Archives of Sexual Behavior, Vol. 34, No. 1, February 2005, pp. 117–122 (C ?2005)
Family Size in White Gay and Heterosexual Men
Michael King, M.D.,1,3John Green, Ph.D.,2David P. J. Osborn, Ph.D.,1Jamie Arkell,
M.R.C. Psych.,1Jacqueline Hetherton, M.Sc.,2and Elizabeth Pereira, M.Sc.2
Received December 9, 2003; revisions received May 25, 2004 and August 4, 2004; accepted August 7, 2004
There is some evidence for a genetic influence on sexual orientation. However, gay men have
fewer children than heterosexual men. Increased fecundity in the biological relatives of gay men
could offset this selection pressure. We measured family size in gay (n = 301) and heterosexual
(n = 404) men, attending clinics for sexually transmitted infections. The main outcome measure
was the number of each man’s uncles and aunts, first cousins, siblings, nephews and nieces,
and his own children. With the exception of the participants’ own offspring, mean family size
of each category of relatives was significantly larger for gay men (paternal and maternal total
OR = 1.02, CI = 1.01–1.03). This remained the case after adjustment for other predictors of family
size (paternal and maternal total OR = 1.02, CI = 1.00–1.03). We found increased fecundity in the
relatives of gay men and this is one explanation of how a genetic influence might persist in spite
of reduced reproductive fitness in the gay phenotype. There are, however, a number of alternative
explanations for our finding, including unknown psychological and social factors, which might
mediate the association between family size and sexual orientation.
KEY WORDS: homosexuality; family size.
How sexual orientation in men and women is
of factors have been reported in recent years, but two in
particular are supported by considerable evidence. The
first is that gay men are significantly more likely to come
higher in the birth order in families, by which is meant
that they are likely to have greater numbers of older
siblings. This effect has been found mainly for number
of older brothers(Cantor,Blanchard, Paterson, &Bogaert
2003). Only one study to date has reported that gay men
also have greater numbers than expected of older sisters
(Bogaert, 1998). Blanchard (2001) has estimated that the
population attributable “risk” for homosexual orientation
1Department of Mental Health Sciences, Royal Free and University
College Medical School, London, United Kingdom.
2Department of Clinical Psychology, Paterson Centre, London, United
Health Sciences, Royal Free and University College Medical School,
Rowland Hill Street, London NW3 2PF, United Kingdom; e-mail:
due to older brothers is 24% for gay men with one older
brother, 43% for gay men with two older brothers, and
over 50% for those with three or more older brothers.
Explanations offered for this consistent finding is the
possibility of a maternal immune response by mothers to
possible sexual contact with their older brothers (Jones &
Blanchard, 1998). Less rigid parenting of younger sons
which enabled them to admit to their sexuality must also
remain a possibility.
The second main line of research is genetic. Male
homosexuality has higher concordance in monozygotic
than dizygotic twins, suggesting a possible genetic in-
fluence on sexual orientation (Bailey & Pillard, 1991;
Kendler, Thornton, Gilman, & Kessler, 2000; Pillard
& Bailey, 1998). However, gay men have far fewer
children than heterosexual men (Bell & Weinberg, 1978)
and thus selection pressure against any genetic influence
would be considerable. One suggestion to explain this
anomaly is that of a polygenetic trait in men that results
in a personality more conducive to reproducing and
supporting offspring (Miller, 2000). Although this trait
in larger dose may produce male homosexuality (Miller,
0004-0002/05/0200-0117/0 C ?2005 Springer Science+Business Media, Inc.
118 King, Green, Osborn, Arkell, Hetherton, and Pereira
2000), greater reproductive fitness or fecundity in the
biological relatives of gay than heterosexual men would
offset the selection pressure against homosexuality. Our
objective was to test the hypothesis that gay men have
larger families than heterosexual men.
We asked consecutive male attenders to two central
London clinics for sexually transmitted infections (STI)
to complete anonymous questionnaires. We did not pay
the 1061 men who participated. Constraints imposed by
the ethical committees approving the study meant that we
were unable to collect any information on non-responders
and thus the percentage of men who did not participate is
Measures and Procedure
After providing demographic information, partic-
ipants were asked whether they identified themselves
as “homosexual/gay, bisexual, or heterosexual/straight.”
