ArticlePDF AvailableLiterature Review

Abstract and Figures

Ongoing political controversies around the world exemplify a long-standing and widespread preoccupation with the acceptability of homosexuality. Nonheterosexual people have seen dramatic surges both in their rights and in positive public opinion in many Western countries. In contrast, in much of Africa, the Middle East, the Caribbean, Oceania, and parts of Asia, homosexual behavior remains illegal and severely punishable, with some countries retaining the death penalty for it. Political controversies about sexual orientation have often overlapped with scientific controversies. That is, participants on both sides of the sociopolitical debates have tended to believe that scientific findings—and scientific truths—about sexual orientation matter a great deal in making political decisions. The most contentious scientific issues have concerned the causes of sexual orientation—that is, why are some people heterosexual, others bisexual, and others homosexual? The actual relevance of these issues to social, political, and ethical decisions is often poorly justified, however.
Content may be subject to copyright.
Psychological Science in the
Public Interest
2016, Vol. 17(2) 45 –101
© The Author(s) 2016
Reprints and permissions:
DOI: 10.1177/1529100616637616
This article has two related goals. The first is to provide a
current summary of scientific findings regarding sexual
orientation. Although we focus most on causation, we
also address other scientific issues concerning sexual ori-
entation, including its meaning and measurement, sex
differences in its expression, its development, and its
expression across time and place. Regarding causation,
we provide a taxonomy of causal hypotheses and review
evidence for them. These include hormonal, genetic,
social environmental, and nonsocial environmental influ-
ences. Our second goal is less scientific and more analyti-
cal: to criticize and improve common but incorrect
reasoning in this domain. For example, the commonly
phrased question of whether sexual orientation is “a
choice” is a poor one for advancing either scientific
understanding or policy. A more meaningful formulation
is whether sexual orientation is socially influenced.
Our review has led us to the following conclusions.
Sexual orientation refers to relative sexual attraction
to men, to women, or to both. People who are sexually
attracted to the same sex (whom we denote as “nonhet-
erosexual”) represent a minority of adults. Those with
predominantly same-sex attractions comprise fewer than
5% of respondents in most Western surveys. Data from
non-Western cultures are consistent with this conclusion.
There is no persuasive evidence that the rate of same-sex
attraction has varied much across time or place.
Male and female sexual orientations differ in several
respects. Women are more likely to report a bisexual
than an exclusively same-sex orientation; men show the
opposite pattern. Men’s sexual orientations are closely
637616PSIXXX10.1177/1529100616637616Bailey et al.Sexual Orientation, Controversy, and Science
Corresponding Author:
J. Michael Bailey, Northwestern University, Swift Hall 303B, 2029
Sheridan Rd., Evanston, IL 60208-2710
Sexual Orientation, Controversy,
and Science
J. Michael Bailey1, Paul L. Vasey2, Lisa M. Diamond3,
S. Marc Breedlove4, Eric Vilain5,6,7,8, and Marc Epprecht9,10
1Department of Psychology, Northwestern University; 2Department of Psychology, University of Lethbridge;
3Department of Psychology, University of Utah; 4Neuroscience Program, Michigan State University; 5Department
of Human Genetics, University of California, Los Angeles; 6Department of Pediatrics, University of California, Los
Angeles; 7Department of Urology, University of California, Los Angeles; 8Joint International Unit on Epigenetics,
Data, and Politics, Centre National de la Recherche Scientifique, Paris, France; 9Department of History, Queen’s
University; and 10Department of Global Development Studies, Queen’s University
Ongoing political controversies around the world exemplify a long-standing and widespread preoccupation with the
acceptability of homosexuality. Nonheterosexual people have seen dramatic surges both in their rights and in positive
public opinion in many Western countries. In contrast, in much of Africa, the Middle East, the Caribbean, Oceania, and
parts of Asia, homosexual behavior remains illegal and severely punishable, with some countries retaining the death
penalty for it. Political controversies about sexual orientation have often overlapped with scientific controversies. That
is, participants on both sides of the sociopolitical debates have tended to believe that scientific findings—and scientific
truths—about sexual orientation matter a great deal in making political decisions. The most contentious scientific
issues have concerned the causes of sexual orientation—that is, why are some people heterosexual, others bisexual,
and others homosexual? The actual relevance of these issues to social, political, and ethical decisions is often poorly
justified, however.
sexual orientation, causes, sex differences, social implications
by guest on July 5, 2016psi.sagepub.comDownloaded from
46 Bailey et al.
linked to their pattern of sexual arousal to male versus
female erotic stimuli; women’s are not. Women appear
more likely than men to experience same-sex attraction
in the context of close affectionate relationships, and
their patterns of sexual attraction appear more likely to
exhibit change over time.
Across many different cultures, male and female non-
heterosexuality in adulthood tends to be preceded by
childhood gender nonconformity: a pattern of behavior
somewhat like that of the other sex. Childhood gender
nonconformity is a matter of degree, and it can range
from subtle to extreme. Often, it emerges at an early age,
despite conventional gender socialization. Among some
adults, childhood gender nonconformity appears to per-
sist into adulthood.
Individuals’ political attitudes about sexual orientation
tend to correlate with their views of the causes of sexual
orientation. Those who hold positive attitudes (i.e., that
there is nothing inherently wrong with nonheterosexual-
ity or its open expression) have tended to believe that
sexual orientation is due to nonsocial causes such as
genetics. Those who hold negative attitudes (i.e., that
nonheterosexuality is undesirable or immoral and that
society should restrict its free expression) have tended to
believe that homosexuality has social causes, such as
early sexual experiences and cultural acceptance of non-
heterosexuality. We refer to these as the “nonsocial” and
“social” hypotheses, respectively. Both hypotheses
require direct scientific support; neither can claim confir-
mation solely because support for the other is weak.
No causal theory of sexual orientation has yet gained
widespread support. The most scientifically plausible
causal hypotheses are difficult to test. However, there is
considerably more evidence supporting nonsocial causes
of sexual orientation than social causes. This evidence
includes the cross-culturally robust finding that adult
homosexuality is strongly related to childhood gender
nonconformity; moderate genetic influences demon-
strated in well-sampled twin studies; the cross-culturally
robust fraternal-birth-order effect on male sexual orienta-
tion; and the finding that when infant boys are surgically
and socially “changed” into girls, their eventual sexual
orientation is unchanged (i.e., they remain sexually
attracted to females). In contrast, evidence for the most
commonly hypothesized social causes of homosexual-
ity—sexual recruitment by homosexual adults, patterns
of disordered parenting, or the influence of homosexual
parents—is generally weak in magnitude and distorted
by numerous confounding factors.
With respect to scientific and social policy, we offer
three general conclusions:
1. Scientifically, sexual orientation is an important,
fundamental trait that has been understudied
because it is politically controversial. This is a mis-
take. In fact, the more politically controversial a
topic, the more it is in the public interest to illumi-
nate it in a revealing and unbiased manner. Our
article is offered in the spirit of progress toward
that end.
2. Scientists, activists, and policy makers should rea-
son more carefully regarding potential ethical or
policy implications of scientific findings. For
example, the issue of whether sexual orientation
is chosen represents intellectual confusion, and
no scientific finding will illuminate this issue in
any interesting way. Although clumsy reasoning
may advantage a particular political position in
the short term, in the long term, clear thinking is
best for everyone.
3. The most common meaningful controversy across
time and place has concerned the extent to which
homosexuality is socially influenced and, more
specifically, whether or not it spreads as a result of
contagion and social tolerance. There is no good
evidence that either increases the rate of homo-
sexual orientation, although tolerance may facili-
tate behavioral expression of homosexual desire.
Suppressing homosexual behavior imposes an
immense burden on homosexually oriented peo-
ple and serves no apparent legitimate social goal
that cannot be reached in other ways.
The political rights of lesbian, gay, and bisexual (LGB)
men and women have dramatically improved in many
Western countries during the past 50 years. In the United
States, for example, the Supreme Court ruled in June
2015 that individual states must allow marriages between
same-sex couples. This result would have been unthink-
able in 1965, when homosexual behavior was illegal,
homosexual inclinations were a source of shame, and
most Americans believed homosexuality was a mental
illness (e.g., “The Homosexual in America, 1966). As of
May 2015, 118 nations do not criminalize homosexual
behavior (Carroll & Itaborahy, 2015).
The trajectory of LGB rights has been quite different in
many other parts of the world, however. Currently, 75
countries legally proscribe homosexual behavior. Eleven
countries—all in Africa, Asia and the Middle East—retain
the death penalty as a possible sanction for homosexual
acts1 (International Lesbian and Gay Association, 2015;
Stewart, 2015).
It might be tempting to assume that much of the world
is lagging behind but will ultimately follow the more
accepting Western nations toward tolerance. That out-
come is not assured, however. In some nations, tolerance
of homosexuality appears to be decreasing. For example,
Uganda has been struggling with the issue of whether to
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 47
increase penalties for criminal offenses related to homo-
sexuality. Same-sex relationships are already illegal in
Uganda (as they are in most, but not all, African coun-
tries), with penalties as severe as 14 years in prison, but
many Ugandan lawmakers have sought to increase pen-
alties substantially. Early versions of proposed legislation
included the death penalty for certain offenses (e.g.,
homosexual acts committed by a person with HIV).
It is important to emphasize that the Ugandan
approach to homosexuality is not uniform across Africa—
some African nations are quietly moving toward increased
LGB rights (e.g., Cowell, 2013), as are parts of the Middle
East, Asia, the Caribbean, and Oceania (e.g., “Nepal Court
Rules on Gay Rights,” 2007; Brodie, 2014; Hu, 2015;
Lavers, 2012; Littauer, 2014). Equally, it is important to
note that Uganda is not alone in its attempts to curtail
LGB rights (e.g., Canning, 2013; Garcia, 2014). Despite
their different political trajectories, there are important
similarities among almost all modern nations. All have
histories of anti-homosexual prejudice.
Another similarity between pro- and anti-homosexual
forces worldwide has been intense debate about scien-
tific findings regarding sexual orientation. In the West,2
there has been a strong correlation between political and
moral positions about homosexuality and certain scien-
tific positions about the causes of homosexuality. For
example, in the United States, the National Association
for Research & Therapy of Homosexuality, whose mem-
bers believe that homosexuality is an undesirable and
treatable condition, maintains a website that includes
material attempting to debunk claims that sexual orienta-
tion is inborn. Scientific findings have often been used to
either support or attack gay rights (Horton, 1995; Pitman,
2011; Sprigg, 2012). In Uganda, anti-homosexual activists
from the United States have been influential (Gettleman,
2010; Throckmorton, 2010), and science has also figured
prominently in their campaigns for anti-homosexual leg-
islation. President Yoweri Museveni initially refused to
sign the Anti-Homosexuality Act because he was unsure
whether scientific evidence indicated that sexual orienta-
tion is either inborn or acquired. Only in the latter case
did he believe that the anti-homosexual legislation could
be justified (Mugerwa, 2014). Accordingly, he asked for a
summary of the relevant scientific evidence. The result-
ing statement provided by the Uganda Ministry of Health
(Aceng, 2014) removed Museveni’s doubts, and he signed
the bill. The Constitutional Court in Uganda subsequently
struck down the bill, and currently it is uncertain whether
legislators will try again (Feder, 2014; Williams, 2015).
The Present Article
After President Museveni’s call for scientific evidence
about the causes of homosexuality, some of us were
approached by persons hoping to affect the course of
events in Uganda. They requested that we provide a sci-
entific statement regarding the current status of the sci-
ence of sexual orientation (Throckmorton, 2014). The
delay between Museveni’s reconsideration and his deci-
sion was brief, however, and not conducive to a thor-
ough and considered scientific review. Subsequently, we
decided to write this article, with the aim of providing
one. Collectively, we are well positioned to write such an
article, given that the five first authors conduct comple-
mentary programs of research that inform on the science
of sexual orientation and the last author has published
extensively on nonheterosexuality in Africa. We have not
attempted to time the completion of the article to coin-
cide with any particular decision by the Ugandan govern-
ment, but we hope that policymakers in Uganda and
elsewhere will find it useful. We expect some of the
issues we write about to be discussed in political debates
about homosexuality—both in Uganda and in the rest of
the world—for the foreseeable future.3
Our article has two main goals: first, to review the cur-
rent science of sexual orientation, and second, to con-
sider the relevance of scientific findings to political
debates about homosexuality. Regarding the first goal,
we have focused most effort on the question of causa-
tion. The question of whether sexual orientation is influ-
enced—and to what degree—by specific aspects of
nature and nurture is the most important and contentious
scientific question at issue. Yet we have also addressed
several related scientific issues concerning sexual orien-
tation, including its meaning, prevalence, sex differences,
development, and universality.
Regarding the second goal of this article, we believe
that vast amounts of time have been wasted through the
use of imprecise language and dubious arguments con-
cerning the linkage of scientific findings on sexual orien-
tation to political conclusions regarding LGB rights. We
mean to correct the most common and serious linguistic
and logical mistakes in this arena. More specifically, we
argue below that the links between scientific findings
and desirable social policies have often been overstated
and misidentified. We expect our review will clarify
which research questions are potentially politically
important and which are “merely” scientifically impor-
tant. We hope to eliminate, or at least to reduce, long-
standing arguments that mix the wrong scientific and
political questions. If we can do so, perhaps more prog-
ress can be made in resolving unavoidable rather than
unnecessary conflict. Thus, we have ambitions to convey
basic science accurately and to influence political discus-
sions rationally.
The science of sexual orientation is in the public
interest for at least two reasons. First, as we have noted,
such science is frequently—if not always correctly—
used to support political, social, and moral conclusions
regarding homosexuality. Second, the science of sexual
by guest on July 5, 2016psi.sagepub.comDownloaded from
48 Bailey et al.
orientation—basic inquiry into why some people are
sexually attracted to men and others to women—is
intrinsically interesting. Because there is so much public
interest in the science of sexual orientation, much rele-
vant literature on this topic comes from sources outside
of conventional scientific journals, such as news media,
political advocacy writings, and blogs. We have freely
cited these unconventional sources where relevant.
The science of sexual orientation comprises a very
large body of empirical findings, and so we must narrow
it in two ways. First, we focus on research areas we
believe are most relevant to the public interest, in both of
the respects mentioned above. Second, acknowledging
current valid concerns about the excess of statistically
significant—but incorrect—scientific findings (Simmons,
Nelson, & Simonsohn, 2011), we have tried to focus on
the most well-established findings. In addition, we focus
on research that has garnered considerable public atten-
tion, such as Hamer, Hu, Magnuson, Hu, and Pattatucci’s
(1993) genetic linkage study, LeVay’s (1991) brain study,
and Regnerus’s (2012a) study of children with nonhetero-
sexual parents. Although we are limiting our review in
some ways, we are expanding it in others. Namely, we
want to address not only the preoccupations of scientists
but also those of non-academics interested in sexual-ori-
entation science. With these goals in mind, we com-
mence our scientific review.
What Is Sexual Orientation?
Four related phenomena fall under the general rubric of
sexual orientation, but they are conceptually and empiri-
cally distinguishable. They are listed here not in order of
importance but in an order that reflects their degree of
historical attention. The first phenomenon, sexual behav-
ior, consists of sexual interactions between persons of
the same sex (homosexual), the other sex (heterosexual),
or both sexes (bisexual). The second phenomenon, sex-
ual identity, is one’s self-conception (sometimes dis-
closed to others and sometimes not) as a homosexual,
bisexual, or heterosexual person. The third phenomenon
of sexual orientation is one’s degree of sexual attraction
to the same sex, both sexes, or the other sex. The fourth
phenomenon is one’s relative physiological sexual arousal
to men versus women (or to male vs. female erotic stim-
uli), which is more closely related to other aspects of
sexual orientation in men than in women.4
Terminology also differs among the different phe-
nomena of sexual orientation. People identify as “gay,
“lesbian,” “bisexual,” or “straight.” Scientists primarily
concerned with the consequences of same-sex behavior
may refer to “men who have sex with men.” In this
report, we refer to sexual attraction patterns as “homo-
sexual,” “bisexual,” or “heterosexual.”5 Alternatively, it is
sometimes more relevant to distinguish individuals not
according to whether they are attracted to same-sex or
other-sex partners but according to whether they are
attracted to men (androphilic) or women (gynephilic).
In this usage, both heterosexual women and homosex-
ual men would be considered androphilic because both
groups are attracted to men; both heterosexual men and
homosexual women would be considered gynephilic
because both groups are attracted to women.
Although the four aforementioned phenomena of
sexual orientation (behavior, attraction, identity, and
arousal) tend to go together—homosexually oriented
persons tend to identify as gay or lesbian and to have
sex with same-sex partners—they do not always. For
example, some men who identify as straight/heterosex-
ual have sex with other men and appear to be most
strongly attracted to men. Some adolescents engage
in homosexual activity yet grow up to identify and
behave as heterosexuals. Similarly, some individuals pur-
sue same-sex relationships in sex-segregated environ-
ments, such as boarding schools, prisons, or the military,
but resume heterosexual relationships once other-sex
partners are available. Moreover, the degree of associa-
tion among homosexual attraction, behavior, and iden-
tity varies across individuals in different cultural contexts.
For example, in some cultures and communities, homo-
sexually attracted men regularly engage in same-sex
behavior while still maintaining a heterosexual identity.
In other cultures and communities, such a pattern may
be less common, and homosexually attracted men may
find it difficult to find male partners without identifying
themselves as homosexual or bisexual.
Sexual orientation is defined here as attraction to
members of the same sex, both sexes, or the other sex.
Most researchers studying sexual orientation focus on
self-reported patterns of sexual attraction rather than sex-
ual behavior or identity, because sexual behavior and
identity can be extremely constrained by local culture
and because sexual attraction motivates behavior and
identity, rather than vice versa.
Measurement of sexual orientation
Many scientific studies related to sexual orientation have
compared subjects who have been recruited on the basis
of identifying as either exclusively homosexual/gay or
exclusively heterosexual/straight. That is, such studies
have ignored or even excluded bisexually attracted indi-
viduals. On the one hand, this approach to studying sex-
ual orientation is incomplete. On the other hand, findings
from studies using this approach need not be misleading,
provided we acknowledge their limitations.
Two general approaches to measuring sexual orienta-
tion have dominated scientific research. By far the most
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 49
common approach uses self-report measures. Less com-
mon are psychophysiological measures—in particular,
measures of genital arousal in men. Table 1 lists a variety
of measures that have been used.
Self-report measures. The Heterosexual-Homosexual
Rating Scale, commonly referred to as the Kinsey Scale, is
the best-known self-report measure of sexual orientation.
The scale ranges from 0 (representing entirely hetero-
sexual orientation) to 6 (representing entirely homosex-
ual orientation); the middle score of 3 represents a
bisexual orientation with equal attraction to men and
women.6 The other scores represent gradations between
those anchors. Kinsey famously justified his measure of
sexual orientation by asserting,
Males do not represent two discrete populations,
heterosexual and homosexual. The world is not to
be divided into sheep and goats. Not all things are
black nor all things white. It is a fundamental of
taxonomy that nature rarely deals with discrete
categories. Only the human mind invents categories
and tries to force facts into separated pigeon-holes.
The living world is a continuum in each and every
one of its aspects. The sooner we learn this
concerning human sexual behavior, the sooner
we shall reach a sound understanding of the
realities of sex. (Kinsey, Pomeroy, & Martin, 1948,
pp. 638–639).
Although Kinsey was ultimately making an empirical
claim requiring scientific evidence, his scale provided a
simple way of collecting useful data, and it has become
widely used. It is important to note that the Kinsey Scale
is a self-report instrument. This makes sense and is often
optimal—who better to ask about people’s sexual orien-
tations than the people we want to know about?—but
self-reported Kinsey Scale scores can sometimes be inac-
curate or incomplete, especially when people want to
conceal their sexual orientations or when they are con-
fused or conflicted about their sexual feelings.
Genital measures. Pioneering sex researcher Kurt
Freund invented a technique for assessing penile erec-
tion in response to different kinds of sexual stimuli as a
window on men’s sexual orientation (Freund, 1963).
General terms for the measurement of penile erection
include penile plethysmography (PPG). Freund’s particu-
lar method used pictures of nude men and women as
stimuli and a barometer-like contraption placed over
men’s genitals to measure their erections via changes in
air pressure. Freund’s technology is sensitive to small
increases in penile erection, but it is onerous—for exam-
ple, it requires a technician to place the instrument. Thus,
most research on male sexual orientation with genital
measurement has employed circumferential PPG mea-
sures, such as the penile strain gauge, which are some-
what less sensitive to small changes in penile erection
but less difficult to use (Janssen, 2002; Kuban, Barbaree,
& Blanchard, 1999). Increasingly, researchers have begun
to use videos rather than still images as stimuli, because
the former evoke more arousal and thus allow better
PPG-measured arousal patterns are considered homo-
sexual (or androphilic) when a man’s arousal to adult
male stimuli substantially exceeds his arousal to adult
female stimuli and heterosexual (or gynephilic) when the
opposite pattern occurs. Bisexual arousal patterns do not
necessarily imply equal levels of arousal to male and
female stimuli. Rather, a man is considered to have a
bisexual arousal pattern when the absolute difference
between his arousal to male versus female stimuli is
smaller than the absolute difference observed among
heterosexual and homosexual men. To clarify, a homo-
sexual man is typically much more aroused by male than
female stimuli, and a heterosexual man is typically much
more aroused by female than male stimuli. Both of these
patterns yield large absolute differences between arousal
to female versus male stimuli. A bisexual man, in con-
trast, should have levels of arousal to female and male
stimuli that are not as discrepant, yielding a smaller abso-
lute difference.
Although Freund invented PPG to assess male sexual
orientation as we mean it here (i.e., homosexual vs. het-
erosexual orientation), that application was never com-
mon and has become less so. Use of PPG to assess typical
male sexual orientation is almost exclusively done in the
context of basic scientific research (i.e., research aiming
to answer scientific questions as opposed to applied
research on clinical populations, which has the goal of
developing useful treatments or assessments), and below
we review some important findings from research using
that technique. The most common practical applications
of PPG assessment have been in the area of diagnosis
and treatment of erectile dysfunction (Broderick, 1998)
and detection of pedophilia, typically among men
accused or convicted of sex offenses (e.g., Blanchard,
Klassen, Dickey, Kuban, & Blak, 2001). One general
implication of this work is that PPG can be useful in
assessing the sexual interests of men who wish to hide
them. In studies employing PPG of normal men who
vary in sexual orientation—men who presumably have
nothing to hide—very high correlations are generally
obtained between genital and self-report measures (e.g.,
Chivers, Rieger, Latty, & Bailey, 2004).
When a man’s PPG-assessed arousal pattern differs
from his self-reported sexual orientation, what should we
conclude? In some cases, PPG measurement is poor,
by guest on July 5, 2016psi.sagepub.comDownloaded from
Table 1. Measures of Sexual Orientation.
Measure Description Strengths Weaknesses
Kinsey Scale 7-point scale of relative attraction to men
and women
Relatively non-invasive
Reduced Kinsey Scale 5-point scale of relative attraction to men
and women
Relatively non-invasive
Used in more recent surveys
Sexual identity Self-reported orientation as “gay/
homosexual,” “bisexual,” or “straight/
Simple and widely understood Self-report
No gradation within categories
Some “other” identities (e.g., queer) are
ambiguous with respect to sexual orientation
Sexual behavior Sexual activity with men and/or women Avoids labels and the need for
The same sexual behavior pattern may reflect a
variety of motivations
Can be assessed only for sexually active persons
Cannot be observed (usually), so assessment
relies on self-report
Klein Sexual
Orientation Grid
Multidimensional grid that assesses seven
different aspects of past, present, and
“ideal” romantic and sexual variables
Assesses more dimensions than other
Little research has been conducted on the
validity of distinctions among different
More onerous than single-item measurements
Genital measurement Blood flow to genitals, as measured using a
psychophysiological apparatus
Measure most closely captures male
sexual orientation
Not very voluntarily manipulable
In males, can sometimes be more
accurate than self-report
Onerous and invasive
Sometimes response is too weak for good
Does not provide a good measure of female
sexual orientation
by guest on July 5, 2016psi.sagepub.comDownloaded from
Measure Description Strengths Weaknesses
fMRI Neural activation while viewing erotic male
and female stimuli
Less invasive than genital measurement
In principle (though this has not yet
been demonstrated), can sometimes
exceed the accuracy of self-report
May be used with both sexes
Viewing time Relative time spent looking at pictures of
attractive men versus attractive women in
a series of photographs
Less onerous and invasive than genital
In principle (though this has not yet
been demonstrated), can sometimes
exceed the accuracy of self-report
May be used with both sexes
Can be voluntarily manipulated
Pupil dilation Relative eye pupil dilation while viewing
(typically erotic) pictures or movies
featuring males versus females
Less invasive than genital assessment
In principle (though this has not yet
been demonstrated), can sometimes
exceed the accuracy of self-report
May be used with both sexes
Potential for voluntary manipulation is unclear
Investigated by few studies so far
Implicit measures (e.g.,
priming, Implicit
Association Task)
Timing of classifications of male and female
stimuli (generally pictures)
Less invasive than genital assessment
In principle (though this has not yet
been demonstrated), can sometimes
exceed the accuracy of self-report
May possibly be used with both sexes,
although to date has been used only
with men
Investigated by few studies so far
Table 1. (continued)
by guest on July 5, 2016psi.sagepub.comDownloaded from
52 Bailey et al.
generally as a result of low levels of induced erection,
and in such cases it would be rash to dismiss a man’s self-
report and better to ignore the PPG data. However, in
cases in which PPG arousal is very well measured—
meaning at least that a man produces a clear pattern of
arousal that is potentially repeatable—PPG arousal is the
better measure. This is especially true when there is rea-
son to doubt a man’s self-report. For this reason, PPG
arousal measures are important, both conceptually and
scientifically, in examining certain issues related to sexual
Measures of female genital arousal also exist. The most
common uses vaginal photoplethysmography (VPP), a
technique that is sensitive to changes in blood flow in the
vagina. Women’s genitals, like men’s, have increased
blood flow during sexual arousal. Women’s patterns of
genital arousal to male versus female sexual stimuli do
not mirror those of men, and they are different in a way
that prevents VPP from being useful in measuring female
sexual orientation. We discuss this evidence below in the
“Sex Differences in Expression of Sexual Orientation”
Other non-self-report measures. During the past decade,
several other measures of sexual orientation have been
studied that do not rely on self-report. These include
viewing time (time spent viewing pictures of males
vs. females or rating them for attractiveness; Israel &
Strassberg, 2009; Rullo, Strassberg, & Israel, 2010), fMRI
activation (activation of relevant brain areas in response
to viewing male versus female erotic images; Safron
etal., 2007), implicit attitudes (Snowden, Wichter, & Gray,
2008), and pupil dilation (pupil dilation while viewing
pictures or videos of males vs. females; Rieger & Savin-
Williams, 2012). Studies using these measures have
yielded strong correlations with self-reported sexual ori-
entation in samples of individuals with no apparent rea-
son to give inaccurate self-reports. These measures are
more onerous to administer than self-report measures,
but some have the advantage of being less onerous than
genital arousal measures. Many of these newer measures
are probably less objectionable than invasive genital
measures to many potential research subjects, and in that
sense they are easier to use. All require considerably
more research to understand their strengths and weak-
nesses compared with genital and self-reported arousal
and with each other.
