Content uploaded by Yin Xu
Author content
All content in this area was uploaded by Yin Xu on Feb 21, 2019
Content may be subject to copyright.
Developmental Psychology
Early Life Conditions and Adolescent Sexual Orientation:
A Prospective Birth Cohort Study
Yin Xu, Sam Norton, and Qazi Rahman
Online First Publication, February 14, 2019. http://dx.doi.org/10.1037/dev0000704
CITATION
Xu, Y., Norton, S., & Rahman, Q. (2019, February 14). Early Life Conditions and Adolescent Sexual
Orientation: A Prospective Birth Cohort Study. Developmental Psychology. Advance online
publication. http://dx.doi.org/10.1037/dev0000704
Early Life Conditions and Adolescent Sexual Orientation:
A Prospective Birth Cohort Study
Yin Xu, Sam Norton, and Qazi Rahman
King’s College London
This study tested the association between multiple prenatal and postnatal early life factors and adolescent
sexual orientation in a longitudinal birth cohort. Factors included birth weight, gestational age, parental
age at birth, number of older brothers and sisters, breastfeeding, maternal anxiety/depression, family
socioeconomic position, parent– child relationships, parental absences, pubertal body mass index, and
housing issues. We used data on 5,007 youth from the Avon Longitudinal Study of Parents and Children
(ALSPAC). Sexual orientation was assessed using a 5-point scale of sexual attraction at 15.5 years. Early
life factors were separated into three developmental periods: prenatal (n⫽9), before 7 years (n⫽5), and
after 7 years (n⫽5). We controlled for childhood gender nonconformity, handedness, and digit ratio as
markers of prenatal androgen exposure. Gender nonconformity was strongly associated with later male
and female nonheterosexuality, and higher right-hand digit ratio was associated with later male nonhet-
erosexuality. Boys with low birth weight and shorter breastfeeding duration were more likely to have a
later nonheterosexual orientation. Boys with parental absence before 7 years of age were more likely to
be nonheterosexual, but this effect disappeared after entering all early life history factors. Parental
absence since birth, low prenatal family socioeconomic position, and poorer parent– child relationship
were associated with later nonheterosexuality among girls. The results are discussed in the context of a
life history framework for understanding human sexual orientation development in males and females.
Keywords: life history, sexual orientation, birth weight, gender nonconformity, ALSPAC
Supplemental materials: http://dx.doi.org/10.1037/dev0000704.supp
Human sexual orientation is likely to be multifactorial in its
origins. Biological and psychosocial factors may influence the life
course of sexual orientation (affecting sexual attractions, identity,
and sexual behaviors differently; Bailey et al., 2016). This is
consistent with multifactorial influences on other developmental
traits such as personality. However, scholars often imply multiple
influences on sexual orientation but rarely, if ever, test them.
Research suggests that genetic factors explains approximately one
third of the variation in sexual orientation (Bailey et al., 2016).
Thus, most of the differences between people in their sexual
orientation are due to environmental factors (often nonshared)
pointing to multiple etiology. Causal pathways are rarely tested in
prior research because of the use of cross-sectional designs, sur-
veys of adult heterosexuals and nonheterosexuals which are sus-
ceptible to reporting biases, and problems accounting for depen-
dencies in data (Bailey et al., 2016). Few data sets lend themselves
to testing a range of biological and psychosocial variables. We
therefore examined the role of multiple prenatal and postnatal
early life factors on the development of adolescent sexual orien-
tation in a longitudinal, birth cohort from England.
Life History Theory
Life history theory aims to explain the diversity in life trajec-
tories of organisms, especially in their reproductive histories (Del
Giudice, Gangestad, & Kaplan, 2015). It can be conceived of as a
“biopsychosocial” model, integrating genetic and nongenetic (e.g.,
socialization, environment, and learning) processes in accounting
for variation in development and reproduction. In humans, this
framework has focused on the influence of early life conditions
(those characterized by high mortality and morbidity, environmen-
tal unpredictability, low parental investment, and resource scar-
city) on sexual behavior traits across the life span. Prior research
suggests that humans will adopt “fast” or “slow” sexual life history
strategies according to the early environment they experience
(Ellis, 2004). The fastness or slowness is often defined by the age
at which reproductive or sexual behaviors begin. For example,
Yin Xu, Sam Norton, and Qazi Rahman, Department of Psychology,
Institute of Psychiatry Psychology & Neuroscience, King’s College Lon-
don.
The authors are extremely grateful to all the families who took part in
this study, the midwives for their help in recruiting them; and the whole
ALSPAC team, which includes interviewers, computer and laboratory
technicians, clerical workers, research scientists, volunteers, managers,
receptionists and nurses. The U.K. Medical Research Council and the
Wellcome Trust (Grant ref: 102215/2/13/2) and the University of Bristol
provide core support for ALSPAC. Yin Xu is supported by a King’s-China
Scholarship.
Correspondence concerning this article should be addressed to Yin Xu or
Qazi Rahman, Department of Psychology, Institute of Psychiatry Psychol-
ogy & Neuroscience, King’s College London, 5
th
floor Bermondsey Wing,
Guys Hospital Campus, London SE1 9RT, United Kingdom. E-mail:
yin.xu@kcl.ac.uk or qazi.rahman@kcl.ac.uk
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Developmental Psychology
© 2019 American Psychological Association 2019, Vol. 1, No. 999, 000
0012-1649/19/$12.00 http://dx.doi.org/10.1037/dev0000704
1
studies have reported associations between low birth weight, or
other indicators of reduced parental investment during the early
years (such as father absence) and accelerated pubertal maturation
among human females (Quinlan, 2003). Various explanatory mod-
els propose that differences in early life experiences may influence
subsequent sexual behavior by constraining an individual’s growth
or by acting as forecasts of ecological conditions (Ellis, 2004).
This makes responding to variable early life conditions with
changes in sexual behavior a facultative, adaptive response. Al-
though much research has focused on female sexual life history,
one meta-analysis revealed that low family socioeconomic posi-
tion (SEP), maternal absence, and paternal absence were associ-
ated with early sexual debut, early first birth, and early marriage
among men (Xu, Norton, & Rahman, 2018).
Life History Theory and Sexual Orientation
Life history theory is generally silent on the trait of sexual
orientation. We have argued previously that life history approaches
are relevant to sexual orientation (Xu et al., 2018). We hypothesize
that heterosexuality is an optimal and adaptive (in fitness terms)
life history strategy given the species-typical pattern of sexual
interests in humans is toward the opposite sex. Nonheterosexuality
represents a departure from that generally fitness-enhancing pat-
tern of sexual behavior. Nonheterosexuality (homosexuality/bisex-
uality) may represent a “fast” life history or be a byproduct of
other life history traits. Epidemiological research suggests that
people with infrequent homosexual behavior and attractions are
more common than those with substantial or exclusive homosex-
uality (Gates, 2011). This may permit some direct reproduction to
occur because heterosexual contacts still dominate in the former
group and help transmit alleles related to nonheterosexuality in a
population. A broader phenotype of “nonheterosexuality” may
comprise reproductive and nonreproductive sexual contacts.
A life history approach also has the potential to integrate bio-
logical and psychosocial accounts of the development of human
sexual orientation. The dominant theories regarding sexual orien-
tation development are biological. For example, prenatal androgen
theory proposes that high prenatal androgen exposure during crit-
ical periods may be related to heterosexuality in men and homo-
sexuality in women, whereas low androgen exposure may be
associated with homosexuality in men and heterosexuality in
women (Ellis & Ames, 1987). Another model, the maternal im-
munity hypothesis, proposes that carrying successive male fetuses
triggers a maternal immune response that feminizes brain devel-
opment and results in nonheterosexuality in later born sons
(Blanchard, 2018; Bogaert et al., 2018).
Twin studies have shown that sexual orientation has a modest
heritability (about 30%; Alanko et al., 2010). However, the twin
concordance rate for sharing the same sexual orientation is esti-
mated as 24% across men and women (Bailey et al., 2016). A large
proportion of monozygotic twins who are nonheterosexual have
heterosexual cotwins (approximately 76%). As these twins share
the same genotype, prenatal and early postnatal environment fac-
tors must play a large role in their different sexual orientations.
The environmental influences appear nonshared rather than shared
(Bailey et al., 2016). We have previously proposed that early life
conditions suggested by life history theory may constitute one
source of these nonshared factors in the development of male
sexual orientation (Xu et al., 2018). They may also interact with
the proposed biological mechanisms. For example, prenatal early
life factors (e.g., poor maternal condition) may alter hormonal
processes during fetal development, which then bias subsequent
brain sex differentiation affecting circuitry related to sexual ori-
entation (Rahman, 2005). Alternatively, early hormonal processes
may themselves constitute a third factor which links early life
factors and later sexual orientation. Several candidate early life
factors may be important in sexual orientation development.
Birth Weight
Variation in birth weight may serve as an indicator of adverse
early environments or maternal condition and predict later sexual
life history strategy. Low birth weight is associated with earlier
first pregnancy in women (and mediated by SEP; Nettle, Coall, &
Dickins, 2011). In general, low birth weight offspring of both
sexes may attract less parental investment. Birth weight may also
be associated with male sexual orientation and interact with sibling
sex composition. Specifically, it is suggested that maternal im-
mune response triggered by carrying successive male fetuses re-
sults in nonheterosexuality and low birth weight in later born sons
(Blanchard & Ellis, 2001), but studies are inconsistent. One study
reported that homosexual men with older brothers had lower
birthweights than their heterosexual counterparts (Blanchard &
Ellis, 2001). Another study found no difference in birthweights
between feminine and control boys with fewer than two older
brothers (Blanchard et al., 2002). Blanchard (2012) also proposed
that mothers of firstborn homosexual sons were more likely to
develop an immune response to a fetus, which may lower their
birth weight (see also Bogaert et al., 2018). Again, results are
inconsistent (Skorska, Blanchard, VanderLaan, Zucker, & Bo-
gaert, 2017). It is also not clear why mothers of firstborn homo-
sexual men without younger siblings would be more likely to
develop an immune response.
Older Brothers
The number of older brothers is robustly associated with male
sexual orientation (Blanchard, 2018). Homosexual men are usually
born after their siblings, and this late position in birth order is
associated with the number of older brothers but not with the
number of older sisters, younger sisters, or younger brothers
(Blanchard, 2018). This phenomenon has been termed the fraternal
birth order effect or FBO. Here, each male fetus the mother carries
appears to increase the odds of homosexuality in later born males
by about 33% (see Blanchard, 2018). It has been hypothesized that
carrying successive male fetuses triggers a maternal immune re-
sponse (possibly toward Y-linked minor histocompatibility anti-
gens from male fetuses), and this feminizes the brain development
of later born males increasing the odds of expressing a homosexual
orientation (Bogaert et al., 2018). However, FBO cannot explain
the sexual orientation of all nonheterosexual men and studies thus
far are cross-sectional.
Parental Involvement and Psychosocial Factors
Low parental investment characterized by parental absence,
shorter breastfeeding duration, and other forms of psychosocial
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
2XU, NORTON, AND RAHMAN
hardships appear to be associated with fast reproductive life his-
tories in women and men (Nettle et al., 2011; Xu et al., 2018). For
example, paternal absence before 5 years of age is associated with
earlier menarche in girls (Quinlan, 2003 cf. Sear, Sheppard, &
Coall, 2018; Sohn, 2017). Again, few studies are longitudinal in
design and associations are generally weak. Although the sugges-
tion of psychosocial or family influences in reproductive life
history is widely studied, its role in sexual orientation is socially
and scientifically controversial. The history of research in this area
has tended to see psychosocial explanations as more stigmatizing
of nonheterosexuality than biological ones (Bailey et al., 2016).
