004); in (presumably pre-homosexual) boys with gender identity disorder (Zucker et al., 1997); in homosexual paedophiles (Bogaert et al., 1997; Blanchard & Bogaert, 1998; Blanchard et al., 2000); and bisexual paedophiles (Bogaert et al., 1997) but not heterosexual paedophiles (Bogaert et al., 1997; Blanchard & Bogaert, 1998; Blanchard et al., 2000). These findings were interpreted by Blanchard et al. (2000) as consistent with the hypothesis that fraternal birth order was associated with homosexuality but not paedophilia. However, others have reported fraternal birth order effects in sex offenders expressing deviant sexual preferences (viz. for rape and/or paedophilia; Lalumiere et al., 1998; Cote et al., 2002); and in mentally disordered offenders showing deviant (as defined above) sexual behaviour (MacCulloch et al., 2004). Lastly, fraternal birth order effects have been found in self-reported rapists (Lalumiere et al., 2000) and in men who showed sexual arousal to depictions of non-sexual violence (Cote et al., 2002). But MacCulloch et al. (2004) were unable to find a birth order effect in mentally disordered men who had committed non-sexual violent offences. Thus, as these latter authors conclude (p. 469), the available empirical evidence suggests a fraternal birth order effect in ‘homosexuality, paedophilia and sexual preference for rape’, though these authors felt that the evidence regarding ‘paedophilic preference is inconclusive’.

My reason for introducing these further topics is that most of these authors dealing with the causes of rape and paedophilia invoke maternal immunization by male fetuses as if it were an established phenomenon that maternal anti-male antibodies have specific effects on the developing fetal brain. Yet as far as I know (as noted above) no experiments have been specifically designed to test the maternal immune hypothesis—let alone shown that it is substantially true. Neither is it clear that fraternal birth order effects in rape and paedophilia (if they exist) would add to the evidence that immune processes underlie them all. Instead, one might wonder whether some male sexual transgressors were hormonally fuelled opportunists. This suggestion is strengthened by the conclusions of Smallbone & Wortley (2004) who had a sample of 221 adult males convicted of sexual offences against children. These authors found that most of these men’s previous convictions (86%) were for non-sexual offences. These authors wrote: ‘The results are better explained by a general theory of crime than by traditional clinical conceptions linking sexual offences with sexual psychopathology’. Similar findings were offered by Parkinson et al. (2004) who wrote: ‘Any theory concerning the dynamics of sex offending against children needs to account for the levels of non-sex offences committed by child molesters’.

Again it is worth emphasizing that since there is almost certainly a multiplicity of causes of paedophilia, the process of establishing one such cause does not, ipso facto, falsify another.

Summary of the evidence relating to the maternal immune hypothesis

Blanchard (2004, p. 182) wrote: ‘There is, in summary, a variety of antigens or potential antigens that might play the role required by the maternal immune hypothesis. At the present time, however, there is no evidence that any specific one of them actually plays it’. It seems fair to add that indeed, there is no strong evidence that maternal immune processes are involved in homosexuality or paedophilia at all.

Of men who prefer sex with adults, the percentage who prefer it with men is of the order of 2-5% (Wellings et al., 1994, pp. 185-190). This value is partly dependent on definition; however the figure used here is merely intended to be illustrative and is not essential to the present argument.

Among male paedophiles, about 25% are exclusively or primarily interested in boys (Blanchard et al., 2000).

Fagan et al. (2002) reported a review based on all Medline references to paedophilia since 1965. These authors failed to find any published estimate of the prevalence of paedophilia in the general population. However, Bagley et al. (1994) ascertained eight apparently active paedophiles in a stratified random sample of 750 males aged 18-27 in Calgary, Canada, using a computer response system that assured anonymity. Thus the percentage of paedophiles of this age in that society may very roughly be estimated at 1%. That value is taken here.

Evidence relating to the postnatal learning hypothesis

It is convenient to consider paedophilia and homosexuality jointly here because there is strong evidence that the age and sex of men’s sexual targets interact. The point is illustrated in Table 1, which gives hypothetical percentages of men by the age and sex of their preferred sexual partner. As will be seen, among male paedophiles, about 25% are exclusively or primarily interested in boys (Blanchard et al., 2000). This is about ten times higher than the proportion of homosexuals among men whose sexual interests are directed to adults. This contrast stands in need of explanation. So it has been proposed that male homosexual orientation and paedophilia share a common cause, viz. sexual or quasi-sexual experience at an early age. It cannot be emphasized too strongly that it is not suggested here that such experience is a necessary (let alone sufficient) precondition for either homosexuality or paedophilia.