They were then asked to complete three 7-point Likert
scale(Kinsey, Pomeroy, & Martin, 1948). Participants
then gave information on numbers of individual family
members, allowing us to calculate family size for each
each man’s uncles and aunts, first cousins, siblings,
nephews and nieces, and his own children. Parents and
participants were not included since these values are
constant for each man. Participants were asked to specify
the numbers of younger and older siblings and their
gender. The study was passed by the research ethics
committees for each site.
Multiple response variables such as occupation and
religion were examined for their distribution by sexual
orientation and family size. Categories similar on these
measures were amalgamated, to facilitate the analysis
and the interpretation of results. Occupations were coded
according to five social classes prior to data entry (Office
of Population Censuses and Surveys, 1990).
A total of 1061 men reported whether they were
gay, bisexual or heterosexual as well as completing
Kinsey-type scales of sexual orientation. Self-identified
sexual orientation was checked against the Kinsey scales
to ensure that men identifying as gay or heterosexual
were not actually reporting marked sexual attraction or
experience which differed from their reported orientation.
In fact, Kinsey scale responses wholly concorded with
the reported sexual orientation. A total of 24 men were
adopted and 15 did not answer the adoption question.
since information about their biological family was likely
to be unavailable or less reliable.
Seven of the 1022 remaining respondents did not
and were excluded. We explored whether cultural differ-
and to family size might make comparisons across these
groups invalid. We grouped ethnicity as white (white UK,
white Irish, and white “other”) and non-white (Black
Chinese, Mixed, and other). Of the 294 non-white men,
(55.3%) of the 721 white men (χ2(1) = 26.9, p < .001).
Of the 294 non-white men, 63 (21.4%) had missing
occupation data compared to 72 (10.0%) of the 721 white
men (χ2(1) = 23.7, p < .001). Family size also varied
family size in non-whites was 35.0 while that in whites
was 19.9 (t = 9.27, p < .0001). Because of these
findings and missing data, we decided not to include the
294 non-white men in the main analyses of associations
with sexual orientation. Of the remaining 721 men, 16
self-identified as bisexual and scored between 2 and 4 on
the Kinsey sexual attraction, sexual history, and sexual
dreams scales. Since these bisexual men may differ from
both homosexual and heterosexual men, they were also
excluded from the analysis, leaving a sample of 705 men.
10, of whom seven were aged under five, suggesting
that these generations were largely complete. However,
the low number of children of participant’s siblings
suggests that they were unlikely to have completed their
own families when they participated in the study. Of
the 11 ethnic categories included in the questionnaire,
only White-Irish, White-British, and White-other ethnic
groups contained more than 20 respondents and at least
30% gay men. Thus, to reduce variance in family size
attributable to culture in the parents’ and participant’s
generation, our analysis concentrated on these three
Participants were less likely to complete the ques-
(e.g., providing number of siblings but omitting their
children). However, for the most complex variables, there
was no significant difference between gay (30.2%) and
heterosexual men (31.4%) failing to complete all required
information (χ2(1) < 1, ns).
Family Size in Gay Men 119
Table I. Sociodemographic Characteristics and Clinic Attendance
Age (in yrs)
College and higher
Note. Numbers vary due to missing data.
The gay men were older and more likely to be unem-
ployed, Protestant, and in non-manual occupations than
the heterosexual men (Table I). Fewer adopted men (8/23,
34.8%) self-identified as heterosexual than non-adopted
men (619/1016; 60.9%) (χ2(1) = 6.4, p = .01). Sexual
orientation was also analyzed by Kinsey scales, and we
were able to redefine bisexuals as those scoring 2–4 on
were heterosexual compared with 630/994 (63.5%) non-
adopted men (χ2(1) = 5.7, p = .02).
Gay men had more older brothers and older sisters
than heterosexual men, a difference that remained sig-
nificant after adjustment for age, occupational class, and
religion (Table II).
In order to explore the sibship composition more
fully and to test whether the finding for older sisters
was not merely due to a correlation between number
of older brothers and older sisters, we undertook a
logistic regression analysis in which number of younger
sisters, younger brothers, older sisters, and older brothers
were entered into the regression as predictors of sexual
orientation, and those that were not significant were
removed in a step wise procedure (Table III). While
number of older brothers and older sisters independently
predicted a gay sexual orientation, younger brothers and
younger sisters did not.