However, theories of causation may not be directly relevant to
specific social and moral positions. Monozygotic twin differences
clearly point to nonshared environmental influences. This contrib-
utes to the complexity of sexual orientation development, com-
prising both biological and nonbiological influences. The role of
social factors is likely to be small based on the existing evidence,
with few studies of good methodological quality (Bailey et al.,
2016). Importantly, the role of environmental factors is hypothe-
sized to be stronger in female than male sexual orientation given
its greater malleability. Women report more bisexuality, report
greater change in sexual identity and behavior patterns over time,
and have a sexual orientation that is less category-specific com-
pared to men (Bailey et al., 2016). The hypothesis of greater
environmental involvement in female sexual orientation is thus
uncontroversial.
Proposed psychosocial factors have predominantly included
parent– child relationships. Poorer parent– child relationships may
be indicators of low parental investment and thus be associated
with nonheterosexuality. Bell, Weinberg, and Hammersmith
(1981) found weak correlations between retrospective ratings of
parent– child relationship traits and sexual orientation, but this
effect became either nonsignificant or weak when childhood gen-
der nonconformity was controlled. This highlights important prob-
lems in testing psychosocial factors. Nonheterosexuality may pre-
cede poor parent– child relationship. First as prehomosexual
children are likely to be gender nonconforming, this may nega-
tively influence relationships with parents (Kane, 2006). Homo-
sexual men may also show elevated rates of separation anxiety in
childhood and this could also place negative pressure on parent–
child relationships (VanderLaan, Gothreau, Bartlett, & Vasey,
2011). The parents of prehomosexual children may also differ
from parents of preheterosexual children in ways that influence
parent– child relations. Strong evidence against parental psychos-
ocial influences comes from studies of children raised by nonhet-
erosexual parents. These show no differences in sexual orientation
between children reared by nonheterosexual parents and those
reared by heterosexual parents (see Bailey et al., 2016). However,
these studies rely on small convenience samples and may be
confounded by third factors, such as genetic parental effects. There
is no robust work examining early familial SEP or other forms of
parental investment on the development of sexual orientation to
our knowledge. Nonheterosexual youth are more likely to experi-
ence early disruption such as homelessness but it not clear whether
this is related to family environment or other social factors such as
discrimination and stigma (Gattis, 2009). Studies using longitudi-
nal designs may be better able to test the causal role of these and
confounding factors (such as childhood gender nonconformity).
Other Early Life Factors
Prior studies have found a range of other indicators of adverse
early life conditions including preterm birth, having older parents,
shorter duration of breastfeeding, and maternal anxiety/depression
are associated with faster sexual life history strategies (James,
Ellis, Schlomer, & Garber, 2012; Wehkalampi et al., 2011). Such
factors may also be associated with later sexual orientation but are
poorly studied. Often, they interact with each other making asso-
ciations with outcomes difficult to disentangle. Older maternal age
has been reported to be associated with nonheterosexuality (Xu et
al., 2018) or homosexual marriage in men (Frisch & Hviid, 2006),
but results are inconsistent (e.g., Blanchard & Bogaert, 1996).
Developmental Sensitive Periods
Several developmental stages may be important sensitive peri-
ods for the influences of early life conditions and their impact upon
later sexuality. Life history scholars have suggested that the first 5
or 7 years of life may be the most important (Simpson, Griskevi-
cius, Kuo, Sung, & Collins, 2012). This may be due to prepubertal
hormonal changes including adrenarche, the 5- to-7-year-old shift
in cognition, language, and social skills; the responsivity of devel-
oping neural systems to adverse environments; and increased
self-sufficiency during this time (Ellis, 2004). Middle childhood (6
to 11 years) may also be important, especially because sex differ-
ences in physiology (strength and muscularity), social behavior,
and aggression intensify during this stage (Del Giudice, 2014).
Biological models of sexual orientation development (e.g., pre-
natal androgen) suggest that prenatal periods are critical (Rahman,
2005). However, there may be more than one critical period for
males, and more sensitive periods for females, during which sex
hormones act (McCarthy, Herold, & Stockman, 2018). Psychoso-
cial approaches to sexual orientation are silent on which develop-
mental periods are important. This lack of specificity is a weakness
of social explanations of sexual orientation. Given the role of
developmental stages in life history and in models of sexual
orientation, the present study investigated early life factors in
prenatal, postnatal before 7 years, and postnatal after 7 years
periods.
Childhood Gender Nonconforming Behavior and
Prenatal Hormonal Exposure
Three important developmental correlates of sexual orientation
require comment here. Nonheterosexual men and women report
more childhood gender-nonconforming behavior (GNCB), on av-
erage, than heterosexual adults. This pattern has been found in
both prospective and retrospective studies (Bailey & Zucker, 1995;
Li, Kung, & Hines, 2017). Digit ratio (2D:4D) is a marker ascribed
to the actions of prenatal androgens. This ratio shows a moderate
sex difference (smaller in men than women; Hönekopp & Watson,
2010). Women with congenital adrenal hyperplasia, a condition
characterized by high prenatal androgen exposure, exhibit smaller
(more masculine) digit ratios (Brown, Hines, Fane, & Breedlove,
2002). Nonheterosexual women have more masculine digit ratios
than heterosexual women but there is no difference in digit ratios
between heterosexual and nonheterosexual men (Grimbos, Da-
wood, Burriss, Zucker, & Puts, 2010). Finally, handedness is a
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
3
EARLY LIFE CONDITIONS AND SEXUAL ORIENTATION
somatic trait that appears robustly associated with sexual orienta-
tion. Nonheterosexual men and women are significantly more
likely to be nonright-handed than heterosexual men and women
(Lalumière, Blanchard, & Zucker, 2000).
The role of these variables in the life history framework is
speculative but offer plausible hypotheses. Early life factors could
promote GNCB which then cascades into later nonheterosexuality
(because early life factors are causally closer to GNCB than to
sexual orientation). Or GNCB, handedness, and digit ratio could
have independent influences on both early life conditions and
sexual orientation. Early life conditions may alter prenatal andro-
gen exposure, and then influence later sexual orientation. Or,
GNCB could act as a behavioral proxy for a common underlying
mechanism, such as prenatal androgen exposure (with handedness
and digit ratio acting as more direct markers; Skorska & Bogaert,
2017). As prenatal androgen theory is the dominant model of
sexual orientation development, the present study examined these
three indicators of androgen exposure as potential covariates.
The Present Study
Here we use data from a prospective birth cohort in England to
test whether a range of early life factors (in prenatal, pre-7 years of
age, and post-7 years developmental stages) were associated with
adolescent sexual orientation at age 15.5 years. It appears to be
appropriate to begin measuring sexual orientation at 15.5 years
old. Studies also show that men and women recall first having
feelings of sexual attraction at approximately age 10, on average
(McClintock & Herdt, 1996). One study reported a mean age of
self-reported first awareness of same-sex attraction at approxi-
mately 15 years (Calzo, Antonucci, Mays, & Cochran, 2011).
Some studies report even earlier recalled mean age of awareness of
same-sex attractions (e.g., Floyd & Bakeman, 2006). Changes in
reported sexual orientation identity were also found to occur at a
similar rate throughout adolescence and into emerging adulthood
(Ott, Corliss, Wypij, Rosario, & Austin, 2011). This is the first
study of its kind and includes a range of early life factors never
studied before in relation to sexual orientation. The longitudinal
design will permit better tests of causal pathways. Here we test the
extent to which prenatal factors and postnatal early life factors
predict later sexual orientation in boys and girls separately. We
hypothesized that prenatal early life factors (e.g., low birth weight)
would be associated with nonheterosexuality in both boys and
girls, and postnatal early life factors (e.g., poorer parent– child
relationship) would be associated with nonheterosexuality in girls
since their sexual orientation is more socially influenced. We
further controlled for the influences of GNCB, handedness, and
digit ratio in our analyses.
Method
Participants
Participants were part of the Avon Longitudinal Study of Par-
ents and Children (ALSPAC). All pregnant women with an ex-
pected date of delivery between April 1, 1991 and December 31,
1992 in the Bristol area of the South West of the United Kingdom
were eligible and invited to attend the ALSPAC. The initial sample
recruited 14,541 (71.81% of the eligible sample) pregnant women
who delivered 14,062 live-born children and 13,988 were alive at
1 year. Additional recruitment attempting to bolster the original
sample with eligible cases who had failed to join the study at the
beginning resulted in 15,458 fetuses with data collected from the
age of 7 onward. Of this total sample of 15,458 fetuses, 14,775
were live births and 14,701 were alive at 1 year of age. Fifty-nine
percent of the cohort attended the “Teen Focus” sessions and have
been followed four times between the age of 12.5 years old and 17
years old. For more details, see Boyd et al. (2013). The study
website contains details of all the data, which are available through
a searchable data dictionary: http://www.bris.ac.uk/alspac/researc
hers/data-access/data-dictionary/. Ethical approval for the study
was obtained from the ALSPAC Law and Ethics Committee and
the Local Research Ethics Committees, and King’s College Lon-
don Psychiatry, Nursing & Midwifery research ethics subcommit-
tee (protocol reference number: LRS-16/17– 4194; “Testing a Life
History Approach to the Study of Variation in Human Sexual
Orientation”). We analyzed ALSPAC data reported by parents and
children across different time points. Adolescents who reported a
valid response of sexual orientation and sexual behavior (see
Supplemental Text S1 in the online supplemental material) at 15.5
years old were included here, N⫽5,007 (2,349 boys and 2,658
girls).
Measures
Sexual orientation. At 15.5 years old, adolescents were re-
quired to answer the question: “Please choose the description that
best fits how you think about yourself” on a 5-point Kinsey-like
scale, ranging from 1 (100% heterosexual),2(mostly heterosexual
but also attracted to the same sex), 3 (bisexual), 4 (mostly homo-
sexual but also attracted to the opposite sex),5(100% homosex-
ual),6(not sexually attracted to either sex), and 7 (not sure). This
was done via computer to promote disclosure of sensitive personal
information. Adolescents who chose “not sexually attracted to
either sex” (n⫽17) or “not sure” (n⫽91) were excluded from the
analyses. This is because we had no a priori predictions about the
role of early life conditions in adolescents with ambiguous or no
sexual attractions and many such adolescents identify as hetero-
sexual later in life (Ott et al., 2011; Savin-Williams & Joyner,
2014). Such 5-point scales of sexual attractions have been used in
large studies of adolescents (Austin et al., 2009; Ott et al., 2011;
Remafedi, Resnick, Blum, & Harris, 1992; Saewyc, Skay, Bear-
inger, Blum, & Resnick, 1998). Saewyc et al. (1998) also pilot
tested their items with a youth sample before full-scale implemen-
tation. The 5-point measures show good stability (i.e., test–retest
over 2-year intervals) in adolescents (Ott et al., 2011), expected
associations with sex of sexual partners among adolescents (Re-
mafedi et al., 1992; Saewyc et al., 1998), and low nonresponse
rates compared with measures of other components of adolescent
sexual orientation (Saewyc et al., 2004). As bisexuals may differ
from gay/lesbian individuals in some components of sexual orien-
tation, we treated them as separate groups. Accordingly, adoles-
cents who chose 100% heterosexual or mostly heterosexual but
also attracted to the same sex were coded as heterosexual, those
who chose bisexual were coded as bisexual, and those who chose
mostly homosexual but also attracted to the opposite sex or 100%
homosexual were coded as homosexual. As a result, 2,290 hetero-
sexual boys (45.73%), 29 bisexual boys (0.58%), 30 homosexual
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
4XU, NORTON, AND RAHMAN
boys (0.60%), 2,585 heterosexual girls (51.63%), 56 bisexual girls
(1.12%), and 17 homosexual girls (0.34%) were included.
