Similarity in transgender and cisgender children’s

gender development
Selin Gülgöza,1, Jessica J. Glaziera, Elizabeth A. Enrighta, Daniel J. Alonsoa, Lily J. Durwooda, Anne A. Fasta,b,
Riley Lowea,c, Chonghui Jia, Jeffrey Heerd, Carol Lynn Martine, and Kristina R. Olsona
a
Department of Psychology, University of Washington, Seattle, WA 98195; bDepartment of Psychology, Western Washington University, Bellingham, WA
98225; cDepartment of Psychology, University of Michigan, Ann Arbor, MI 48109; dDepartment of Computer Science and Engineering, University of
Washington, Seattle, WA 98195; and eT. Denny Sanford School of Social and Family Dynamics, Arizona State University, Tempe, AZ 85287

Edited by Susan T. Fiske, Princeton University, Princeton, NJ, and approved October 22, 2019 (received for review May 30, 2019)

Gender is one of the central categories organizing children’s social sense of their gender identities and their expressions of that identity
world. Clear patterns of gender development have been well- may differ (or not) from gender identities and expressions of their
documented among cisgender children (i.e., children who identify cisgender peers, and whether the time that a transgender child has
as a gender that is typically associated with their sex assigned at spent living and being treated by others as their current gender
birth). We present a comprehensive study of gender development predicts their gender development.
(e.g., gender identity and gender expression) in a cohort of 3- to
12-y-old transgender children (n = 317) who, in early childhood, are Early Gender Development
identifying and living as a gender different from their assigned sex. Previous research on gender has primarily focused on cisgender
Four primary findings emerged. First, transgender children strongly children—children whose gender identities align with their sex
identify as members of their current gender group and show assignment at birth. By their third birthdays, nearly all cisgender
gender-typed preferences and behaviors that are strongly associ- children label their gender according to their assigned sex (10–
ated with their current gender, not the gender typically associated 13), and by ages 3–5 y, most children believe that their gender
with their sex assigned at birth. Second, transgender children’s gen- will remain the same in adulthood (14–16). Throughout the
der identity (i.e., the gender they feel they are) and gender-typed preschool and elementary school years, children typically view
preferences generally did not differ from 2 comparison groups: cis- themselves as highly similar to others of the same gender, and as
gender siblings (n = 189) and cisgender controls (n = 316). Third, different from members of another gender (17–19). Cisgender
transgender and cisgender children’s patterns of gender develop- children often show preferences and behaviors that are highly
ment showed coherence across measures. Finally, we observed min- stereotypical of their gender. For example, they show strong
imal or no differences in gender identity or preferences as a function preferences for same-gender playmates by age 3 (20–27), pref-
of how long transgender children had lived as their current gender. erences for gender-typed toys throughout early and middle
Our findings suggest that early sex assignment and parental rearing
Downloaded from https://www.pnas.org by 47.61.168.101 on April 13, 2023 from IP address 47.61.168.101.

childhood, and gender-typed clothing preferences (28).

based on that sex assignment do not always define how a child Most theories of gender development explain these patterns
identifies or expresses gender later. through a combination of influences of biological factors (e.g.,
assigned sex, prenatal hormone exposure; ref. 29), children’s
gender development | gender identity | transgender children gender cognitions (e.g., self-socialization; refs. 30 and 31), and
the cultural and interpersonal socialization they experience (e.g.,

A lthough identities similar to the Western concept of

“transgender” have likely existed for centuries in locations
across the globe (1–5), for what is likely the first time in Western
Significance