How prevalent is nonheterosexual
One of the most common questions scientists are asked
concerning sexual orientation concerns the popula-
tion prevalence of homosexuality and bisexuality. It is
impossible to provide precise estimates, for several rea-
sons. First, the different phenomena associated with
homosexuality and bisexuality—behavior, identity, and
sexual orientation—vary in frequency. For example, peo-
ple who identify as heterosexual may still engage in
homosexual sex and admit homosexual attraction, and
hence one would find different population estimates
depending on which phenomenon one assessed. Sec-
ond, the different phenomena associated with homosex-
uality and bisexuality may vary over the life course, and
hence one would find different population estimates
depending on whether one assessed individuals’ current
patterns of behavior and attractions versus their total life-
time history of behavior or attractions. For example, the
percentage of people who have ever had a homosexual
experience is larger than the percentage of people who
have had one during the past year, which, in turn, is
larger than the percentage of people who have had only
homosexual experiences for their entire lives. Third,
homosexuality remains stigmatized to some degree even
in the most liberal nations (Kohut, 2014), and thus some
individuals may be motivated to underreport homosex-
ual attractions, identity, and behavior. Fourth, because
some aspects of homosexuality—especially homosexual
identity and exclusively homosexual attractions—are
uncommon, precision requires large and representative
samples, which are expensive and difficult to survey.
Finally, there is no good reason to expect that a single set
of estimated frequencies applies to all places and times.
This is especially likely to be true for sexual identity and
behavior, which seem to us far more culturally malleable
than sexual attraction.
Kinsey conducted the first large surveys of homosexu-
ality in the United States during the 1940s (Kinsey etal.,
1948). His results shocked readers because they made
homosexual behavior and attractions appear so common.
For example, 37% of men surveyed admitted having had
a homosexual experience. Most of these occurred during
adolescence, perhaps indicating brief experimentation.
Approximately 10% of the men had been more or less
exclusively homosexual for at least 3 years during adult-
hood—this is the origin of the “10% of people are homo-
sexual” assertion that was commonly made until recent
and more representative surveys supported lower rates.
About 4% of his male respondents had been homosexual
for their entire lives.
Starting in the 1980s and motivated by epidemiolo-
gists’ need for better numbers to monitor the AIDS epi-
demic, several large and careful surveys of sexual
behavior have been conducted, primarily in North Amer-
ica, Europe, and Australia. Most of these assessed aspects
of homosexuality as well as heterosexuality, and most
focused on sexual behavior rather than sexual attraction.
Results of these studies are generally consistent in
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 53
providing far lower numbers than Kinsey’s 10%. Asking
about sexual identity—whether respondents consider
themselves homosexual/gay/lesbian, bisexual, or hetero-
sexual/straight—is perhaps the simplest way to survey
people about sexual orientation. A recent survey of
34,557 U.S. adults yielded rates of 96.6% heterosex-
ual, 1.6% gay or lesbian, and 0.7% bisexual (Ward,
Dahlhamer, Galinsky, & Joestl, 2014). Additionally, 1.1%
of respondents identified as “something else” or said they
“don’t know the answer. These numbers are in reason-
ably close agreement with a recent review of nine large,
careful studies conducted in Western populations (Gates,
2011), which concluded that approximately 3.5% of U.S.
adults identified as gay, lesbian, or bisexual. The only
careful estimation of nonheterosexual orientation for a
non-Western culture focused on Samoan males, and the
resulting estimate of 1.4% to 4.7% for androphilia is
similar to Western estimates (VanderLaan, Forrester,
Petterson, & Vasey, 2013).
Depending on what one means by “nonheterosexual,
nonheterosexual identity may be a conservative measure
of nonheterosexual orientation. In the recent review,
three studies assessed same-sex attraction as well as
identity. The percentage of adults who admitted to “any
homosexual feelings” ranged from 1.8% to 11%, exceed-
ing the percentage identifying as “homosexual” or “bisex-
ual” by factors ranging from 1.5 to 3.1 (Gates, 2011); note
the wide range of the estimates. But what does it mean
to say that one has experienced same-sex attraction “at
least once” (Smith, Rissel, Richters, Grulich, & Visser,
2003) or that one’s sexual attractions are “mostly” directed
toward one sex (Chandra etal., 2011)? In the study that
yielded a figure of 11% for a history of any homosexual
attractions, only 3.3% of respondents said they were as
attracted to the same sex as to the other sex. Are people
who say that they have had at least one but possibly very
few same-sex attractions intermediate between exclu-
sively heterosexual and homosexual people on a con-
tinuum of sexual orientation? Or, alternatively, are they
understanding the term “same-sex attractions” differently
than most people? We defer consideration of these ques-
tions to our section on bisexuality. The percentage of
adults who have ever had a homosexual experience is
also larger than the percentage of adults who identify as
homosexual or bisexual. In the recent review by Gates,
three studies assessed both nonheterosexual identity and
behavior. The percentage of adults reporting a history of
any same-sex sexual interaction ranged from 6.9% to
8.8%, exceeding those reporting a nonheterosexual iden-
tity by ratios ranging from 2:3 to 3:3 (Gates, 2011).
Clearly, no one number can provide an estimate of the
prevalence of nonheterosexual orientation. Even assess-
ments of highly specific aspects of homosexuality have
yielded a fairly wide range of estimates across recent
studies conducted with similar populations. Several con-
clusions are possible, however, with a high degree of
confidence. First, Kinsey’s famous survey likely overesti-
mated the frequencies of nonheterosexual attractions and
experiences. Second, individuals with incidental homo-
sexual feelings and contacts are much more common
than those with substantial (i.e., persistent and strong)
feelings and frequent same-sex experiences. Third, indi-
viduals with substantial homosexual feelings comprise a
small, albeit nontrivial, minority of adults in Western
developed nations—a smaller percentage than suggested
by Kinsey’s data from more than half a century ago—
despite the remarkable increase in tolerance of homo-
sexuality (Voeten, 2012).
Although there may be scientific value in conducting
future surveys of Western subjects to increase the preci-
sion of estimates related to the prevalence of nonhetero-
sexual (and, necessarily, heterosexual) orientation, we do
not see this as a high priority. There have already been a
sufficient number of carefully sampled Western surveys
related to sexual orientation, and hence future meta-
analyses of these data may reveal interesting systematic
patterns. We worry, however, that variation in prevalence
estimates between studies may primarily reflect measure-
ment error, both systematic and random. Asking increas-
ingly detailed questions and perhaps even including
non-self-report measures related to sexual orientation
have the potential to reveal more than yet another care-
fully sampled self-report survey. Additionally, rather than
continuing to survey the same, very similar, Western pop-
ulations, it would be more scientifically useful to survey
more non-Western populations.
Finally, we note that there has been a tendency for
LGB individuals and their advocates to favor higher prev-
alence figures, whereas those opposed to LGB rights
have tended to favor lower figures (e.g., LaBarbera, 2014).
To the extent that this dispute is political, it makes little
sense. If homosexuality is wrong, then it is wrong even if
it is common; if it is not wrong, then nonheterosexual
people deserve their rights regardless of how rare they
are. As we noted earlier, the conflation of political motives
and scientific findings has been common in debates
related to sexual orientation, and when this conflation is
mistaken, it is to the detriment of both politics and
Sex Differences in Expression of
Sexual Orientation
In summarizing recent prevalence figures for nonhetero-
sexual orientation, we deferred discussing a consistent,
large, and interesting sex difference. Among those who
identify as nonheterosexual, women tend to identify as
bisexual whereas men tend to identify as homosexual
by guest on July 5, 2016psi.sagepub.comDownloaded from
54 Bailey et al.
(Gates, 2011). Studies that have assessed sexual attrac-
tions have found similar results. Figure 1 summarizes the
distributions from a recent review of sexual orientation
among 71,190 adult men and 117,717 adult women who
reported their sexual orientations using a 5-point scale:
1= entirely heterosexual/straight, 2 = mostly heterosexual/
straight, 3 = bisexual, 4 = mostly homosexual/gay/lesbian,
5 = entirely homosexual/gay/lesbian (Savin-Williams &
Vrangalova, 2013). The figure reveals markedly different
shapes for the two sexes. The large majority of both sexes
rated themselves as “entirely heterosexual,” although
men (93.2%) were considerably more likely to do so than
women (86.8%). Many more women than men rated
themselves as “mostly heterosexual.” For women, there is
a steep decline in the frequency of “mostly heterosexual”
to “bisexual” ratings, and then a more gradual decline to
“mostly homosexual” and “entirely homosexual” ratings,
which have similar frequencies. For men, the least fre-
quently self-rated category is “bisexual,” with a slight rise
in the frequency of “mostly homosexual” and “entirely
homosexual” ratings. The male distribution is bimodal.
Bimodal distributions are rare and potentially interesting,
suggesting that two groups are different in kind and
notmerely degree. For example, a bimodal distribution
would be expected if there were a single cause underly-
ing the distinction between two groups, whereas a con-
tinuous distribution would be expected if there were
multiple causes. The classic human example is biological
sex, which is caused by a dichotomous genotype (XX vs.
XY) that leads to a bimodal distribution of prenatal tes-
tosterone, generating categorical sex differences in some
anatomical features, such as genitalia. One conclusion
from the sex difference in distributions of sexual orienta-
tion is clear, however: Among the Western populations
that have been studied, bisexual patterns of attraction are
rarer in men than in women, and exclusively homosexual
attractions are rarer in women than in men. The reasons
for this sex difference are not known. It might reflect
prenatal biological sex differences; it might reflect the
influence of cultural factors that create different social
contexts for the development of female versus male
same-sex sexuality; it might reflect differences in wom-
en’s and men’s susceptibility to such social influences
(Baumeister, 2000); or it might simply result from mea-
surement error. If there are innate sex differences in the
causation of sexual orientation, then we should expect to
find that the correlates of sexual orientation differ consid-
erably between men and women. This is the topic we
address next.
Sex differences in category-specific
sexual arousal
Sexual arousal comprises both subjective and genital
response to a sexual stimulus. In men, sexual orientation
is closely related to a pattern of genital sexual arousal.
Specifically, homosexual men show genital arousal to
men (or, in the lab, sexual stimuli depicting men),
whereas heterosexual men show genital arousal to
women (or, in the lab, sexual stimuli depicting women).
This pattern has been called “category-specific” because
men’s genital sexual arousal is specific to the category of
person to whom they are most attracted (Chivers etal.,
2004). Indeed, one of us has argued that male sexual
orientation can be precisely defined as a pattern of sex-
ual arousal, for two reasons (Bailey, 2009). First, as we
have mentioned, sexual arousal patterns reliably reflect
male sexual orientation. In cases where a man’s self-
reported sexual orientation diverges from his sexual
arousal pattern (and in which his sexual arousal pattern
is robust and dichotomous, consisting of strong erections
to one sex but not the other), we believe his sexual
arousal pattern best represents his sexual orientation—if
not necessarily his self-chosen sexual identity. Second,
and more important, is the direction of causation. Sexual
arousal pattern motivates sexual behavior, which pro-
vides most individuals with the fundamental basis for
their sexual self-identification. In contrast, there is no evi-
dence to suggest that individuals can consciously alter
their genital arousal to fit a certain sexual identity label.
Women show a much different pattern of linkage
between sexual orientation and genital arousal (i.e., geni-
tal blood flow as measured by VPP). In the laboratory,
heterosexual women show approximately equal levels of
genital arousal to male and female stimuli (despite some-
times reporting that they find male sexual stimuli more
arousing). Homosexual women show a pattern of genital
sexual arousal that is slightly category-specific (i.e., they
show stronger genital arousal for female than male stim-
uli), but their genital responses are not as starkly cate-
gory-specific as men’s (Chivers etal., 2004; Chivers, Seto,
& Blanchard, 2007). One potential interpretation of the
sex difference in genital arousal patterns is that female
sexual orientation is fundamentally less category-specific
than male sexual orientation (i.e., that homosexual
women retain some potential for attraction/arousal to
men and heterosexual women retain some potential for
attraction/arousal to women, even if they are unaware of
this potential). This explanation presumes that a woman’s
pattern of genital arousal provides a reliable measure of
her sexual orientation (as is the case for men). An alter-
native explanation is that genital arousal is less reliably
linked to sexual orientation in women than in men. This
explanation remains agnostic as to whether female sex-
ual orientation is less categorical in women than in men;
it simply suggests that the measure of genital arousal
commonly employed—namely, genital blood flow as
assessed via VPP—cannot definitively answer this ques-
tion, because this measure has a weaker pattern of asso-
ciation with female than male sexual orientation.
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 55
There are important differences between measures of
male and female genital arousal that need to be consid-
ered before accepting the idea that women are less cate-
gory-specific than men in their sexual arousal patterns.
Female genital arousal is measured much differently than
male genital arousal (specifically, a device is inserted into
the vagina to measure changes in blood flow, in contrast
to the erectile measure used for men). Nonetheless, sev-
eral findings argue against the possibility that the sex
difference in category-specific genital arousal is a mea-
surement artifact. First, a study of postoperative male-to-
female transsexuals (assessed with VPP) produced a
category-specific pattern of genital arousal similar to that
of men (Chivers etal., 2004). This would not have been
expected to occur if the vaginal measure of arousal was
not effective in detecting category-specific arousal
responses. Second, women also show substantially less
category-specificity in their self-reported patterns of
sexualarousal to male and female stimuli. Third, other
correlates of sexual interest have been shown to be
less category-specific in women than in men, such as
viewing time for female and male sexual stimuli (Lippa,
Patterson, & Marelich, 2010), pupil dilation in response to
such stimuli (Rieger & Savin-Williams, 2012), and fMRI
responses to such stimuli (Sylva etal., 2013).7
Accepting for now the validity of the hypothesis that
women’s sexual arousal patterns are less category-spe-
cific than men’s, the question arises as to why this is the
case. One possibility is cultural—namely, Western cul-
tures produce omnipresent depictions of female beauty,
which are often sexualized, and exposure to these images
from an early age may sensitize both men and women to
experience sexual arousal to the female body. Although
this explanation might account for heterosexual women’s
genital arousal to female stimuli, it cannot account for the
fact that homosexual men, who have experienced just as
much exposure to sexualized images of women, do not
experience sexual arousal to female stimuli. Similarly,
exposure to sexualized images of women fails to account
for the fact that homosexual women show more genital
arousal to male sexual stimuli than homosexual men
show to female sexual stimuli.
Prevalence (%)
Fig. 1. Average prevalence for each of five categories of sexual orientation in recent Western
population surveys. Data are from Table 1 of Savin-Williams and Vrangalova (2013).
by guest on July 5, 2016psi.sagepub.comDownloaded from
56 Bailey et al.
Another possibility is that the difference is based in
fundamental evolved differences between female and
male sexuality (Baumeister, 2000; Chivers & Bailey, 2005),
although currently there is no direct evidence speaking
to this question. Intriguingly, Goy and Goldfoot (1975)
showed over 30 years ago that in many different mam-
malian species, bisexuality is an intrinsically dimorphic
trait that develops (through prenatal hormonal pathways)
in either the male or the female of a species, but never
both. This suggests the provocative possibility that in
humans, women are “the more bisexual sex,” whereas
males are more likely to be exclusively heterosexual or
Regardless of the reason for sex differences in cate-
gory-specificity, this sex difference is likely to produce
additional sex differences in sexuality. Sexual arousal is a
powerful motivating force, and so men’s greater cate-
gory-specificity might be expected to produce more cat-
egorical patterns of sexual behavior (i.e., exclusively
heterosexual behavior among heterosexual men and
exclusively homosexual behavior among homosexual
men). In contrast, the fact that women’s genital arousal is
less category-specific might be expected to produce less
categorical (and hence more “bisexual”) patterns of sex-
ual behavior, which may help to explain the aforemen-
tioned sex difference in bisexual identification.
Sexual fluidity
A related sex difference is the phenomenon of sexual
fluidity, which appears to be especially common among
women (L. M. Diamond, 2009). Sexual fluidity is situa-
tion-dependent flexibility in a person’s sexual respon-
siveness, which makes it possible for some individuals to
experience desires for either men or women under cer-
tain circumstances regardless of their overall sexual ori-
entation. Evidence that some women are especially
sexually fluid includes results from a longitudinal study
of 80 women first interviewed at 16 to 23 years of age
(L.M. Diamond, 2000, 2003a, 2008). At the first interview,
none of the women identified as “heterosexual”; rather,
their reported identities were “lesbian,“bisexual,” or
unlabeled. Many of the women’s sexual feelings toward
women versus men changed over time, although typi-
cally the changes were not large (about 1 Kinsey Scale
point, on average). Yet changes in sexual identity were
common. Two years after the initial interviews, approxi-
mately one-third of the participants changed their sexual
identities (L. M. Diamond, 2000); between the second
and third interview, another quarter of the participants
changed their sexual identities (L. M. Diamond, 2003a);
and between the third and fourth interviews, another
third of the participants changed their sexual identities
(L. M. Diamond, 2008). Usually, these changes were
between adjacent categories (e.g., “heterosexual” and
“bisexual”) rather than larger changes (e.g., from “hetero-
sexual” to “lesbian”). Perhaps more importantly, over
time, women were significantly more likely to adopt
identities that permitted sexual attractions and/or behav-
ior with both sexes (i.e., “bisexual” or unlabeled) than to
adopt exclusively lesbian identities. For these women,
their range of potential attractions was limited by their
sexual orientations, but fluidity allowed movement within
that range. Obviously, these results generalize to nonhet-
erosexual-identified women more readily than to hetero-
sexual-identified women, but Diamond’s findings are
consistent with the findings of non-category-specificity in
heterosexual women’s genital arousal (Chivers & Bailey,
2005; Chivers etal., 2007) as well as with the growing
body of research on “mostly heterosexual” women, who
perceive their underlying orientation to be heterosex-
ual despite experiencing periodic same-sex attractions
(Thompson & Morgan, 2008; Vrangalova & Savin-Williams,
2010). It is an open question whether the nonheterosexual
women in Diamond’s sample were more likely to experi-
ence sexual fluidity or whether they were simply more
aware of it and willing to discuss it openly.
What accounts for the ability of some women to be
sexually fluid? L. M. Diamond (2003b) has theorized that
to an extent, romantic love and sexual desire rely on dif-
ferent motivational systems: the former more on the
attachment or pair-bonding system, the latter more on
the sexual mating system. Although separate, these two
systems may affect each other, and the bidirectional links
(permitting feelings of strong emotional attachment to
give rise to sexual desire) may be especially strong in
women. Another possibility concerns sex differences in
the reproductive context of female versus male sexual
behavior. Whereas all of men’s sexual interactions can
result in reproduction, women’s sexual behavior can
result in reproduction only if it coincides with ovulation.
However, women are capable of becoming sexually
aroused at any point in the menstrual cycle, and although
their sexual motivation appears to increase as a function
of the rising estrogen levels that accompany ovulation,
their sexual arousability during the rest of the month
appears unrelated to estrogen levels (reviewed in Wallen,
1995). Women’s capacity for arousability independent of
ovulation permits them to experience sexual desires and
behaviors with same-sex individuals at no evolutionary
cost. In fact, researchers have argued that sexual behav-
ior with nonreproductive (and even same-sex) partners
can confer multiple evolutionary advantages on females,
such as alliance formation (de Waal, 1987; Hohmann &
Fruth, 2000). Hence, one possibility is that women’s
capacity for sexual fluidity evolved in concert with (or is
an artifact of) their capacity for nonreproductive arous-
ability during nonfertile parts of the menstrual cycle.
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 57
Development and Psychological
Correlates of Sexual Orientation
Having reviewed basic sex differences in orientation, we
now turn to the question of when and how sexual orien-
tation emerges. Behavioral correlates of sexual orienta-
tion are present during childhood, long before a child
has any apparent sexual feelings.
Childhood gender nonconformity
Childhood gender nonconformity—behaving like the
other sex—is a strong correlate of adult sexual orienta-
tion that has been consistently and repeatedly replicated
(Bailey & Zucker, 1995). More specifically, childhood
gender nonconformity comprises the following phenom-
ena among boys: cross-dressing, desiring to have long
hair, playing with dolls, disliking competitive sports and
rough play, preferring girls as playmates, exhibiting ele-
vated separation anxiety, and desiring to be—or believ-
ing that one is—a girl. In girls, gender nonconformity
comprises dressing like and playing with boys, showing
interest in competitive sports and rough play, lacking
interest in conventionally female toys such as dolls and
makeup, and desiring to be a boy. Childhood gender
nonconformity usually emerges by preschool age. It is
important to add that children are usually considered to
be gender-nonconforming only if they persistently
engage in a variety of these behaviors, as opposed to
engaging in a single behavior once or twice. Further-
more, childhood gender nonconformity is not an either/
or trait but, rather, a dimensional one, so the differences
we discuss are a matter of degree and not of kind.
Children who grow up to be nonheterosexual are sub-
stantially more gender nonconforming, on average, than
children who grow up to be heterosexual. There are two
types of studies supporting this. In retrospective studies,
homosexual and heterosexual (and sometimes bisexual)
adults are asked about their childhood behavior. In pro-
spective studies, extremely gender nonconforming chil-
dren are followed into adulthood to assess their sexual
orientations. Both kinds of studies have limitations that in
isolation might lead to skepticism about their validity.
Retrospective studies rely on childhood memories, which
might be distorted in a way that supports stereotypes
about homosexuality and heterosexuality. Prospective
studies often focus on highly unusual children, and one
might worry about generalizing findings from these stud-
ies to less atypical children. Nonetheless, the two kinds
of studies have produced highly convergent findings that
support a strong association between childhood gender
nonconformity and adult nonheterosexuality.
A 1995 review of retrospective studies found large dif-
ferences in retrospectively reported childhood gender
nonconformity between homosexual and heterosexual
men (32 studies; d = 1.3) and women (16 studies; d = 1.0;
Bailey & Zucker, 1995). The review estimated that with
respect to male sexual orientation, 89% of homosexual
men exceeded the heterosexual median score, whereas
2% of heterosexual men exceeded the homosexual
median. For female sexual orientation, the respective fig-
ures were 81% and 12%. Although this review was pub-
lished 20 years ago, there is little reason to suspect that its
results would change much if updated. Furthermore,
despite the skepticism of some writers (e.g., Fausto-
Sterling, 2014), no study has persuasively demonstrated
that retrospective studies exaggerate orientation differ-
ences associated with childhood gender nonconformity.
Retrospective self-report measures have been supple-
mented by childhood home videos provided by hetero-
sexual and homosexual adults, with similar results:
Viewers of the videos could tell, at far better than chance
levels, which children would grow up to be homosexual
and which would grow up to be heterosexual (Rieger,
Linsenmeier, Gygax, & Bailey, 2008). The retrospective
studies suggest that some degree of childhood gender non-
conformity is a common precursor of adult homosexuality
in both sexes. This conclusion holds for both Western and
a wide array of non-Western cultures, including those in
Asia, Latin America, Polynesia, and the Middle East (Bartlett
& Vasey, 2006; Cardoso, 2005, 2009; Vasey, VanderLaan,
Gothreau, & Bartlett, 2011; Whitam & Mathy, 1986).
Several prospective studies of highly gender noncon-
forming children have been conducted, most focusing on
boys (for reviews, see Bailey & Zucker, 1995; Zucker,
2014). These studies have identified children through clin-
ical interventions to help the children accept their birth
sex identity, as opposed to obtaining sex reassignment
surgery in adulthood. Thus, in general, these children
exhibited extreme gender nonconformity as well as dis-
comfort with or confusion about their gender identity. For
males, all studies have found most men to be homosexu-
ally or bisexually oriented at follow-up,8 with rates rang-
ing from 64% (N = 129; Singh, 2012) to 80% (N = 66;
Green, 1987). The two studies including females were
smaller and had more variable outcomes: 32% of 25 sub-
jects (Drummond, Bradley, Peterson-Badali, & Zucker,
2008) and 100% of 10 subjects (Wallien & Cohen-Kettenis,
2008) endorsed bisexual or homosexual feelings. Not all
of the studies had control groups, but all the rates of non-
heterosexual outcomes far exceed plausible epidemio-
logical estimates for the general population, which might
be generously estimated at 5% (see Fig. 1). Thus, the pro-
spective studies also suggest a large association between
(extreme) childhood gender nonconformity and adult
nonheterosexuality. In nearly all the studies, there was
a significant tendency for subjects self-reporting as
heterosexual to be younger than those self-reporting as
by guest on July 5, 2016psi.sagepub.comDownloaded from
58 Bailey et al.
nonheterosexual. This suggests that the prospective stud-
ies may underestimate the likelihood of a nonheterosex-
ual outcome because younger nonheterosexual individuals
are more likely to provide false claims that they are het-
erosexual (Bailey & Zucker, 1995).
Both retrospective and prospective studies support a
strong association between childhood gender nonconfor-
mity and adult nonheterosexuality. This does not mean,
of course, that all nonheterosexual individuals were more
gender nonconforming than average, much less that they
were all extremely gender nonconforming. There are at
least three potentially important implications, however.
First, long before sexual attraction emerges, some chil-
dren who will become nonheterosexual are markedly
different than other children. Indeed, childhood gender
nonconformity is often evident by age 2 (Cohen-Kettenis
& Pfäfflin, 2003). Second, there is little evidence that gen-
der nonconforming children have been encouraged or
taught to behave that way; rather, childhood gender non-
conformity typically emerges despite conventional social-
ization. Third, the differences between children who will
become nonheterosexual and those who will become
heterosexual are related to sex-typed behavior and gen-
der identity (Bailey & Zucker, 1995; Zucker & Bradley,
1995). The content of these differences provides some
potential clues about the causes of sexual orientation,
which we discuss further below.
Onset of sexual attraction
Children cannot know their sexual orientations until they
experience sexual attraction toward the same or the other
sex. When does this happen? At present in North Amer-
ica, it is impossible to study actual children as they
become sexually aware. Such a study would be a politi-
cal minefield if not an ethical one. Therefore, the onset of
sexual attraction has been studied using retrospective
reports of adolescents and adults.
Studies have shown that subjects recall first having
feelings of sexual attraction at age 10, on average
(McClintock & Herdt, 1996). Male and female subjects
report similar ages. Importantly, so do homosexual and
heterosexual subjects. Age 10 is several years before the
typical age of onset of sexual activity (Cavazos-Rehg
etal., 2009). This is consistent with an earlier retrospec-
tive study showing that homosexual men and women
recalled their first homosexual feelings as preceding their
first homosexual experiences by 3 years (Bell, Weinberg,
& Hammersmith, 1981). Although this gap may have
changed since that publication, the important point—to
which we return—is that the large majority of nonhetero-
sexual people recall that homosexual desires preceded
homosexual experiences.
How do nonheterosexual people experience the rec-
ognition of homosexual feelings? In a retrospective study,
homosexual men reported that their early sexual att-
ractions were not necessarily a source of any distress
(Savin-Williams, 1996). Same-sex attractions were often
experienced as an obsession with being near masculine,
often older, same-sex teenagers and adults, such as male
teachers and coaches. Eventually, these men recognized
that their same-sex desires were rarely shared by others,
were not condoned, and should be hidden. Although first
homosexual male experiences can occur in the context
of romantic relationships, it is not uncommon for these
experiences to occur with strangers (Savin-Williams &
Diamond, 2000).
On average, homosexual women describe their early
same-sex attractions as more emotionally than sexually
charged. Lesbian or female bisexual self-identification
often precedes the onset of homosexual activity, and first
same-sex sexual attractions and behaviors are often
experienced within the context of a romantic relationship
(Savin-Williams & Diamond, 2000).