Body size. Birthweights (kg) were taken from birth notifica-
tion and/or obstetric data and/or recorded by ALSPAC measurers
in the delivery room. For data recorded by more than one method,
we used the following criteria. If birth weight values from each
measurement method were identical, then the value was accepted;
if disagreement between birth weight values from different mea-
surement methods were less than 100 g, then the lowest value was
accepted; if disagreement between birth weight values from dif-
ferent measurement methods were greater than 100 g, then the
value was coded as missing (no observations in our sample had
disagreement greater than 100 g). At 14 years and 7 months,
adolescents gave their height and weight, to generate body mass
index.
Gestational age. This was based on the date of mother’s last
menstrual period, pediatric or obstetric assessment, and ultrasound
assessment. Adolescents were categorized into three gestational
age groups: preterm birth (⬍37 weeks’ gestational age), term birth
(37– 41 weeks’ gestational age), and postterm birth (⬎41 weeks’
gestational age; Savitz et al., 2002).
Parental age. Maternal age was recorded as the age at the last
menstrual period. When adolescents were at 12 weeks’ gestation,
the partner of the mother was required to report whether he was the
father of the child. If he reported “yes,” his age at completion of
questionnaire was coded as paternal age; otherwise, paternal age
was coded as missing.
Maternal anxiety and depression. When adolescents were
18 weeks’ gestational age and 8 weeks old, two subscales of
Crown-Crisp Experiential Index (CCEI) were used to measure
maternal anxiety and depression (Ross & Hafner, 1990). These
have acceptable reliability and validity (Alderman, Mackay, Lu-
cas, Spry, & Bell, 1983; Burgess, Mazzocco, & Campbell, 1987).
The test–retest reliabilities over a year period were .77 and .72 for
anxiety and depression, respectively (Crown, Duncan, & Howell,
1970). Patients diagnosed with anxiety disorder and depression
scored very highly on corresponding CCEI subscales (Crisp,
Jones, & Slater, 1978). Each subscale consists of eight items, rated
on a 4-point scale from 1 (never)to4(very often). An example
item is “Do you worry a lot?”. We recoded this into four variables:
prenatal maternal anxiety/depression and postnatal maternal anx-
iety/depression. Because prenatal maternal anxiety and depression
were correlated (r⫽.77), the average of prenatal maternal anxiety
and depression was used in the analysis. We did similarly for
postnatal maternal anxiety and depression (r⫽.73).
Older siblings. At 6 months, adolescents’ mothers reported
the numbers of older brothers and older sisters who live with the
adolescents, including maternal and paternal half-brothers and
half-sisters, stepbrothers and stepsisters, fostered children, and
adopted children.
Family structure changes. At six different points in time,
adolescents’ mothers answered the questions: “Is the present
live-in father-figure/mother-figure the biological father/mother of
the study child?” and “How old was the child when the biological
father/mother stopped living with the child?”. Father absence and
mother absence were recoded into two variables (father absence
and mother absence) with four categories: never with father/
mother, father/mother absence before 7 years of age, father/mother
absence since 7 years of age, and father/mother present. Because
of low rates of mother absence, it was necessary to combine these
variables to indicate parental absence (see missing data section).
Duration of breastfeeding. When adolescents were 1 year
and 3 months old, their mothers reported whether their children
were breast-fed (yes or no) and the duration of breastfeeding in
months. Duration of breastfeeding in months was used in the analysis,
and adolescents who were not breast-fed received a score of 0 on this
variable. Maternal reports of breastfeeding initiation and duration
appear accurate and reliable (Li, Scanlon, & Serdula, 2005).
House moves. At six different points in time, adolescents’ moth-
ers reported how many times they have moved home since last
interview. We recoded this information into two variables: the number
of house moves before 7 years of age and since 7 years of age.
Parent– child relationship. When adolescents were 9 years 7
months old, they rated their relationship with their parents on a 5-
point scale ranging from 1 (not true)to5(true). Nine items
developed by the ALSPAC study team were used. An example
item is “I have a parent who I have a lot of fun with.” Cronbach’s
alpha for the scale in our sample was .82. Exploratory factor
analysis yielded one factor (eigenvalue ⫽3.78) accounting for
42.03% of the variance in our sample (Supplemental Table S1 in
the online supplemental material). The total score of the nine items
was used in the analysis, with a higher score indicating a better
relationship with parents.
Family SEP. Family SEP was assessed via parents’ education,
parents’ occupation, household income, and family financial difficul-
ties. When adolescents were 32 weeks’ gestation old, their mothers
reported their own and their partner’s highest educational qualifica-
tions (CSE, vocational, O level, A level, and bachelor’s degree).
At six different points in time, adolescents’ mothers reported
their own occupation. At five different points in time, the mothers’
partners reported their own occupation. The Office for National
Statistics (2000) was used to categorize occupation type. We
recoded this into four variables, mother’s lowest occupation before
adolescents were born, father’s lowest occupation before adoles-
cents were born, mother’s lowest occupation before adolescents
were 7 years old, and father’s lowest occupation before adoles-
cents were 7 years.
At four different points in time, adolescents’ mothers also an-
swered the question: “On average, about how much is the take-
home family income each weak?” Participants were required to
choose from Less than £100, £100 –199, £200 –299, £300 –399,
£400 or more, and do not know. This was recoded into two
variables, lowest family income before adolescents were born, and
lowest family income before adolescents were 7 years old.
At six different points in time, adolescents’ mothers answered
the question: “How difficult at the moment do you find it to afford
these items? (e.g., food, clothing, and heating)” on a 4-point scale
from 0 (not difficult)to3(very difficult). Five items were used to
measure financial difficulties. This was recoded into three variables:
the worst financial difficulties before adolescents were born, the worst
financial difficulties before adolescents were 7 years old, and the
worst financial difficulties since adolescents were 7 years old.
Because these indicators of family SEP are correlated (poly-
choric correlation from .19 to .63), summary scores incorporating
these indicators were constructed: prenatal family SEP, family
SEP before 7 years of age, and family SEP since 7 years of age.
We applied polychoric principal component analysis and used the
loadings on the first principal component as item weightings to
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
5
EARLY LIFE CONDITIONS AND SEXUAL ORIENTATION
generate a summary score for each developmental stage. A higher
score indicates lower family SEP (Supplemental Table S2 in the
online supplemental material for factor loadings). The first com-
ponent explained 49.20%, 50.27%, and 55.32% of the variation in
prenatal family SEP, family SEP before 7 years of age, and family
SEP since 7 years of age, respectively.
Childhood gender-nonconforming behavior. When adoles-
cents were 2 years 6 months, 3 years 6 months, and 4 years 9
months old, mothers rated their children’s gender-nonconforming
behaviors (GNCB) using the Preschool Activities Inventory
(PSAI; Golombok & Rust, 1993). PSAI is a validated self-report
questionnaire (Golombok & Rust, 1993). PSAI also has acceptable
reliability (test–retest reliability over a year ⫽.64; split-half reli-
ability ⫽.88; Golombok & Rust, 1993). The PSAI consists of 12
male-typical items and 12 female-typical items assessing chil-
dren’s toy preferences (e.g., jewellery), activity preferences (e.g.,
playing at fighting), and characteristics (e.g., avoid getting dirty).
Each item was rated using a 5-point scale ranging from 1 (never)
to5(very often). The PSAI is scored via deducting the total score
for female-typical items from the total score for the male-typical
items, then transforming to a pseudo-Tscale by multiplication with
1.10 and adding 48.25 (Golombok & Rust, 1993). A higher score
indicates more male-typical behavior and less female-typical be-
havior for both girls and boys. The average of the three time points
was used (GNCB at these three times significantly and consistently
predicts adolescent sexual orientation; Li et al., 2017).
2D:4D digit ratio. When adolescents were 11 years old, pho-
tocopies of their hands were taken. They were required to place the
ventral surface of both hands flat onto the photocopier, and the
lengths of the second and the fourth digits for each hand were
measured to 0.01 mm using the “Mahr digital caliper16 EX” (from
tip of finger to basal crease). This method has been shown to be
accurate and reliable (Ribeiro, Neave, Morais, & Manning, 2016).
The digit ratio (2D:4D) was calculated as the ratio of the lengths
of the second digit to the fourth digit.
Handedness. At 9 year 7 months, adolescents were asked
which hand they prefer to use for six activities (e.g., “Which hand
to you draw”) rated from 1 ⫽left,2⫽either,3⫽right,and4⫽
do not do this at all. Cronbach’s alpha for the scale in our sample
was .84. Adolescents who chose do not do this at all were coded
as missing. Higher total scores indicated greater right-handedness.
Procedure
Missing data. The variables had 2.78 –56.80% missing infor-
mation within the analysis sample (Table 1 and 2). These missing
data were handled using multiple imputation stratified by sex.
Prior to imputation, we examined the potential missing data mech-
anisms using logistic regression to assess whether the observed
variables predict missingness. The results indicated that the data
were unlikely to be missing completely at random (e.g., duration of
breastfeeding and house moves before 7 years predicted the miss-
ingness of family SEP before 7 years).
For the imputation model, recommendations for longitudinal
studies are that all variables in the analysis should be included
(White, Royston, & Wood, 2011). Thus, the outcome variable
(sexual orientation), predictors (early life conditions), covariates
(e.g., GNCB), and an auxiliary variable (sexual behavior; see
Supplemental Text S1 in the online supplemental material) that
independently related to the outcome were included. Recommen-
dations also instruct that the number of imputations should be at
least as large as the percentage of missing data (White et al., 2011).
Thus, we used 57 imputations. We used the chained equations
algorithm (MICE) model since we have a combination of contin-
uous and categorical variables. The continuous variables included
in the current study were not normally distributed (Shapiro-Francia
test showed that all ps⬍.001). Consequently, we used predictive
mean matching since this approach makes no distributional as-
sumption.
Imputation for mother absence failed to converge due to small
cell sizes. Thus, we were forced to combine father absence and
mother absence into one variable labeled parental absence (never
with father or mother, either parent absence before 7 years, either
parent absence since 7 years, and both parents presence). Trace
plots and other diagnostics provided no cause for concern regard-
ing the imputed values. Typically, sensitivity analysis comparing
analyses based complete-case and imputed data would be under-
taken, however, due to the proportion of missing data in the sample
this was not possible.
Data analysis. All analyses were performed in Stata 15.0 and
carried out separately for boys and girls. First, a univariate ordered
logistic regression was estimated with sexual orientation (hetero-
sexual, bisexual, or homosexual) regressed onto each early life
variable to determine the unadjusted association. Then a three-step
hierarchical multivariable ordered logistic regression was esti-
mated with early life factors entered in sequential manner based on
the age period at which they were measured and controlling for
covariates (GNCB, 2D:4D digit ratio, and handedness). In the first
step, prenatal early life factors (e.g., birth weight and gestational
age) were entered. In the second step, early life factors before 7
years were entered. In the third step, early life factors since 7 years
were entered.