The reported rates of child sexual abuse in cases and controls

Paedophilia. Quinsey (2003) wrote: ‘The evidence supporting the early sexual experiences hypothesis is very weak, mostly consisting of the retrospective reports of identified offenders contrasted with inadequate comparison groups (Garland & Dougher, 1990)’. This assessment by Quinsey seems idiosyncratic in view of the evidence that has been published since his chosen citation (Freund & Kuban, 1994; Bagley et al., 1994; Howitt, 1995; Worling, 1995; Glasser et al., 2001; Cohen et al., 2002; Fagan et al., 2002; Lee et al., 2002; Salter et al., 2003). These authors are unanimous in reporting that child sexual abuse (CSA) constitutes a significant risk factor for the abused male child to become a paedophile. One of these studies was based on a stratified random sample of 750 men (Bagley et al., 1994), one study was longitudinal (Salter et al., 2003) and one was a review based on all the Medline references to paedophilia since 1965 (Fagan et al., 2002). It may be acknowledged that the majority of men who were sexually abused in childhood do not later engage in paedophilic behaviour (Salter et al., 2003): but the risk that they do so is reportedly significantly higher than in non-abused controls. And it is difficult to resist the notion that there is a (weak) causal relationship here. The grounds for this claim will later be outlined in reference to such correlational studies relating child sexual abuse to subsequent homosexuality. Of course, it is usually possible to find methodological imperfections in such correlational studies. However, the weaknesses alleged by Quinsey (2003) seem not as pervasive as his words suggest. Indeed, it would be useful if he were to attempt to substantiate them in a detailed analysis of the nine cited studies.

Male homosexuality. If the postnatal learning hypothesis were correct, this would be seen as having profound consequences for society. This is so because—even in the absence of criminal behaviour—the (male) children of lesbian or gay parents might be expected to have an enhanced probability of becoming homosexual. And if that were established, some legislatures would probably be minded to reverse current legislation supporting gay and lesbian parenting. Passions in the United States run high on this issue, and it is worth citing evidence for this. Without wishing to offend either, it is possible to apply the terms ‘liberal’ and ‘conservative’ to the two wings of the present debate. As representative conservatives, Stacey & Biblarz (2001) reviewed 21 studies and wrote: ‘Most research in psychology concludes that there are no differences in developmental outcomes between children raised by lesbigay (sic) parents and those raised by heterosexual parents. The analysis here challenges this defensive conceptual framework and analyses how heterosexism has hampered intellectual progress in this field’. These authors refer to L. D. Wardle, a Brigham Young University law professor who, in a paper in the University of Illinois Law Review ‘impugns the motives, methods and merits of social science research on lesbian and gay parenting. He charged the legal profession and social scientists with an ideological basis favouring gay rights that has compromised most research in this field and the liberal, judicial and policy decisions it has informed’.

However, it is not only the ‘conservatives’ who have complained of US governance in this respect. Rind et al. (1998) published a meta-analysis of 59 studies and concluded that the assumed harmfulness of child sexual abuse had been overstated. In July 1999, the US Congress passed a formal resolution condemning this article. Rind et al. (2000) complained that the resolution followed ‘months of attacks by social conservatives and by mental health professionals specializing either in curing homosexuality or in treating patients by inducing them to recover memories of child sexual abuse’. It is unfortunate that Rind et al. (1998) paid no explicit attention to the question of whether CSA generates homosexuality and/or paedophilia in some of its victims.

With warnings like these, it is necessary to tread warily. So, at the risk of seeming long-winded, I shall first consider each of the three studies that Blanchard (2004) marshalled to justify his conclusion (p. 177) that the ‘available data argue against’ the postnatal learning hypothesis. I shall later consider some data that he did not cite.

The data cited by Blanchard (2004)