Bisexual-identified individuals typically experience
later recognition of same-sex attractions compared with
homosexual men and women. This may reflect the fact
that bisexual identity usually follows the establishment of
a heterosexual identity (Fox, 1995; Weinberg, Williams, &
Pryor, 1994). Homosexual people are atypical in two
ways: their lack of attraction to the other sex and their
increased attraction to their own. In contrast, bisexual
people are atypical in only the latter sense, and this may
delay their recognition of their own difference.
Adult gender nonconformity
It would be surprising if the large orientation differences
in childhood gender nonconformity disappeared without
a trace by adulthood. Several differences between hetero-
sexual and nonheterosexual adults persist. They include
patterns of occupational and recreational interests (Lippa,
2005a, 2005b). Research indicates that heterosexual men
have greater interest in occupations and hobbies focus-
ing on things and less interest in those focusing on peo-
ple, compared with heterosexual women (Lippa, 2005a;
Su, Rounds, & Armstrong, 2009). In contrast, homosexual
men show a somewhat feminine pattern of interests, and
homosexual women a somewhat masculine one. The ori-
entation differences are large, although smaller than the
sex differences. They are consistent with stereotypes
about occupational differences between homosexual and
heterosexual people. This makes them more difficult to
interpret than childhood differences because both homo-
sexual and heterosexual adults may—either consciously
or unconsciously—mold their behavior in accordance
with societal expectations.
Other differences between homosexual and hetero-
sexual adults include patterns of movement (i.e., gestures
and walking) and speech (i.e., articulation), physical
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 59
presentation (i.e., clothing choices and hairstyles; Bailey,
2003; Rieger et al., 2008; Rieger, Linsenmeier, Gygax,
Garcia, & Bailey, 2010), and even facial appearance (Rule,
Ambady, Adams, & Macrae, 2008). These differences also
tend to be large and help explain the phenomenon by
which individuals can sometimes judge a target’s sexual
orientation by observing superficial aspects of the target’s
nonsexual behavior (“gaydar”). As with gender noncon-
formity in interests, gender nonconformity in movement,
speech, and physical presentation might be influenced
by societal expectations. Homosexual and heterosexual
individuals also tend to differ in their performance on
cognitive and personality tests that show sex differences.
For example, studies of visuospatial abilities and verbal
fluency show that homosexual individuals are shifted in
the direction of the other sex (M. Peters, Manning, &
Reimers, 2007; Rahman, Abrahams, & Wilson, 2003). Sim-
ilarly, studies of neuroticism9 have indicated that homo-
sexual individuals are shifted in a sex-atypical direction
(Lippa, 2008). It is important to note, however, that the
effect sizes for cognitive ability and personality are much
smaller—both for sex and sexual orientation—than the
effect sizes for patterns of interest.
Before leaving the topic of gender nonconformity,
we address a commonly raised question: Might the gen-
der-atypicality of adult homosexual men and women
simply reflect a culturally influenced self-fulfilling
prophecy? In other words, given that society expects
homosexual individuals to be gender atypical, and given
that LGB communities often support and facetiously
celebrate such gender atypicality, perhaps some homo-
sexual people adopt gender-atypical characteristics to
conform to their own stereotypes. Because of the evi-
dence we have reviewed—indicating that gender non-
conformity often begins before a prehomosexual child
even has a sexual orientation or is aware of cultural
stereotypes, and that the link between gender noncon-
formity and nonheterosexual orientation has been found
in a wide variety of cultures—we think it is highly
unlikely that gender nonconformity in LGB populations
represents a self-fulfilling prophecy due to cultural
beliefs. It is possible, however, that cultural stereotypes
sometimes amplify gender nonconformity among LGB
people. Many LGB individuals report that they have
always been fairly gender-typical in dress, appearance,
and interests. It is possible that as these individuals
come to identify as LGB and participate in the LGB
community, they adopt aspects of gender-atypicality.
The past decade has seen a surge of research about
bisexuality. This has reflected (a) growing awareness that
a nontrivial proportion of nonheterosexual people iden-
tify as bisexual; (b) the widespread acknowledgment that
bisexuals have often been both socially and scientifi-
callymarginalized; (c) scientific controversy concerning
bisexual orientation; and, notably, (d) appreciable
researchfunding by the American Institute of Bisexuality
(Denizet-Lewis, 2014). Until recently, research on sexual
orientation often ignored bisexuality, as reflected in the
two most common approaches to dealing with poten-
tially bisexual subjects: excluding them or combining
them with monosexual subjects.
The topic of bisexuality requires special precision in
discussing identity, attractions, behavior, arousal, and ori-
entation, given the prevalence of discrepancies among
these domains. Research has found that some individuals
who identify as bisexual show patterns of sexual arousal
(and sometimes patterns of sexual behavior) that appear
to be predominately heterosexual or homosexual,
whereas some individuals who identify as heterosexual
or homosexual show bisexual patterns of genital arousal,
attraction, or behavior. Such discrepancies reflect wide-
spread variability in individuals’ and communities’ defini-
tions of “bisexual,” as well as variability in individuals’
motivations for identifying as bisexual. Although it seems
reasonable to presume that bisexual-identified individu-
als exhibit both same-sex and other-sex attractions and
sexual behavior, other patterns are possible. Some indi-
viduals identify as bisexual because they have previously
engaged in sexual activity or intimate relationships with
both men and women, even if their current sexual attrac-
tions are exclusively toward the same sex or the other
sex. Other individuals identify as bisexual because they
periodically experience sexual attraction toward both
sexes, even if their sexual behavior and identity is exclu-
sively homosexual. As a result of this diversity, the popu-
lation of individuals reporting bisexual attractions,
behavior, or identity contains individuals with a range of
different orientations.
Transitional versus persistent
Some individuals who will eventually identify as homo-
sexual temporarily adopt a bisexual identity before doing
so (Lever, 1994). Transitional bisexual identification
appears to be more common in men than in women. In
a large U.S. national survey conducted by the LGB-
focused news magazine The Advocate (Lever, 1994), 40%
of gay-identified men reported having previously identi-
fied as bisexual. Two recent longitudinal studies of non-
heterosexual-identified youth found that changes of
identity from bisexual to homosexual were particularly
common among males (Rosario, Schrimshaw, Hunter, &
Braun, 2006; Savin-Williams, Joyner, & Rieger, 2012).
Transitional bisexual identity and behavior may also
occur among women on their way to a homosexual iden-
tity (Lever, 1995), though women’s sexual identities also
by guest on July 5, 2016psi.sagepub.comDownloaded from
60 Bailey et al.
not infrequently change from homosexual to bisexual
(L.M. Diamond, 2008).
Researchers of bisexuality have often acknowledged
the existence of transitional bisexuality (Fast & Wells,
1975; Harwell, 1976; Klein, 1993; H. L. Ross, 1971; M. W.
Ross, 1991), but it remains poorly understood. The very
limited existing evidence supports the idea that at least
among men, transitional bisexuality identity does not
necessarily imply a history of bisexual attractions
(Guittar, 2013). Rather, men may adopt transitional bisex-
ual identities in the process of trying to make sense of
divergent parts of their current and previous attractions
and histories, such as the fact that they may have had
emotionally satisfying romantic relationships with women
despite feeling sexual attractions only toward men, or the
fact that their previous heterosexual encounters may
have been unsatisfying but not distasteful. Others may
feel that it is easier to admit one’s homosexual feelings if
they are not appearing to “rule out” the possibility of
heterosexual attractions and relationships. Finally, some
men may initially identify as bisexual because they have
bisexual patterns of attraction but, eventually, switch to a
gay identity because most of their attractions and all of
their sexual behaviors involve men.
Among women, bisexuality appears to be a more sta-
ble identity pattern, as well as a more stable pattern of
self-reported sexual attraction (L. M. Diamond, 2008;
Savin-Williams & Ream, 2007). This is consistent with the
fact (reviewed earlier) that women are more likely than
men to report bisexual patterns of attraction. Specifically,
a woman who comes out as bisexual is more likely to be
“correct” in this assessment than is a man, and hence
more likely to retain that label over time and to find that
it provides an adequate representation of her attractions
and behavior. Among both men and women, negative
stereotypes about bisexuality, perpetuated within the gay
community as well as in the culture at large, make it dif-
ficult to interpret individuals’ adoption or rejection of
bisexual labels. Just as some individuals with homosex-
ual patterns of attraction may identify as bisexual because
they perceive it to be an easier transition to make than a
direct transition to homosexual, some individuals with
bisexual patterns of attraction may identify as homosex-
ual because they are aware that some members of the
gay community view bisexuals as untrustworthy, clos-
eted, or promiscuous (Tania & Mohr, 2004).
Sexual arousal patterns of bisexual men
It is because of these complexities in interpreting bisex-
ual identities that researchers have turned to more objec-
tive measures of bisexual orientations. Specifically,
researchers have used measures of genital arousal to
determine whether men who identify as bisexual also
have bisexual arousal patterns (given that some men may
identify as bisexual despite having homosexual arousal
To assess bisexual patterns of genital arousal in men,
it is first necessary to decide what such an arousal pattern
should look like, relative to a monosexual pattern of
arousal. One possibility is that individuals with bisexual
arousal patterns should show less of a difference (com-
pared to monosexuals) between their arousal responses
to men versus women (Rosenthal, Sylva, Safron, & Bailey,
2012). Another approach emphasizes the overall degree
of arousal to one’s “less preferred” sex (Rieger, Chivers, &
Bailey, 2005). Consider the first approach. To the extent
that bisexual-identified men have bisexual arousal pat-
terns, they should have smaller absolute differences
between their arousal to male and female stimuli com-
pared with monosexual men. That is, a bisexual arousal
pattern implies a relatively similar degree of arousal to
male and female erotic stimuli. The rationale for this
approach, which focuses on arousal to the less-preferred
sex, is that a bisexual man should show greater arousal
to female stimuli than would be shown by homosexual
men as well as greater arousal to male stimuli than gener-
ally shown by heterosexual men. Thus, a bisexual man’s
arousal to stimuli depicting his less-arousing sex—which-
ever that is—should be greater than that for monosexual
men. The dependent variables using the two approaches
are highly correlated, and the two analyses tend to pro-
vide similar results (Rosenthal etal., 2012).
Three studies have examined the genital arousal pat-
terns of bisexual, heterosexual, and homosexual men.
Two found that bisexual men’s self-reported subjective
sexual arousal patterns were relatively bisexual, com-
pared with those of monosexuals, but that their genital
arousal patterns were not (Cerny & Janssen, 2011; Rieger
et al., 2005). The other study found statistically robust
and strong effects for both subjective and genital mea-
sures, supporting the hypothesis that bisexual men have
relatively bisexual arousal patterns (Rosenthal, Sylva,
Safron, & Bailey, 2011, 2012). Which finding is correct?
Likely, both are true, each for a different subset of bisex-
ual-identified men. The studies that failed to find corre-
spondence between bisexual identity and arousal patterns
used relatively liberal inclusion criteria for bisexuality—
namely, that men considered themselves bisexual and
had self-reported Kinsey Scale scores in the bisexual
range. In contrast, the study that found correspondence
required bisexual men to have had both sexual and
romantic experiences with members of both sexes (which
may have more effectively excluded men whose orienta-
tions were more monosexual in nature despite their iden-
tification as bisexual).
The most defensible conclusions from this work are
that some bisexual-identified men have bisexual genital
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 61
arousal patterns and some do not. This provides strong
evidence against the (often stereotyped) notion that male
bisexuality does not exist and that all bisexual-identified
males are misrepresenting their sexual orientation. Yet
basic questions regarding the prevalence of female and
male bisexual phenomena and the specific differences
between bisexual and homosexual phenomena remain
unanswered; this remains an active topic of research.
Causes of Sexual Orientation
The question of what causes different people to be hetero-
sexual, bisexual, or homosexual has evoked intense interest
among the general public (e.g., Finkelstein, 2006; Swidey,
2005). However, the question is unusual among interesting
scientific questions because it has also provoked intense
controversy that in part reflects sociopolitical and moral dif-
ferences among disputants (Pitman, 2011; for examples, see
the relevant sections of the websites for the National Asso-
ciation for the Research and Therapy of Homosexuality,!gay—born-that-way/cm6x, and
Truth Wins Out,
The politics of sexual-orientation
There is a strong correlation between beliefs about the
origins of sexual orientation and tolerance of nonhetero-
sexuality, a correlation that has been stable for several
decades (Lewis, 2009). Specifically, those who believe
that sexual orientation is “innate,” “biological,” “immuta-
ble,” or “inborn” are especially likely to favor equal rights
for homosexual and bisexual people. All four words are
often used by participants in the political debate about
the causes of sexual orientation, despite the fact that
none of the words has a straightforward, consistent, and
uniform interpretation.
The association between political attitudes about
homosexuality and beliefs about the causes of sexual ori-
entation is largely misplaced (Greenberg & Bailey, 1993).
It is based on faulty reasoning about causation and about
the relevance of causation to moral judgments. It has
intruded unhelpfully into social controversies that affect
the lives of homosexual and bisexual people. Further-
more, it has had a harmful effect on the science of sexual
orientation. For example, in the United States, funding for
basic research on sexual orientation has been limited
because of political controversy. We know of several
cases in which applicants for federal U.S. grants were
asked to change grant titles in order to hide their topic of
study from those hostile to such research. Politically con-
troversial topics are precisely where science-based infor-
mation should be sought, when it is relevant. But it is
important to think clearly about what is, and what is not,
relevant. We thus preface our review of scientific evi-
dence on the causes of sexual orientation with a concep-
tual critique—and we hope, a correction—of the way
that people often reason about how these causes relate
to social, political, and moral questions.
The question of choice
Do people choose to be homosexual or heterosexual?
This question is perhaps the most common causal ques-
tion asked in the sociopolitical context. It is asked much
less commonly in scientific contexts because, as we shall
see, it is a bad question. This is partly because there are
at least two different, mutually inconsistent meanings of
“choice” that are often conflated.
Choice as uncaused action. The correlation between
one’s beliefs about the causes of sexual orientation and
one’s degree of tolerance of nonheterosexuality appears
to be based on the following logic: If there are causes—
other than free will—that lead certain people to be non-
heterosexual, then those people were never entirely
free to be heterosexual and hence cannot be held
responsible for their nonheterosexuality. For example,
finding a gene that increases the chance a man will be
homosexual would mean that the man is not completely
free to choose to be heterosexual. To the extent that the
gene causes his homosexuality, we should neither
blame him nor discriminate against him. This is the
essence of the argument regarding sexual orientation
and choice.
Yet this is a bad argument, and the word “choice” (and
associated concepts such as freedom and responsibility)
lies at the root of the problem. Why would discovery of
a gene for sexual orientation imply that homosexuality is
not freely chosen? It would do so only if we could assume
that free will is the null hypothesis on which causal stud-
ies chip away. This assumption is not scientific, however,
and is not intellectually defensible (Dennett, 1984; Pinker,
2003). For instance, to the extent that a trait is not genetic,
it is caused by the environment, not by free will. If a trait
is not present from birth, then it is caused by events
occurring after birth, not by free will.
Choice as decision to act. There is an alternative sense
of “choice” that is more meaningful: the sense of making
a decision. This ordinary-language sense of “choice” is
something that is commonly understood. “I chose to raise
my hand,” “I chose to eat broccoli,“I chose to rob a
bank,” and “I chose to have sex with that person” are all
meaningful sentences. It is this sense of “choice” that
people likely mean when they debate whether sexual
orientation is a lifestyle choice. Note that cause has noth-
ing to do with it. The four sentences all make sense even
by guest on July 5, 2016psi.sagepub.comDownloaded from
62 Bailey et al.
though the respective choices to which they refer could
all be determined by a combination of genetic and envi-
ronmental factors. People may choose to do things for
environmental reasons—perhaps a woman has been
offered a million dollars to raise her hand. Or they may
choose them for genetic reasons—perhaps a man chose
a blue car because his genetically determined color
blindness made red and green cars unappealingly gray.
In a deterministic world (which behavioral scientists
assume), decisions have causes.
The meanings of words can be illuminated by how
people use them, and an important regularity in the way
people use “choice” concerns the distinction between
behavior and feelings. We choose our actions, but we do
not choose our feelings. Consider the following two
1. “I choose to have sex with partners of my own
2. “I choose to desire to have sex with partners of
my own sex.”
The first sentence is conventional and sensible; the
second sentence is neither. Einstein summarized Scho-
penhauer’s famous argument appropriately and thusly:
“Man can do what he wills, but he cannot will what he
wills” (as quoted in Planck, 1933, p. 201). Applied to
sexual orientation, it makes sense to say that people
choose their sexual partners, but it doesn’t make sense to
say that they choose their desires. Sexual orientation is
defined as relative desire for same-sex or other-sex sex
partners. Thus, it makes no sense to say that one chooses
one’s sexual orientation. One does, however, choose to
behave consistently or inconsistently with one’s sexual
orientation. That is a lifestyle choice.10
A flawed question. The question of whether or not
people choose their sexual orientations has clouded
rather than clarified thinking. We can answer the ques-
tion without knowing anything scientific about the causes
of sexual orientation, because the answer depends
entirely on what we mean by “choice” and “choose.” If a
“choice” requires a lack of causes, then people cannot
choose their sexual orientation because all human behav-
ior has causes. If “to choose” means “to make a decision,”
then we do not choose our sexual orientations because
sexual orientations are patterns of sexual desire, and we
do not choose our sexual desires. Of course, advocates of
therapeutic attempts to change sexual orientation might
argue that even if sexual desires are not initially chosen,
individuals can choose to alter their desires through pro-
cesses of conditioning and reinforcement. As we discuss
in greater detail below, however, there is no evidence
that such attempts are successful.
Is causation relevant at all?
Even if the question of whether people choose their sex-
ual orientations is a bad one, this does not necessarily
mean that all questions about the causes of sexual orien-
tation are irrelevant to political, social, and ethical con-
cerns. Unfortunately, such relevance is usually assumed
rather than argued for. Furthermore, in cases where the
argument is spelled out, it typically amounts to an argu-
ment that any discernable causes negate free will, which
we have seen is invalid.
A more comprehensible basis upon which people dif-
fer in their moral and political responses to homosexual-
ity is in their beliefs about its consequences rather than its
causes. People who believe that homosexuality has nega-
tive effects on psychological, moral, or social functioning
will disapprove of homosexuality and may seek to dis-
courage it by restricting its expression and limiting the
rights of homosexuals. In contrast, people who believe
that there are no negative consequences of homosexual-
ity will favor the rights of homosexuals to live openly and
to enjoy the same rights as heterosexuals. These grounds—
rather than questions of causation and choice—are the
appropriate grounds on which the battle for equal rights
for nonheterosexual people should be fought.11
This is not to dismiss the importance of basic research
on sexual orientation, including research on causation.
Despite our pessimism about causal research’s impor-
tance in resolving social questions, there are some spe-
cific limited cases in which it is relevant. One example
concerns the belief, expressed by President Museveni in
Uganda as well as anti-homosexual activists worldwide,
that homosexual people attempt to recruit others into
homosexuality. Equally important is the question of
whether individuals’ sexual orientation can be changed
through “conversion,” “reparative,” or “reorientation”
therapies and whether such therapies should be deemed
safe and effective by psychological organizations. Claims
about both homosexual recruitment and the success of
conversion therapies are causal claims, and we examine
them in a later section.
In our view, the single best justification for studying the
causes of sexual orientation is scientific, not sociopolitical.
Quite simply, sexual orientation is a basic human trait that
influences identity and behavior at both the individual
and the group level, and hence it is fundamentally impor-
tant and interesting to understand its causes and develop-
ment. Toward this end, in the following sections, we
summarize the current state of this understanding.
A taxonomy of causal questions
Non-experts are exposed to scientific findings primarily
through the news media. Journalists try to make a
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 63
judicious compromise between scientific accuracy and
simple comprehensibility; sometimes they fail. Some-
times even scientists are not as precise and accurate as
they should be. A plethora of terms have been used as
alternatives in the two sides of the nature-nurture debate.
For nature: “biological,” “genetic,” “hereditary,” “herita-
ble,” “innate,” “inborn,” “natural,” and “essential.” For nur-
ture: “chosen,” “learned,” “acquired,” “environmental,”
“socialized,” “unnatural,” and “socially constructed.” Some
of these terms have originated from scientists and have
been accurately applied to particular studies but over-
generalized to other, inapplicable studies. Other words,
such as “chosen,” are rarely used by scientists but com-
monly (and erroneously) used by laypersons. Impor-
tantly, words that are treated as synonyms in informal
usage often have quite different formal definitions. Thus,
it is important to correct some common misunderstand-
ings about causal terminology before proceeding.
The first two columns of Table 2 list some words most
commonly used in the nature-versus-nurture debate.
Adjacent words in the same “Nature” or “Nurture” col-
umns represent synonyms—conceptual distinctions that
are dissociable but not completely independent. Words
in the same row but different columns represent concep-
tual opposites (or complements). The final two columns
of Table 2 convey our judgments of whether the concep-
tual distinctions implied by a given word are valid and
whether those distinctions are precise or require further
We have already addressed the conceptual vacuity of
“choice.” Another word that obscures more than it clarifies
is “biological.” Because all behavior requires participation
of the brain and body, and such participation is at least
in principle measurable, all behavior is “biological”
(Greenberg & Bailey, 1993). Thus, the word denotes noth-
ing unique about any trait or behavior. Although “biologi-
cal” is often used synonymously with other words in the
“Nature” column of Table 2, that practice should stop, and
more accurate and precise words should be used instead.
The extent to which a trait is “genetic” or “environmen-
tal” is a conceptually meaningful and precise question,
although it may often be practically difficult to provide a
precise answer. The extent to which people differ in a
trait as a result of genetic or environmental differences
among them can be estimated as heritability, which is
expressed as a proportion ranging from 0 (only environ-
mental differences matter) to 1 (only genetic differences
matter), using methodology including twin and adoption
The idea that hormonal differences may lead to sex-
ual-orientation differences has been especially influential
(Ellis & Ames, 1987; LeVay, 2011). The extent to which
trait differences reflect differences in hormones is gener-
ally meaningful and entirely separate from the question
of genetic influence. If genes for sexual orientation exist,
they may or may not operate via hormonal pathways.
Conversely, hormonal differences may not depend on
genetic differences (aside from the initial role of the Y
chromosome in triggering androgens in males). Thus,
evidence regarding heritability provides no evidence
regarding hormonal influence, and vice versa. The extent
to which a trait is hormonally influenced is a less precise
question compared with the extent to which it is geneti-
cally influenced. A trait might be hormonally influenced
via levels of hormones currently present in relevant tis-
sue; alternatively, it may have been influenced via levels
of hormones present during the organization of the tis-
sue, or it may be influenced by receptivity to hormones,
either at present or during organization; there are likely
other possibilities. Nor do assessments of hormone levels
yield a numeric index comparable to heritability.
Innateness is a famously difficult concept (Elman etal.,
1998), in large part because it has a number of different
connotations (Mameli & Bateson, 2006). For sexual orien-
tation, the question of innateness is most meaningfully
asked as follows: When people have different sexual ori-
entations, is this because they had different postnatal
social experiences, or did the different sexual orientations
emerge despite their social experiences? If differences in
social experiences matter, then sexual orientation is not
entirely innate; if they do not matter, then it is.12 For
humans, “social experiences” mean experiences involving
Table 2. Causality-Related Terms Commonly Used in Scientific and Political Debates
About Sexual Orientation
Nature terms Nurture terms Meaningful? Precise?
Chosen No No
Biological No No
Genetic, inherited, heritable Environmental Yes Yes
Hormonal Not hormonal Yes No
Innate, inborn, congenital Learned, acquired, socialized Uncertain No
Natural Unnatural Uncertain No
Essential Socially constructed Yes No
by guest on July 5, 2016psi.sagepub.comDownloaded from
64 Bailey et al.
other people, and the mechanisms through which such
experiences are thought to influence development gener-
ally comprise diverse forms of learning and reinforce-
ment. In contrast, mechanisms of innate development
include practically anything else: genes, hormones, ran-
dom prenatal developmental perturbations, infections,
and other factors too numerous to list. Thus, although
genetic and hormonal influences on sexual orientation
constitute innate influences, not all innate influences
require either genes or hormones. Arguably, the innate-
ness debate is best focused on whether particular candi-
date social influences matter rather than whether any
social influence matters.
The question of whether sexual orientation is “natural”
or “unnatural” has tended to be asked in the context of
religious debates, with anti-homosexual advocates insist-
ing that homosexuality is unnatural (e.g., TFP Committee
on American Issues, 2004) and their opponents insisting
it is natural (e.g., A. Jackson, 2013). What can this ques-
tion mean? It has at least three different interpretations.
First, does homosexuality occur in nonhuman animals?
Second, is homosexuality a result of human evolution?
Third, is homosexuality consistent with natural law? The
third question is philosophical rather than scientific
(Pickett, 2011), and we do not consider it here. In con-
trast, the first two interpretations of the question are sci-
entifically meaningful and interesting. Answers to both
are of no moral consequence, however.
The final row of Table 2 refers to a debate that is nei-
ther conducted mainly by scientists nor by members of
the general public. Rather, whether sexual orientation is
an essential or a socially constructed categorization of
people has largely been the preoccupation of scholars of
human culture and society, including philosophers, histo-
rians, social theorists, cultural anthropologists, and soci-
ologists (e.g., Norton, 1997; Stein, 1992). One of the main
issues in this debate concerns the degree to which sexual
orientation manifests, and is recognized, in a similar man-
ner across cultures, both geographically and historically.
Those who believe that sexual orientation is socially con-
structed emphasize cultural variation, whereas those who
believe that it is an essential human trait are impressed
with its cross-cultural regularities. We address the cross-
cultural evidence in the following section.
Sexual orientation across culture and
In a speech at Columbia University, Iranian President
Mahmoud Ahmadinejad asserted, “In Iran, we don’t have
homosexuals like in your country” (“‘We Don’t Have Any
Gays in Iran,’” 2007, para. 1). The idea that homosexual-
ity exists in some times or places but not others—or, less
radically, that it exhibits different frequencies and forms
across different times and places—requires data that are
difficult to obtain. Establishing facts about sexuality is
complicated enough in the relatively open contemporary
Western world, where representative surveys and even
objective measures can be employed. Inferring facts
about taboo behavior in different cultures or time periods
is much more difficult. Such inference making is gener-
ally the job of anthropologists and historians, who do the
best they can. On the one hand, it is important to recog-
nize the limitations of the data and to be skeptical of
surprising claims that would seem to depend on better
data than currently exist. On the other hand, these data,
with all their limitations, are all we have. We expect that
in all cultures the vast majority of individuals are sexually
predisposed exclusively to the other sex (i.e., heterosex-
ual) and that only a minority of individuals are sexually
predisposed (whether exclusively or non-exclusively) to
the same sex. Data that might speak to this issue are
exceedingly limited but consistent with this conclusion
(e.g., VanderLaan etal., 2013; Whitam & Mathy, 1986).
In contrast, sexual identity and behavior are far more
susceptible to cultural variation. Indeed, the sexual-iden-
tity categories of “gay,“lesbian,” and “homosexual” are
historically and culturally specific, and they do not neces-
sarily translate to other times and places (Blackwood &
Wieringa, 1999; Boswell, 1982/1983; Murray, 2000; Nanda,
2014; Norton, 1997). Consequently, the ways in which
many non-Western same-sex-attracted individuals and
Western gay men and lesbians think about themselves
and pattern their lives (sexual or otherwise) can differ in
important respects. Thus, when individuals from non-
Western cultures say that there are no “gays,” “lesbians,
or “homosexuals” in their societies, they are not necessar-
ily inaccurate or dishonest: Based on their understanding
of what it means to be “gay,“lesbian,” or “homosexual,
no one in their society identifies or behaves as such, and
neither do they identify other members of their culture in
that way. However, this does not mean that their culture
contains no same-sex-attracted individuals.