cultures, thousands of young transgender children are living as a Questions of nature and nurture have dominated efforts to
gender that differs from their sex assignment at birth. Unlike understand human gender development. Today’s transgender
past generations of Western transgender individuals, who pri- children provide a unique window into gender development:
marily socially transitioned to live in line with their asserted They have been treated as 2 different genders (1 gender before
gender identity in late adolescence or adulthood (6), these transition and 1 gender after their social transition) and are the
children are socially transitioning—changing pronouns from first sizable group of children living as a gender that differs from
those associated with their sex assigned at birth to binary pro- their assigned sex. As such, their experiences enable insight into
nouns associated with a different gender (often accompanied by gender development that is otherwise not possible. The current
changes in first names, clothing, and hairstyles)—in the pre- study provides the largest report to date of the experiences of
school and primary school years (7–9). Unlike future generations these early-transitioning children’s gender development.
of transgender children, these children are often the first in their
neighborhoods, schools, and religious communities to socially Author contributions: S.G. and K.R.O. conceptualized paper, acquired funding, and su-
transition, thereby facing challenges and breaking new ground as pervised project; K.R.O. designed research; S.G., J.J.G., E.A.E., D.J.A., L.J.D., A.A.F., R.L.,
and K.R.O. performed research; D.J.A., R.L., and C.J. curated data; S.G. and J.J.G. analyzed
they assert their identities on sports teams, at sleep-away camps, data; S.G., E.A.E., A.A.F., and J.H. visualized data; S.G., J.J.G., E.A.E., and L.J.D. prepared
and in legal battles for bathroom access, passports, and birth original drafts; C.L.M. provided background on theories; and S.G., J.J.G., E.A.E., D.J.A., L.J.D.,
certificates. These young transgender children are different from A.A.F., R.L., J.H., C.L.M., and K.R.O. wrote and reviewed the paper.
their cisgender peers in their unique gender socialization expe- The authors declare no competing interest.
rience, having lived part of their childhoods treated as members of This article is a PNAS Direct Submission.
one gender (before transitioning) and part of their childhoods Published under the PNAS license.
treated as members of another gender (after transitioning). Such a Data deposition: The data presented in this article were deposited on Open Science
unique gender trajectory and socialization experience raises several Framework, https://osf.io/q2kuw/.
important questions about gender development (e.g., identity, self- 1
To whom correspondence may be addressed. Email: [email protected].
perceptions, attitudes, behaviors) that could not be answered This article contains supporting information online at https://www.pnas.org/lookup/suppl/
without this cohort of transgender children. In this paper, we ex- doi:10.1073/pnas.1909367116/-/DCSupplemental.
amine the extent to which young transgender children’s (3–12 y) First published November 18, 2019.

24480–24485 | PNAS | December 3, 2019 | vol. 116 | no. 49 www.pnas.org/cgi/doi/10.1073/pnas.1909367116

 direct parental and teacher socialization involving rewarding and to cisgender children. Alternatively, transgender children could
punishing particular behaviors associated with the child’s assigned show weaker association with their current gender than their
sex and indirect exposure to social stereotypes and norms about cisgender peers. Given the hypothesized role of early socializa-
gender more broadly; ref. 32), although theories differ in their tion on children’s gender development (41–43), and insofar as
emphases of the relative contributions of these forces. Crucially, transgender children were initially viewed as and socialized as
for the cisgender children who participated in the research cited members of a different gender than they currently live as (i.e.,
above, sex assigned at birth, gender identity, and socialization are were rewarded and punished for acting in accordance with that
all typically aligned. Therefore, these various factors likely work to other gender)*, they might show weaker association with their
encourage and reinforce stereotypical behaviors, making it diffi- current gender (i.e., more association with the other binary
cult to isolate or rule out the impact of different forces. gender), compared to cisgender children who have always been
To illustrate, due to biological factors, a prototypical girl has treated and socialized as their current gender.
XX sex chromosomes and feminized genitalia. Because of these Although some research has examined gender development in
attributes, she is labeled as a girl at birth, and is treated by people children whose sex is neither clearly male nor female (e.g., in-
around her as a girl. She likely will develop an internal sense of tersex children; refs. 44 and 45), and in children who show some
being a girl (i.e., gender identity). She may engage in cognitive cross-gender preferences and behaviors (e.g., tomboys; refs. 46–
self-socializing in which she will attend to and emulate social 48), to date, only a few studies have examined gender develop-
norms related to being a girl, thereby increasing the likelihood ment in socially transitioned transgender children (14, 49–52).
that she will engage in behaviors stereotyped “for girls,” and to Using smaller samples, these early studies have found that
develop preferences for gender-stereotypic activities and ex- transgender children differ significantly from cisgender children
pression (i.e., appearance), while also experiencing direct posi- of the same assigned sex in terms of their gender development
tive or vicarious reinforcement from socialization agents (e.g., but have not found significant differences between transgender
parents, peers, media) for engaging in those behaviors. There- and cisgender children of the same gender. One concern with
fore, by the time she is a young child, it becomes difficult to

PSYCHOLOGICAL AND
small studies, however, is that statistical power is low, making