• Blanchard (2004) reiterated his claim that the postnatal learning hypothesis is impugned by the data of Wellings et al. (1994) on the proportion of homosexuals among the past pupils of all-male British boarding schools. In James (2004a) I gave grounds for suggesting that ‘these data point in the direction opposite to that suggested, but in any case are far too flimsy to bear any interpretation of the sort put on them both by their authors, and by Ellis & Blanchard (2001)’.
• Blanchard (2004) also cited Dawood et al. (2000) who studied 37 pairs of male homosexual brothers. These authors reported that the large majority of them knew about their own homosexual feelings before they learned about their brothers’ feelings. The authors inferred that fraternal interaction was not an important cause of male homosexuality. The point may be admitted. But the study lacked the power to detect what must—if it exists—only be a weak effect. An absence of evidence is not evidence of absence: before an absence can reasonably be inferred, a power analysis should be offered. Moreover, sibships ascertained by two homosexual brothers must be regarded as markers for established genetic causes, and so less likely to be informative on environmental ones.
• Blanchard (2004) cited Bogaert (2000) who studied a probability sample of 1511 men. They answered questions on birth order, and on whether they had learned about sex from a brother, and whether they had been sexually touched by a brother. The beta coefficients from a weighted least-squares multiple regression analysis were presented: that for ‘birth order’ was highly significant (confirming previous studies). Those for ‘learned about sex from brother’ and ‘older brother sexually touched’ were not significant. However, the quantity of data underlying these findings bears scrutiny. Only 50 of these men reported that they were mainly or only attracted to men; and only six that they had been sexually touched by an older brother. I could find no indication of how many of those six were included in the 50.

My point may be understood as follows: we want to know whether boys who have been sexually touched are more likely to be or become homosexual. How many such subjects do we need in adulthood to test that proposition? Suppose that we want to stand eight chances in ten of detecting significance in a one-way test at the 0^05 level that sexual touching has the effect of doubling an assumed homosexuality rate of 5% (viz. from 5% to 10%). Then standard power-analytic methods suggest that 341 subjects are needed in each of two equal groups (those who were touched in childhood and those who were not; Snedecor & Cochran, 1967). Yet the authors above based their conclusion on six subjects!

As Blanchard correctly writes, these authors ‘found no evidence that homo­sexuality was induced by incestuous sexual relations between brothers’. However the data form no firm basis for the denial that such relations occasionally do cause homosexuality. The data are simply too exiguous to allow detection of even a moderate-sized effect, let alone one a weak one.

The data not cited by Blanchard (2004)

As I noted in my earlier pieces (James, 2004a, in press), there is very strong evidence that male homosexuals report higher rates of child sexual abuse (however defined) than male heterosexuals (Finkelhor, 1979, 1984; Johnson & Shrier, 1987; Wyre, 1990; Cunningham et al., 1994; Glasser et al., 2001; Rind, 2001; Garcia et al., 2002). To these citations may be added Arreola et al. (2005), Beitchman et al. (1991), Jinich et al. (1998), Laumann et al. (1993), Lenderking et al. (1997), Paul et al. (2001) and Tomeo et al. (2001). It should be acknowledged that the reported associations between the sexual abuse of boys and subsequent adult homosexuality do not (in the words of Bartholomew et al. 1994) ‘warrant attributions of causality as they are correlational in nature and confounded by sampling methodology. Antecedent dispositions such as effeminate behaviour, lack of secondary sex characteristics, or lack of peer and familial support during sexual identity development may be more reasonable explanations for the reported associations between sexual abuse and homosexual identity because such dispositions may prompt gay youth to seek contact in environments where the likelihood of sexual abuse increased’. However, it is one thing to identify methodological traps: and quite another to conclude that each of the above studies fell into one or other of these traps. So I want to outline some points suggesting that part, at least, of the reported association is causally driven.

1. A consideration of desirability scores suggests that the high self-reported prevalence of CSA by paedophiles has not been faked (Freund & Kuban, 1994). Moreover Lee et al. (2002) tried to establish which of several related risk factors (childhood emotional abuse, childhood behaviour problems and CSA) were related to each of several types of sexual offending (paedophilia, exhibitionism, rape). Both these sets of authors reported that CSA was a specific risk factor for paedophilia. The inference of a causal relationship between CSA and paedophilia is strengthened by the demonstration that the other associated factors (emotional abuse and behaviour problems) were not so closely associated with paedophilia.
2. The evidence (noted above) that ‘sissy-boys’ are prettier than controls (Zucker et al., 1993) argues strongly that (some of them) were ‘chosen’ rather ‘born’, and that therefore there is a learned component in some cases of homosexuality.
3. The evidence (see Table 1) for interaction between the sex and age of men’s sexual targets suggests that they share a cause. This poses no difficulty for proponents of the postnatal learning hypothesis. (Both paedophiles and homosexuals report high levels of CSA). However, proponents of the immune hypothesis have no alternative than to burden their hypothesized as-yet-unidentified anti-male antibodies with the further function of (sometimes) affecting the age, as well as the sex, of men’s preferred sexual targets.

The sex ratios of sibs of male homosexuals

The sex ratios (proportions male) of sibs of male homosexuals have been reported to be significantly high in some samples (Jensch, 1941; Lang, 1960; Suarez & Przybeck, 1980; Blanchard & Sheridan, 1992; Blanchard et al., 1995, 1996; Bogaert, 1998) though not all (Blanchard & Bogaert, 1994; Blanchard & Zucker, 1994; Zucker & Blanchard, 1994).