One approach to studying same-sex sexuality cross-
culturally involves focusing on its deep structure rather
than its culturally specific constructions and meanings
(Vasey & VanderLaan, 2014). The deep structure of same-
sex sexuality can be thought of as a set of traits that reli-
ably characterizes same-sex-attracted males and another
set that reliably characterizes same-sex-attracted females,
regardless of the cultural context in which these individu-
als are found. To this end, a focus on individuals’ sexual
attractions facilitates cross-cultural comparisons, given
that individuals in all cultures have been observed to
report experiencing sexual attractions, and hence sexual
attractions can be considered a universal human phe-
nomenon. In contrast, a focus on sexual identities or
behaviors does not facilitate cross-cultural comparisons
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 65
of same-sex sexuality, given that not all cultures have a
notion of sexual identity and not all cultures permit indi-
viduals to pursue sexual behavior that concords with
their desires.
“Androphilia” and “gynephilia” are useful terms for
denoting basic human sexual attractions. Androphilia
refers to sexual attraction and arousal to adult males,
whereas gynephilia refers to sexual attraction and arousal
to adult females. Homosexual men and heterosexual
women are androphilic, and heterosexual men and
homosexual women gynephilic. The terms “androphilia”
and “gynephilia” are preferable to terms such as “homo-
sexual” and “heterosexual” because the latter terms have
been used and defined specifically in a Western cultural
context and often do “triple duty” by serving as descrip-
tions of attractions, behavior, and identities. The terms
“androphilia” and “gynephilia” avoid these problems and
therefore provide for more accurate descriptions.
Although same-sex-attracted individuals can differ in dra-
matic ways from one culture to the next, they can all be
accurately described as either androphilic biological
males or gynephilic biological females.
The manner in which same-sex sexual attraction is
publicly expressed varies across time and cultural space
(Blackwood & Wieringa, 1999; Murray, 2000; Norton,
1997; Whitam & Mathy, 1986). Same-sex sexuality
between adults typically takes one of two cross-culturally
recurrent forms, which are related to gender-role enact-
ment and gender identity. These two forms are cisgen-
der13 and transgender14 male androphilia and female
Cisgender male androphiles and female gynephiles
occupy the gender role typical of their sex and identify as
“men” and “women,” respectively. This is the form of
homosexuality that is nearly universal in the contempo-
rary West. In contrast, transgender male androphiles and
female gynephiles do not occupy the gender role typical
of their sex. Not only do they behave in a highly gender-
atypical manner, but they often identify, and are identi-
fied by others, as neither “men” nor “women,” but rather,
as a member of some alternative gender category. Con-
temporary examples of transgender male androphiles
include the kathoey of Thailand (P. Jackson, 1997), the
xanith of Oman (Wikan, 1977), the muxes of Mexico
(Chiñas, 1992), and the fa’afafine of Samoa (Vasey &
VanderLaan, 2014). Some contemporary examples of
transgender female gynephiles include the tombois of
Sumatra (Blackwood, 2010) and the mahu of Tahiti
(Elliston, 1999).
In some cultures, transgender male androphilia and
female gynephilia are linked to particular institutional-
ized labor practices, which often involve specialized reli-
gious activities. This type of transgender male androphilia
has been referred to as “profession defined” (Adams,
1986). For example, on the Indian subcontinent, trans-
gender male androphiles known as hijra bestow bless-
ings from Hindu gods and goddesses for luck and fertility
at weddings and at the births of male babies (Nanda,
1998). In Sulawesi, Indonesia, transgender androphilic
males known as bissu are shamans who bless people for
good health and successful journeys and who play
important ritual roles in weddings (Peletz, 2009). These
institutionalized religious roles sometimes carry with
them the expectation of asceticism, but often this ideal is
not realized (e.g., Nanda, 1998; Peletz, 2009). In general,
same-sex-attracted individuals self-select to fill these
roles, probably because they are recognized as socially
acceptable niches (Murray, 2000).
Cisgender male androphiles and female gynephiles
behave in a relatively gender-typical manner when com-
pared with their transgender counterparts. However, they
are relatively gender-atypical when compared to gyne-
philic cisgender men and androphilic cisgender women
(Bailey, 2003; Cardoso, 2013; Lippa, 2008; Whitam &
Mathy, 1986; L. Zheng, Lippa, & Zheng, 2011). Thus,
regardless of the form they take, male androphilia and
female gynephilia are associated with gender-atypicality.
However, the strength of this association varies with the
manner in which same-sex sexuality is publicly expressed.
Both the cisgender and transgender forms of same-sex
sexuality may occur within a given culture, but typically
one or the other predominates (Murray, 2000; Whitam &
Mathy, 1986). For example, the cisgender form tends to
be much more common in many Western cultures. In
contrast, the transgender form appears to be more com-
mon in many non-Western cultures. In places where the
two forms coexist, their members often consider each
other to be part of the same subculture (Whitam & Mathy,
1986). Margaret Mead observed a meeting in which an
Omaha minquga (i.e., a transgender male androphile)
and a Japanese homosexual man (i.e., a cisgender male
androphile) who visited her field site in 1961 instantly
recognized each other. Within an hour of the Japanese
man’s arrival, the sole minquga in the tribe turned up
and tried to make contact with him (Mead, 1961). Simi-
larly, sociologist Fredrick Whitam (1995) noted that, in
São Paulo, travesti (transgender male androphiles) are an
especially conspicuous presence in gay clubs and are
treated with a high degree of respect.
In contemporary Western cultures, cisgender male
androphiles typically engage in sexual interactions with
each other; the same is true of cisgender female gyne-
philes. That is, in the West, homosexual relationships are
typically between two homosexual individuals. Such
individuals comprise the Western gay and lesbian com-
munities. This type of same-sex sexual relationship has
been referred to as “egalitarian” and is characterized by
partners who are not markedly different in age15 or
by guest on July 5, 2016psi.sagepub.comDownloaded from
66 Bailey et al.
gender-related characteristics (B. D. Adams, 1986). Within
such relationships, partners tend not to adopt special
social roles, and they treat each other as equals. In con-
trast, this pattern appears to be relatively uncommon in
non-Western cultures and has emerged only recently in
certain non-Western urban centers (e.g., P. Jackson,
Although transgender male androphiles are same-sex
attracted, they rarely, if ever, engage in sexual activity
with each other; the same is true of transgender female
gynephiles (Blackwood & Wieringa, 1999; Murray, 2000;
Nanda, 2014). Rather, these individuals engage in sexual
activity with same-sex cisgender partners who self-
identify, and are identified by others, as “men” or
“women.” For example, in Samoa, very feminine natal
males called fa’afafine (which means “in the manner of
women”) have sex with masculine Samoan men. The
fa’afafine would be aghast at the idea of having sex with
one another.
Little research has focused on the cisgender sexual
partners of same-sex-attracted transgender males and
females. Blackwood (2010, p. 137) noted that, in Sumatra,
the cisgender female partners (femmes) of tombois “assert
an uncomplicated attraction to men, [but] position them-
selves (if temporarily) under the label ‘lesbi’”—a deriva-
tive of “lesbian. This suggests an episodic pattern of
bisexual attraction on the part of femmes. In many cul-
tures, same-sex sexual interactions between transgender
and cisgender persons are not considered “homosexual”
because they are understood to be hetero-gendered
(Murray, 2000; Nanda, 2014; Norton, 1997). In other
words, if a cisgender androphilic male and a transgender
androphilic male engage in sex, the former individual is
often understood to be “the male partner” in the interac-
tion, whereas the latter individual is often understood to
be “the female partner.” Accordingly, the interaction is
understood as male-female rather than male-male. The
degree to which cisgender individuals who have sex with
transgender persons of their same biological sex (i.e.,
men who have sex with female-appearing men and
women who have sex with male-appearing women) are
perceived as different from those whose sexual behavior
is only with the other sex (i.e., conventional heterosexu-
als) remains an open question (Norton, 1997).
Same-sex sexual activity appears to have existed
throughout human history. Mesolithic rock art depicts
male-male sexual activity (e.g., Nash, 2001). Graves con-
taining male skeletal remains and female-typical artifacts
have been interpreted as evidence for transgender andro-
philic males in the prehistoric past (Hollimon, 1997). The
earliest written documentation of male-male sexual activ-
ity involving humans (as opposed to gods) comes from
ancient Egypt and dates to the late New Kingdom (1292–
1069 BCE). The story includes a sexual tryst between the
pharaoh Neferkare and the military general Sisene,
wholived during the 6th Dynasty of Egypt’s Old King-
dom (2345–2181 BCE; Meskell, 2001). As early as the
7th century BCE, the Greek lyrical poet Sappho wrote
about romantic love and infatuation between females
(Campbell, 1982). Although one can question whether
these examples reflect actual events, they do at the very
least suggest that some prehistoric and ancient peoples
understood same-sex sexual and romantic activity to be
within the realm of possibility.
The historical record also indicates that at many times
and in many places, androphilic males and gynephilic
females have been recognized by others, and by them-
selves, as a unique class of individuals owing to their
sexual orientations (e.g., Boswell, 1982/1983; Murray,
2000; Norton, 1997). Typologies for same-sex-attracted
individuals can be traced back to ancient times. In Plato’s
Symposium (c. 385–370 BCE), Aristophanes described his
theory for exclusively androphilic and gynephilic men
and women, asserting that all individuals, in seeking lov-
ers, are trying to reunite with a primordial conjoined
twin—in some cases male and in others female—from
whom they were severed. Ancient Indian medical texts
from the 1st century contain typologies for same-sex-
attracted individuals that are strikingly similar to modern
ones (M. Sweet & Zwilling, 1993; Vanita & Kidwai, 2001).
Male same-sex sexual activity appears to have existed
in most cultures for which data are available (e.g.,
Murray, 2000; Nanda, 2014), and the population preva-
lence rate for male androphilia appears to be similar
(approximately 1.5%–5%) across a range of different con-
temporary cultural settings (e.g., Smith et al., 2003;
VanderLaan etal., 2013; Whitam & Mathy, 1986). Claims
that male androphilia is “absent” in a particular culture
are often demonstrably false, even when the culture
in question seemingly lacks words to describe male
androphilia or male same-sex sexual activity (Boswell,
1982/1983; Norton, 1997; VanderLaan, Garfield, et al.,
2014). Although male-male sexuality may truly be absent
in a minority of cultures (e.g., Hewlett & Hewlett, 2010)—
especially if population size is small—these exceptions
do not invalidate the conclusion that male androphilia
appears to be a predictably and reliably reoccurring phe-
nomenon in the large majority of human cultures.
Female gynephilia has also been reported in many
cultures worldwide, but less often than male androphilia
(Blackwood, 1999; Blackwood & Wieringa, 1999; Elliston,
1999; Murray, 2000). Based on this evidence, some schol-
ars have concluded that female gynephilia is less com-
mon than male androphilia across cultures and, when it
occurs, is less enduring over the life span (e.g., Murray,
2000). It is possible that romantic relationships between
women have been more common historically and cross-
culturally than relationships involving genital sexuality
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 67
(Faderman, 1981; Murray, 2000; Sahli, 1979). Arguably,
the exclusivity and longevity of female gynephilia may
relate to whether the individual in question is a “butch”
(masculine) or “femme” (feminine) female gynephile.
Research suggests that gynephilia is expressed in a more
exclusive and enduring manner in butch/masculine as
opposed to femme/feminine women (Blackwood, 2014;
L. M. Diamond, 2009). Another possibility is that because
female gynephilia often occurs within close relationships
between women or among young women who are not of
marriageable age, it has been less visible and less con-
cerning and therefore has escaped accurate documenta-
tion (Blackwood & Wieriga, 1999; Kendall, 1999).
Historical and cross-cultural evidence suggests that if a
sufficient number of people exist in a society, subcultures
populated by same-sex-attracted individuals will appear
within it (Gaudio, 2009; Norton, 1997; Whitam & Mathy,
1986). The formation of these subcultures appears to be
more common among androphilic men, compared with
gynephilic women. In Europe, recognizable centers of
homosexual activity can be traced at least as far back as
the 11th-century Anglo-Norman court of King William II
(Goodich, 1979).
Shared interests and personality characteristics beyond
a common sexual orientation likely facilitate the forma-
tion of such subcultures. Same-sex-attracted individuals
often have more in common with each other, even when
they come from disparate cultures, than they do with
their larger culture, in part because of gender nonconfor-
mity (Norton, 1997; Whitam & Mathy, 1986). For exam-
ple, across cultures, androphilic men tend to be more
female-typical and “people-oriented” in their interests
compared to gynephilic men; conversely, gynephilic
females tend to be more male-typical and “thing-ori-
ented” than androphilic females (Cardoso, 2013; Lippa,
2008; L. Zheng etal., 2011). Not surprisingly, androphilic
males in many cultures worldwide share interests per-
taining to the house and home, decoration and design,
language, travel, helping professions, grooming, and the
arts and entertainment (Whitam & Mathy, 1986).
Even in small, isolated, and traditional cultures, the
rudiments of such subculture become apparent. Whitam
and Mathy (1986) described one such example:
In the Guatemalan Indian town of Chimaltenango, two
men lived together as lovers, wearing typical Indian
clothing in an outwardly traditional Indian adobe
house. The house, however, was decorated in a manner
strikingly different from the other Indians. It was
meticulously and elaborately decorated, a characteristic
frequently found in homosexual subcultures. . . . The
occupation of the lovers was that of stringing pine
needles in decorative strands, traditionally used in
Guatemala for holidays and other festive occasions,
and supplying flowers for weddings. In essence
these two men were florists, involved in the arts of
embellishment, which in larger societies areuniversally
linked with homosexual subcultures. (p.84)
Homosexuality in Africa. Because there have been
proposals to increase legal penalties against homosexual-
ity in several African nations, we briefly address the cul-
tural history of homosexuality in Africa. The transgender
form of same-sex sexuality predominated in traditional
societies in Africa (Epprecht, 2008). Notwithstanding
commonplace claims today that “homosexuality is un-
African,” same-sex-attracted transgender male andro-
philes and female gynephiles were observed from the
earliest recorded times and in hundreds of discrete cul-
tures across the continent (Epprecht, 2006; Murray &
Roscoe, 1998). European observers, with all their assump-
tions, prejudices, and linguistic limitations, were not nec-
essarily reliable witnesses, and, for this reason, many of
the accounts produced in the colonial periods can be
misleading. There are, however, a wide range of ethnog-
raphies that document exceptions to heterosexual norms.
Same-sex relationships typically involved a male-bod-
ied person said to be possessed by a female ancestral or
other spirit. To avoid offending that spirit, the man
needed to avoid sex with women and perhaps, to win the
favor of the female spirit, actively seek out sex with a
male partner. Male spirits could similarly be embodied in
living females, demanding that a woman remain celibate
or take female sex partners. Sometimes this was a source
of mocking humor to the majority population, but often
the “possessed” person had a respected, even cherished
role in society as a healer or seer (Epprecht, 2008;
Nkabinde, 2008). In some cases, particularly in Islamic
Africa, same-sex-attracted transgender people have
occupiedspecialized occupations, much like the stereo-
typical ones observed in the West and described above
(Gaudio, 2009). In urban centers, communities of same-
sex-attracted males have developed out of friendship
networks and around shared locations for meeting or
work (Gaudio, 2009; Moodie & Ndatshe, 1994).
In addition to being highly gendered, same-sex rela-
tionships were frequently age-stratified as well, with the
elder partner assuming the masculine role (in conven-
tional terms) and the younger partner the feminine role.
With maturity, the younger partner would commonly
graduate into a heterosexual marriage. Because the
younger partners were not recognized socially as fully
adult “men” (a moniker that comes with marriage), they
were often referred to as “boys.” It is important to note,
however, that these “boys” were usually postpubertal and
thus, although these relationships often involved age dis-
crepancies, they were not pedophilic. In a strongly
homosocial environment, same-sex sexual bonds could
by guest on July 5, 2016psi.sagepub.comDownloaded from
68 Bailey et al.
continue into adulthood without being explicitly named
as such (Carrier & Murray, 1998; Gay, 1985; Kendall,
Aspects of gay identity as lived and performed in the
West since “gay liberation,” including the egalitarian
model of overt same-sex relationships, are attractive to
many Africans who chafe under traditional, heterosexual
family norms and obligations. As a result, small numbers
of Africans began to come out in politically recognizable
ways as lesbian and gay beginning in the 1990s (Hoad,
Martin, & Reid, 2005; Morgan & Wieringa, 2005).
Homosexuality in nonhuman animals
Animal studies have informed the science of human sex-
ual-orientation causation in several ways. Experimental
studies of nonhuman animals have helped generate mod-
els of sexual-orientation development. Field studies of
animals have allowed tests of hypotheses regarding the
situations that are most likely to be associated with
increased homosexual behavior. Although same-sex
interactions involving genital contact have been reported
in hundreds of animal species (Bagemihl, 1999), they are
routinely manifested in only a few (Sommer & Vasey,
2006). In this sense humans are rare, but we are not
Same-sex genital interactions between nonhuman ani-
mals can take a variety of forms, including mounting
(with and without pelvic thrusting), oral-genital contact,
and manual-genital contact. In many instances, same-sex
genital interactions can be accurately described as
“homosexual behavior” because they involve genital
arousal, stimulation, and orgasm. For example, it is not
uncommon for males to exhibit erections during these
interactions (e.g., Japanese macaques [Macaca fuscata],
Leca, Gunst, & Vasey, 2014), and ejaculation occurs
occasionally (e.g., mountain gorillas [Gorilla beringei],
Yamagiwa, 1987). Intromission between males has also
been reported (e.g., bottlenose dolphins [Tursiops sp.],
Mann, 2006), but rarely, perhaps in part because of the
challenges associated with recording such behavior. In
some species, females actively stimulate their genitals
during same-sex mounts (e.g., bonobos [Pan paniscus],
de Waal, 1987; Japanese macaques, Vasey & Duckworth,
2006). Orgasm between female partners is difficult to
document unambiguously but has also been reported
(stump-tailed macaques [Macaca arctoides], Goldfoot,
Westerborg-van Loon, Groenevelde, & Slob, 1980). In
addition to these genital interactions, same-sex courtship
has been reported in numerous animal species, and some
same-sex individuals form temporary or even long-last-
ing pair-bonds (e.g., male greylag geese [Anser anser],
Kotrschal, Hemetsberger, & Weiss, 2006).
There is abundant evidence that nonhuman animals
engage in same-sex genital interactions under free-ranging
conditions, and therefore this behavior cannot be explained
as a result of captivity in abnormal environments. Skewed
sex ratios can facilitate the expression of same-sex genital
interactions among the more numerous sex (Poiani, 2010),
but such conditions are not necessary for such behavior to
be manifested (Bagemihl, 1999; Sommer & Vasey, 2006).
Even when opportunities to engage in heterosexual con-
tact are available, individuals sometimes prefer to engage
in homosexual behavior (Leca, Gunst, Huffman, & Vasey,
2015). Quantitative research indicates that species’ mating
systems, rates of development, and disparities in parental
care all influence the expression of homosexual behavior
in birds. For example, the frequency of male-male sexual
behavior in birds increases with the degree of contextual
polygyny, whereas female-female sexual behavior is more
frequently expressed in monogamous species that exhibit
precocial development (MacFarlane, Blomberg, Kaplan, &
Rogers, 2007).
Many of the same-sex genital interactions exhibited by
animals are, at least in part, sociosexual in character—
that is, sexual in terms of their outward form, but primar-
ily enacted to facilitate adaptive social goals (Wickler,
1967). The context-specific manner in which such behav-
ior is expressed often cues researchers to its adaptive
function. For example, in female bonobos, a close tem-
poral relationship exists between same-sex genital inter-
actions and food sharing. Females entering a food patch
that is monopolized by a same-sex competitor are more
likely to acquire food if they first engage in genital rub-
bing with that competitor (de Waal, 1987; Hohmann &
Fruth, 2000). In this way, same-sex genital rubbing func-
tions to reduce inter-individual tension and facilitate food
sharing. In male savanna baboons (Papio cynocephalus
anubis), there is a close temporal relationship between
same-sex mounting/genital fondling and successful alli-
ance formation. Males will often engage in these behav-
iors just prior to challenging a same-sex rival (Smuts &
Watanabe, 1990). Sociosexual interactions can be devoid
of sexual arousal, but more often than not, they are char-
acterized by some mix of both sexual and social motiva-
tion (Wickler, 1967). However, in some species, same-sex
mounting appears to be entirely sexual, with no discern-
able sociosexual component whatsoever. For instance, in
certain populations, female Japanese macaques will
sometime choose other females as sexual partners despite
the presence of sexually motivated male mates (Vasey &
Duckworth, 2006). Female Japanese macaques will even
compete intersexually with males for exclusive access to
female sexual partners (Vasey, 1998).
Exclusive same-sex sexual orientation across the life
course is, however, extremely rare among animals. The
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 69
only conclusively documented example is among males
(rams) of certain breeds of domestic sheep (Ovis aries,
Perkins & Fitzgerald, 1997). Approximately 6% to 10% of
rams in these domestic breeds choose to court and mount
other rams, but never ewes, when given a choice (Roselli,
Larkin, Resko, Stellflug, & Stormshak, 2004). During some
mounts between rams, penile-anal intromission and ejac-
ulation occurs (Perkins & Fitzgerald, 1997). In all other
animal species, with the exception of humans, individu-
als that engage in same-sex genital interactions engage in
heterosexual ones as well.
The previous two sections have focused on observable
expressions of same-sex sexuality in humans and ani-
mals. In the remainder of our review, we focus on various
mechanisms that have been hypothesized to explain vari-
ation in human sexual orientation, particularly the pro-
pensity to experience same-sex sexual attraction.
The possibility that people differ in sexual orientation
because of hormonal differences has been the most
influential causal hypothesis involving a specific mecha-
nism (Dörner, 1976; Ellis & Ames, 1987; LeVay, 2011;
Meyer-Bahlburg, 1984; Money & Ehrhardt, 1972; Zucker,
2005). On the one hand, several considerations point to
this hypothesis; on the other hand, it has difficulties
explaining some key facts. In addition, there is little
direct evidence for this hypothesis and, of course, such
evidence is unlikely to be forthcoming, given that it
would be unethical to undertake direct testing on human
If we conceptualize sexual orientation as a sexual pre-
disposition toward males or toward females (androphilia
or gynephilia) instead of a sexual predisposition for the
same sex or the other sex, then we can immediately see
that there are large sex differences in sexual orientation.
Specifically, far more women than men are androphilic,
and far more men than women are gynephilic. We can
therefore think of androphilic men and gynephilic
women as being gender-atypical with regard to their sex-
ual orientation, since each shows a pattern of sexual
interest that is more characteristic of the other sex. We
also know, as reviewed earlier, that androphilic men and
gynephilic women are gender-atypical in a variety of
other ways, sometimes even in childhood. Thus, it seems
reasonable to wonder whether gender-atypical sexual
orientations (i.e., homosexuality and bisexuality) are
caused by the same factors that drive differences between
male and female development more generally. Two fac-
tors are known to create these differences: hormones and
sex-role socialization. We discuss the possible role of
socialization in a later section, and begin with the orga-
nizing role of prenatal hormones.
The organizational hypothesis. The development of
most physical sex differences strongly depends on hor-
mones. Some of these differences (e.g., patterns of mus-
cle and fat distribution) are caused by circulating
hormones and are thus reversible to some extent. Others
(e.g., the internal and external sex organs) are caused by
hormonal differences during a critical period and are
mostly irreversible (Sisk, Lonstein, & Gore, 2013). Early,
irreversible effects of hormones are organizational, and
later, reversible effects are activational. The organiza-
tional-activational distinction is conceptually useful,
although some effects do not fit neatly into either cate-
gory; for example, some later effects of circulating hor-
mones are irreversible (Arnold & Breedlove, 1985).
Early studies of possible hormonal influences on sex-
ual orientation focused on circulating levels of the sex
hormones testosterone and estrogen. In general, homo-
sexual and heterosexual men do not differ in levels of
these hormones (Meyer-Bahlburg, 1984). In contrast,
studies have tended to find higher testosterone levels in
homosexual women compared with heterosexual women
(Meyer-Bahlburg, 1984; Singh etal., 1999). The meaning
of such differences among females is unclear, however.
Circulating hormone levels are affected by many factors,
such as body weight, and lesbians may be taller and
heavier, on average, than heterosexual women (Bogaert,
1998; Singh, Vidaurri, Zambarano, & Dabbs, 1999).
Although some studies have statistically controlled for
known correlates of circulating testosterone, they may
have unknowingly neglected important confounds. If the
results were valid (i.e., not due to some third variable), it
would remain unclear how having higher testosterone
might cause a woman to be sexually attracted to other
women. Perhaps, instead, the testosterone differences are
the results of brain differences that cause sexual-orienta-
tion differences. Scant recent research has addressed
sexual-orientation differences in levels of circulating sex
The possibility that sexual orientation reflects the
organizational effects of early hormones, especially tes-
tosterone, has been far more influential. This hypothesis
is a special application of a general hypothesis: the orga-
nizational hypothesis of mammalian sexual differentia-
tion, according to which the same prenatal factors that
shape the body as male-typical or female-typical also
shape the brain to perform in a male- or female-typical
fashion (Phoenix, Goy, Gerall, & Young, 1959). The gen-
eral hypothesis has been confirmed for a remarkably
diverse range of behaviors in every mammalian species
examined so far (De Vries & Simerly, 2002). The sexuality
of the body and behavior of mammals does not appear
to be directly affected by their genetic sex (XX vs. XY).
Rather, it is an indirect consequence of genetic sex—
namely, whether ovaries or testes develop prenatally.
by guest on July 5, 2016psi.sagepub.comDownloaded from
70 Bailey et al.
Testicular release of steroid hormones, specifically andro-
gens such as testosterone, act early in life to induce the
formation of male genitalia (penis, scrotum, etc.) and
alter the developing brain to promote male sexual behav-
ior in adulthood. The absence of these testicular secre-
tions leads to the development of female-typical genitalia
(clitoris, labia, vagina, etc.) and brain organization that, in
adulthood, is more likely to drive female-typical sexual
The organizational hypothesis applied to human sex-
ual orientation holds that during an early—probably pre-
natal—sensitive period, the brains of homosexual people
are subject to atypical influences of androgens. Specifi-
cally, homosexual men’s fetal brains were androgenized
less than heterosexual men’s; homosexual women’s were
androgenized more than heterosexual women’s. Two
lines of research have been influential in convincing
researchers that the organizational hypothesis of sexual
orientation is plausible, and these include research on
the consequences of early hormonal manipulation in
mammals, especially rats, mice, and ferrets, as well as
research on the psychosexual outcomes of individuals
with atypical hormonal development.
In studies of mice, rats, and ferrets, experimentally
depriving males of testosterone or exposing females to
male-typical levels of testosterone early in development
dramatically alters their adult sexual behavior. That is,
males can be made to exhibit aspects of female-typical
behavior and females to exhibit aspects of male-typical
behavior (Henley, Nunez, & Clemens, 2011).