COGNITIVE SCIENCES
disentangle the role each of these factors plays in her current null results especially easy to obtain; even a medium or large
gender identity as a girl and her expression of gender via pref- difference may be obscured by the lack of statistical power (53,
erences for girl-typed clothing and toys. 54). In contrast, the present work examines the largest sample of
In contrast, studying transgender children (as well as other transgender children to date—more than 300 transgender chil-
gender-diverse samples, such as intersex individuals, or children dren aged 3–12—to ask 4 research questions.
with less-common socialization such as children raised without a Our first 2 questions aim to extend findings from prior pre-
gender; ref. 33) can help us begin to separate some of these liminary research with this much larger sample. First, what does
contributors to gender development. For example, a transgender transgender children’s gender development look like? If findings
girl may have male external genitalia and be assigned male at from the preliminary studies with transgender children (14, 49–
birth, may be assumed to identify as a boy, and may be treated by 52) are generalizable to a larger sample, transgender children in
others as a boy, but may nonetheless internally feel like a girl and
Downloaded from https://www.pnas.org by 47.61.168.101 on April 13, 2023 from IP address 47.61.168.101.

this study might show clear identification and gender-typing

self-socialize accordingly (e.g., seeking out information about aligned with their current gender, rather than the gender asso-
what girls like and how they look). If such a child then favors ciated with their sex assigned at birth. Assuming this is the case
feminine-typed toys and clothing, these preferences are unlikely
(as in past work), our second question addresses how trans-
due to being assumed a boy at birth or due to direct socialization
gender children’s gender development compares to the gender
as an (assumed) boy. Instead, the child’s internal sense of
development of their cisgender peers. Again, if past research is
identity (e.g., identifying as a girl), paired with broad gender
replicated with the current larger sample, transgender children in
socialization regarding how boys and girls typically act, could
this study might not differ from cisgender children. Such a
better explain the child’s gender preferences and expression. In
finding would strengthen interpretation of previous null findings
this way, examining the gender development of transgender
due to larger sample size and even more importantly, would be
children can help us to disentangle these different forces and
their role in governing children’s gender development. theoretically significant in suggesting a particularly strong impact
of children’s self-socialization and/or (unknown) biological fac-
Gender Development in Transgender vs. Cisgender Children tors on gender expression.
Historically, development among children who showed persistent The large sample size of the current study allows us to also
and strong cross-gender identification and preferences were often address previously unanswered questions. Central to these ques-
studied as a marker of problematic development that needed to be tions is whether transgender children show coherence across
fixed (34–37), being considered a sign of developmental delay (38), measures in their gender identity and expression. Some theories
or a sign of inherent psychopathology (34). More recently, however, of gender development have argued for the importance of self-
some researchers, clinicians, and parents are viewing children’s socialization of gender: that once children identify with a gender,
cross-gender identification and behaviors as part of a spectrum of they are motivated to behave consistently with that identity, and as a
normal gender variation rather than a clinical concern (7–9). result, gender-typed behaviors and preferences tend to become
In the current research, we are answering the question of stronger (13, 30, 31, 55, 56). These theories further argue that self-
whether transgender children’s gender development resembles socialization leads to coherence between identity, preferences, and
that of cisgender children by comparing these groups on stan- behavior (17, 57–60), meaning that children show preferences and
dard measures of gender development. Although this research is behaviors that are stereotypically associated with the gender they
exploratory, there are a few possible findings we might observe. identify themselves as. If there is coherence in measures of identity,
Previous research on gender and ethnic identity suggests that a preference, and behavior in transgender children, then this rein-
person whose identity is questioned or denied may show stronger forces the idea that self-socialization is an important piece of the
association with and assertion of that identity and associated
behaviors (39, 40). By similar logic, transgender children could
feel threat (e.g., because of that treatment and/or because they *We tested this assumption by coding early childhood (e.g., first day home after birth,
first Halloween) photos of a subset of transgender participants, prior to socially transi-
have bodies more often associated with another gender identity) tioning. Our analyses of these photos showed that transgender children were initially
and respond by showing stronger identification with their current reared in line with their sex assigned at birth, not the gender they are living as today.
gender and more extreme gender-typed preferences compared The full description of these analyses can be found in SI Appendix.