Finally, we used the mimrgns command to calculate the average
predicted probability of being homosexual for each significant
early life factor and covariate in the final model generated by the
third step of the multivariable ordered logistic regression. The
predicted probabilities for each continuous variable were com-
puted using its 25th/50th/75th percentile and observed values for
the remaining variables in the model. We also calculated the
predicted probability of being homosexual for adolescents with all
the significant factors and covariates present, and for adolescents
with none of these significant factors and covariates. The signifi-
cant continuous variables were set to the 25th/75th percentile for
present (75th for variables with odds ratios [ORs] greater than 1
and 25th for variables with ORs less than 1) and 50th percentile for
absent, and the remaining variables were set to the observed
values.
The ordered logistic regression analyses assume that the out-
come measure is ordinal in nature and that the association between
the predictor and outcome is equivalent across the levels of the
outcome (proportional odds). That is, the log-OR for the predictor
in a logistic model where the outcome is heterosexual versus
bisexual/homosexual is equivalent to the log-OR for the model
where the outcome is heterosexual/bisexual versus homosexual.
The Brant test was used iteratively to assess the likelihood that this
assumption held for each predictor (Brant, 1990). Where the test
was significant at the 5% level the assumption was relaxed for that
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
6XU, NORTON, AND RAHMAN
Table 1
Descriptive Statistics for Early Life Conditions and Covariates Among Boys
Variable
Sexual orientation
Heterosexual Bisexual Homosexual
Prenatal early life factors
Birthweight (in kilograms)
n2,159 29 25
M(SD) 3.47 (.59) 3.28 (.62) 3.20 (.52)
Gestational age (n)
Preterm birth 131 — —
Term birth 1,887 25 23
Postterm birth 168 — —
Maternal age (in years)
n2,186 29 26
M(SD) 28.59 (4.55) 27.83 (4.78) 27.85 (3.56)
Paternal age (in years)
n1,571 21 16
M(SD) 31.53 (5.41) 30.29 (6.94) 31.75 (4.88)
Prenatal maternal depression/anxiety
a
n1,867 25 24
M(SD) 4.50 (2.72) 4.58 (2.69) 5.34 (3.13)
Prenatal family socioeconomic position
b
n1,618 22 18
M(SD) 4.84 (2.72) 5.21 (2.76) 5.75 (2.43)
Number of older brothers
n2,138 29 28
M(SD) .38 (.62) .17 (.38) .50 (.64)
Number of old sisters
n2,123 29 27
M(SD) .36 (.61) .41 (.68) .22 (.42)
Early life factors before 7 years
Duration of breastfeeding (in months)
n1,704 23 20
M(SD) 6.07 (4.71) 5.04 (4.87) 3.60 (3.52)
Postnatal maternal depression/anxiety
a
n1,989 26 24
M(SD) 3.24 (2.78) 3.12 (2.85) 4.17 (3.60)
Number of house moves before adolescents were 7
n1,498 19 20
M(SD) 1.06 (1.44) 1.26 (1.63) 1.20 (1.61)
Family socioeconomic position before adolescents were 7
c
n1,007 14 14
M(SD) 5.08 (2.70) 5.17 (3.28) 5.01 (2.63)
Early life factors since 7 years
Family socioeconomic position since adolescents were 7
d
n1,588 23 16
M(SD) 2.73 (1.98) 3.22 (2.18) 3.07 (2.21)
Parent–child relationship
e
n1,790 22 22
M(SD) 41.30 (4.46) 41.36 (4.26) 39.86 (7.83)
Early life factors since 7 years
Pubertal body mass index
n1,277 13 13
M(SD) 20.21 (2.98) 21.29 (4.28) 20.13 (3.89)
Number of house moves since adolescents were 7
n1,984 26 23
M(SD) .32 (.59) .62 (.90) .43 (.59)
Father absence (n)
Never with father 54 — —
Father absence before adolescents were 7 192 7 —
Father absence since adolescents were 7 64 — —
Father presence 1,706 16 20
Mother absence (n)
Never with mother — — —
Mother absence before adolescents were 7 — — —
Mother absence since adolescents were 7 — — —
(table continues)
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
7
EARLY LIFE CONDITIONS AND SEXUAL ORIENTATION
predictor. Thus, one OR is presented where the proportional odds
assumption held and two are presented where it did not.
Because bisexual and homosexual individuals may differ in
some components of sexual orientation, it is possible that the
outcome in the ordered regressions is not ordinal in nature. To
assess the robustness of our estimates to varying conceptualiza-
tions of sexual orientation, we also estimated logistic regression
where the outcome was heterosexual versus bisexual/homosexual,
and multinomial logistic regression where the outcome was het-
erosexual versus bisexual and heterosexual versus homosexual. In
these models no assumption is made about the outcome being
ordinal or the proportionality of the odds (Supplemental Table
S3–S6 in the online supplemental material.).
Although no formal power calculation was conducted, it is
possible to consider power to detect a meaningful effect. At the 5%
level, the sample size allows for the detection of OR of greater than
approximately 1.8 for a binary predictor variable with 80% power,
where the proportional odds assumption holds for the predictor. As
such, despite the low number of nonheterosexuals in the sample,
power to detect meaningful effects is acceptable.
Results
Boys
GNCB (OR ⫽0.888 to 0.902, all ps⬍.001) and right 2D:4D
digit ratio (OR ⫽1.185 to 1.283, all ps⬍.01) were significantly
associated with nonheterosexual orientation in both univariate and
multivariable regressions, although there were no significant as-
sociations between left 2D:4D digit ratio, handedness, and sexual
orientation (Table 3 and 4). When the OR was transformed to the
percentage change in the ratio for one-unit increase in the predictor
using the formula: 100 ⫻(OR ⫺1), the results suggested that boys
who displayed more GNCB had 10.90% to 12.60% greater odds of
being nonheterosexual, and boys with higher right 2D:4D digit
ratio had 18.50% to 28.30% greater odds of being nonhetero-
sexual.
Birth weight (OR ⫽0.458 to 0.570, all ps⬍.05) and duration
of breastfeeding (OR ⫽0.904 to 0.921, all ps⬍.05) were strong
predictors of nonheterosexual orientation in both regression mod-
els. Boys with low birth weight had 75.40% to 118.30% greater
odds of being nonheterosexual, and boys with shorter duration of
breastfeeding had 8.60% to 10.60% greater odds of being nonhet-
erosexual. Greater number of older brothers (OR ⫽2.069 to 2.254,
all ps⬍.05) was significantly associated with homosexual orien-
tation in multivariable regressions, indicating that boys with
greater number of older brothers had 106.90% to 125.40% greater
odds of being homosexual. Greater number of house moves since
7 and parental absence before 7 years of age were also signifi-
cantly associated with nonheterosexual orientation in univariate
regression, but these disappeared in the multivariable regression
when all early life factors were entered into the model.
To further aid understanding of the results, the significant ORs
(relative difference) from the third step of the multivariable or-
dered logistic regression were transformed to average marginal
effects (absolute difference). Boys who displayed more GNCB
(25th percentile) had a 1.73%, 95% confidence interval (CI) ⫽
1.10 –2.36% probability of being homosexual, boys with higher
right 2D:4D digit ratio (75th percentile) had a 1.65%, 95% CI ⫽
1.02–2.29% probability of being homosexual, boys with lower
birth weight (25th percentile) had a 1.57%, 95% CI ⫽0.99 –2.16%
probability of being homosexual, and boys with shorter duration of
breastfeeding (25th percentile) had a 1.80%, 95% CI ⫽
1.08 –2.51% probability of being homosexual. Boys with more
GNCB, higher right 2D:4D digit ratio, lower birth weight, and
shorter duration of breastfeeding had a 3.74%, 95% CI ⫽
2.00 –5.49%, probability of being homosexual, while boys with
none of these had a 0.78%, 95%CI ⫽0.36 –1.20% probability of
being homosexual (see Figure 1).
Girls
Consistent with the results for boys, GNCB was a strong pre-
dictor of nonheterosexual orientation among girls in both regres-
sion models, with ORs ranging from 1.072 to 1.097, all ps⬍.01,
Table 1 (continued)
Variable
Sexual orientation
Heterosexual Bisexual Homosexual
Mother presence 2,002 27 23
Covariates
Childhood gender nonconforming behavior
f
n1,706 25 21
M(SD) 61.93 (7.28) 55.96 (10.30) 54.64 (7.73)
Left 2D:4D
n2,128 27 24
M(SD) .96 (.03) .96 (.04) .97 (.02)
Right 2D:4D
n2,125 27 25
M(SD) .96 (.03) .97 (.04) .99 (.03)
Handedness
n1,882 23 22
M(SD) 15.85 (3.13) 16.48 (1.95) 16.00 (2.79)
Note. Dashes means 5 or less. Cell counts 5 or less are not presented in order to comply with ALSPAC publication requirements.
a
The range is from 0 to 16.
b
The range is from 0 to 15.32.
c
The range is from 0 to 16.40.
d
The range is from 0 to 12.25.
e
The range is from 9
to 45.
f
The range is from ⫺4.55 to 101.05.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
8XU, NORTON, AND RAHMAN
Table 2
Descriptive Statistics for Early Life Conditions and Covariates Among Girls
Variable
Sexual orientation
Heterosexual Bisexual Homosexual
Prenatal early life factors
Birthweight (in kilograms)
n2,423 53 16
M(SD) 3.37 (.50) 3.43 (.44) 3.43 (.42)
Gestational age (n)
Preterm birth 96 — —
Term birth 2,185 49 15
Postterm birth 178 — —
Maternal age (in years)
n2,459 53 16
M(SD) 28.37 (4.52) 28.92 (5.94) 28.06 (4.17)
Paternal age (in years)
n1,758 37 10
M(SD) 31.22 (5.41) 31.65 (5.77) 32.20 (5.25)
Prenatal maternal depression/anxiety
a
n2,042 40 11
M(SD) 4.50 (2.74) 5.22 (3.09) 4.66 (2.20)
Prenatal family socioeconomic position
b
n1,824 35 11
M(SD) 4.93 (2.73) 5.60 (2.73) 6.66 (3.05)
Number of older brothers
n2,391 49 16
M(SD) .38 (.63) .31 (.62) .63 (.81)
Number of old sisters
n2,395 49 16
M(SD) .37 (.60) .33 (.72) .31 (.60)
Early life factors before 7 years
Duration of breastfeeding (in months)
n1,859 41 11
M(SD) 6.36 (4.75) 6.85 (4.94) 8.36 (6.22)
Postnatal maternal depression/anxiety
a
n2,287 43 15
M(SD) 3.27 (2.74) 3.91 (3.53) 3.00 (2.01)
Number of house moves before adolescents were 7
n1,637 31 8
M(SD) 1.14 (1.47) 1.26 (1.71) 1.63 (1.51)
Family socioeconomic position before adolescents were 7
c
n1,101 20 7
M(SD) 5.21 (2.80) 5.46 (2.49) 5.43 (2.13)
Early life factors since 7 years
Family socioeconomic position since adolescents were 7
d
n1,745 31 10
M(SD) 2.84 (2.04) 2.98 (1.97) 3.02 (1.64)
Parent–child relationship
e
n2,085 34 14
M(SD) 42.22 (3.77) 40.82 (4.79) 39.79 (4.73)
Early life factors since 7 years
Pubertal body mass index
n1,334 21 6
M(SD) 20.83 (3.36) 22.26 (3.90) 22.27 (5.04)
Number of house moves since adolescents were 7
n2,177 43 10
M(SD) .33 (.72) .49 (1.10) .40 (.70)
Father absence (n)
Never with father 62 — —
Father absence before adolescents were 7 239 10 —
Father absence since adolescents were 7 53 — —
Father presence 1,879 28 9
Mother absence (n)
Never with mother — — —
Mother absence before adolescents were 7 — — —
Mother absence since adolescents were 7 — — —
(table continues)
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
9
EARLY LIFE CONDITIONS AND SEXUAL ORIENTATION
indicating that girls who displayed more GNCB had 7.20% to
9.70% greater odds of being nonheterosexual. There were no
significant associations between left or right 2D:4D digit ratios,
handedness, and sexual orientation (Table 3 and 5).