Furthermore, in some well-studied mammalian spe-
cies (e.g., rats, mice, ferrets, sheep), particular brain
regions show sex differences in size, and this sexual
dimorphism depends entirely on exposure to early sex
hormones. The largest and best studied of these is a
brain region called the sexually dimorphic nucleus of
the preoptic area (SDN-POA), which is located in the
hypothalamus. This sexually dimorphic area has been
identified in a wide range of mammalian species and is
always larger in males than it is in females (Henley etal.,
2011). The SDN-POA sex difference is caused by sex dif-
ferences in perinatal hormones: By controlling testoster-
one levels just before and after birth, researchers can
induce the SDN-POA to be as masculine or as feminine
as they like in adulthood. This brain region appears to
play important roles in both the appetitive and consum-
matory behavior of male mammals (Balthazart & Ball,
2007). Furthermore, in sheep, the best animal model for
human homosexuality—a minority (approximately 7%)
of male sheep (rams) exclusively mount other males—
the SDN-POA is half the size in male-mounting rams as
it is in female-mounting rams (Roselli et al., 2004).
Clearly, organizational androgens have important influ-
ences on the brains and sexual behavior of nonhuman
animals, and there is every reason to expect this to be
true in humans as well.
Atypical prenatal androgen exposure in humans.
Some humans are exposed to atypical levels of prenatal
hormones, most often because they have unusual genetic
syndromes. For example, in congenital adrenal hyperpla-
sia (CAH), both male and female fetuses are exposed to
high levels of testosterone in utero. Usually, after birth,
they receive medication to reduce testosterone; thus,
later differences are likely attributable to organizational
effects of testosterone. As adults, women with CAH report
homosexual orientation at elevated rates, compared with
unaffected women. Despite this elevation, most women
with CAH report exclusively heterosexual attractions
(Meyer-Bahlburg, Dolezal, Baker, & New, 2008)—thus,
the high levels of prenatal androgens do not ensure
homosexuality in these women. Similar to unaffected,
homosexual women, women with CAH tend to exhibit
gender nonconformity throughout their lives, on average
(Berenbaum & Beltz, 2011). Although other syndromes
(e.g., 5-alpha-reductase deficiency) provide relevant evi-
dence to the organizational hypothesis, CAH remains the
best instantiation of the theory.
Finger length ratios. The animal and human studies
just discussed at best establish the plausibility of the
organizational hypothesis of human sexual orientation.
Direct evidence would include studies showing that
homosexual people had a variety of signs of atypical
early androgen exposure. Besides gender nonconformity
(which might in theory reflect the influences of socializa-
tion), there are few replicated findings of this type.
One exception is the 2D:4D digit ratio, or the ratio of
index to ring finger length. This index shows a moderate
sex difference (δ = 0.46),16 with women having larger
ratios and men having smaller ones. Several lines of evi-
dence pertaining to digit ratios are consistent with the
organizational hypothesis. First, women with CAH exhibit
smaller (i.e., more masculine) 2D:4D ratios (Hönekopp &
Watson, 2010). Second, genetic male (XY) individuals
with androgen insensitivity syndrome display a feminine
pattern of digit ratios, which supports a hormonal rather
than a genetic mechanism for the sex difference
(Berenbaum, Bryk, Nowak, Quigley, & Moffat, 2009).
Third, the same sex difference is present in mice, in
which experimental manipulations have confirmed that
prenatal androgens act directly upon osteoblasts in the
digit to engender the sex difference (Z. Zheng & Cohn,
2011). Fourth, a meta-analysis (Grimbos, Dawood,
Burriss, Zucker, & Puts, 2010) confirmed that homosexual
and heterosexual women differ, on average, in terms of
their 2D:4D ratios, with homosexual women having a
more masculine ratio. Taken together, this evidence
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 71
suggests that prenatal androgens are responsible for the
development of gynephilia in at least some human
females (Williams etal., 2000). These results also suggest
that most men are gynephilic in orientation because of
their exposure to prenatal androgens, an idea we return
to below in our discussion of cloacal exstrophy. Interest-
ingly, there is no difference in digit ratios between homo-
sexual and heterosexual men, which indicates that
variance in prenatal androgen does not account for vari-
ance in sexual orientation in men. Alternatively, variance
in brain responsiveness to prenatal androgen may con-
tribute to variance in male sexual orientation.
Although the 2D:4D evidence supports the organiza-
tional hypothesis of female (but not male) sexual orienta-
tion, it is not decisive evidence. The sex difference in
2D:4D is not large, and therefore 2D:4D cannot be an
especially accurate marker of prenatal androgen expo-
sure. On the one hand, this could mean that the female-
sexual-orientation findings are especially impressive,
given the weak 2D:4D signal. On the other hand, varia-
tion in 2D:4D surely has many causes unrelated to prena-
tal androgens, and these could contribute to the
association with female sexual orientation.
The human SDN-POA. The brain is the fundamental
cause of all impulses, desires, thoughts, and behaviors.
Hence, any difference between the thoughts and feelings
of two individuals can be called a neurological differ-
ence. But where exactly does the difference reside? Does
it reflect the same basic brain structures functioning in
different ways (through different patterns of neural con-
nections), or does it reflect differences in brain structures
In 1991, neuroscientist Simon LeVay of the Salk Institute
for Biological Studies created an international furor by
publishing a paper in Science with the finding that 19
homosexual and 17 presumed heterosexual men17 differed
in the size of the third interstitial nucleus of the anterior
hypothalamus (INAH-3), a brain region in the preoptic
area, such that the INAH-3 was larger in heterosexual men.
Heterosexual men’s INAH-3 volume was more than twice
that of homosexual men’s, whose volume was similar to
that of a control group of presumed heterosexual women.
LeVay’s study required autopsied brains, and all his homo-
sexual subjects had died of AIDS. However, LeVay ruled
out AIDS as the cause of the difference, because the differ-
ence remained significant when he restricted his hetero-
sexual male sample to those who had also died of AIDS.
Also, among heterosexual subjects, there was no correla-
tion between brain volume and AIDS. The common asser-
tion that LeVay’s findings may have reflected AIDS rather
than sexual orientation has no apparent merit.
LeVay did not choose brain areas for analysis haphaz-
ardly. Rather, he focused on four clumps of cells in the
hypothalamus: the INAH-1, INAH-2, INAH-3, and INAH-
4. These cell groups had been previously compared
between men and women of unspecified orientation; two
of the cell groups—INAH-2 and INAH-3—were sexually
dimorphic, such that the INAH-3 was two to three times
larger in men (Allen, Hines, Shryne, & Gorski, 1989).
Thus, the importance of LeVay’s finding was magnified
by the fact that the brain region that differed between
homosexual and heterosexual men was sexually dimor-
phic; indeed, LeVay replicated this finding. A second
important fact is that the INAH-3 is in the hypothalamus.
It appears to be similar in some ways to the previously
mentioned SDN-POA, which shows sexual dimorphism
due to organizational hormonal effects in most mammals
that have been studied (Byne etal., 2000; Byne et al.,
2001; LeVay, 2011). This argues against the common criti-
cism of LeVay’s findings—that they could be a conse-
quence of the “homosexual lifestyle.” LeVay’s study was
important precisely because it supported the organiza-
tional hypothesis of human sexual orientation with direct
neuroscience data.
LeVay’s findings were from a small sample. Although
they were highly statistically significant, it is obviously
important that the findings also be replicable. Replicating
the study was difficult because it required dissecting
brains of people with known (or well-inferred) sexual
orientations. The tragic AIDS epidemic was waning when
neuroscientist William Byne began collecting specimens
for a replication. He assembled a sample of brains of 14
HIV-positive homosexual men, 34 presumed heterosex-
ual men (10 of whom were HIV positive), and 34 pre-
sumed heterosexual women (9 of whom were HIV
positive). Byne’s laboratory replicated the sex difference
in INAH-3 size among heterosexual subjects (Byne etal.,
2000; Byne et al., 2001), such that heterosexual men’s
INAH-3 had both more neurons and larger volume com-
pared with heterosexual women’s. Furthermore, Byne
argued on the basis of cytoarchitectonic evidence that
the human INAH-3 (as well as the dorsocentral portion of
the anterior hypothalamic nucleus, or AHdc, in rhesus
macaques [Macaca mulatta]) is a likely homologue of
the SDN-POA that has been so well studied in other spe-
cies (Byne, 1998).18
LeVay’s key sexual-orientation finding was not clearly
replicated (Byne et al., 2000; Byne et al., 2001). Byne
etal. found a statistical trend (i.e., .05 < p < .10) whereby
homosexual men had a smaller INAH-3, but they found
no difference in the number of neurons. To be sure, Byne
et al.’s sample was also small, with even fewer homo-
sexual men than LeVay’s. Scientifically, Byne et al.’s
research should not be the last word on the INAH-3 and
sexual orientation. Practically, however, it is likely to be
the last word for the foreseeable future. The INAH-3 is so
small—about the size of a grain of sand—that it cannot
by guest on July 5, 2016psi.sagepub.comDownloaded from
72 Bailey et al.
be accurately measured without dissecting brain tissue.
With effective treatments for HIV, there is no longer a
steady stream of brain tissue from homosexual men—an
entirely happy development, but one that prevents fur-
ther study of the human INAH-3 and sexual orientation,
for now. If the true INAH-3 difference between hetero-
sexual and homosexual men is intermediate between
LeVay’s and Byne etal.’s results, it would still be consis-
tent with the organizational hypothesis of sexual orienta-
tion. However, even if this were true, it would be unlikely
that the INAH-3 size would be a key factor regulating
sexual orientation. This is because there would be too
many exceptions—homosexual men with a large INAH-3
and heterosexual men with a small INAH-3—to believe
that INAH-3 size is crucial.
There have been several other reports of brain differ-
ences between homosexual and heterosexual subjects,
some of which were detected using brain imaging (e.g.,
Ponseti etal., 2007). Although we find those studies inter-
esting, we do not review them in this article. None of the
brain areas in these articles has been studied as inten-
sively as the INAH-3, and none is as relevant to the orga-
nizational hypothesis of sexual orientation.
Limitations of the hormonal evidence. Both experi-
mental nonhuman-animal studies and clinical human
studies of the effects of atypical early hormone exposure
support the possibility that such exposure can influence
sexual orientation. However, both lines of evidence have
important limitations. The main limitation of the animal
research is the lack of a close correspondence between
the animal behaviors studied and human sexual orienta-
tion. For example, one of the main behaviors hormonally
manipulated in laboratory mammals is the propensity to
either mount (a male-typical behavior) or exhibit lordosis
(i.e., move the body in a way that permits penile penetra-
tion; a female-typical behavior). Neither of these behav-
iors maps well onto human sexual orientation.
Limitations of clinical studies of humans include the
lack of a model of male homosexuality analogous to the
CAH model of androgen influences on female homosex-
uality. Furthermore, atypical early hormonal exposure
typically results in atypical genital development. For
example, girls with CAH are often born with masculin-
ized genitalia including enlarged clitorises. However,
there is no conclusive evidence that homosexual and het-
erosexual people differ in their genital anatomy. In prin-
ciple, atypical androgen exposure could cause human
homosexuality without also causing abnormal genital (or,
more generally, gross anatomical) development if there
were different critical periods for brain and genital dif-
ferentiation and if atypical androgen exposure occurred
only during the former. This suggestion is plausible,
given that different sensitive periods for genital and
behavioral masculinization have been demonstrated for
some nonhuman species, such as rhesus macaques (Goy,
Bercovitch, & McBrair, 1988).
Given the limited direct evidence for the organiza-
tional hypothesis that prenatal androgens cause human
sexual orientation, some may be tempted to reject the
hypothesis. We think that would be premature. Some-
times good hypotheses are difficult to test. Furthermore,
a remarkable natural experiment whose results are
strongly consistent with the organizational hypothesis
has been underappreciated. We review this natural
experiment in the next section.
The near-perfect quasi-experiment. Imagine that sci-
entists wanted to conduct a single study to determine the
degree to which sexual orientation is influenced by
nature versus nurture. Imagine also that scientists were
not ethically constrained. The following study would be
difficult to beat: Take newborn boys, surgically feminize
them (i.e., castrate them and surgically transform their
penises into plausible facsimiles of female genitalia), and
present them as girls to naive adoptive parents. (An anal-
ogous study of natal females would also be desirable, but
it is currently impossible to transform female genitalia
into a facsimile of a penis.) Follow these children into
adulthood and assess their sexual orientations. If only
nurture is important in determining whether people are
sexually attracted to men or women, then these individu-
als should all be attracted to men, because they were
raised female. In contrast, if only nature is important,
then they should all be attracted to women, because
before birth, and consistent with the organizational
hypothesis, they were exposed to typical male hormonal
and genetic influences and the large majority of men are
attracted to women. Of course, intermediate results and
appropriate interpretation would also be possible.
An unintentional version of this study has been con-
ducted on a few unfortunate subjects, all born male,
because they either were born with malformed penises
or lost their penises in surgical accidents. Between 1960
and 2000, many doctors in the United States believed that
such males would be happier being socially and surgi-
cally reassigned female. This belief has changed, as has
medical practice (Diamond etal., 2011), but not before a
generation of these individuals were made part of this
natural experiment. Besides surgical accidents, the other
medical reason for early sex reassignment that we con-
sider here has been cloacal exstrophy, a severe abdomi-
nal malformation that includes penile malformation in
boys. Importantly, neither cloacal exstrophy nor surgical
accidents are associated with abnormalities of prenatal
androgens. Thus, the brains of these individuals were
male-organized at birth, or at least as male-organized as
those of typical boys are at birth, on average.
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 73
Far more attention has been paid to the eventual gen-
der identities of these cases than to their sexual orienta-
tions. Indeed, sexual orientation has often not been
assessed, presumably because of the sensitivity of asking
the question (Meyer-Bahlburg, 2005). But the few pub-
lished cases that have provided sexual-orientation infor-
mation have shown remarkably consistent results (Table
3). In all seven cases, sexual orientation was predomi-
nantly or exclusively gynephilic (i.e., attraction was
toward women).19 This is the result we would expect if
male sexual orientation were entirely due to nature, and
it is opposite of the result expected if it were due to nur-
ture, in which case we would expect that none of these
individuals would be predominantly attracted to women.
These results comprise the most valuable currently
available data concerning the broad nature-versus-nurture
question for sexual orientation. They show how difficult it
is to derail the development of male sexual orientation by
psychosocial means. If one cannot reliably make a male
human become attracted to other males by cutting off his
penis in infancy and rearing him as a girl, then what other
psychosocial intervention could plausibly have that effect?
These cases establish a strong prima facie case that in
males, heterosexual orientation is at least partly estab-
lished before birth. And if male heterosexual orientation
is so difficult to alter after birth, there is no reason to
doubt the same is true of male homosexual orientation.
However, these cases fall short of the perfect experi-
ment. Most obviously, assignment to condition is not ran-
dom, so we call it a “quasi-experiment” rather than an
experiment. Another imperfection is that parents know
the birth status of these children (i.e., male) and might
conceivably treat them differently than natal girls and in a
way that influences their sexual orientation. Although this
possibility cannot be presently excluded, it is vague and
ad hoc. After all, the parents have agreed to surgical femi-
nize these children and to raise them as girls. What subtle
parental signals could trump these massive interventions?
In any case, the available evidence indicates that in such
instances, parents are deeply committed to raising these
children as girls and in as gender-typical a manner as pos-
sible (Colapinto, 2000).
A second limitation of this evidence is the small num-
ber of cases: only seven published, although several
more cases are potentially available (Reiner & Gearhart,
2004). For two reasons, rejecting these findings because
of their small sample size would be a mistake, even
though a larger sample would obviously be desirable.
First, the findings comprise the closest test of nature ver-
sus nurture with respect to sexual orientation, and right
now they are all we have. Second, the seven cases have
remarkably similar outcomes with respect to sexual ori-
entation. There is no good reason to suspect that this
would change if additional cases similar to these were
added. Because of the limited sample size, however, the
data provide much stronger support for the importance
of nature than they do for the lack of importance of
A third limitation pertains to gender identity. Gender
self-reassignment in these cases—boys reassigned and
reared as girls renouncing their sex of rearing and declar-
ing that they are male—is not unusual. It is also not uni-
versal. A disproportionate number of the cases for which
we have sexual-orientation information—six of seven—
identified as male at follow-up. Several of the other cloa-
cal exstrophy cases were still living as female. Notably,
none of them would discuss their sexual attractions, in
contrast to two adolescents with cloacal exstrophy who
were born female, and who discussed their attractions to
males. We suspect that the former cases—natal males
reassigned as female and still living in the female role—
are highly likely to be gynephilic. Certainly, however, it is
scientifically important to obtain data from such cases. It
would also be desirable to support any self-reported
claims of androphilia (i.e., attraction to males) with objec-
tive measures. This is because those individuals still in
Table 3. Sexual Orientations of Hormonally Typical Males Reassigned as Females Early in Life
Study Condition
Age at
Age at
Gender identity
at follow-up Sex attracted to
M. Diamond and Sigmundson (1997) Surgical accident 17 months 32 Male Females
Bradley, Oliver, Chernick, and Zucker
Surgical accident 7 months 26 Female “Predominantly”
Reiner and Gearhart (2004; Subject 9)aCloacal exstrophy 0 months (birth) 21 Male Females
Reiner and Gearhart (2004; Subject 10)aCloacal exstrophy 0 months (birth) 11 Male Females
Reiner and Gearhart (2004; Subject 14)aCloacal exstrophy 0 months (birth) 20 Male Females
Reiner and Kropp (2004)aCloacal exstrophy 0 months (birth) NR Male Females
Reiner and Kropp (2004)aCloacal exstrophy 0 months (birth) NR Male Females
Note: NR = not reported.
aOriginal publications included only summaries of a series of patients. Specific information for this table obtained from William G. Reiner.
by guest on July 5, 2016psi.sagepub.comDownloaded from
74 Bailey et al.
the female role may feel social pressure to claim attrac-
tion to males even if none exists.
The importance of these cases applies only to male
sexual orientation. As noted, there is no close analogue
of this natural experiment in which hormonally normal
girls are reassigned and reared as boys. Female sexual
orientation could be different—less fixed by early hor-
monal and genetic factors—although evidence from CAH
and digit-ratios studies supports some early influences.
Two very different kinds of empirical studies have been
used to investigate genetic influences on sexual orienta-
tion: twin studies and molecular genetics studies. The
aim of twin studies is to estimate the general importance
of genetic versus environmental influences on sexual ori-
entation. Molecular genetics studies aim to identify par-
ticular genes that influence sexual orientation.
Twin studies. Twin studies quantify the magnitude of
genetic compared with environmental influences. Con-
ceptually, the most straightforward version of a twin
study involves identical (monozygotic, or MZ) twins sep-
arated shortly after birth and reared in separate, uncor-
related environments. Any similarity between the
separated twins must be due to the fact that they are
genetically identical and shared the same intrauterine
environment, and any differences must reflect postnatal
environmental differences. Unfortunately for science,
early-separated twins are quite rare, and separated twin
pairs in which at least one twin is homosexual are still
rarer. Thus, our knowledge of sexual orientation among
separated twins is limited to a few case reports, insuffi-
cient in number to draw firm conclusions (Eckert,
Bouchard, Bohlen, & Heston, 1986).
In contrast, the classical twin design relies on jointly
reared twins, who are far more common. This design
depends on the fact that there are two kinds of twins: MZ
and dizygotic (DZ) twins (also known as fraternal twins).
MZ twins are genetically identical, and DZ twins share
half their genes.20 The classical twin design allows esti-
mation of the magnitudes of genetic influences and two
kinds of environmental influences: shared and nonshared
environment. The respective estimates are heritability
(h2), shared environmentality (c2), and nonshared envi-
ronmentality (e2). (Shared environment causes siblings
reared together to be similar; nonshared environment
causes them to differ.) Each estimate is bounded by zero
and one, and the sum of the estimates equals one; they
are interpreted as the proportion of trait variance attribut-
able to variation in the respective influences. These esti-
mates require knowing the correlations of trait values
separately for MZ and DZ twins.21
It should be noted that existing twin studies all
assessed sexual orientation via self-report. It is possible
that some reports are false, and false reports seem most
likely to be denial of homosexual interests. It is possible
that true discordance is less common than implied by
these studies. Research supporting the validity of twin
discordance—that twins who report different sexual ori-
entations truly have them—is currently lacking, and
would be most desirable for MZ pairs. In males, persua-
sive evidence for valid twin discordance would consist of
different patterns of PPG-measured arousal to male and
female erotic stimuli. In females, viewing time for male
versus female stimuli would provide a potentially objec-
tive measure.
Table 4 includes the twin concordances from studies
of sexual orientation.22 Two general categories of twin
studies are distinguished: those using targeted sampling
and those using twin registries or probability sampling
(e.g., random telephone sampling). In targeted sampling,
homosexual (or, more generally, nonheterosexual) twins
are recruited explicitly, via advertisements or word of
mouth. This method of recruitment may be especially
vulnerable to sampling bias, because homosexual twins
not only must hear about the study, they must also decide
whether to participate, likely knowing that the researcher
is studying homosexuality. When deciding whether to
participate, the potential participants of such studies may
consider their twins’ orientations (in order to avoid suspi-
cion or tension that may be more likely between siblings
of different orientations), thus leading to an overrepre-
sentation of concordant pairs (Kendler & Eaves, 1988).
Furthermore, concordant MZ pairs may be especially
likely to be recruited, because they are especially likely
to both meet potential referrers (Torrey, 1992). Such bias
would inflate MZ relative to DZ concordances, leading to
spuriously high estimates of heritability. Studies using
twin registries or probability sampling are less suscepti-
ble to sampling bias because the population of interest
has been assembled without regard to sexual orientation.
Table 4 confirms the likely increased bias in the targeted
samples. Median MZ and DZ concordances are .52 and
.17, respectively, in the targeted samples, compared with
.24 and .15 in the registry/probability samples. Note that
the median concordances also show a larger MZ-DZ dif-
ference for the targeted samples than for the registry/
probability samples. It is likely that the latter figures pro-
vide better estimates of the true concordances.
What can we conclude from the results of the better
samples? We provide only a rough summary here, rather
than engaging in rigorous estimation and hypothesis test-
ing. There are too few registry/probability samples of
either sex to provide separate analyses, and therefore we
consider male, female, and mixed samples together. The
MZ concordance rate exceeds the DZ rate in six of seven
by guest on July 5, 2016psi.sagepub.comDownloaded from
Table 4. Probandwise Concordances of Published Twin Studies of Sexual Orientation
Recruitment Location Study
Definition of
nonheterosexuality Sex
of MZ
Number of
MZ cotwins
of DZ
Number of
DZ cotwins
Rate of
in sample
Targeted United States Kallmann
Kinsey score of 3 or
Male 37 37 26 3 1.00 0.12 503.57 NA
Targeted United States Bailey and
Pillard (1991)
Kinsey score of 2 or
Male 56 29 54 12 0.52 0.22 3.65 NA
Targeted United States Bailey, Pillard,
Neale, and
Agyei (1993)
Kinsey score of 2 or
Female 71 34 29 3 0.48 0.10 6.97 NA
Targeted United States Whitam,
and Martin
Kinsey score of
4 or greater for
probands, 2
or greater for
Male 34 22 14 4 0.65 0.29 4.20 NA
Targeted United Kingdom King and
sexual identity
(“homosexual” or
Mixed 17 4 24 4 0.24 0.17 1.52 NA
— — 0.52 0.17 4.20 NA
United States Kendler,
and Kessler
sexual identity
(“homosexual” or
Mixed 19 6 15 2 0.32 0.13 2.60 0.03
Twin registry United States Hershberger
Uncertain Male 16 4 8 2 0.25 0.25 0.94 0.04
Twin registry United States Hershberger
Uncertain Female 11 6 8 2 0.55 0.25 3.07 0.03
Twin registry Sweden Långström,
Any lifetime same-
sex partners
Male 78 14 56 6 0.18 0.11 1.75 0.06
Twin registry Sweden Långström
etal. (2010)
Any lifetime same-
sex partners
Female 240 52 153 26 0.22 0.17 1.34 0.08
Twin registry Finland Alanko etal.
Score of at least
1.5 on the Sell
Assessment of
Sexual Orientation
Mixed 16 6 23 0 0.38 0.00 29.10 0.02
Twin registry Australia Zietsch etal.
Self-reported sexual
preference for
same sex, other
sex, or both sexes
Mixed 134 32 154 20 0.24 0.13 2.08 0.04
— — 0.25 0.13 2.08 0.04
Note: Only studies with at least 10 monozygotic (MZ) and 10 dizygotic (DZ) probands were included. Probandwise concordance represents the probability that an index twin with the trait (in this case, nonheterosexuality) has a twin who also has
the trait. Both members of a twin pair are probands if they were ascertained independently, as they typically would be in a twin registry study. Probandwise concordance contrasts with pairwise concordance, which is the probability that a twin
pair will be concordant. Although probandwise concordance tends to be the less intuitive index, it is the correct one (McGue, 1992). NA = not applicable.
aConcordance data were not in the original publication but were obtained from the authors.
by guest on July 5, 2016psi.sagepub.comDownloaded from
76 Bailey et al.
studies (with the exception a statistical tie), and the aver-
age effect size (log odds ratio) significantly differs from
zero (z = 3.14, p = .0017). This supports the hypothesis
that heritability exceeds zero, although the alternative
hypothesis is not resoundingly rejected. The median MZ
and DZ concordances are .24 and .15, respectively, and
the median rate of nonheterosexual orientation is .04.
These figures translate into tetrachoric correlations of .57
and .41, and heritability is estimated as 2 × (.57 − .41) =
.32, meaning that about a third of variation in sexual ori-
entation is attributable to genetic differences. The best
estimate of the nonshared environmentality is simply one
minus the MZ correlation,.43, and shared environmental-
ity is the remainder, .25.
Twin studies that have included retrospective child-
hood gender nonconformity have yielded some addi-
tional intriguing findings (Bailey, Dunne, & Martin, 2000;
Bailey & Pillard, 1991; Bailey, Pillard, Neale, & Agyei,
1993). First, the probability that either male or female MZ
twin pairs were concordant for sexual orientation was
unrelated to the degree of homosexual twins’ childhood
gender nonconformity. Thus, there is no evidence that
homosexuality associated with childhood gender non-
conformity is especially heritable. Second, among discor-
dant MZ twin pairs, there were significant and substantial
differences in childhood gender nonconformity, with the
homosexual twin recalling much more gender noncon-
formity compared with the heterosexual twin. This sug-
gests that nonshared environmental influences on sexual
orientation begin early, at latest by childhood. Third,
among concordant MZ pairs, there was a high correlation
for childhood gender nonconformity: Both twins tended
to have shown high, moderate, or low levels of gender
nonconformity in childhood. This finding has been repli-
cated in a study of nontwin brothers (Dawood, Pillard,
Horvath, Revelle, & Bailey, 2000). It suggests that genes
or shared environment may affect the particular develop-
mental expression of sexual orientation.
In conclusion, the evidence supporting a genetic influ-
ence on sexual orientation is consistent, although sam-
pling biases remain a concern even for the best available
studies. Our best estimate of the magnitude of genetic
effects is moderate—certainly not overwhelming. In con-
trast, the evidence for environmental influence is unequiv-
ocal, given that MZ twin concordances tend to be far less
than 100%,23 assuming that the MZ twin pairs are truly
discordant. It is important to recognize that the limits of
this evidence reflect the difficulty of conducting twin
studies of sexual orientation. Nonheterosexual orientation
is relatively rare, and so even a large twin registry will
have an insufficient number of twin pairs with nonhetero-
sexual members to provide very good estimates. Twin
studies of some rare phenomena such as schizophrenia
have succeeded because some nations keep good records
of both who is a twin and who has the trait, allowing the
assembly of a sufficient sample (Cardno & Gottesman,
2000); no analogous situation for nonheterosexual orien-
tation is likely for the foreseeable future.