Family Size Analysis
Associations between sexual orientation, sociode-
mographic variables, and family size were used to guide
a logistic regression analysis. Odds ratios (OR) for being
homosexual were firstly adjusted for age and clinic site
in case of any cluster effect. Other variables significant
in the univariate analyses were then added to the model
and the effect on associations with sexual orientation was
noted. Continuous variables, such as age and family size,
were added into the model in their continuous form but as
a check on the influence of outliers continuous variables
were subsequently entered into the model re-coded as
deciles and as square roots.
mean family size of each category of relatives was
consistently larger for gay men (Table II). The OR
linking family size to self-identification as gay was
also consistently larger. There were several possible
than heterosexual men to have attended one particular
clinic, work in non-manual occupations, and be from
Protestant backgrounds. Gay men were also older than
we used a form of mean sibling age, taking the mean age
of the participant and up to his eldest four siblings. This
aggregate age measure was a more powerful confounder
A limit of the four eldest siblings was set to minimize
the number of participants excluded due to missing
After adjustment, total family size remained signif-
icantly larger for gay men only on the paternal or on
the paternal plus maternal, sides (Table II). This was
true for family sizes only at the parents’ and partici-
pant’s generations (completed generations) or when the
offspring/nieces and nephews generation was included.
Although the trends were also for larger family sizes on
the maternal side alone, they were not significant after
Our main results remained robust after transforming
continuous variables with large outliers (e.g., siblings)
into their deciles or square roots. This ensured that
individual participants with very large families were not
we conducted a multiple linear regression of sexual
120 King, Green, Osborn, Arkell, Hetherton, and Pereira
Table II. Associations of Family Size with Sexual Orientation
Paternal aunts & uncles
Maternal aunts and uncles
Paternal ‘Complete generation’c
Maternal ‘Complete generation’c
Pat + Mat ‘Complete generation’c
Nieces & nephews
Paternal + Maternal totald
.019 1.12 (1.03–1.22).012
.074 1.03 (1.00–1.07).045
.074 1.01 (.96–1.12)
.002 1.31 (1.03–1.65).026
.034 1.02 (1.00–1.04)
.0171.02 (1.00–1.03) .025
0.47 (0.34–0.65) <.001 0.31 (0.21–0.47)
.0011.02 (1.00–1.03) .029
aOdd ratios for identifying as homosexual, p value for Wald test. OR represents the increase in odds per additional family member.
bAge variable = mean age of (proband + up to four eldest sibs) (see text).
cCompleted generation = aunts + uncles + cousins + siblings, but excludes the offspring and nephews/nieces generation who may not
be complete yet.
dTotals = (Mat/Pat) aunts + (Mat/Pat) uncles + (Mat/Pat) cousins + siblings + nieces & nephews + offspring; Mat = maternal. Pat =
Family Size in Gay Men 121
Table III. Analysis of Sibships as Predictors of a Homosexual
Steps Siblings OR
CI of OR
Step 1 Older brothers
orientation on total family size, as a continuous variable.
Once again, sexual orientation significantly predicted
family size after adjustment for confounders.
Our main finding was that gay men were members
of significantly larger families than heterosexual men. We
also found that gay men had more older brothers and
sisters than heterosexual men and that men who reported
This was a relatively large sample of gay and
study in STI clinics provides an unrepresentative sample
heterosexual and homosexual men. Therefore, although
selection bias in our sample remains a possibility, it is
difficult to see how it might work. All the men were
clinic attenders and it seems implausible that gay men
with bigger families systematically attend STI clinics
more than those with smaller families. The gay men were
more likely to be in non-manual occupations and from
Protestant backgrounds, factors that might predict lower
family sizes. However, these confounders were adjusted
for our findings.
We do not know what proportion of men who today
might define themselves as gay conducted heterosexual
relationships and fathered children over past millennia.
However, our results must be viewed in the context of the
past 50 years, an era in Western society in which gay men
argument might be put forward that selection pressures in
the past against homoerotic orientation were non-existent
because in the past men with little or no attraction to
women married and were equal in fecundity to men who
this although we are unaware of any evidence to support
such a proposition.