Low prenatal family SEP (OR ⫽1.110 to 1.343, all ps⬍.05),
parental absence since birth (OR ⫽2.501 to 4.494, all ps⬍.05),
and poorer reported relationship with parents (OR ⫽0.928 to
0.931, all ps⬍.05) were also significantly associated with non-
heterosexuality in both regression models. These indicated that
girls with a low prenatal family SEP had 11.00% to 34.30% greater
odds of being nonheterosexual, girls with parental absence since
birth had 150.10% to 349.40% greater odds of being nonhetero-
sexual, and girls with a more negative relationship with parents
had 7.40% to 7.80% greater odds of being nonheterosexual.
In terms of average marginal effects, girls who displayed more
GNCB (75th percentile) had a 0.74%, 95% CI ⫽0.38 –1.10%
probability of being homosexual, girls with low prenatal family
SEP (75th percentile) had a 1.24%, 95% CI ⫽0.35–2.12% prob-
ability of being homosexual, girls with parental absence before 7
had a 1.35%, 95% CI ⫽0.35–2.35% probability of being homo-
sexual, and girls with poorer reported relationship with parents
(25th percentile) had a 0.65%, 95% CI ⫽0.34 –0.96% probability
of being homosexual. The predicted probability of being homo-
sexual for girls with more GNCB, low prenatal family SEP,
parental absence before 7, and poorer reported relationship with
parents was 3.36%, 95% CI ⫽0.03–6.69%, whereas girls with
none of these had a 0.29%, 95%CI ⫽0.09 –0.49% probability of
being homosexual (Figure 2 and 3).
Discussion
This study in a prospective birth cohort produced three main
findings. First, boys with low birth weight and shorter duration of
breastfeeding were more likely to be nonheterosexual (bisexual
and homosexual), and boys with greater number of older brothers
were more likely to be homosexual. Second, boys with parental
absence before 7 were more likely to be nonheterosexual, but this
association disappeared after entering all early life factors into the
statistical models. Finally, parental absence since birth, low pre-
natal family SEP, and low parent– child relationship scores pre-
dicted nonheterosexual orientation among girls. These results were
found while controlling for GNCB, handedness, and 2D:4D digit
ratio.
Early Life Conditions and Sexual Orientation in Boys
The findings regarding birth weight and older brothers is con-
sistent with prior work in cross-sectional samples (Blanchard,
2018; Skorska et al., 2017). However, caution must be exercised in
interpreting these findings. The number of older brothers in the
present cohort included half-brothers, stepbrothers, fostered broth-
ers, and adopted brothers. The maternal immunity theory behind
the FBO effect predicts that only previously carried biological
male siblings should increase the probability of homosexuality of
later born male fetuses (Bogaert, 2006). The birth weight findings
support growing evidence showing low birth weight predicts faster
sexual life history (Nettle et al., 2011). In addition, low birth
weight among later nonheterosexual boys offers support for ma-
ternal immunity hypothesis linking birth weight and FBO (al-
though here we find those two effects to be independent;
Blanchard, 2012). None of the early life conditions since boys
were 7 years were significantly related to sexual orientation in the
multivariate regressions, adding further support to the importance
of the prenatal or early postnatal developmental period for male
sexuality (Xu et al., 2018).
The association between breastfeeding duration and male non-
heterosexuality is novel. Low breastfeeding duration may be re-
lated to birth weight, although the direction reported here is op-
posite to what is typically found (low birth weight infants usually
receive more breastfeeding as per medical advice). At this stage,
the link between breastfeeding and male sexual orientation is
unclear but does potentially indicate lower maternal somatic in-
vestment in later nonheterosexual boys. Trade-offs underlying
breastfeeding decisions (physiological as well as behavioral) by
Table 2 (continued)
Variable
Sexual orientation
Heterosexual Bisexual Homosexual
Mother presence 2,215 44 11
Covariates
Childhood gender nonconforming behavior
f
n1,857 33 14
M(SD) 37.50 (7.74) 41.71 (9.63) 45.63 (9.60)
Left 2D:4D
n2,355 50 14
M(SD) .97 (.03) .97 (.03) .96 (.02)
Right 2D:4D
n2,359 50 14
M(SD) .97 (.03) .97 (.03) .96 (.02)
Handedness
n2,172 35 12
M(SD) 15.93 (2.84) 15.57 (3.01) 15.58 (2.75)
Note. Dashes means 5 or less. Cell counts 5 or less are not presented in order to comply with Avon Longitudinal Study of Parents and Children (ALSPAC)
publication requirements.
a
The range is from 0 to 16.
b
The range is from 0 to 15.32.
c
The range is from 0 to 16.40.
d
The range is from 0 to 12.25.
e
The range is from 9
to 45.
f
The range is from ⫺4.55 to 101.05.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
10 XU, NORTON, AND RAHMAN
Table 3
Univariate Ordered Logistic Regressions for Sexual Orientation Stratified by Sex
Variable
Boys (n⫽2,349) Girls (n⫽2,658)
OR OR
a
OR
b
OR OR
a
OR
b
Covariates
Childhood gender nonconforming
behavior .902
ⴱⴱⴱ
(.871, .935) 1.074
ⴱⴱⴱ
(1.040, 1.108) 1.097
ⴱⴱⴱ
(1.045, 1.151)
Handedness 1.036 (.931, 1.152) .968 (.881, 1.063)
Left 2D:4D 1.077
ⴱ
(1.002, 1.157) 1.014 (.939, 1.094)
Right 2D:4D 1.185
ⴱⴱⴱ
(1.092, 1.287) 1.245
ⴱⴱⴱ
(1.132,1.368) .979 (.909, 1.056)
Early life conditions
Gestational age (Ref ⫽term birth)
Preterm birth .992 (.311, 3.165) .739 (.179, 3.054)
Postterm birth .936 (.328, 2.671) .575 (.180, 1.841)
Birthweight .570
ⴱⴱ
(.382, .851) 1.237 (.761, 2.010)
Maternal age .968 (.913, 1.027) 1.018 (.967, 1.072)
Paternal age .985 (.933, 1.040) 1.016 (.971, 1.062)
Prenatal family socioeconomic position 1.059 (.959, 1.170) 1.110
ⴱ
(1.015, 1.214)
Number of older brothers .902(.574, 1.417) 1.784 (.896, 3.553) 1.059 (.717, 1.565) 1.600 (.943, 2.716)
Number of older sisters .887 (.559, 1.408) .879 (.564, 1.372)
Prenatal maternal anxiety/depression 1.063 (.967, 1.167) 1.077 (.988, 1.173)
Parental absence (Ref ⫽parents presence)
Never with mother or father 3.063
ⴱ
(1.052, 8.920) 2.870 (.915, 8.997)
Either parent absence before 7 2.139
ⴱ
(1.050, 4.357) 2.572
ⴱⴱ
(1.309, 5.054)
Either parent absence since 7 1.354 (.317, 5.784) 4.135
ⴱ
(1.358, 12.585)
Duration of breastfeeding before 7 .921
ⴱ
(.853, .994) 1.043 (.989, 1.099)
Postnatal maternal anxiety/depression 1.049 (.958, 1.150) 1.051 (.968, 1.142)
Number of house moves before 7 1.073 (.896, 1.285) 1.078 (.898, 1.293)
Family socioeconomic position before 7 1.000 (.997, 1.004) 1.001 (.998, 1.004)
Number of house moves since 7 1.491
ⴱ
(1.036, 2.146) 1.186 (.944, 1.492)
Parent–child relationship .971 (.918, 1.027) .931
ⴱ
(.881, .984)
Pubertal body mass index 1.046 (.935, 1.170) 1.095
ⴱ
(1.006, 1.192)
Family socioeconomic position since 7 1.087 (.947, 1.248) 1.065 (.938, 1.210)
Note. We applied Brant test to test the proportional odds assumption. If the proportional odds assumption is not violated, we reported one odds ratio (OR) in the column OR; if it is violated, we applied
the generalized ordered logit model (gologit2) and reported two ORs in the columns OR
a
and OR
b
.
a
Heterosexual boys/girls versus bisexual and homosexual boys/girls.
b
Heterosexual and bisexual boys/girls versus homosexual boys/girls.
ⴱ
p⬍.05.
ⴱⴱ
p⬍.01.
ⴱⴱⴱ
p⬍.001.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
11
EARLY LIFE CONDITIONS AND SEXUAL ORIENTATION
Table 4
Three-Step Hierarchical Multivariable Ordered Logistic Regressions for Sexual Orientation Among Boys (N ⫽2349)
Variable
Step 1 Step 2 Step 3
OR ORaORbOR ORaORbOR ORaORb
Covariates
Childhood gender nonconforming
behavior .898ⴱⴱⴱ (.865, .932) .890ⴱⴱⴱ (.856, .925) .888ⴱⴱⴱ (.853, .924)
Handedness 1.034 (.922, 1.160) 1.033 (.920, 1.159) 1.035 (.920, 1.164)
Left 2D:4D .953 (.850, 1.070) .954 (.849, 1.072) .956 (.850, 1.076)
Right 2D:4D 1.213ⴱⴱ (1.085, 1.356) 1.273ⴱⴱⴱ (1.120, 1.446) 1.224ⴱⴱ (1.092, 1.371) 1.283ⴱⴱⴱ (1.127, 1.461) 1.219ⴱⴱ (1.087, 1.368) 1.278ⴱⴱⴱ (1.122, 1.457)
Early life conditions
Gestational age (Ref ⫽term birth)
Preterm birth .294 (.067, 1.283) .307 (.068, 1.384) .315 (.070, 1.425)
Postterm birth 1.233 (.412, 3.692) 1.229 (.406, 3.724) 1.305 (.424, 4.014)
Birthweight .458ⴱⴱ (.261, .805) .489ⴱ(.276, .866) .463ⴱⴱ (.259, .826)
Maternal age .987 (.897, 1.087) .992 (.899, 1.094) 1.000 (.906, 1.105)
Paternal age 1.013 (.936, 1.098) 1.017 (.937, 1.104) 1.019 (.938, 1.106)
Prenatal family socioeconomic position 1.045 (.922, 1.185) 1.063 (.906, 1.247) 1.060 (.844, 1.331)
Number of older brothers 1.078 (.671, 1.731) 2.069ⴱ(1.034, 4.141) 1.117 (.690, 1.807) 2.176ⴱ(1.080, 4.383) 1.145 (.699, 1.877) 2.254ⴱ(1.105, 4.598)
Number of older sisters .731 (.424, 1.261) .759 (.439, 1.313) .752 (.433, 1.307)
Prenatal maternal anxiety/depression 1.038 (.937, 1.151) 1.046 (.903, 1.212) 1.043 (.899, 1.209)
Parental absence (Ref ⫽parents presence)
Never with mother or father 3.336 (.998, 11.155) 3.708ⴱ(1.051, 13.074) 3.449 (.873, 13.623)
Either parent absence before 7 2.051 (.902, 4.664) 1.946 (.794, 4.771)
Either parent absence since 7 1.052 (.218, 5.087)
Duration of breastfeeding before 7 .907ⴱ(.836, .983) .904ⴱ(.834, .981)
Postnatal maternal anxiety/depression .994 (.864, 1.143) .994 (.863, 1.146)
Number of house moves before 7 1.053 (.868, 1.278) 1.041 (.852, 1.272)
Family socioeconomic position before 7 .998 (.993, 1.003) .998 (.993, 1.003)
Number of house moves since 7 1.360 (.882, 2.097)
Parent–child relationship .979 (.922, 1.039)
Pubertal body mass index 1.079 (.962, 1.210)
Family socioeconomic position since 7 .988 (.722, 1.352)
Note. We applied Brant test to test the proportional odds assumption. If the proportional odds assumption is not violated, we reported one odds ratio (OR) in the column OR; if it is violated, we applied
the generalized ordered logit model (gologit2) and reported two ORs in the columns OR
a
and OR
b
.
a
Heterosexual boys versus bisexual and homosexual boys.
b
Heterosexual and bisexual boys versus homosexual boys.