Based on the evidence from twin studies, we believe
that we can already provide a qualified answer to the
question “Is sexual orientation genetic?” That answer is:
“Probably somewhat genetic, but not mostly so.” On the
one hand, that answer is not surprising, given the evo-
lutionary pressure against genes that diminish repro-
duction, as genes for homosexuality likely do, especially
in males (Vasey, Parker, & VanderLaan, 2014). On the
other hand, we expect many people will find the con-
clusion surprising, mainly because they have miscon-
strued the meanings of “genetic” and “environmental.
There can be little doubt that sexual orientation is envi-
ronmentally influenced. However, to acknowledge this
does imply that the social environment shapes sexual
orientation. There is a social environment, but there is
also a vast and largely unexplored nonsocial environ-
ment. Thus, the conclusion that sexual orientation is
socially influenced requires evidence in addition to that
produced by twin studies. In a later section, we exam-
ine the evidence for social environmental effects on
sexual orientation.
Molecular genetics studies. Twin studies provide
information about the broad question of how important
all genes are, together, in causing a trait. In contrast,
molecular genetics studies have the potential to identify
particular genes and elucidate the causal pathway from
gene to trait. One of the most famous studies of the
causes of sexual orientation was a molecular genetics
study published in 1993 by the geneticist Dean Hamer of
the National Cancer Institute (Hamer etal., 1993). Hamer
used a sibling-pairs linkage design: First, he recruited
pairs of homosexual brothers, from whom he collected
DNA. In the data-analysis phase, he looked for chromo-
somal segments that were shared more often by these
brothers than one would expect by chance (i.e., on aver-
age, any particular segment should be identical by
descent [IBD] 50% of the time). Any segment with
increased sharing may have a gene inside it affecting the
trait. (It is often not understood that in linkage analysis,
detected chromosomal segments have many different
genes, and so linkage analysis is in this sense a precursor
to more precise gene mapping.) Hamer found one chro-
mosomal region with increased sharing: Xq28, which is
located at the tip of the X chromosome. The finding
seemed especially interesting because Hamer also
reported evidence that homosexual men were more
likely to have homosexual male relatives on their moth-
ers’ sides than on their fathers’ sides, a finding consistent
with X-linkage.
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 77
Hamer’s linkage finding was controversial, in part
because the linkage sample contained only 40 pairs of
brothers. Although Hamer’s lab reported a successful
replication (Hu etal., 1995), another lab failed to repli-
cate the results with a sample of 52 pairs of homosexual
brothers (G. Rice, Anderson, Risch, & Ebers, 1999). Later,
Hamer’s own lab published a linkage study from an
expanded sample, including their original subjects (from
1993 and 1995), in which they failed to find significant
linkage evidence for Xq28 but instead found evidence
for linkage to regions on chromosomes 7, 8, and 10
(Mustanski et al., 2005). No significant genetic linkage
has been reported for female sexual orientation.
During the years after Hamer’s original report, it
became clear that molecular genetics findings were
not replicating well (Ioannidis, Ntzani, Trikalinos, &
Contopoulos-Ioannidis, 2001), a conclusion that has
extended to most of science (Ioannidis, 2005). Partly this
reflected genetic phenomena such as heterogeneity—
that is, the fact that different genes can lead to the same
outcome. Partly it reflected Type 1 statistical error due to
(massive) multiple testing. There has been universal rec-
ognition of the importance of using much larger samples
than Hamer’s. The largest linkage study of male sexual
orientation to date, which included 409 pairs of homo-
sexual brothers, was recently published (Sanders etal.,
2014). The study’s main positive findings included both
the detection of linkage for the pericentromeric (i.e., situ-
ated near the center) region on chromosome 8 and rep-
lication of the Xq28 linkage. Although it is encouraging
that the largest study replicated some findings from
smaller studies, the case is not closed. Still larger studies
are needed to provide the degree of certainty now
expected in molecular genetics, and in any case, the
mapping case is not closed in molecular genetics until
one has identified the genes that affect a trait.
To this end, there has also been a shift from linkage
studies, which look at patterns of DNA sharing within
families, to genome-wide association studies (GWAS),
which compare DNA between individuals with and
without a trait variant of interest. In the only completed
GWAS study of sexual orientation to date, researchers
from the genomics company 23andMe examined genetic
markers from nearly 24,000 people who provided self-
reports of their sexual orientations (Drabant et al.,
2012). The sample comprised 1,181 exclusively homo-
sexual men and 10,679 exclusively heterosexual men, as
well as more than 1,500 men of intermediate orienta-
tions, and 733 women whose degree of nonheterosexu-
ality was at least bisexual, 7,599 exclusively heterosexual
women, and more than 2,000 women of intermediate
orientation. GWAS studies examine group differences in
thousands of markers, and so statistical corrections must
be made to minimize Type 1 error. After these
corrections, the study yielded no significant effects. Still,
the genetic marker closest to statistical significance in
men was located on pericentromeric chromosome 8 in
the same region identified by Mustanski et al. (2005)
and Sanders etal. (2014). (Results for women did not
reach conventional levels of statistical significance after
necessary corrections for multiple testing.) Although
these findings are interesting, we note that even the
23andMe sample was modestly sized—particularly the
nonheterosexual portion—and potentially underpow-
ered in the present era of GWAS (Moonesinghe, Khoury,
Liu, & Ioannidis, 2008). Based on what we know about
molecular genetics findings in general (Manolio etal.,
2009), we expect that any sexual-orientation genes will
have small effects individually. Thus, consortia of coop-
erating research teams may be necessary to obtain suf-
ficiently large samples.
Molecular genetics findings can also be used to illumi-
nate environmental influences. In particular, the concept
of epigenetics—chemical modifications of the genome,
such as DNA methylation—is consistent with environ-
mental influences on gene expression. Epigenetics could
play a role in some MZ twin discordance (Petronis etal.,
2003). In the case of sexual orientation, indirect evidence
for epigenetic influence on male sexual orientation
includes the high MZ discordance rate, the fraternal-
birth-order effect, and some interesting but as yet unrep-
licated molecular genetic associations (Ngun & Vilain,
2014).24 However, a preliminary report of a study of 34
male MZ twin pairs discordant for sexual orientation
revealed no support for this hypothesis (Bocklandt etal.,
Evolution and homosexuality. Human homosexuality
has rightly been called an evolutionary paradox. Both
male and female sexual orientation appear to be moder-
ately heritable, and homosexuality is known to have
existed for many generations. Yet it tends to reduce
reproductive success (Bell & Weinberg, 1978) and, in one
non-Western sample, has reduced it to zero (Vasey etal.,
2014). Although it is likely that in some cultural contexts,
nonheterosexuality decreases reproduction to a lesser
extent (Murray, 2000), a genetic contribution to sexual
orientation is paradoxical unless there is no difference in
reproduction rates between heterosexual and nonhetero-
sexual people. Thus, the paradox concerns how any
gene that increases the likelihood of homosexuality could
persist. Furthermore, even if heritability of sexual orienta-
tion were zero, we would still need to understand how
the processes regulating the development of sexual ori-
entation allow a nontrivial rate of variants associated with
reduced reproduction to exist.
Most research on the evolution of same-sex sexuality
in humans has focused on explaining the origin and
by guest on July 5, 2016psi.sagepub.comDownloaded from
78 Bailey et al.
persistence of male androphilia. (We refer here to andro-
philia and gynephilia rather than homosexuality because
some of this work was conducted with populations for
whom the term “homosexuality” does not make sense.)
The evolution of female gynephilia remains underre-
searched (but see L. M. Diamond, 2006).
To date, research has addressed two main hypotheses
that attempt to explain the persistence of genes for male
androphilia: the kin selection hypothesis (KSH) and the
sexually antagonistic gene hypothesis (SAGH). The KSH
holds that genes for male androphilia could be main-
tained to the extent that having androphilic males as
close relatives helps people to reproduce more than they
otherwise would have (Wilson, 1975). In theory, andro-
philic males could increase their indirect fitness—a mea-
sure of an individual’s impact on the fitness of kin (who
share some identical genes by virtue of descent), weighted
by the degree of relatedness (Hamilton, 1963)—by direct-
ing altruistic behavior toward kin, which, in principle,
would allow kin to increase their reproductive success.
Consistent with the predictions of the KSH, research
conducted in Samoa on transgender androphilic males
(fa’afafine) has repeatedly demonstrated that they show
elevated avuncular (uncle-like) tendencies compared
to Samoan women and gynephilic men. (This is mea-
suredvia a 9-item scale measuring willingness to care for,
and togive resources to, nieces and nephews.) Further-
more, this finding does not appear to reflect a general
tendency to help others, but a specific preference for kin
(Vasey &VanderLaan, 2010). In contrast, research on cis-
genderandrophilic males in Western populations (Abild,
VanderLaan, & Vasey, 2014; Bobrow & Bailey, 2001;
Rahman & Hull, 2005) and non-Western industrialized cul-
tures (Vasey & VanderLaan, 2012) has garnered virtually
no support for the KSH. It is possible that elevated avun-
cularity is not expressed unless male androphilia takes on
the transgender form. More research is needed to ascertain
whether other populations of transgender male andro-
philes exhibit elevated kin-directed altruism or not.
The SAGH for male androphilia holds that genes asso-
ciated with the development of androphilia result in
decreased reproduction in male carriers but increased
reproduction in female carriers. The female relatives of
androphilic males are especially likely to carry these
genes, and so their increased reproductive success helps
balance the diminished reproductive success of their
androphilic male relatives. Sexual-antagonism-balancing
selection works especially well for genes on the X chro-
mosome. This is because males have only one copy of
any gene on the X chromosome, whereas females have
two copies. Thus, in theory, females have twice the
chance of benefit by fitness-enhancing X-linked genes
compared with their androphilic male relatives. (This is
one reason why scientists have found the idea of a gene
on Xq28 so exciting.) Over time, modifier genes should
be selected that suppress male androphilia but allow the
adaptive effects of the genes when carried by females
(W.R. Rice, 1984).
The strongest support for the SAGH has been gar-
nered exclusively in Western Europe, where the mothers
and maternal aunts of male androphiles have been
repeatedly shown to exhibit elevated reproductive out-
put compared to those with no androphilic sons or neph-
ews (e.g., Camperio-Ciani, Corna, & Capiluppi, 2004).
The dearth of such studies in non-Western cultures is
unfortunate, because reproductive output in the West is
artificially low because of family planning and birth con-
trol. For example, the contemporary birth rate in Italy,
where several studies have been conducted, is 1.4 births
per woman during her lifetime (World Bank, 2014). In
contrast, the typical hunter-gatherer female will give birth
to 6 children (Pennington, 2001). For this reason, studies
of evolutionary hypotheses that focus on reproductive
rates in contemporary Western populations are of very
limited value for explaining the origin and persistence
of traits (Perusse, 1993). Several studies conducted in
Samoa, a non-Western culture with higher fertility rates,
have indicated that the mothers and grandmothers of
androphilic males exhibit elevated reproductive output
compared to those with only gynephilic male rela-
tives (VanderLaan, Forrester, Petterson, & Vasey, 2012;
VanderLaan & Vasey, 2011; Vasey & VanderLaan, 2007).
However, elevated reproduction among mothers and
grandmothers furnishes inconclusive evidence for the
SAGH. This is because reproductive output among these
categories of female kin could be influenced by the
genetic contributions from fathers and grandfathers
(VanderLaan, Garfield, etal., 2014). As with the KSH, fur-
ther work is needed to clarify the degree to which the
SAGH can explain the evolution of male androphilia.
An important limitation of this evolutionary research is
that it remains unclear how either kin selection or sexu-
ally antagonistic selection could offset the reproductive
costs associated with male androphilia. This is because
the direct reproductive costs of male androphilia are
severe (Bell & Weinberg, 1978; Vasey etal., 2014), and
the indirect benefits to kin via the androphilic male’s
investment (KSH) or genes (SAGH) would need to be
very high indeed to completely balance the costs. Accord-
ing to the KSH, each child an androphilic male could
have via direct reproduction, if he were gynephilic,
would need to be balanced by four extra nieces or neph-
ews who survive to reproduce as a result of the andro-
philic male’s provision of resources, including protection
and child care. Conservatively assuming that gynephilic
males father two children who survive to reproductive
age, androphilic males would have to contribute an extra
eight surviving nieces and nephews. Similar calculations
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 79
could be made for the SAGH, although the presumed
mechanism underlying the calculations would differ. Of
course, both theories may help explain a portion of male
androphilia, even if they cannot explain it completely.
Further research is needed to quantify the relative inclu-
sive fitness of androphilic and gynephilic males.
Societies in which transgender male androphilia pre-
dominates exhibit a significantly greater presence of
human ancestral sociocultural conditions compared to
societies in which the cisgender form predominates
(VanderLaan, Ren, & Vasey, 2014). This suggests that the
transgender form of male androphilia was likely the
ancestral form. As such, transgender male androphilia
likely represents the best model for testing evolutionary
hypotheses, given that more derived forms of this trait
may reflect recent cultural/historical influences that might
obscure the outcome of evolutionary processes. Conse-
quently, the most promising results from tests of both the
KSH and SAGH are from studies of Samoan fa’afafine.
The evidence would be much stronger if other popula-
tions of transgender androphilic males showed similar
The nonsocial environment
The most persuasive evidence that environment affects
sexual orientation is the fact that the MZ twin of a non-
heterosexual person is usually heterosexual. What is the
nature of the environmental influence that causes the
sexual orientations of genetically identical twins to
diverge?25 We have already noted that the environment
comprises more than social influences. One striking
example of the importance of the nonsocial environment
comes from a review of twin concordances for the con-
genital brain defects hydrocephaly and anencephaly
(Torrey, Bowler, & Taylor, 1994).26 Among MZ twin pairs
in which one twin was born with one of these condi-
tions, the other twin was usually normal—although MZ
concordances were also higher than those for DZ twins.
The prenatal, nonsocial environment is clearly implicated
in these differences, and just as clearly the social environ-
ment has nothing to do with them. The causes of MZ
differences are a source of exciting speculation, but so
far they are poorly understood (Matias, Silva, Martins,
& Blickstein, 2014; van Dongen, Slagboom, Draisma,
Martin, & Boomsma, 2012). In this section, we review the
evidence for nonsocial environmental influences on sex-
ual orientation.
The fraternal-birth-order effect. One of the most
consistent findings in sexual-orientation research is the
fraternal-birth-order effect for male sexual orientation.
This is the finding that homosexual men tend to have
a greater number of older brothers compared with
heterosexual men, heterosexual women, and homosexual
women (Blanchard, 2004; Bogaert & Skorska, 2011). Plau-
sible confounds that cannot explain this effect include
parental age and other kinds of siblings. After statistically
controlling for number of older brothers, there is no dif-
ference in number of younger brothers or sisters or older
sisters. Nor is there any effect of the number of older or
younger siblings of either sex on female sexual orienta-
tion. The fraternal-birth-order effect applies only to male
sexual orientation. Notably, it has also been detected
among the Samoan fa’afafine (VanderLaan & Vasey, 2011;
Vasey & VanderLaan, 2007) and among androphilic trans-
gender males (who self-identify as transwomen) in West-
ern nations (Blanchard & Sheridan, 1992).
The effect is almost certainly causal, with each addi-
tional older brother causing an increase in the chances of
a man’s being homosexual. The effect is also large, in the
sense that it increases the odds of a man’s homosexuality
appreciably, by an estimated 33% to 34% (Cantor,
Blanchard, Paterson, & Bogaert, 2002; VanderLaan &
Vasey, 2011). Assuming that a man without any older
brothers has a 2% chance of being homosexual, a man
with one older brother has a 2.6% chance; with two,
three, and four older brothers, the chances are 3.5%,
4.6%, and 6.0%, respectively. Estimates of the proportion
of homosexual males who owe their sexual orientation to
the fraternal-birth-order effect have ranged from 15.1%
(Cantor et al., 2002) to 28.6% (Blanchard & Bogaert,
The fraternal-birth-order effect appears to depend
only on biological, not social, older brothers. We know
this because of a study by Bogaert (2006) that compared
homosexual men who had biological older brothers
(defined as those with the same biological mother as the
index subject) with a specially recruited sample of homo-
sexual men who had nonbiological older brothers (i.e.,
half-, step-, or adoptive siblings not from the same bio-
logical mother). The fraternal-birth-order effect was
found only among homosexual men with biological
older brothers. The crucial factor appears to be the num-
ber of older brothers one’s mother has given birth to,
whether one is reared with them or not.
Blanchard (2001) has offered a specific, immunologi-
cal hypothesis that can account for the fraternal-birth-
order effect as a maternal effect. Male fetuses carry
male-specific proteins on their Y chromosome, called
H-Y antigens. Blanchard hypothesized that some of these
antigens promote the development of heterosexual ori-
entation in males (it is likely that such development also
requires circulating testosterone at a critical period).
Because these H-Y antigens are not present in the moth-
er’s body, they trigger the production of maternal anti-
bodies. These antibodies bind to the H-Y antigens and
prevent them from functioning, which, in turn, impedes
by guest on July 5, 2016psi.sagepub.comDownloaded from
80 Bailey et al.
sexual differentiation of brain centers mediating male
sexual orientation. The degree of this reaction, the likeli-
hood of its happening, or both increase with successive
pregnancies with male offspring. Blanchard and col-
leagues have identified candidate male-specific antigens
that are preferentially expressed in the brain and hope to
test for antibodies to them in both mothers of homosex-
ual men and mothers of only heterosexual men.
Blanchard’s theory predicts that the former should have
an excess of antibodies, whereas the latter should not.
This maternal immune hypothesis is also consistent with
the possibility that genes on the Y chromosome play a
role alongside sex hormones in sexual differentiation
(Ngun, Ghahramani, Sánchez, Bocklandt, & Vilain, 2011).
Blanchard’s theory could help reconcile a combination
of findings that are difficult to explain with the theory
that sexual-orientation differentiation relies only on orga-
nizational androgens. Both females with CAH and female-
reared males with cloacal exstrophy are exposed to high
levels of prenatal androgens, but only the latter seem
especially likely to become attracted to women. Like all
females with the normal complement of sex chromo-
somes, females with CAH lack a Y chromosome, which is
hypothetically necessary for male heterosexuality. Like all
males, natal males with cloacal exstrophy have a Y chro-
mosome. Thus, perhaps it is the Y chromosome—in the
presence of prenatal androgens, which both groups
have—that accounts for the different sexual-orientation
outcomes in the two groups. A Y chromosome is not suf-
ficient to cause male heterosexual orientation, however.
Individuals who have a Y chromosome but are com-
pletely insensitive to the effects of androgen usually are
attracted to men (Wisniewski etal., 2000).
The fraternal-birth-order effect is at least partly envi-
ronmental, because birth order is obviously not genetic.
It could conceivably have a genetic component, how-
ever, because some mothers or sons may be genetically
more or less likely to cause or experience the effect. The
environmental component causes some brothers to have
different sexual orientations (i.e., firstborn heterosexual
sons vs. later-born sons who are homosexual as a result
of the effect), but it also causes other brothers to have the
same sexual orientations (i.e., multiple younger brothers
who are homosexual as a result of the effect). In this
sense, it is part of both the shared environment and the
nonshared environment. Two phenomena the fraternal-
birth-order effect seems incapable of explaining are
homosexuality in firstborn sons and sexual-orientation
discordance in MZ twins. With respect to the former, the
limitation of the theory is obvious. As for the latter, both
MZ twin brothers have the same number of older broth-
ers that might have induced a fraternal-birth-order effect.
Thus, additional nonshared environmental factors must
affect male sexual orientation. We have already noted
that the nonshared environment causing twin discor-
dance for sexual orientation tends also to affect child-
hood gender nonconformity, and hence is an early
influence; perhaps the nonshared environmental influ-
ence is prenatal like the fraternal-birth-order effect. Given
the nonsocial nature of the fraternal-birth-order effect, as
well as the results of the near-perfect quasi-experiment
regarding the outcome of sexually reassigned hormon-
ally normal boys, we hypothesize that these additional
factors may be nonsocial.
The social environment
Several hypotheses regarding potential social influences
on sexual orientation have a long history of social and
scientific controversy. These include the following ideas:
that homosexuality can be caused by “recruitment,” or
the sexual seduction of a younger, sexually naive person
by an older homosexual person; that the children of non-
heterosexual parents have increased rates of nonhetero-
sexuality due to social influences; and that psychotherapy
can change homosexuality. All of these are hypotheses of
We have reviewed evidence from twin studies show-
ing that environmental influence on sexual orientation is
considerable. As we have tried to make clear, MZ twin
differences generally signal environmental influence, but
the environment causing the differences need not com-
prise social influence. That is, MZ twins may differ in
their sexual orientation not because of different social
experiences but because of nonsocial differences, includ-
ing those that took effect before they were born. In other
words, environment can be part of what is generally
understood to be “nature” rather than solely of
We have also pointed to reasons why nonsocial envi-
ronmental influences on sexual orientation appear espe-
cially likely. These include the facts that many discordant
MZ twins tend to remember being treated differently in
childhood; that the fraternal-birth-order effect is at least
partly environmental and appears entirely nonsocial; and
that, for males, the results of the near-perfect quasi-
experiment suggest that even the most severe social envi-
ronmental pressure (i.e., early anatomical and social
transformation of males into females) does not affect
eventual sexuality.
We find this evidence strongly suggestive of nonsocial
rather than social influences, but if one disagrees, one
cannot merely assume instead that social influences are
important; one must make the scientific case that they
are. Hypothetically, the best way to provide evidence for
a particular social environmental hypothesis would be
with an experiment: Randomly assign people to have, or
not to have, a particular social experience and then later
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 81
assess their sexuality. Of course, in this case, the hypoth-
esized social influences—including sexual seduction
and/or abuse, psychotherapy to change orientation, and
parental sexual orientation—cannot be assigned at ran-
dom given ethical considerations. This means that we
must evaluate these hypotheses non-experimentally, with
all attendant limitations.
As any undergraduate student knows, correlation does
not necessarily imply causation. A major threat to the
validity of inferring causation from a correlation is selec-
tion bias. That is, people with different social experiences
may differ prior to having those experiences. In this case,
subsequent differences might reflect the preexisting dif-
ferences rather than be caused by the social experiences.
For example, children who are physically punished by
their parents are more likely than other children to have
later behavior problems. But it would be a mistake to
conclude that all, or even most, of this effect is due to the
causal effect of punishment. Much of the effect is due to
the fact that children tend to receive physical punishment
for especially severe misbehavior, and this tendency to
misbehave persists (Larzelere, Kuhn, & Johnson, 2004).
At the very least, one must attempt to statistically control
for the potential confound of initial misbehavior level.
This is easier said than done, because doing so requires
an accurate, objective measure of child misbehavior,
which would be onerous and expensive. And even after
controlling for obvious confounds, it is possible that one
will have omitted important, less obvious additional vari-
ables. In examining each of the specific social environ-
mental hypotheses below, therefore, we take pains to
consider selection effects.
Before we review particular hypotheses, we note that
some people sympathetic to homosexual rights are
uncomfortable with research on whether certain social
factors increase the likelihood of homosexuality. They
argue that because homosexual people are as worthy as
heterosexual people, no one should care if certain expe-
riences increase that likelihood. Although we are entirely
sympathetic to homosexual rights and also believe that
homosexual people are as worthy as heterosexual peo-
ple, we believe it is important to address putative social
environmental causes of sexual orientation for three rea-
sons. First, some existing social environmental hypothe-
ses about homosexuality are intrinsically repugnant—for
example, the idea that it is caused by older people seduc-
ing younger people. Second, showing that a particular
social hypothesis is false helps avoid unnecessary argu-
ments with people who do not share the same values
about the equal worth of homosexual and heterosexual
people. For example, showing that having a homosexual
teacher is unlikely to cause schoolchildren to become
homosexual may be easier than convincing some parents
that even if it did, they should not care. Third, scientists
should study controversial topics with the aim of finding
what is true, rather than declaring some topics off-limits
because they are sensitive. Indeed, it is in controversial
arenas where a scientific approach is most needed.
Recruitment/seduction. In an address before his
signing of the Anti-Homosexuality bill, President Musev-
eni of Uganda decried “those who were promoting
homosexuality and recruiting normal people into it”
(“President Museveni’s Full Speech, 2014, para. 5). The
idea that homosexual people recruit heterosexual peo-
ple into being homosexual has also been promulgated
in the West by a variety of anti-homosexual forces,
including anti-gay activists Anita Bryant (Fejes, 2008),
Judith Reisman (Blumenthal, 2004), and Paul Cameron
(Klingenschmitt, 2014), the anti-homosexual organiza-
tions Abiding Truth Ministries and Traditional Values
Coalition, and the Nazis of prewar Germany (Oosterhuis,
1997). To mention these groups, even the Nazis, is not
sufficient to dismiss the validity of the “recruitment
hypothesis,” which must be evaluated scientifically.
Rather, we do so to make the point that a belief in the
recruitment hypothesis has often been associated with
strongly negative attitudes toward homosexual people.
In general, those endorsing the recruitment hypothesis
have not also explicated the published empirical basis
for their belief. But establishing the scientific plausibility
of a belief requires evidence. Is there any? And what
would supportive evidence look like?
It would surely not be persuasive to show, as seems
generally true, that nonheterosexual people are much
more likely than heterosexual people to have a first sex-
ual partner of the same sex (e.g., Bell etal., 1981). This
result would be expected under two opposing hypothe-
ses: that nonheterosexuality is spread by recruitment, and
that nonheterosexual orientation is present upon sexual
awakening, so nonheterosexual people are more likely to
choose same-sex partners—no recruitment necessary.
Consistent with the idea that early nonheterosexual expe-
rience is the result, not the cause, of sexual orientation
(even if sexual orientation not acknowledged until later
in life), the large majority of nonheterosexual people
recalled experiencing homosexual feelings preceding
homosexual behavior by an average of 3 years in one
study (Bell etal., 1981).
It may also be more common for nonheterosexual
people—especially nonheterosexual males—to have
early sex partners who are older than they are. For exam-
ple, Holmes and Slap’s (1998) review suggested that boys
who had age-discrepant early sexual experiences, espe-
cially with men, were up to seven times more likely
to identify as homosexual or bisexual as adults. (Age-
discrepant sexual experiences tend to be operationalized
as having a partner at least 5 years older when one is
by guest on July 5, 2016psi.sagepub.comDownloaded from
82 Bailey et al.
below the age of 18.) This pattern would be predicted
from the recruitment hypothesis; would it also be pre-
dicted from the alternative hypothesis that recruitment is
not a cause of homosexual orientation?
For males, the answer is surely “yes.” To see why,
imagine two hypothetical 16-year-old male adolescents
named Jerome and Stephen. Jerome is heterosexually
oriented and Stephen is homosexually oriented. Who is
more likely to have sex with a significantly older person?