An excess of older brothers is a consistent finding
in gay men(Cantor et al., 2002), but the excess of older
sisters is novel. There was no imbalance in gender in any
category of relatives across the three generations, making
an X-linked, sex-ratio altering genetic effect unlikely
(Bailey et al., 1999; Turner, 1995). The birth order effect
for gay males has been interpreted as a possible mater-
nal immune reaction against Y-linked histocompatibility
antigens (Blanchard, 2001) that results in homosexuality,
feminization, and lower birth weight (Blanchard, 2001;
Blanchard et al., 2002). However, it is difficult to see
how this might be so if our finding about older sisters is
between familial reproductive success and a behavioral
phenotype in man. This study does not prove that there is
findings we report leave such a possibility open. If there
were no such effect, then it would be hard to explain how
a heritable component to a homosexual orientation could
persist. It would also be hard to explain why, even if there
orientation have not evolved given the markedly reduced
fecundity of gay men. The existence of genes which
may reduce the individual’s reproductive success but
increase that of their close relatives has been invoked
to explain, for instance, “altruistic” behaviors (Hamilton,
1964; Wilson, 1975). If a genetic influence on sexual
orientation is associated with increased reproductive
fecundity in the biological relatives of gay men, this
suggests how a genetic influence might persist in spite
of reduced reproductive fitness in the gay phenotype.
One can speculate that inheritance of a number of genes
that affect personality, nurturing, and masculine behavior
make heterosexual males that carry such alleles more
attractive to females and more successful fathers (Miller,
2000). How our findings on adoption fit into this picture
is less clear since giving up children for adoption might
be regarded as one indication of poorer skills in bringing
children to adulthood.
or social reasons for gay men being members of larger
families. Indeed, if the explanation is genetic, it is
striking that the effect is measurable over only three
complete generations. Weak effects would only be seen
over many generations. Whatever the underlying cause of
this association, our findings are a novel addition to the
literature regarding sexual orientation and families.
122 King, Green, Osborn, Arkell, Hetherton, and Pereira
We are grateful to the participating staff and patients
in the Marlborough Clinic, Royal Free Hospital, London
in particular, Dr. Daniel Ivens and Dr. David Goldmeier
for their advice and help. JA was funded as a research
fellow by the Priory Hospital, North London. DPJO was
funded by a Medical Research Council clinical training
Bailey, J. M., & Pillard, R. C. (1991). A genetic study of male
sexual orientation. Archives of General Psychiatry, 48, 1089–
Bailey, J. M., Pillard, R. C., Dawood, K., Miller, M. B., Farrer, L. A.,
Trivedi, S., et al. (1999). A family history study of male sexual
orientation using three independent samples. Behavior Genetics,
Bell, A. P., & Weinberg, M. S. (1978). Homosexualities: A study of
diversity among men and women. New York: Simon & Schuster.
Blanchard, R. (2001). Fraternal birth order and the maternal immune
hypothesis of male homosexuality. Hormones and Behavior, 40,
Blanchard, R., Zucker, K. J., Cavacas, A., Allin, S., Bradley, S. J., &
Schachter, D. C. (2002). Fraternal birth order and birth weight in
probably prehomosexual feminine boys. Hormones and Behavior,
Bogaert, A. F. (1998). Birth order and sibling sex ratio in homosexual
and heterosexual non-white men. Archives of Sexual Behavior, 27,
Cantor, J. M., Blanchard, R., Paterson, A. D., & Bogaert, A. F. (2002).
How many gay men owe their sexual orientation to fraternal birth
order? Archives of Sexual Behavior, 31, 63–71.
Hamilton, W. (1964). The genetical evolution of social behavior: I.
Journal of Theoretical Biology, 7, 1–16.
Jones, M. B., & Blanchard, R. (1998). Birth order and male homosexu-
ality: An extension of Slater’s index. Human Biology, 70, 775–787.
Kendler, K. S., Thornton, L. M., Gilman, S. E., & Kessler, R. C. (2000).
Sexual orientation in a US national sample of twin and non-twin
sibling pairs. American Journal of Psychiatry, 157, 1843–1846.
Kinsey, A. C., Pomeroy, W. B., & Martin, C. E. (1948). Sexual behavior
in the human male. Philadelphia: W. B. Saunders.
Miller, E. M. (2000). Homosexuality, birth order, and evolution toward
an equilibrium reproductive economics of homosexuality. Archives
of Sexual Behavior, 29, 1–34.
Office of Population Censuses and Surveys. (1990). Standard occupa-
tional classification. London: Her Majesty’s Stationery Office.
Pillard, R. C., & Bailey, J. M. (1998). Human sexual orientation has a
heritable component. Human Biology, 70, 347–365.
Turner, W. J. (1995). Homosexuality type 1: An Xq28 phenomenon.
Archives of Sexual Behavior, 24, 109–134.
Harvard University Press.