ⴱ
p⬍.05.
ⴱⴱ
p⬍.01.
ⴱⴱⴱ
p⬍.001.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
12 XU, NORTON, AND RAHMAN
mothers are well documented (Tully & Ball, 2013). Cross-cultural
work suggests that economically wealthier mothers will breastfeed
sons more frequently than daughters, whereas poorer mothers will
breastfeed daughters more frequently than sons (Fujita et al.,
2012). This is consistent with the prediction that selection pres-
sures favor parental investment in daughters when times are hard
and in sons when times are easy. It is not clear how this plasticity
in breastfeeding response during resource rich or poor conditions
applies to sexual orientation.
Early Life Conditions and Sexual Orientation in Girls
The pattern of results for girls is consistent with the notion that
female sexual orientation is more influenced by environmental
factors, perhaps over the life course (rather than restricted to the
prenatal or early postnatal period; Baumeister, 2000). This is
supported by cross-sectional studies reporting associations be-
tween environmental factors and female sexual orientation. These
include education level, religion affiliation, peer influence, and
parental influences (e.g., parents age and divorce; Baumeister,
2000). Women report more sexual fluidity over the life-course,
more bisexuality, and a less category-specific sexual orientation
compared to men (Bailey et al., 2016). However, it is not clear why
greater female malleability in sexual orientation would be causally
related to social factors. Behavioral and identity components of
female sexual orientation may fluctuate as a result of situational
factors, but attractions may remain stable (Diamond, 2008). Other
studies have found no associations between changes in same-sex
attractions among women and parental divorce, family SES, edu-
cation level, or family disapproval of one’s sexuality (Diamond,
2008).
The present study covaried GNCB, handedness, and digit ratios
as proxy markers of prenatal androgen exposure. Thus, the asso-
ciations found here cannot be explained by GNCB, which is often
a confounding variable between putative causal factors (such
parent– child relationships) and sexual orientation (Bell et al.,
1981). Our results replicate the finding that boys and girls who
showed more GNCB before 4.75 years old were more likely to be
nonheterosexual. This finding is consistent with a large body of
work in retrospective and prospective studies (Bailey & Zucker,
1995; Li et al., 2017). In this study, GNCB was used a proxy
behavioral marker of androgen exposure. We encourage future
studies to test GNCB as a formal mediator variable. This will
provide a better test of its role of a third factor explaining the
pattern of the results.
We found that boys with higher (more feminine) right digit ratio
were more likely to be nonheterosexual later in life, consistent with
the prediction that low levels of prenatal androgens influence male
homosexuality. However, this is inconsistent with a meta-analysis
showing no differences in digit ratio between heterosexual and
nonheterosexual men (Grimbos et al., 2010). That study only
found an association between low digit ratio (more masculine) and
female nonheterosexuality, which we did not find here. We did not
replicate an association between greater nonright-handedness and
later nonheterosexuality, which also contradicts prior studies
(Lalumière et al., 2000). Some studies suggest that sexual orien-
tation difference in digit ratio and handedness vary according to
criteria used to classify individuals as nonheterosexual (Xu &
Zheng, 2016, 2017). This could be one explanation for our results.
The longitudinal nature of our design (handedness and digit ratio
measured before sexual orientation) may have contributed to the
inconsistency. Although it is difficult to see why this would be the
case given handedness and digit ratios appear stable over time.
A Life History Perspective
We have suggested that early life conditions documented in life
history research may constitute one source of nonshared environ-
mental factors in the development of sexual orientation. Our find-
ings on sexual orientation are consistent with known associations
between similarly measured early conditions and a range of other
“fast” reproductive and sexual behavior outcomes (James et al.,
2012; Xu et al., 2018). Such factors may interact with processes
hypothesized by other models of sexual orientation, such as pre-
natal androgen theory. For example, prenatal early life factors may
influence the levels of sex hormones during fetal or early postnatal
brain development (Rahman, 2005). The present findings suggest
that life history models should pay attention to the specificity of
early conditions, developmental stages, and sexual behavior out-
comes. The current results also suggest that the experience of early
life conditions may operate in a facultative manner depending on
sex with potential greater plasticity of responses among females
across developmental periods.
In general, the associations between early life conditions and
sexual orientation in both boys and girls were small. This may
indicate that other causal factors influence both our predictor and
outcome variables, including genetic correlations (Barbaro,
Boutwell, Barnes, & Shackelford, 2017). Twin research shows a
small influence of the shared environment (including familial
influences) on a range of behavioral traits, including sexual be-
Figure 1. Predicted probability of being homosexual for significant early
life factors and covariates in the final model among boys.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
13
EARLY LIFE CONDITIONS AND SEXUAL ORIENTATION
Table 5
Three-Step Hierarchical Multivariable Ordered Logistic Regressions for Sexual Orientation Among Girls (N ⫽2,658)
Step 1 Step 2 Step 3
Variable OR ORaORbOR ORaORbOR ORaORb
Covariates
Childhood gender nonconforming
behavior 1.074ⴱⴱⴱ (1.038, 1.110) 1.096ⴱⴱ (1.038, 1.157) 1.074ⴱⴱⴱ (1.039, 1.111) 1.097ⴱⴱ (1.039, 1.158) 1.072ⴱⴱⴱ (1.036, 1.110) 1.095ⴱⴱ (1.035, 1.157)
Handedness .979 (.888, 1.079) .977 (.885, 1.078) .968 (.873, 1.073)
Left 2D:4D 1.051 (.943, 1.170) 1.053 (.945, 1.173) 1.057 (.948, 1.180)
Right 2D:4D .958 (.861, 1.066) .959 (.861, 1.067) .950 (.851, 1.061)
Early life conditions
Gestational age (Ref ⫽term birth)
Preterm birth 1.158 (.248, 5.412) 1.220 (.256, 5.813) 1.101 (.218, 5.553)
Postterm birth .617 (.187, 2.033) .623 (.188, 2.064) .663 (.195, 2.250)
Birthweight 1.393 (.818, 2.374) 1.416 (.824, 2.433) 1.376 (.795, 2.380)
Maternal age 1.050 (.966, 1.140) 1.039 (.955, 1.130) 1.049 (.963, 1.143)
Paternal age 1.006 (.941, 1.075) 1.012 (.946, 1.082) 1.014 (.946, 1.087)
Prenatal family socioeconomic position 1.126ⴱ(1.011, 1.254) 1.149ⴱ(1.006, 1.311) 1.343ⴱⴱ (1.109, 1.627)
Number of older brothers .771 (.497, 1.196) 1.162 (.667, 2.025) .727 (.465, 1.136) 1.095 (.624, 1.922) .698 (.443, 1.098) 1.055 (.600, 1.855)
Number of older sisters .808 (.505, 1.292) .785 (.488, 1.262) .756 (.465, 1.229)
Prenatal maternal anxiety/depression 1.035 (.943, 1.136) 1.031 (.909, 1.170) 1.027 (.904, 1.166)
Parental absence (Ref ⫽parents presence)
Never with mother or father 2.002 (.593, 6.755) 1.584 (.419, 5.991) 2.620 (.561, 12.248)
Either parent absence before 7 2.501ⴱ(1.201, 5.212) 2.990ⴱⴱ (1.361, 6.570)
Either parent absence since 7 4.494ⴱ(1.368, 14.760)
Duration of breastfeeding before 7 1.056 (.997, 1.119) 1.063ⴱ(1.001, 1.129)
Postnatal maternal anxiety/depression .998 (.885, 1.126) 1.003 (.887, 1.135)
Number of house moves before 7 1.062 (.867, 1.300) 1.055 (.853, 1.305)
Family socioeconomic position before 7 1.000 (.996, 1.004) 1.000 (.996, 1.005)
Number of house moves since 7 1.161 (.865, 1.558)
Parent–child relationship .928ⴱ(.873, .986)
Pubertal body mass index 1.078 (.978, 1.189)
Family socioeconomic position since 7 .744ⴱ(.564, .981)
Note. We applied Brant test to test the proportional odds assumption. If the proportional odds assumption is not violated, we reported one odds ratio (OR) in the column OR; if it is violated, we applied
the generalized ordered logit model (gologit2) and reported two ORs in the columns OR
a
and OR
b
.
a
Heterosexual boys versus bisexual and homosexual boys.
b
Heterosexual and bisexual boys versus homosexual boys.
ⴱ
p⬍.05.
ⴱⴱ
p⬍.01.
ⴱⴱⴱ
p⬍.001.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
14 XU, NORTON, AND RAHMAN
havior (Polderman et al., 2015). Twin research on sexual orienta-
tion conforms to this pattern, even for female sexual orientation
where some studies show small shared environmental influences
and others none (Bailey et al., 2016). Thus, the influence of
genetic, shared, and nonshared environmental factors may also
depend on the type of early life condition, developmental stage,
and sexual orientation outcome (attractions, behavior, and iden-
tity).
Strengths and Limitations
The present study has particular strengths including the use of a
prospective design in a well-characterized cohort, with early life
factors (and covariates) measured before sexual orientation. This is
important because most studies on the development of sexual
orientation rely on cross-sectional or correlational designs. Pro-
spective studies reduce the risk of recall biases and get closer to
possible causal pathways. Our design also makes the possibility
of reverse causation between early life conditions and later
sexual orientation less likely. The range of early life conditions
has never been tested before and their choice was theoretically
and empirically motivated (using a life history perspective and
models of sexual orientation development). We also controlled
for GNCB, one possible confounding variable between hypoth-
esized causal factors and later sexual orientation (Bailey et al.,
2016).