Jerome is openly heterosexual and interacts with many
female adolescents who might be sexually attracted to
him. In contrast, women at least 5 years older than
Jerome tend not to be interested in 16-year-old sex part-
ners, even in states in which having sex with them would
be legal (Silverthorne & Quinsey, 2000). Thus, Jerome
will most likely have sex with adolescent females around
his own age. What about Stephen? There is a good
chance that Stephen knows of few other homosexually
oriented adolescent males. This is partly due to popula-
tion base rates—even in a large high school, we would
not expect many of the male students to be homosex-
ual—but it is also partly due to the fact that in most
places, homosexual adolescents like Stephen are likely
to be hiding their homosexuality from others. Hence,
Stephen is especially likely to want to have sex with
someone outside of his social circle. Where might he
find someone? In many cities, there are places where
homosexual males meet and socialize, and these males
will tend to be older than Stephen. Assuming that
Stephen’s body is sexually mature (i.e., that he is in the
final stage of puberty), he will be sufficiently attractive
to homosexual men—just as 16-year-old females in the
final stage of puberty are sufficiently sexually attractive
to heterosexual men. Whether or not Stephen is above
the legal age of consent, some men at least 5 years older
may take the chance and have sex with him; the likeli-
hood that they would do so is much higher than the
likelihood that older women would have sex with Jerome
(Silverthorne & Quinsey, 2000).
Thus, an array of forces—some social and some likely
due to putatively universal human mating preferences
(Buss, 1989)—ensure that age-discrepant early sexual
relationships are more common among homosexual than
heterosexual adolescent males. Furthermore, given simul-
taneous interest in sexual activity, we might expect Ste-
phen to have sex earlier than Jerome, because his desired
partners (males) have higher sex drives (Baumeister,
Catanese, & Vohs, 2001) and a greater desire for casual
sex (Bailey, Gaulin, Agyei, & Gladue, 1994) compared
with Jerome’s desired partners (females). The hypothesis
of homosexual recruitment is not needed to explain the
fact that homosexual males tend to have earlier, and
more age-discrepant, sexual interactions compared with
heterosexual males.
What about females? The same argument predicts that
homosexual females should be much more likely than
heterosexual females to have early sexual experiences
with other females, even if the recruitment hypothesis is
false. Homosexual, but not heterosexual, females are
attracted to their own sex, and thus the former are more
likely to be open to the idea of same-sex sexual and
romantic interaction. The recruitment hypothesis is not
necessary to explain why homosexual women are more
likely than heterosexual women to have early same-sex
Some studies have found that nonheterosexual females
also have a higher rate of early age-discrepant sexual
interactions compared to heterosexual females (e.g.,
Tomeo, Templer, Anderson, & Kotler, 2001), although
some other studies have failed to find this (e.g., Bell
etal., 1981). Among the various factors that make homo-
sexual males more likely than heterosexual males to have
early age-discrepant sexual experiences, some but not all
also apply to homosexual females. For example, female
homosexuality is also stigmatized and relatively rare, and
thus young homosexual females do not generally have
sexual and romantic opportunities with similar-aged
peers that are equivalent to those of heterosexual females.
This would increase the chances that early experiences
of nonheterosexual women would be age-discrepant.
However, adult homosexual women are less likely than
adult homosexual men to be sexually interested in sub-
stantially younger partners (Silverthorne & Quinsey,
2000), making age-discrepant relationships less likely for
young homosexual females. Another possibly relevant
factor concerns personality. Homosexual women’s scores
on the personality trait openness to experience are about
half a standard deviation higher than heterosexual wom-
en’s (Lippa, 2005a). Openness to experience is related to
erotophilia (Wright & Reise, 1997), a more specific per-
sonality trait related to a person’s tendency to react to
sexual cues in a positive (erotophilic) versus negative
(erotophobic) way. Erotophilic individuals tend to have
sex earlier than other individuals, among other sex-
related differences (Fisher, White, Byrne, & Kelley, 1988),
and are likely to be more open to age-discrepant sexual
This is not to suggest that all early sexual experiences
of nonheterosexual people are positive and consensual.
In a recent, carefully sampled, and very large (N = 33,902)
study, T. Sweet and Welles (2012) found that nonhetero-
sexual adults of both sexes were much more likely than
heterosexual adults to have experienced childhood sex-
ual abuse, defined as “sexual experiences with an adult
or any other person younger than 18 years when the
individual did not want the sexual experience or was too
young to know what was happening” (p. 401). As is
common in research on childhood sexual abuse, this
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 83
definition combines a number of different experiences
likely to have different causes and effects—for example,
sexual experiences of children too young to have under-
stood what was happening and sexual experiences of
late adolescents who understood those experiences but
did not want them, as well as abusive experiences with
the same sex and with the other sex. Thus, it is difficult
to know what to make of the findings. Still, the risk was
much higher for nonheterosexual respondents: 38.1% for
lesbians, 43.5% for bisexual women, and 14.2% for het-
erosexual women; 18.6% for gay men, 19% for bisexual
men, and 4.6% for heterosexual men. The authors rejected
the interpretation of their data according to which abuse
caused nonheterosexual orientation. Instead, they specu-
lated that childhood gender nonconformity makes pre-
homosexual children especially vulnerable to predation.
For example, pre-homosexual children are to some
extent identifiable, and they may be especially suscepti-
ble to same-sex experiences even at young ages. (For
example, they may be recognized by older opportunistic
individuals with same-sex desires. Alternatively, they may
be victimized by others who dislike gender nonconfor-
mity.) Another possibility is systematic reporting bias,
whereby heterosexual respondents may underreport or
nonheterosexual respondents may overreport sexual
abuse; there are doubtless many other plausible hypoth-
eses. Although these results merit attention, concern, and
further study in order to understand them, they provide
no additional reason to consider the recruitment hypoth-
esis to be a promising one.
Pedophilia. An idea sometimes related to homosex-
ual recruitment is that homosexual men are especially
likely to molest children. The main data leading some
(e.g., Dailey, 2002) to this conclusion show that about
one-third of child molestation victims are male (Freund,
Watson, & Rienzo, 1989). Homosexual men comprise a
much smaller percentage of the population, perhaps 3%.
Assuming it is homosexual men who are molesting boys,
then they are much more likely to molest children than
heterosexual men, assuming it is heterosexual men who
are molesting girls.
The italicized assumptions are wrong, however. This is
a circumstance in which the word “homosexual” can be
misleading. We have been using the term “homosexual
men” to refer to men attracted to sexually mature males;
their sexual orientation is more precisely called “andro-
philic.” Androphilia contrasts with homosexual pedo-
philia, which is sexual attraction to sexually immature
males.27 Male androphiles and homosexual pedophiles
do not have the same sexual orientations; nor do male
gynephiles (i.e., heterosexual males oriented toward sex-
ually mature females) and heterosexual pedophiles. We
know this from basic research on sexual-arousal patterns.
Androphilic men have far less arousal to sexually imma-
ture than to sexually mature males; homosexually pedo-
philic men have the opposite pattern. The same is true of
gynephilic men and heterosexually pedophilic men
(Blanchard etal., 2012). In a recent survey of men sexu-
ally attracted to children, those men rated their attraction
to children as 9.5 on a 10-point scale. In contrast, they
rated their attraction to adults as 4.2 (Bailey, Hsu, & Ber-
nhard, 2013). Androphilic and gynephilic men have little
motivation to have sex with children. Thus, androphilic
men are not disproportionately responsible for the sexual
molestation of boys; homosexual pedophiles are.
Parent-child relationships. The idea that homosexu-
ality resulted from pathological parent-child relationships
became prominent because of psychoanalysis (Bayer,
1981). In general, psychoanalytic theorists came to blame
a dysfunctional relationship between children and their
parents for children’s homosexuality, which they saw as
a pathological outcome. Culprits included emotionally
distant fathers and overbearing mothers. These hypothe-
ses had the empirical limitations of most psychoanalytic
hypotheses (e.g., Grünbaum, 1986)—namely, that they
stemmed from therapists’ observations filtered through
the lens of highly speculative theory rather than from
systematic scientific studies. The landmark study by Bell
etal. (1981) seemingly disposed of the idea that homo-
sexuality resulted from the quality of parent-child rela-
tionships. That study yielded relatively small correlations
between retrospective ratings of parent-child relationship
characteristics and the child’s eventual sexual orientation
in adulthood. Furthermore, when other variables—espe-
cially childhood gender nonconformity—were covaried
in path analyses, the causal paths between parent-child
relationship characteristics and child’s sexual orientation
were either nonsignificant or quite weak.
The idea that male homosexuality is caused by bad
relationships with parents has been revived by the repar-
ative therapist Joseph Nicolosi (1997; Nicolosi & Nicolosi,
2012). In particular, he has focused on the father-son
relationship. According to Nicolosi, repairing that rela-
tionship and its alleged consequences can help change
sexual orientation. We defer discussing therapy to change
sexual orientation until later. For now, we note only that
Nicolosi’s hypothesis is based on data no better than
those of psychoanalysts.
What might account for the modest associations
between retrospectively assessed parent-child relation-
ship quality and adult sexual orientation? There are sev-
eral alternative hypotheses to the possibility that
parent-child relationship quality influences later sexual
orientation. First, pre-homosexual children tend to be
relatively gender nonconforming, and this may some-
timesstrain relationships with parents—especially fathers
by guest on July 5, 2016psi.sagepub.comDownloaded from
84 Bailey et al.
(Kane, 2006). Second, on average, homosexual men score
slightly higher than heterosexual men on trait neuroticism
(d = 0.20; Lippa, 2005a). Assuming the neuroticism differ-
ences are apparent during childhood, they could contrib-
ute to differences in negative interactions between fathers
and sons. Neuroticism is also related to biased recall of
negative events (Larsen, 1992); thus, the retrospective dif-
ferences in relationship quality could partly reflect mem-
ory biases. Third, on average, same-sex attraction in males
is associated with elevated traits of separation anxiety in
childhood (VanderLaan, Gothreau, Bartlett, & Vasey, 2011;
Vasey etal., 2011; Zucker, Bradley, & Sullivan, 1996), and
this could further strain father-and-son relationships. Of
course, there are many other possibilities that are not
causally related to sexual orientation, including the pos-
sibility that parents of pre-homosexual children are differ-
ent from those of pre-heterosexual children in ways that
affect the parent-child relationship.
The hypothesis that pathological parent-child relation-
ships cause homosexuality has generated little scientific
research, and almost no recent research. We believe that
this is primarily because the hypothesis has little scien-
tific promise.
Rearing by nonheterosexual parents. Two separa-
ble questions are of interest here regarding the outcome
of children reared by nonheterosexual parents: First, do
these children have an increased likelihood of becoming
homosexual, and second, do they differ in other impor-
tant ways from children reared by heterosexual families?
In evaluating research on these questions, it is important
to realize the very strong possibility of selection effects.
The ideal study to investigate the influence of parental
sexual orientation would involve randomly assigning
children (via adoption) to heterosexual versus homosex-
ual parents well matched on factors such as income.
Adoption would be necessary to avoid confounding out-
come due to rearing with that due to genetics. We would
expect, for example, that homosexual parents should be
more likely than heterosexual parents to have homosex-
ual children on the basis of genetics alone. Matching
parental characteristics is necessary if one wishes to
exclude variables correlated with parental sexual orienta-
tion that may influence child outcome, regardless of
whether parental sexual orientation does. Matching (or
statistical control) is important but still imperfect, because
there may be important but unrecognized differences
between heterosexual and homosexual parents that are
not caused by sexual orientation but contribute to chil-
dren’s outcomes.
For various reasons we will elucidate, studies of the
effects of rearing by homosexual parents do not yetallow
strong conclusions. These reasons include specific study
limitations but also a general concern about parental
effects. An important, wide-ranging observation is that
parents tend to have limited environmental effects on
their children’s behavioral traits. More specifically, resem-
blance between parents and offspring tends to be due to
genes rather than environment, and, surprisingly, envi-
ronment tends to make siblings reared together more dif-
ferent than similar (Harris, 1995; Plomin, 2011; Plomin &
Daniels, 1987; Turkheimer, 2000). We know this through
studies that use genetically informative designs, such as
twin and adoption studies. (For an especially compelling
adoption example, see Sacerdote, 2007.) To the extent
that parents systematically and environmentally affect
their children, the shared environment is important. The
more important shared environment is for a trait, the
more similar adoptive siblings (i.e., genetically unrelated
children reared together) should be. In general, however,
evidence indicates that adoptive siblings are quite dis-
similar. Although some shared environmental effects
exceed zero, these tend to be modest and to diminish
with age (e.g., Buchanan, McGue, Keyes, & Iacono, 2009;
Burt, 2009).
Researchers have increasingly recognized that accu-
rately measuring the effect of parents on children requires
genetically informative designs (Harris, 1995; Plomin,
2011; Plomin & Daniels, 1987). Other kinds of correla-
tional studies—whether of resemblance between parents
and children or of family characteristics predicting child
outcomes—will generally and greatly overestimate the
importance of the parental environment. One example of
a variable whose effects have been exaggerated in non-
genetically informative designs is parental divorce. Peo-
ple who divorce are different from those who do not.
Personality, among other factors, is important—and partly
heritable (Jockin, McGue, & Lykken, 1996). Indeed, the
tendency to divorce one’s spouse is strongly heritable
(McGue & Lykken, 1992). Children of divorce share the
experience of divorce as well as their parents’ genes. This
means that correlations between the experience of paren-
tal divorce and a child’s eventual outcome will be inflated
without controlling for genetics. Of course, in principle,
parental sexual orientation could be different from other
parental variables such as divorce and affect children’s
outcomes far more. But one cannot simply assume that it
does; one must prove it with data. Unfortunately, collect-
ing the requisite data is difficult, and no one has come
close to doing so.
Parental “effects” in convenience samples. Most stud-
ies of children raised by nonheterosexual parents have
depended on small convenience samples. Reviews of data
from convenience samples have suggested that children
reared by nonheterosexual parents have outcomes simi-
lar to those of children reared by heterosexual parents,
with respect to both sexual orientation and quality of life
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 85
(Patterson, 2006; Tasker, 2005). These types of studies
are not, however, without their limitations. Convenience
samples are recruited nonsystematically, through tech-
niques including advertisements and snowball sampling
(i.e., having participants refer other potential participants
they know). Participants recruited in such ways are likely
to be unusual in some respects—having high motivation
and curiosity, for example. Consequently, outcome dif-
ferences between children of heterosexual and homo-
sexual parents recruited nonsystematically may also be
biased. Because most studies using convenience samples
have been small, statistical power to detect small, or even
moderate, differences has typically been low. Finally, by
their very nature, convenience samples are not amenable
to the kind of sophisticated quantitative analyses neces-
sary to disentangle environmental from genetic parental
effects. (That goal would require data from informative
pairs of relatives, such as twins or adoptive siblings.)
Parental “effects” in a probability sample. In probabil-
ity sampling, the attempt is made to recruit respondents
in a representative way. To the extent that the attempt
succeeds, numeric estimates will more accurately reflect
population values compared with estimates obtained
using convenience sampling. A recent large study using
this approach (Regnerus, 2012a) has generated consider-
able controversy (Carey, 2012a; Cheng & Powell, 2015;
Davidson, 2012; Gates etal., 2012; Oppenheimer, 2012;
Regnerus, 2013), and we focus on it here.
The Regnerus (2012a) study used a sample of nearly
3,000 adults, including 175 who reported that their
mother had ever had a same-sex romantic relationship
and 73 who reported that their father had. Respondents
provided information about a number of demographic
characteristics, attitudes, behaviors, personality traits, and
psychological symptoms. The main analyses compared
variables between adult children raised in still-intact bio-
logical families (henceforth CIBFs) and other types,
including those with homosexually experienced parents.
Separate comparisons were made for those with homo-
sexually experienced mothers (henceforth, CHEMs) and
those with homosexually experienced fathers (CHEFs).
Both CHEMs and CHEFs showed small to moderate sta-
tistically significant differences with CIBFs. Moreover, to
the extent that the differences could be judged to be
desirable or undesirable, they were generally in favor of
the CIBFs. For CHEMs, these differences included, among
others, higher rates of welfare assistance, unemployment,
nonheterosexual identification, having experienced coer-
cive sex, depression, marijuana use, tobacco smoking,
and having been arrested and lower levels of education,
income, and physical health. The smaller sample of
CHEFs showed fewer statistically significant differences,
with examples including higher rates of welfare
assistance, depression, and arrests and lower educational
attainment. Most of these differences remained statisti-
cally significant after controlling for several demographic
As has been widely noted by critics of the study, the
families with homosexually experienced parents were
almost all disrupted (e.g., because of parental divorce).
Only 2 of the 248 children with homosexual mothers or
fathers had lived with homosexually partnered parents
for their entire childhoods (Regnerus, 2012b). Thus,
parental homosexual experience was almost perfectly
confounded with family disruption, and differences
attributed to the former might be more accurately attrib-
uted to the latter. It is an important and controversial
question why Regnerus’s study was able to recruit so few
intact families with two nonheterosexual parents. Assum-
ing this was not due to sampling bias, there are at least
two possibilities. First, stigmatization of homosexual rela-
tionships may make them more fragile; if so, their fragil-
ity will diminish with the ongoing decline of stigma, at
least in the West. Second, other factors unrelated to
stigma may cause the fragility. These could include, for
example, characteristics of nonheterosexual people or
characteristics of nonheterosexual relationships.
A recent reanalysis has raised serious questions about
the validity of Regnerus’s results, suggesting potential
misclassification of respondents’ family types (Cheng &
Powell, 2015). Clearly, in order for us to take the afore-
mentioned differences seriously, it is crucial that respon-
dents’ parents were accurately classified as heterosexual
or not. Results of the reanalysis suggested that a high
proportion of Regnerus’s respondents classified as having
nonheterosexual parents may have been misclassified.
The basis for these concerns included inconsistencies in
some respondents’ answers, some extremely unlikely
responses suggesting mischief, and evidence that some
respondents did not live long with their nonheterosexual
parents. When problematic cases were excluded, differ-
ences associated with family type largely vanished.
These recent concerns about Regnerus’s family classi-
fications appear to be sufficiently serious that no valid
conclusions can be drawn about any association between
having a homosexual parent and adult adjustment. But
even ignoring those problems, causal attribution on the
basis of the Regnerus study is inappropriate. Regnerus
has correctly noted this, although his reason for doing so
is not entirely correct. He has stated that the study “is not
a longitudinal study, and therefore cannot attempt to
broach questions of causation” (Regnerus, 2012a, p. 755).
However interesting a longitudinal component might be,
the lack of a genetically informative design, or of some
other causally informative design such as the quasi-
experimental instrumental-variables regression, remains
problematic.28 We simply cannot know from Regnerus’s
by guest on July 5, 2016psi.sagepub.comDownloaded from
86 Bailey et al.
study whether the differences he observed were caused
by children’s being reared by parents with or without
homosexual experience, whether the differences were
caused by the presence versus absence of family disrup-
tion, or whether rearing was mostly irrelevant to the
observed differences. Based on what we know from
studies of parental influence that have included the nec-
essary designs, the latter possibility is most likely.
Therapy to change sexual orientation. Conversion
therapy is the attempt to change homosexually oriented
people to heterosexually oriented people via psychother-
apy or behavioral therapy. Reparative therapy is a rela-
tively recent form of conversion therapy associated with
the psychologist Joseph Nicolosi (1997). Early forms of
conversion therapy were often offered by conventional
mental health professionals including psychiatrists and
psychologists, because homosexuality was generally
viewed as a form of psychopathology. Both that view and
the practice of conversion therapy have changed. For
example, both the American Psychiatric Association
(2000) and the American Psychological Association
(2009) oppose conversion/reparative therapy for both
ethical and scientific reasons. Present-day conversion
therapy, including reparative therapy, is most likely
offered by, and to, the religiously observant. We use the
term “conversion therapy” to refer to any therapy attempt-
ing to change sexual orientation.
Two important methodological concerns oppose all
existing claims that conversion therapy can change sexual
orientation: the likelihood of very strong selection bias
and reliance on self-report data. Because nonheterosex-
ual-oriented persons are not assigned randomly to either
receive or not receive therapy to become heterosexual,
those who receive therapy are likely to be unusual. Most
obviously, they are especially motivated to change. This
makes their data especially vulnerable to the second con-
cern regarding self-report. Specifically, individuals who
undergo conversion therapy may be especially suscepti-
ble to believing and reporting that therapy has succeeded
regardless of its true effectiveness. Therefore, reliance on
self-reported sexual feelings by such individuals renders
even systematically collected data highly questionable as
a gauge of sexual-orientation change.
The most notable study claiming that persistent, mean-
ingful changes in sexual orientation are possible was
conducted by Spitzer (2003), who later changed his mind
about the validity of his findings (Carey, 2012b). Partici-
pants were 143 men and 57 women who had undergone
reparative therapy and believed that they had shifted at
least 10 points in the heterosexual direction on a 100-
point scale of sexual orientation. On the basis of inter-
views, Spitzer initially concluded that the self-reported
changes were credible. Later, he came to believe that self-
report data are simply inadequate, because such reports
are often inaccurate, subject to both intentional decep-
tion and self-deception.
Designing a valid empirical study of the effectiveness
of reparative therapy is straightforward, at least for men.
Because an objective measure of male sexual orientation
exists that is not easily manipulated, a study of sexual
orientation in men need not rely on self-report: Measures
of men’s genital arousal patterns to a variety of male and
female erotic stimuli, both before and after conversion
therapy, would provide highly relevant evidence. Control
groups of heterosexual and homosexual men not receiv-
ing conversion therapy would clarify the interpretation of
results. For example, it appears easier to suppress genital
arousal to arousing stimuli than it is to enhance arousal
to non-arousing stimuli (H. E. Adams, Motsinger,
McAnulty, & Moore, 1992). Thus, it would be important
to look for increases in homosexual men’s genital arousal
to female stimuli and not just decreases in their arousal to
male stimuli.
Data that might speak to this issue are meager. Freund
(1960) found that clients’ claims of sexual reorientation
were not supported by phallometric assessments. Conrad
and Wincze (1976) found that physiological arousal mea-
surements did not support the positive reports of those
who had participated in sexual-reorientation therapy.
Non-self-report measures also exist to measure wom-
en’s sexual orientation—for example, relative viewing
time for pictures of attractive men versus attractive
women. But changes in these measures after conversion
therapy would be more difficult to interpret than changes
in male sexual arousal patterns. For example, conversion
therapy might encourage the individual to avoid looking
at same-sex targets and to look more at other-sex targets.
If effective, this would change viewing-time patterns but
not necessarily sexual orientation.
It is, of course, possible to change one’s public sex-
ual-orientation identity, and one can certainly make
choices about whether one will or will not engage in
same-sex or opposite-sex sexual behavior or become
celibate. These sorts of choices likely explain claims by
ex-gays and ex-lesbians that they are no longer leading
a “homosexual lifestyle” (see Beckstead, 2001, pp. 92–
109, for examples). There is no good evidence, however,
that sexual orientation can be changed with therapy, and
we strongly doubt it can be. Even many therapists sym-
pathetic to the desire of some homosexual people to live
heterosexual lives have shifted their efforts from chang-
ing sexual orientation to helping homosexual people
live as they prefer under the unchangeable constraint of
homosexual orientation (Schwartz, 2011; Throckmorton
& Yarhouse, 2006).
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 87
The causes of sexual orientation: An
interim summary
No specific theory of what causes people to be attracted
to men, to women, or to both has received enough sup-
port to win the backing of all reasonable scientists, most
of whom remain open-minded to a large extent. This
does not mean, however, that any hypothesis is just as
good as any other. In general, we know enough to con-
clude that most hypotheses are constrained, or at least
not equally likely.
Two contrasting views of the development and causes
of homosexuality tend to divide those with pro- and anti-
homosexual attitudes. The first, associated with positive
attitudes toward homosexuality, is that a small percent-
age of people are homosexual for so-far-unspecified rea-
sons of nature rather than social nurture (Knauer, 2000).
The second, associated with negative attitudes toward
homosexuality, is that homosexuality is to some extent
socially contagious and can spread either through sexual
recruitment or through the relaxation of moral and legal
prohibitions of homosexual behavior (Eskridge, 2005,
2008; Knauer, 2000). We believe scientific evidence sup-
ports the first view much more strongly than it does the
Several dependable findings of large effect are consis-
tent with the former but not with the latter. Children who
will become homosexual often differ in noticeable ways
from those who will become heterosexual. These differ-
ences often emerge long before the children clearly have
anything like a sexual orientation. Furthermore, these dif-
ferences in childhood gender nonconformity emerge
despite socialization, which works to enforce gender
norms, not because of it. Homosexual attractions emerge
prior to homosexual behavior in most people.29 This
should not be surprising because it follows the same pat-
tern by which most heterosexual people’s lives unfold.
Perhaps the most extreme and plausibly effective social
manipulation possible—changing boys into girls, socially
and physically—has been attempted, with no apparent
alteration of sexual orientation. These individuals—natal
males changed into females—typically grow up to be
attracted to women, based on limited available evidence.
There is good evidence for both genetic and nonsocial
environmental influences on sexual orientation. The lim-
ited evidence we have about the prevalence of nonhet-
erosexuality across cultures and time suggests that
homosexual orientation does not increase in frequency
with social tolerance, although its expression (in behav-
ior and in open identification) may do so.
In contrast, evidence that might be marshaled to sup-
port a socialization explanation is equally consistent with
the alternative (nonsocial) view. For example, the fact
that homosexual people are much more likely to have
early same-sex experiences, including experiences that
are age-discrepant, is consistent with a recruitment
hypothesis. But it is also consistent with the hypothesis
that young people who already have homosexual attrac-
tions are more likely to have these experiences. The fact
that the nonsocial hypothesis can explain several find-
ings much better than the social hypothesis should affect
our judgment of which is more likely.
The hypothesis that causal influences on sexual orien-
tation are nonsocial rather than social is better supported
for male than for female sexual orientation, for at least
two reasons. First, the nonsocial fraternal-birth-order
effect applies only to male sexual orientation. Second,
the near-perfect quasi-experiment, in which seven infants
were socially and physically reassigned to the other sex,
involved only natal males. Indeed, sexual orientation
may generally be better understood for males than for
females (Bailey, 2009). We would be surprised if differ-
ences in social environment contributed to differences in
male sexual orientation at all. We would be less surprised
if the social environment affected the expression of male
sexual orientation, including the likelihood that a homo-
sexually oriented male would choose to act on his feel-
ings. Although it would also be less surprising to us (and
to others; see Baumeister, 2000) to discover that social
environment affects female sexual orientation and related
behavior, that possibility must be scientifically supported
rather than assumed.
Policy Revisited
Because our article was initially motivated to help policy-
makers navigate the relevant science of sexual orienta-
tion, we concluded with an analysis of the likely effects
of proposed policy changes. Our projections would
apply, with appropriate adjustments, to any nation con-
sidering an increase in criminal or social sanctions against
homosexual behavior. At the time of writing, such discus-
sions were ongoing in various African, Asian, and Eastern
European nations (Carroll & Itaborahy, 2015). For the
purposes of this discussion, we will focus on recent,
highly publicized policy proposals in Uganda.
Recent attempts to increase criminal penalties for
homosexual behavior in Uganda, as well as in other
countries, have been partly justified by the belief that the
social environment is a powerful influence on sexual ori-
entation. Specifically, President Museveni has said he
believes that homosexual people recruit heterosexual
people and that homosexuality can be successfully dis-
couraged via social means, such as long prison sentences.
In order to check his assumptions, he sought scientific
support for—and subsequently rejected—the idea that
homosexual orientation is inborn. Museveni believed that
the inconclusive evidence he was presented showed that
by guest on July 5, 2016psi.sagepub.comDownloaded from
88 Bailey et al.
sexual orientation is socially malleable (“President Musev-
eni’s Full Speech,” 2014). We have taken considerable
effort to show why this reasoning is erroneous.