Both the maternal immune hypothesis and the postnatal learning hypothesis imply: (a) a high sex ratio in the older sibs of homosexuals and (b) a fraternal birth order effect. So it is interesting to consider how the samples in which these effects are found differ from those in which they are not. A high sib sex ratio has been particularly noted in respect of feminine homosexuals (Jensch, 1941; Zucker & Blanchard, 1994; Blanchard et al., 1995, 1996) and of gender dysphorics (Blanchard & Sheridan, 1992; Cohen-Kettenis et al., 1994) but not among the Kinsey interviewees (Blanchard & Bogaert, 1994). So it is tempting to suggest that feminine homosexuals in particular have excesses of older brothers because these brothers were attracted by their appearance, and interacted with them sexually. I suggest that, like the pre-homosexual ‘sissy-boys’, their appearance attracts sexual attention from others.

The sex ratios of half-sibs of male homosexuals

Lang (1960) reported on the sex ratios of half-sibs of male homosexuals. He noted that in both his own data and those of Jensch (1941) the sex ratios of the paternal half-sibs were high, and the sex ratios of the maternal half-sibs were low. Taken jointly the data are highly significant. Eliot Slater was mystified by these data so he showed them to me and I devised an explanation invoking reporting bias (James, 1971). However, in the present context it would be interesting to know whether that explanation is false, and whether the data may be taken at face value. The point could be tested by examining the sex ratios of maternal and paternal half-sibs of control heterosexual men.

This would be well worth doing because the data on the sex ratio of paternal half-sibs would seemingly constitute a test between the maternal immune hypothesis and the postnatal learning hypothesis. The former would presumably predict a normal sex ratio and the latter a high sex ratio. Meanwhile it may be acknowledged that neither hypothesis readily deals with the data on maternal half-sibs. In short, these data on the sex ratios of the half-sibs of male homosexuals urgently need replication.

Suggestions for future research

This note summarizes the evidence relating to two hypotheses on the aetiology of male homosexuality and paedophilia. I am merely a retired armchair epidemiologist who has never worked in this field, so it may seem impertinent to offer advice to those so engaged. However, I will take this risk.

1. It seems virtually certain that male homosexuality is of multifactorial origin. As noted above, there are established (probably weak) genetic causes; and it seems plausible that pre- and postnatal hormone levels may also play a part (e.g. Alias, 2004; James, in press). Lastly, as argued above, it seems highly likely that postnatal psychosocial factors also are involved in the development of male homosexuality. That being so, it is desirable that proponents of these various hypotheses should collaborate, rather than wrangle, with one another. As I understand it, the problem is to establish the relative weights of their chosen factors, and to ascertain whether these factors interact.

2. It has been argued above that proponents of the maternal immune hypothesis have not yet provided strong direct evidence in support of their hypothesis. It would be useful if they were to collaborate with immunologists in the attempt to devise a decisive test of their hypothesis.

3. I have noted elsewhere (James, in press) that, even at this late date, research on the aetiology of male homosexuality may usefully start with taxonomic considera­tions. Initially, at least, men should be categorized in as diverse a manner as possible. This follows from the facts that we do not have answers to the following questions: (a) Is it scientifically more fruitful to classify men by their actions or by their self-descriptions (or by some combination of these criteria)? So should the category of ‘bisexuals’ contain only those few men who so describe themselves, or should it be expanded to include those who (regardless of self-description) have (with an agreed frequency) had sex with members of both sexes? Indeed, is there an important aetiological distinction between these two categories of men? (b) May homosexual men (however defined) be usefully categorized into ‘active’ and ‘passive’? (Analogous to the apparently valid butch/femme distinction in lesbians.) (c) Do mean measure­ments on bisexual men (however defined) lie, on the average, at the heterosexual end of continua, rather than between those of control homosexuals and heterosexuals (James, in press)? Or is this true only of some parameters? If so, why?

4. There should be further research on the fraternal birth order effect, and the sibling sex ratio of male homosexuals. These two effects are related—but the exact relationship is not established. To what extent is the one due to the other? Moreover, it is important to ascertain whether the strengths of these two effects vary as between different categories of homosexual or, indeed whether there are exceptions to them. In particular, are there such effects with respect to bisexual men?

5. The causes of male and female homosexuality are not identical (Bailey et al., 2000), but they may overlap: so workers in the one field should be familiar with work in the other.