However, several limitations are important to note. There may
be unobserved confounders or “third variables” between early life
factors and later sexual orientation that affect the results. Although
reverse causation is less likely with longitudinal designs, other
sources of confounding remain. Such unmeasured variables in-
clude personality factors, parental awareness or suspicions about
their child’s later sexual orientation (via elevated gender noncon-
formity), or unmeasured family dynamics impacting upon some of
the early life factors (e.g., parental conflict resulting in house
moves). These third factors could also include other indicators of
a “fast” life history strategy, such as sexual maturation and number
of sexual partners, which we could not measure here. Importantly,
we could not control for genetic confounds such as shared genes
between parents and children. Nor could we control for unmea-
Figure 2. Predicted probability of being homosexual for significant con-
tinuous early life factors and covariates in the final model among girls.
Figure 3. Predicted probability of being homosexual for parental absence in the final model among girls.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
15
EARLY LIFE CONDITIONS AND SEXUAL ORIENTATION
sured genetic and environmental confounds that load simultane-
ously on early life conditions, our covariates, and sexual orienta-
tion (Barbaro et al., 2017). This requires further study. Maternal
reports of some of the early life factors may either under- or
overestimate their prevalence among adolescents who later be-
come nonheterosexual. For example, mothers with gender noncon-
forming children could engage in less parental investment (in-
dexed by factors such duration of breastfeeding, parent– child
relationships, and so on), which is then associated with later
nonheterosexuality. There is growing evidence that gender non-
conforming children experience negative reactions and greater
stigmatization from family and peers and so this possibility re-
quires further study.
The nature of the cohort meant that several early life factors
could not be simultaneously measured in all three developmental
periods. Some of the measures of early life conditions (e.g., family
SEP) may also act as mediators between other conditions and later
sexual orientation. SEP, because of its close association with
growing up in a resource rich or poor context, requires careful
study. As does the link between birth weight and duration of
breastfeeding. In addition, of the 17 early life factors two (parental
absence and gestational age) had restricted response categories
which may have reduced power. For some of these measures,
validity and reliability information was not available.
The sample sizes of nonheterosexual boys and girls were small,
especially homosexual girls. The rates for gestational age and
parental absence were low and were not observed in one or more
of the groups. Thus, a small increase in the number of nonhetero-
sexual boys/girls who experienced parental absence, or were pre-
term birth, will result in larger ORs. However, given the low
prevalence of nonheterosexual orientation among the population,
small numbers of nonheterosexuals in longitudinal or other cohort
studies are to be expected. Despite the small number of nonhet-
erosexuals, the power to detect meaningful effects was acceptable
because of the large overall size of the sample. The use of ordered
regression models improves precision gains for the ORs, particu-
larly where some groups are small (or where there are many
categories) because one can carry over power from the larger
groups to the smaller ones. Although case-control studies would
afford greater power this would come at the cost of associated
biases.
The current study measured sexual orientation when adolescents
were 15.5 years of age. Prior research found that the number of
people who identified themselves as nonheterosexual increases
from adolescence to adulthood (Austin et al., 2009). Thus, it is
possible that adolescents may change their sexual orientation re-
ports if we reassess our cohort at later ages, which may even
produce somewhat different results. Adolescents may also misre-
port their sexual orientation (Savin-Williams & Joyner, 2014).
Future studies must investigate such cohorts at later ages and
model any change in sexual orientation outcomes and their impact
on model estimates. The use of a single-item measure of sexual
orientation is also a limitation. Future studies should aim to mea-
sure several components of sexual orientation (e.g., identity, at-
tractions, and behavior) or more fully explore the reliability and
validity of single-item measures over several time points during
adolescence and young adulthood.
Conclusion
The results offer support to the hypothesis that early life factors
influence sexual orientation in adolescent boys and girls. The
developmental stage of these factors appears important to sex
differences in adolescent sexual orientation. Among boys, prenatal
and early life conditions before 7 years of age predicted later
sexual orientation. Among girls, a mix of factors measured prena-
tally and later in childhood predicted later sexual orientation.
These associations were found while controlling for putative mark-
ers of prenatal androgen influence. GNCB was a strong predictor
of sexual orientation as previously shown. Note the role of those
influences that could also be conceptualized as “psychosocial”
appears small across the board. Future longitudinal studies should
test for the role of possible third variables (or genetic confounds),
which may act of mediators for the associations found here.
References
Alanko, K., Santtila, P., Harlaar, N., Witting, K., Varjonen, M., Jern, P.,...
Sandnabba, N. K. (2010). Common genetic effects of gender atypical
behavior in childhood and sexual orientation in adulthood: A study of
Finnish twins. Archives of Sexual Behavior, 39, 81–92. http://dx.doi.org/
10.1007/s10508-008-9457-3
Alderman, K. J., Mackay, C. J., Lucas, E. G., Spry, W. B., & Bell, B.
(1983). Factor analysis and reliability studies of the Crown-Crisp Ex-
periential Index (CCEI). British Journal of Medical Psychology, 56,
329 –345. http://dx.doi.org/10.1111/j.2044-8341.1983.tb01565.x
Austin, S. B., Ziyadeh, N. J., Corliss, H. L., Rosario, M., Wypij, D.,
Haines, J.,...Field, A. E. (2009). Sexual orientation disparities in
purging and binge eating from early to late adolescence. Journal of
Adolescent Health, 45, 238 –245. http://dx.doi.org/10.1016/j.jadohealth
.2009.02.001
Bailey, J. M., Vasey, P. L., Diamond, L. M., Breedlove, S. M., Vilain, E.,
& Epprecht, M. (2016). Sexual orientation, controversy, and science.
Psychological Science in the Public Interest, 17, 45–101. http://dx.doi
.org/10.1177/1529100616637616
Bailey, J. M., & Zucker, K. J. (1995). Childhood sex-typed behavior and
sexual orientation: A conceptual analysis and quantitative review. De-
velopmental Psychology, 31, 43–55. http://dx.doi.org/10.1037/0012-
1649.31.1.43
Barbaro, N., Boutwell, B. B., Barnes, J. C., & Shackelford, T. K. (2017).
Genetic confounding of the relationship between father absence and age
at menarche. Evolution and Human Behavior, 38, 357–365. http://dx.doi
.org/10.1016/j.evolhumbehav.2016.11.007
Baumeister, R. F. (2000). Gender differences in erotic plasticity: The
female sex drive as socially flexible and responsive. Psychological
Bulletin, 126, 347–374. http://dx.doi.org/10.1037/0033-2909.126.3.347
Bell, A. P., Weinberg, M. S., & Hammersmith, S. K. (1981). Sexual
preference: Its development in men and women (Vol. 2). Bloomington:
Indiana University Press.
Blanchard, R. (2012). A possible second type of maternal-fetal immune
interaction involved in both male and female homosexuality. Archives of
Sexual Behavior, 41, 1507–1511. http://dx.doi.org/10.1007/s10508-011-
9896-0
Blanchard, R. (2018). Fraternal birth order, family size, and male homo-
sexuality: Meta-analysis of studies spanning 25 years. Archives of Sexual
Behavior, 47, 1–15. http://dx.doi.org/10.1007/s10508-017-1007-4
Blanchard, R., & Bogaert, A. F. (1996). Biodemographic comparisons of
homosexual and heterosexual men in the Kinsey Interview Data. Ar-
chives of Sexual Behavior, 25, 551–579. http://dx.doi.org/10.1007/
BF02437839
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
16 XU, NORTON, AND RAHMAN
Blanchard, R., & Ellis, L. (2001). Birth weight, sexual orientation and the
sex of preceding siblings. Journal of Biosocial Science, 33, 451– 467.
http://dx.doi.org/10.1017/S0021932001004515
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, 41,
321–327. http://dx.doi.org/10.1006/hbeh.2002.1765
Bogaert, A. F. (2006). Biological versus nonbiological older brothers and
men’s sexual orientation. Proceedings of the National Academy of
Sciences of the United States of America, 103, 10771–10774. http://dx
.doi.org/10.1073/pnas.0511152103
Bogaert, A. F., Skorska, M., Wang, C., Gabriea, J., MacNeil, A. J.,
Hoffarth, M. R.,...Blanchard, R. (2018). Male homosexuality and
maternal immune responsivity to the Y-linked protein NLGN4Y. Pro-
ceedings of the National Academy of Sciences of the United States of
America, 115, 302–306. http://dx.doi.org/10.1073/pnas.1705895114
Boyd, A., Golding, J., Macleod, J., Lawlor, D. A., Fraser, A., Henderson,
J.,...Davey Smith, G. (2013). Cohort profile: The “children of the
90s”—the index offspring of the Avon Longitudinal Study of Parents
and Children. International Journal of Epidemiology, 42, 111–127.
http://dx.doi.org/10.1093/ije/dys064
Brant, R. (1990). Assessing proportionality in the proportional odds model
for ordinal logistic regression. Biometrics, 46, 1171–1178. http://dx.doi
.org/10.2307/2532457
Brown, W. M., Hines, M., Fane, B. A., & Breedlove, S. M. (2002).
Masculinized finger length patterns in human males and females with
congenital adrenal hyperplasia. Hormones and Behavior, 42, 380 –386.
http://dx.doi.org/10.1006/hbeh.2002.1830
Burgess, P. M., Mazzocco, L., & Campbell, I. M. (1987). Discriminant
validity of the Crown-Crisp Experiential Index. British Journal of Med-
ical Psychology, 60, 61– 69.
Calzo, J. P., Antonucci, T. C., Mays, V. M., & Cochran, S. D. (2011).
Retrospective recall of sexual orientation identity development among
gay, lesbian, and bisexual adults. Developmental Psychology, 47, 1658 –
1673. http://dx.doi.org/10.1037/a0025508
Crisp, A. H., Jones, M. G., & Slater, P. (1978). The Middlesex Hospital
Questionnaire: A validity study. British Journal of Medical Psychology,
51, 269 –280. http://dx.doi.org/10.1111/j.2044-8341.1978.tb02472.x
Crown, S., Duncan, K. P., & Howell, R. W. (1970). Further evaluation of
the Middlesex hospital questionnaire (MHQ). The British Journal of
Psychiatry, 116, 33–37. http://dx.doi.org/10.1192/bjp.116.530.33
Del Giudice, M. (2014). Middle childhood: An evolutionary-
developmental synthesis. Child Development Perspectives, 8, 193–200.
http://dx.doi.org/10.1111/cdep.12084
Del Giudice, M., Gangestad, S. W., & Kaplan, H. S. (2015). Life history
theory and evolutionary psychology. In D. M. Buss (Ed.), The handbook
of evolutionary psychology (2nd ed., pp. 88 –114). New York, NY:
Wiley. http://dx.doi.org/10.1002/9781119125563.evpsych102
Diamond, L. M. (2008). Female bisexuality from adolescence to adult-
hood: Results from a 10-year longitudinal study. Developmental Psy-
chology, 44, 5–14. http://dx.doi.org/10.1037/0012-1649.44.1.5
Ellis, B. J. (2004). Timing of pubertal maturation in girls: An integrated life
history approach. Psychological Bulletin, 130, 920 –958. http://dx.doi
.org/10.1037/0033-2909.130.6.920
Ellis, L., & Ames, M. A. (1987). Neurohormonal functioning and sexual
orientation: A theory of homosexuality-heterosexuality. Psychological
Bulletin, 101, 233–258. http://dx.doi.org/10.1037/0033-2909.101.2.233
Floyd, F. J., & Bakeman, R. (2006). Coming-out across the life course:
Implications of age and historical context. Archives of Sexual Behavior,
35, 287–296. http://dx.doi.org/10.1007/s10508-006-9022-x
Frisch, M., & Hviid, A. (2006). Childhood family correlates of heterosex-
ual and homosexual marriages: A national cohort study of two million
Danes. Archives of Sexual Behavior, 35, 533–547. http://dx.doi.org/10
.1007/s10508-006-9062-2
Fujita, M., Roth, E., Lo, Y. J., Hurst, C., Vollner, J., & Kendell, A. (2012).