Of course, justifying prison sentences based on the
idea that they deter homosexuality also requires the
assumption that homosexuality should be deterred, an
assumption with which we disagree. But for now, let us
set aside that objection and consider the likely implica-
tions of increasing criminal penalties for homosexuality.
Assume that punitive legislation is passed and stays in
place for a generation. What would change? No one
knows for certain, of course. Certainty would require an
experiment in which randomly assigned countries
enacted laws varying in the severity of their penalties for
homosexual behavior. Nothing like this randomized
experiment will happen. But informed speculation is
possible, based on the evidence we have discussed as
well as the Western experience, with dramatic changes in
attitudes toward homosexuality having occurred during
less than a lifetime.
The following predictions follow from our review:
There would be no change in the proportion of males
strongly sexually attracted to other males, which would
remain about 1 in 25 to 1 in 50. Although we are less
certain that the same would be true for same-sex-attracted
females, there is no scientific evidence that increasing
criminal penalties would reduce the female prevalence of
strong same-sex attraction. The average Ugandan house-
hold contains five people (Uganda National Household
Survey Report, 2010). Approximately 1 in 5 to 1 in 10 of
these households likely contains a nonheterosexual per-
son. The typical Ugandan woman currently expects to
give birth to approximately six children during her life-
time (Uganda Bureau of Statistics, 2011). Thus, approxi-
mately 1 in 4 to 1 in 8 women there gives birth to a
nonheterosexual child. (This calculation assumes that
homosexual children do not aggregate in certain families,
which may not be strictly true, if genes or shared envi-
ronment matter. These factors do not induce much aggre-
gation, however.) Given their large family sizes, compared
with those of their Western counterparts, it is likely that
most Ugandan people will have a sibling, aunt or uncle,
niece or nephew, or cousin who is nonheterosexually
oriented. Our general point here is that the criminaliza-
tion of homosexuality touches many—indeed, probably
most—Ugandan families.
The requirement that no one may engage in homo-
sexual activity has especially harsh implications for Ugan-
dans with exclusively homosexual feelings. (It is useful
for heterosexual people to imagine living in an analo-
gous world, one that outlawed heterosexual relationships
and sexual behavior.) Some of these individuals may live
celibate lives, out of fear. Others—probably far more—
will risk legal penalties in order to live in a way they find
more fulfilling. Some may marry heterosexually, either
because they are genuinely trying to suppress their
homosexuality or because they want to hide their true
sexual orientation. Although some of these marriages will
succeed, others will be unhappy. Regardless, the homo-
sexual spouses’ sexual orientations are unlikely to
Beyond these basic predictions, things get more spec-
ulative. If the Western experience is a guide, enforcement
of strictures against homosexuality is likely to be haphaz-
ard and dependent on local political situations (Eskridge,
2008). This has been especially true of consensual homo-
sexual relationships conducted by adults in private,
which were often tolerated if not embraced in the West
even when they were illegal. Recent news from Uganda
presents a harsher picture, however. According to
Amnesty International, homosexual people in Uganda
have experienced a marked increase in “arbitrary arrests,
police abuse and extortion, loss of employment, evic-
tions and homelessness” related to the punitive legisla-
tion (Amnesty International, 2014, para. 2). Physical
attacks on homosexual people have also drastically
increased (Bowcott, 2014). Aside from threats to their
physical safety, nonheterosexual people in Uganda are
subject to constant disapproval. This is likely to diminish
their experience of self-worth, and increase feelings of
shame, guilt, depression, and anxiety.
Researchers have begun to quantify the economic
costs of anti-homosexual prejudice as a strategy to con-
vince people that there are economic benefits associated
with tolerance toward sexual-orientation and gender
diversity. For example, a World Bank study suggested
that discrimination on the basis of sexual orientation and
gender identity costs India up to $30 billion a year
through lost wages, lower productivity associated with
poor health caused by stress or curtailed education, and
reduced tourism revenues, among other economic
impacts (Patel, 2014). Another study of 39 countries con-
ducted for the United States Agency for International
Development (USAID) estimated the cost of anti-homo-
sexual prejudice—as experienced through educational
and workforce discrimination, police corruption, vio-
lence, and poor health care access—to be as much as 3%
of a country’s gross domestic product (Badgett, Nezhad,
Waaldijk, & van der Meulen Rodgers, 2014). Furthermore,
reducing the scope of anti-homosexual policies and prac-
tices would have the benefit of saving many lives. For
example, a study of the relative cost-effectiveness of dif-
ferent interventions against HIV transmission estimated
that $1 million devoted to HIV education in Kenya could
avert nearly 3,000 deaths per year (i.e., $272 per life
saved; Beyrer etal., 2011).
For these reasons, we urge governments to reconsider
the wisdom of legislation that criminalizes homosexual
behavior. Furthermore, we invite members of govern-
ments considering such legislation to confer with us
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 89
about issues we have raised. Our expertise is theirs for
the asking.
The authors thank Peter Godfrey-Faussett from the Joint United
Nations Programme on HIV and AIDS (UNAIDS) for encourag-
ing this project. We are also grateful to Ray Blanchard, Alan
Sanders, Scott Semenyna, and Brendan Zietsch for their
thoughtful comments on portions of our article.
Declaration of Conflicting Interests
The authors declared that they had no conflicts of interest with
respect to their authorship or the publication of this article.
J. Michael Bailey has been funded by the National Institutes of
Health (NIH; Grants R03 MH62484-01 and R01 HD41563) and
the American Institute of Bisexuality for research reported in
this article and has received general research funding from the
Northwestern University Provost Fund for Sexual Orientation
Research. Paul L. Vasey received an Insight Grant from the
Social Sciences and Humanities Research Council of Canada
and a Discovery Grant from the Natural Sciences and Engineer-
ing Research Council of Canada for research reported in this
article. Lisa M. Diamond has been funded by the American
Institute of Bisexuality for work on this article and has received
general research funding from NIH, the Gay and Lesbian Medi-
cal Association, and the American Psychological Foundation
(Wayne F. Placek Award). S. Marc Breedlove has received gen-
eral research funding from NIH (Grants R21-MH104780 and
T32MH70343). Eric Vilain has received general research fund-
ing from the Disorders of Sex Development Translational
Research Network and the Eunice Kennedy Shriver National
Institute of Child Health and Human Development (Grant R01
1. Brunei is set to adopt anti-gay laws in 2016 that will allow
stoning by death (Garcia, 2014; Mosbergen, 2015).
2. The terms “West” or “Western” are used here to refer to
Euro-American cultures.
3. The Academy of Science of South Africa recently published
a position statement on the science of sexual orientation and
the implications of this science for legal policy (see Academy of
Science of South Africa, 2015).
4. An important current controversy is whether to extend the
term “sexual orientation” to include other stable patterns of
sexual attraction, such as attractions to children (Seto, 2012) or
even to nonhuman animals (Miletski, 2005). We use the term in
the more restrictive sense—that is, to refer only to attraction to
sexually mature human males or females. Both the causes and
consequences of other stable patterns of sexual attraction likely
differ considerably from those of sexual orientation in its usual
sense (e.g., Cantor, 2012).
5. There has been movement to discontinue use of the word
“homosexual,” given that this word was commonly used during
historical periods when nonheterosexual people were generally
seen as deviant (J. W. Peters, 2014). We certainly do not believe
that homosexual people are deviant (except in the statistical
sense of rarity), but we have chosen to use the term “homo-
sexual” (along with “heterosexual” and “bisexual”) for simplic-
ity of communication. Yet we are sensitive to the importance
of word choice when describing phenomena related to sexual
orientation, and we recognize that all descriptors come with
6. Kinsey’s scale measured sexual feelings and behavior in
combination. Nowadays, most researchers who employ Kinsey
scales assess them separately.
7. Although several recent studies have found evidence for
category-specific genital arousal in women using a variety
of measures such as thermal imaging, laser Doppler imag-
ing, and vaginal lubrication (Dawson, Sawatsky, & Lalumière,
2015; Kukkonen, Binik, Amsel, & Carrier, 2010; Waxman &
Pukall, 2009), the studies were small, and none has yet been
8. Here, “homosexual” and “bisexual” mean with respect to
natal sex, even among those males who have chosen to become
transgender women.
9. Personality psychologists define neuroticism (emotional
instability) as the predisposition to experience negative emo-
tions such as anxiety. Individuals who are highly neurotic are
emotionally reactive and vulnerable to stress (McCrae & John,
10. Sexual identity is more complicated, because it has both
a public and a private component. Public sexual identity is
how one presents oneself to others, and this is clearly cho-
sen. Private sexual identity comprises beliefs about oneself,
and these beliefs are sometimes chosen and sometimes not.
For example, if a woman is consciously aware of her same-sex
attractions, then she may have no choice but to believe that she
is lesbian or bisexual. However, some individuals may inten-
tionally adopt alternative beliefs about their same-sex attrac-
tions—for example, that they are signs of confusion or mental
distress—and may therefore choose to believe that they are het-
erosexual despite their same-sex sexuality. It is because of such
complications that questions of choice have no clear interpreta-
tion with respect to sexual identity.
11. We are aware that there are religious objections to tolerance
of homosexuality and that these will need to be factored into
political responses to the issues. For the purposes of this report,
however, we ignore such religious objections. To the extent that
someone believes religious doctrine trumps rationality and sci-
ence, we simply disagree.
12. A common additional understanding of “innate” is “deter-
mined prior to birth.” Although it is simpler here to use the
understanding we have provided, in fact all influential scien-
tific hypotheses that sexual orientation is innate emphasize the
importance of the prenatal element.
13. “Cisgender” refers to individuals whose gender identity
matches the one they were assigned at birth.
14. “Transgender” refers to individuals whose gender identity is
different from the one they were assigned at birth.
15. In egalitarian relationships, both partners are postpubertal,
and if age differences exist, they generally do not exceed one
16. This effect size is highly variable across different samples in
relation to ethnicity, among other characteristics.
by guest on July 5, 2016psi.sagepub.comDownloaded from
90 Bailey et al.
17. The presumed heterosexual subjects were likely heterosex-
ual given what was known about them. In any case, classifica-
tion errors would not create spurious findings.
18. Although we agree with Byne’s argument and would sup-
port renaming the INAH-3 as the human SDN-POA, we con-
tinue to refer to INAH-3 here.
19. How unexpected is this result under a hypothesis of “nur-
ture?” Our review of sexual-orientation prevalence suggests that
1.1% of women are predominately attracted to other women
(Fig. 1). That is the rate we should expect in our near-perfect
quasi-experiment, under the hypothesis that only nurture (i.e.,
sex of rearing) matters. If the fact that these individuals had pre-
natal male biology were irrelevant, then the number expected
to be strongly attracted to women would equal 0.011 × 7 =
0.077; this is a much smaller number than the seven who
were actually strongly attracted to women. The chances that
all seven of these individuals would be strongly attracted to
women, again under the nurture-only hypothesis, would be
.0117, or .0000000000000195. The nurture-only hypothesis is
clearly untenable. Results of the near-perfect quasi-experiment
are exactly as expected via the opposite, alternative, nature-
only hypothesis: seven of seven strongly attracted to women.
Of course, the non-opposite alternative—that both nature and
nurture matter—must also be considered. Because of the small
number of individuals studied (N = 7), fairly sizable sex-of-
rearing influences cannot be statistically excluded. Although
none of the seven individuals had a sexual orientation con-
sistent with a sex-of-rearing hypothesis, the 95% confidence
interval of the proportion 0/7 has an upper bound of .44. This
is nearly 15 times the expected rate assuming that sex of rearing
has no effect (.0295, based on the likelihood that someone born
male will be attracted to males; see Fig. 1). That is, using these
data alone, we cannot statistically exclude the possibility that
rearing someone as female increases the likelihood that she will
be attracted to males by almost fifteen-fold.
20. More accurately, half of their DNA regions are identical by
descent from the same parent, on average. Humans have more
than 99% of their DNA in common on a base-pair level, and so
the half of these variant base pairs that are identical by descent
among DZ twins is an index of their similarity with respect to
DNA (some of which is in or affects genes) that varies among
21. For a trait measured categorically, as sexual orientation usu-
ally is, the correlation is best estimated using the MZ and DZ
twin concordances (the proportion of MZ and DZ homosex-
ual index subjects whose twins are also homosexual), using
the multifactorial threshold model, under the assumption that
a normally distributed trait propensity underlies the observed
trait (Falconer & Mackay, 1996). This is a tetrachoric, rather than
a Pearson, correlation. To ensure the accuracy of heritability
and other estimates, it is crucial that twins are randomly—or
at least representatively—sampled with respect to concordance
status (i.e., whether both twins are homosexual or only one
is). Finally, the logic of the classical twin study requires the
assumption that the trait-relevant environment be equally simi-
lar among members of MZ and DZ pairs. This “equal envi-
ronments” assumption has generally been supported for other
behavioral traits (Flint, Greenspan, & Kendler, 2010).
22. Several published studies were omitted for the following
23. In general, estimates of heritability and shared environ-
mentality are much less stable than those of nonshared envi-
ronmentality, because the former depend on the difference
between two correlations and the latter depend on the differ-
ence between one correlation and unity.
24. The indirect molecular evidence includes findings that
mothers of homosexual men had a pattern of skewed inacti-
vation of the X chromosome (Bocklandt, Horvath, Vilain, &
Hamer, 2006) and that a region of chromosome 10 tended to be
linked with male sexual orientation if transmitted through the
mother (Mustanski etal., 2005).
25. Although MZ twins can differ genetically, such differences
are too rare to account for the high rate of sexual-orientation
26. As usual, our use of medical examples has no implications
for how we view any sexual orientation; it simply reflects the
availability of such examples.
27. Technically, homosexual pedophilia is sexual attraction
to prepubescent males (i.e., pubertal Tanner stage 1; typi-
cal ages= 5–10), and homosexual hebephilia is attraction to
pubescent males (i.e., Tanner stages 2 or 3; typical ages = 11–
14). Androphilic men, in contrast, are most sexually attracted to
males at Tanner stages 5 and, to a lesser degree, 4 (typical ages
= 15 and older). For the sake of simplicity and convention, we
refer to both hebephilic and pedophilic men as “pedophilic.
28. In general, quasi-experimental designs aim to provide more
rigorous tests of causal hypothesis compared with mere tests of
association (Shadish, Cook, & Campbell, 2001).
29. Moreover, because this evidence is retrospective and based
on self-report, there is reason to doubt the accuracy of the
exceptions to this pattern, because such individuals may mis-
remember the sequence of events or be prone to blaming their
stigmatized feelings on experiences initiated by others.
Study Reason
Bailey, Dunne, and Martin
L ikely considerable sample
overlap with Zietsch etal.
Bearman and Brückner (2002) T wins not yet adults with
stable sexual orientations
Burich, Bailey, and Martin
L ikely considerable sample
overlap with Zietsch etal.
Burri, Cherkas, Spector, and
Rahman (2011)
U nable to obtain
concordance data
Eckert, Bouchard, Bohlen,
and Heston (1986)
Too few twin pairs
Heston and Shields (1968) Too few twin pairs
Kirk, Bailey, Dunne, and
Martin (2000)
L ikely considerable sample
overlap with Zietsch etal.
Whitam, Diamond, and Martin
(1993), female twin pairs
Too few female pairs
by guest on July 5, 2016psi.sagepub.comDownloaded from
Sexual Orientation, Controversy, and Science 91
Abild, M., VanderLaan, D. P., & Vasey, P. L. (2014). Does prox-
imity influence the expression of avuncular tendencies
in Canadian androphilic males? Journal of Cognition and
Culture, 14, 40–62.
Academy of Science of South Africa. (2015). Diversity in
human sexuality: Implications for policy in Africa. Pretoria.
Academy of South Africa.
Aceng, R. J. (2014, February 28). Uganda: Scientific statement
from the Ministry of Health on homosexuality. Pambazuka
News. Retrieved from
Adams, B. D. (1986). Age, structure and sexuality. Journal of
Homosexuality, 11, 19–33.
Adams, H. E., Motsinger, P., McAnulty, R. D., & Moore, A. L.
(1992). Voluntary control of penile tumescence among
homosexual and heterosexual subjects. Archives of Sexual
Behavior, 21, 17–31.
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.
Allen, L. S., Hines, M., Shryne, J. E., & Gorski, R. A. (1989). Two
sexually dimorphic cell groups in the human brain. The
Journal of Neuroscience, 9, 497–506.
American Psychiatric Association. (2000). Position statement on
therapies focused on attempts to change sexual orientation
(reparative or conversion therapies). The American Journal
of Psychiatry, 157, 1719–1721.
American Psychological Association. (2009). Report of the
task force on appropriate therapeutic responses to sexual
orientation. Washington, DC: Author. Retrieved from
Amnesty International. (2014, May 15). Uganda: Anti-
homosexuality act prompts arrests, attacks, evictions,
flight. Retrieved from
Arnold, A. P., & Breedlove, S. M. (1985). Organizational and
activational effects of sex steroids on brain and behavior: A
reanalysis. Hormones and Behavior, 19, 469–498.
Badgett, M. V. L., Nezhad, S., Waaldijk, K., & van der Meulen
Rodgers, Y. (2014). The relationship between LGBT inclu-
sion and economic development: An analysis of emerging
economies. Los Angeles, CA: The Williams Institute-
USAID. Retrieve from
Bagemihl, B. (1999). Biological exuberance: Animal homo-
sexual and natural diversity. New York, NY: St. Martin’s
Bailey, J. M. (2003). The man who would be queen: The science
of gender-bending and transsexualism. Washington, DC:
Joseph Henry Press.
Bailey, J. M. (2009). What is sexual orientation and do women
have one? In D. A. Hope (Ed.), Contemporary perspectives
on lesbian, gay, and bisexual identities (pp. 43–64). New
York, NY: Springer.
Bailey, J. M., Dunne, M. P., & Martin, N. G. (2000). Genetic and
environmental influences on sexual orientation and its cor-
relates in an Australian twin sample. Journal of Personality
and Social Psychology, 78, 524–536.
Bailey, J. M., Gaulin, S., Agyei, Y., & Gladue, B. A. (1994).
Effects of gender and sexual orientation on evolutionarily
relevant aspects of human mating psychology. Journal of
Personality and Social Psychology, 66, 1081.
Bailey, J. M., Hsu, K. J., & Bernhard, P. (2013). [Survey of
pedophebephilic men]. Unpublished raw data.
Bailey, J. M., & Pillard, R. C. (1991). A genetic study of male
sexual orientation. Archives of General Psychiatry, 48,
Bailey, J. M., Pillard, R. C., Neale, M. C., & Agyei, Y. (1993).
Heritable factors influence sexual orientation in women.
Archives of General Psychiatry, 50, 217–223.
Bailey, J. M., & Zucker, K. J. (1995). Childhood sex-typed
behavior and sexual orientation: A conceptual analysis and
quantitative review. Developmental Psychology, 31, 43–55.
Balthazart, J., & Ball, G. F. (2007). Topography in the preoptic
region: Differential regulation of appetitive and consumma-
tory male sexual behaviors. Frontiers in Neuroendocrinol-
ogy, 28, 161–178.
Bartlett, N. H., & Vasey, P. L. (2006). A retrospective study of
childhood gender-atypical behavior in Samoan fa’afafine.
Archives of Sexual Behavior, 35, 559–566.
Baumeister, R. F. (2000). Gender differences in erotic plasticity:
The female sex drive as socially flexible and responsive.
Psychological Bulletin, 126, 347–374.
Baumeister, R. F., Catanese, K. R., & Vohs, K. D. (2001). Is
there a gender difference in strength of sex drive? Theo-
retical views, conceptual distinctions, and a review of rel-
evant evidence. Personality and Social Psychology Review,
5, 242–273.
Bayer, R. (1981). Homosexuality and American psychiatry: The
politics of diagnosis. Princeton, NJ: Princeton University
Bearman, P. S., & Brückner, H. (2002). Opposite-sex twins
and adolescent same-sex attraction. American Journal of
Sociology, 107, 1179–1205.
Beckstead, L. (2001). Cures versus choices: Agendas in sexual
reorientation therapy. Journal of Gay & Lesbian Psychother-
apy, 5, 87–115.
Bell, A. P., & Weinberg, M. S. (1978). Homosexualities: A study
of diversity among men and women. New York, NY: Simon
& Schuster.
Bell, A. P., Weinberg, M. S., & Hammersmith, S. K. (1981).
Sexual preference: Its development in men and women
(Vol. 2). Bloomington: Indiana University Press.
Berenbaum, S. A., & Beltz, A. M. (2011). Sexual differentiation
of human behavior: Effects of prenatal and pubertal orga-
nizational hormones. Frontiers in Neuroendocrinology, 32,
Berenbaum, S. A., Bryk, K. K., Nowak, N., Quigley, C. A., &
Moffat, S. (2009). Fingers as a marker of prenatal androgen
exposure. Endocrinology, 150, 5119–5124.
by guest on July 5, 2016psi.sagepub.comDownloaded from
92 Bailey et al.
Beyrer, C., Wirtz, A., Walker, D., Johns, B., Sifakis, F., & Baral,
S.D. (2011). The global HIV epidemics among men who
have sex with men. Washington, DC: World Bank. Retrieved
Blackwood, E. (2010). Falling into the lesbi world: Desire and
difference in Indonesia. Honolulu: University of Hawai’i
Blackwood, E., & Wieringa, S. E. (1999). Female desires: Same-
sex relations and transgender practices across cultures.
New York, NY: Columbia University Press.
Blanchard, R. (2001). Fraternal birth order and the maternal
immune hypothesis of male homosexuality. Hormones and
Behavior, 40, 105–114.
Blanchard, R. (2004). Quantitative and theoretical analyses of
the relation between older brothers and homosexuality in
men. Journal of Theoretical Biology, 230, 173–187.
Blanchard, R., & Bogaert, A. F. (2004). Proportion of homosex-
ual men who owe their sexual orientation to fraternal birth
order: An estimate based on two national probability sam-
ples. American Journal of Human Biology, 16, 151–157.
Blanchard, R., Klassen, P., Dickey, R., Kuban, M. E., & Blak, T.
(2001). Sensitivity and specificity of the phallometric test
for pedophilia in nonadmitting sex offenders. Psychological
Assessment, 13, 118–126.
Blanchard, R., Kuban, M. E., Blak, T., Klassen, P. E., Dickey,
R., & Cantor, J. M. (2012). Sexual attraction to others: A
comparison of two models of alloerotic responding in men.
Archives of Sexual Behavior, 41, 13–29.
Blanchard, R., & Sheridan, P. M. (1992). Sibship size, sibling
sex ratio, birth order, and parental age in homosexual and
nonhomosexual gender dysphorics. Journal of Nervous and
Mental Disease, 180, 40–47.
Blumenthal, M. (2004, December 14). Her Kinsey obses-
sion. AlterNet. Retrieved from
Bobrow, D., & Bailey, J. M. (2001). Is male homosexuality
maintained via kin selection? Evolution & Human Behavior,
22, 361–368.
Bocklandt, S., Horvath, S., Vilain, E., & Hamer, D. H. (2006).
Extreme skewing of X chromosome inactivation in mothers
of homosexual men. Human Genetics, 118, 691–694.
Bocklandt, S., Lin, W., Sehl, M. E., Sánchez, F. J., Sinsheimer, J. S.,
Horvath, S., & Vilain, E. (2011). Epigenetic predictor of age.
PLoS ONE, 6(6), e14821. doi:10.1371/journal.pone.0014821
Bogaert, A. F. (1998). Physical development and sexual orienta-
tion in women: Height, weight, and age of puberty compar-
isons. Personality and Individual Differences, 24, 115–121.
Bogaert, A. F. (2006). Biological versus nonbiological older
brothers and men’s sexual orientation. Proceedings of the
National Academy of Sciences, USA, 103, 10771–10774.
Bogaert, A. F., & Skorska, M. (2011). Sexual orientation, fra-
ternal birth order, and the maternal immune hypothesis: A
review. Frontiers in Neuroendocrinology, 32, 247–254.
Boswell, J. (1982/1983). Revolutions, universals and sexual cat-
egories. Salmagundi, 58-59, 89–113.
Bowcott, O. (2014, May 12). Uganda anti-gay law led to tenfold
rise in attacks on LGBTI people, report says. The Guardian.
Retrieved from
Bradley, S. J., Oliver, G. D., Chernick, A. B., & Zucker, K. J.
(1998). Experiment of nurture: Ablatio penis at 2 months,
sex reassignment at 7 months, and a psychosexual follow-up
in young adulthood. Pediatrics, 102(1), Article e9. Retrieved
Broderick, G. A. (1998). Evidence based assessment of erectile
dysfunction. International Journal of Impotence Research,
10, S64–S73.
Brodie, K. (2014, October 16). Island of Palau decriminalizes
love [Blog post]. Retrieved from
Buchanan, J. P., McGue, M., Keyes, M., & Iacono, W. G. (2009).
Are there shared environmental influences on adolescent
behavior? Evidence from a study of adoptive siblings.
Behavior Genetics, 39, 532–540.
Burri, A., Cherkas, L., Spector, T., & Rahman, Q. (2011). Genetic
and environmental influences on female sexual orientation,
childhood gender typicality and adult gender identity. PLoS
ONE, 6(7), e21982. doi:10.1371/journal.pone.0021982
Burt, S. A. (2009). Rethinking environmental contributions to
child and adolescent psychopathology: A meta-analysis of
shared environmental influences. Psychological Bulletin,
135, 608–637.
Buss, D. M. (1989). Sex differences in human mate preferences:
Evolutionary hypotheses tested in 37 cultures. Behavioral &
Brain Sciences, 12, 1–49.
Byne, W. (1998). The medial preoptic and anterior hypotha-
lamic areas of the rhesus monkey: A comparison with
the human and evidence for sexual dimorphism. Brain
Research, 793, 346–350.
Byne, W., Lasco, M. S., Kemether, E., Shinwari, A., Edgar, M.A.,
Morgello, S., . . . Tobet, S. (2000). The interstitial nuclei of
the human anterior hypothalamus: An investigation of sex-
ual variation in volume and cell size, number and density.
Brain Research, 856, 254–258.
Byne, W., Tobet, S., Mattiace, L. A., Lasco, M. S., Kemether,
E., Edgar, M. A., . . . Jones, L. B. (2001). The interstitial
nuclei of the human anterior hypothalamus: An investiga-
tion of variation with sex, sexual orientation, and HIV sta-
tus. Hormones and Behavior, 40, 86–92.
Campbell, D. A. (1982). Greek Lyric 1: Sappho and Alcaeus
(Loeb Classical Library No. 142). Cambridge, MA: Harvard
University Press.
Camperio-Ciani, A., Corna, F., & Capiluppi, C. (2004). Evidence
for maternally inherited factors favoring male homosexu-
ality and promoting female fecundity. Proceedings of the
Royal Society B: Biological Sciences, 271, 2217–2221.
Canning, P. (2013, September 3). With all eyes on anti-gay
Russian there are three countries with a shocking need for
coverage. Pink News. Retrieved from http://www.pinknews
Cantor, J. M. (2012). Is homosexuality a paraphilia? The evi-
dence for and against. Archives of Sexual Behavior, 41,
237–247. doi:10.1007/s10508-012-9900-3. Retrieved from
by guest on July 5, 2016psi.sagepub.comDownloaded from