In poor families, mothers’ milk is richer for daughters than sons: A test
of Trivers-Willard hypothesis in agropastoral settlements in Northern
Kenya. American Journal of Physical Anthropology, 149, 52–59. http://
dx.doi.org/10.1002/ajpa.22092
Gates, G. J. (2011). How many people are lesbian, gay, bisexual, and
transgender? Retrieved from http://williamsinstitute.law.ucla.edu/
research/census-lgbt-demographics-studies/how-many-people-are-
lesbian-gay-bisexual-and-transgender/
Gattis, M. N. (2009). Psychosocial problems associated with homelessness
in sexual minority youths. Journal of Human Behavior in the Social
Environment, 19, 1066 –1094. http://dx.doi.org/10.1080/10911350
902990478
Golombok, S., & Rust, J. (1993). The measurement of gender role behav-
iour in pre-school children: A research note. Child Psychology & Psy-
chiatry & Allied Disciplines, 34, 805– 811. http://dx.doi.org/10.1111/j
.1469-7610.1993.tb01072.x
Grimbos, T., Dawood, K., Burriss, R. P., Zucker, K. J., & Puts, D. A.
(2010). Sexual orientation and the second to fourth finger length ratio: A
meta-analysis in men and women. Behavioral Neuroscience, 124, 278 –
287. http://dx.doi.org/10.1037/a0018764
Hansen, W. B., Paskett, E. D., & Carter, L. J. (1999). The Adolescent
Sexual Activity Index (ASAI): A standardized strategy for measuring
interpersonal heterosexual behaviors among youth. Health Education
Research, 14, 485– 490. http://dx.doi.org/10.1093/her/14.4.485
Hönekopp, J., & Watson, S. (2010). Meta-analysis of digit ratio 2D:4D
shows greater sex difference in the right hand. American Journal of
Human Biology, 22, 619 – 630. http://dx.doi.org/10.1002/ajhb.21054
James, J., Ellis, B. J., Schlomer, G. L., & Garber, J. (2012). Sex-specific
pathways to early puberty, sexual debut, and sexual risk taking: Tests of
an integrated evolutionary-developmental model. Developmental Psy-
chology, 48, 687–702. http://dx.doi.org/10.1037/a0026427
Kane, E. W. (2006). “No way my boys are going to be like that!” Parents’
responses to children’s gender nonconformity. Gender & Society, 20,
149 –176. http://dx.doi.org/10.1177/0891243205284276
Lalumière, M. L., Blanchard, R., & Zucker, K. J. (2000). Sexual orientation
and handedness in men and women: A meta-analysis. Psychological
Bulletin, 126, 575–592. http://dx.doi.org/10.1037/0033-2909.126.4.575
Li, G., Kung, K. T. F., & Hines, M. (2017). Childhood gender-typed
behavior and adolescent sexual orientation: A longitudinal population-
based study. Developmental Psychology, 53, 764 –777. http://dx.doi.org/
10.1037/dev0000281
Li, R., Scanlon, K. S., & Serdula, M. K. (2005). The validity and reliability
of maternal recall of breastfeeding practice. Nutrition Reviews, 63,
103–110. http://dx.doi.org/10.1111/j.1753-4887.2005.tb00128.x
McCarthy, M. M., Herold, K., & Stockman, S. L. (2018). Fast, furious and
enduring: Sensitive versus critical periods in sexual differentiation of the
brain. Physiology & Behavior, 187, 13–19. http://dx.doi.org/10.1016/j
.physbeh.2017.10.030
McClintock, M. K., & Herdt, G. (1996). Rethinking puberty: The devel-
opment of sexual attraction. Current Directions in Psychological Sci-
ence, 5, 178 –183. http://dx.doi.org/10.1111/1467-8721.ep11512422
Nettle, D., Coall, D. A., & Dickins, T. E. (2011). Early-life conditions and
age at first pregnancy in British women. Proceedings of the Royal
Society of London, Series B: Biological Sciences, 278, 1721–1727.
http://dx.doi.org/10.1098/rspb.2010.1726
Office for National Statistics. (2000). Standard Occupational Classification
2000, Volume 1. Structure and descriptions of unit groups: Government
Statistical Service. London, UK: The Stationary Office.
Ott, M. Q., Corliss, H. L., Wypij, D., Rosario, M., & Austin, S. B. (2011).
Stability and change in self-reported sexual orientation identity in young
people: Application of mobility metrics. Archives of Sexual Behavior,
40, 519 –532. http://dx.doi.org/10.1007/s10508-010-9691-3
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
17
EARLY LIFE CONDITIONS AND SEXUAL ORIENTATION
Polderman, T. J., Benyamin, B., de Leeuw, C. A., Sullivan, P. F., van
Bochoven, A., Visscher, P. M., & Posthuma, D. (2015). Meta-analysis of
the heritability of human traits based on fifty years of twin studies.
Nature Genetics, 47, 702–709. http://dx.doi.org/10.1038/ng.3285
Quinlan, R. J. (2003). Father absence, parental care, and female reproduc-
tive development. Evolution and Human Behavior, 24, 376 –390. http://
dx.doi.org/10.1016/S1090-5138(03)00039-4
Rahman, Q. (2005). The neurodevelopment of human sexual orientation.
Neuroscience and Biobehavioral Reviews, 29, 1057–1066. http://dx.doi
.org/10.1016/j.neubiorev.2005.03.002
Remafedi, G., Resnick, M., Blum, R., & Harris, L. (1992). Demography of
sexual orientation in adolescents. Pediatrics, 89, 714 –721.
Ribeiro, E., Neave, N., Morais, R. N., & Manning, J. T. (2016). Direct
versus indirect measurement of digit ratio (2D:4D): A critical review of
the literature and new data. Evolutionary Psychology, 14, 1– 8. http://dx
.doi.org/10.1177/1474704916632536
Ross, M. W., & Hafner, R. J. (1990). A comparison of the factor structure
of the Crown-Crisp Experiential Index across sex and psychiatric status.
Personality and Individual Differences, 11, 733–739. http://dx.doi.org/
10.1016/0191-8869(90)90259-T
Saewyc, E. M., Bauer, G. R., Skay, C. L., Bearinger, L. H., Resnick, M. D.,
Reis, E., & Murphy, A. (2004). Measuring sexual orientation in adoles-
cent health surveys: Evaluation of eight school-based surveys. The
Journal of Adolescent Health, 35, 345.e1–345.e15. http://dx.doi.org/10
.1016/j.jadohealth.2004.06.002
Saewyc, E. M., Skay, C. L., Bearinger, L. H., Blum, R. W., & Resnick,
M. D. (1998). Sexual orientation, sexual behaviors, and pregnancy
among American Indian adolescents. Journal of Adolescent Health, 23,
238 –247. http://dx.doi.org/10.1016/S1054-139X(98)00016-0
Savin-Williams, R. C., & Joyner, K. (2014). The dubious assessment of
gay, lesbian, and bisexual adolescents of add health. Archives of Sexual
Behavior, 43, 413– 422. http://dx.doi.org/10.1007/s10508-013-0219-5
Savitz, D. A., Terry, J. W., Jr., Dole, N., Thorp, J. M., Jr., Siega-Riz, A. M.,
& Herring, A. H. (2002). Comparison of pregnancy dating by last
menstrual period, ultrasound scanning, and their combination. American
Journal of Obstetrics and Gynecology, 187, 1660 –1666. http://dx.doi
.org/10.1067/mob.2002.127601
Sear, R., Sheppard, P., & Coall, D. (2018, September 11). Cross-cultural
evidence does not support universal acceleration of puberty in father-
absent households. OSF Preprints. http://dx.doi.org/10.31219/osf.io/
89avh
Simpson, J. A., Griskevicius, V., Kuo, S. I., Sung, S., & Collins, W. A.
(2012). Evolution, stress, and sensitive periods: The influence of unpre-
dictability in early versus late childhood on sex and risky behavior.
Developmental Psychology, 48, 674 – 686. http://dx.doi.org/10.1037/
a0027293
Skorska, M. N., Blanchard, R., VanderLaan, D. P., Zucker, K. J., &
Bogaert, A. F. (2017). Gay male only-children: Evidence for low birth
weight and high maternal miscarriage rates. Archives of Sexual Behav-
ior, 46, 205–215. http://dx.doi.org/10.1007/s10508-016-0829-9
Skorska, M. N., & Bogaert, A. F. (2017). Prenatal androgens in men’s
sexual orientation: Evidence for a more nuanced role? Archives of
Sexual Behavior, 46, 1621–1624. http://dx.doi.org/10.1007/s10508-017-
1000-y
Sohn, K. (2017). The null relation between father absence and earlier
menarche. Human Nature, 28, 407– 422. http://dx.doi.org/10.1007/
s12110-017-9299-6
Tully, K. P., & Ball, H. L. (2013). Misrecognition of need: Women’s
experiences of and explanations for undergoing cesarean delivery. Social
Science & Medicine, 85, 103–111. http://dx.doi.org/10.1016/j.socscimed
.2013.02.039
VanderLaan, D. P., Gothreau, L. M., Bartlett, N. H., & Vasey, P. L. (2011).
Recalled separation anxiety and gender atypicality in childhood: A study
of Canadian heterosexual and homosexual men and women. Archives of
Sexual Behavior, 40, 1233–1240. http://dx.doi.org/10.1007/s10508-010-
9695-z
Wehkalampi, K., Hovi, P., Dunkel, L., Strang-Karlsson, S., Järvenpää,
A. L., Eriksson, J. G.,...Kajantie, E. (2011). Advanced pubertal growth
spurt in subjects born preterm: The Helsinki study of very low birth
weight adults. The Journal of Clinical Endocrinology and Metabolism,
96, 525–533. http://dx.doi.org/10.1210/jc.2010-1523
White, I. R., Royston, P., & Wood, A. M. (2011). Multiple imputation
using chained equations: Issues and guidance for practice. Statistics in
Medicine, 30, 377–399. http://dx.doi.org/10.1002/sim.4067
Xu, Y., Norton, S., & Rahman, Q. (2018). Early life conditions, reproduc-
tive and sexuality-related life history outcomes among human males: A
systematic review and meta-analysis. Evolution and Human Behavior,
39, 40 –51. http://dx.doi.org/10.1016/j.evolhumbehav.2017.08.005
Xu, Y., & Zheng, Y. (2016). The relationship between digit ratio (2D:4D)
and sexual orientation in men from China. Archives of Sexual Behavior,
45, 735–741. http://dx.doi.org/10.1007/s10508-015-0535-z
Xu, Y., & Zheng, Y. (2017). Fraternal birth order, handedness, and sexual
orientation in a Chinese population. Journal of Sex Research, 54, 10 –18.
http://dx.doi.org/10.1080/00224499.2015.1104530
Received April 26, 2018
Revision received December 7, 2018
Accepted January 8, 2019 䡲
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
18 XU, NORTON, AND RAHMAN
