Gülgöz et al. PNAS | December 3, 2019 | vol. 116 | no. 49 | 24481

 puzzle of gender development. If, however, the relations between they were to members of their own gender, F(2,684) = 1.52, P =
identity, preference, and behavior are not cohesive in transgender 0.220, ηp2 < 0.01, or members of the other gender, F(2,683) =
children, this would suggest different factors (e.g., hormones, pa- 0.237, P = 0.789, ηp2 < 0.01. The 3 groups did not differ in the
rental socialization) impact different domains of gender develop- extremity of their gender-typed toy preferences, F(2,706) = 0.003,
ment and so empirical work will need to focus on explaining each P = 0.997, ηp2 < 0.01, or the strength of their preferences for
distinct relation (for example, between identity and clothing pref- same-gender peers, F(2,743) = 0.11, P = 0.899, ηp2 < 0.01. Fur-
erences). Thus, our third research question addresses the coherence ther, as can be seen in Fig. 1, participants across groups not only
between children’s responses to standard measures of gender de- showed similarity in their mean responses, but also in terms of
velopment: Do transgender and cisgender children show similar or within-group variability; for example, within both transgender and
different relations among (or coherence in) aspects of gender cisgender groups, while some girls showed strongly feminine re-
development? sponses, others showed much less feminine responses.
Finally, we aimed to test a fourth research question: Is the Although the 3 groups were similar on most measures, we also
extremity of a transgender child’s gender identity and expression observed a few differences. The groups differed on implicit
related to the amount of time in which the child has been treated gender identity (measured using the Gender Identity Implicit
as a member of that gender group? Recent research has found Association Test, which assesses the relative speed of associating
that transgender children tested before socially transitioning do oneself with the categories “boys” vs. “girls”), F(2,547) = 4.74,
not differ in the degree to which they identify and/or express P = 0.009, ηp2 = 0.02, and post hoc Tukey comparisons showed
gender preferences when compared to transgender children that while all groups associated themselves with their current
tested after socially transitioning (52), suggesting that the tran- gender, transgender children did so to a lesser extent than cis-
sition itself may not be changing gender identity and expression. gender controls (P = 0.008, d = 0.28, d values refer to difference
With the current work, we build on this finding by assessing the between groups), and did not differ from siblings (P = 0.056, d =
extent to which time spent living and being treated as their 0.27). The latter 2 groups did not differ from each other (P =
current gender (and not the gender they were assigned at birth) 0.982, d = 0.02). In contrast, for stated clothing preferences,
predicts gender development among transgender children. For transgender participants showed stronger gender-typed clothing
cisgender children, age and the time during which they have been preferences than controls (P = 0.001, d = 0.29) and siblings (P <
treated as a certain gender are perfectly correlated, making it 0.001, d = 0.35); the latter 2 groups did not differ from each
difficult to disentangle the question of how timing of socializa- other (P = 0.784, d = 0.06); overall omnibus: F(2,706) = 9.82, P <
tion relates to development. Transgender children are unique in 0.001, ηp2 = 0.03, comparisons are post hoc Tukey tests. Further,
helping us better understand the role of socialization on gender participants differed in terms of how stereotypically gender-
development, above and beyond age-related development. typed their outfit was rated at the time of appointment,
F(2,760) = 3.44, P = 0.033, ηp2 = 0.01; however, post hoc Tukey
Results comparisons showed no significant differences between groups
We registered our research questions, measures, exclusion criteria, (P ≥ 0.101). Importantly, in the rare case where differences were
scoring, and analyses prior to conducting analyses on the full data observed, they tended to be small differences.
Downloaded from https://www.pnas.org by 47.61.168.101 on April 13, 2023 from IP address 47.61.168.101.

(https://osf.io/q2kuw/; also see the SI Appendix)†, although data

had already been collected (61) and preliminary reports with Do Transgender and Cisgender Children Show Similar Coherence in
smaller numbers of participants had been examined (addressing Gender Development? Our diverse battery of gender measures and
research questions 1 and 2) and, therefore, the registration is not large sample allowed us to answer our third research question. In
a preregistration nor data-independent. all groups, there was coherence among measures of gender de-
velopment (see SI Appendix, Table S10 for correlations for
What Does Transgender Children’s Gender Development Look Like? transgender participants and cisgender controls). The magnitude
On all measures, transgender children showed robust identifica- of these associations was similar across groups‡. This means that,
tion and preferences stereotypically associated with their current for example, children who showed more stereotypical toy pref-
gender, that differed from chance or gender-neutral responding erences also tended to show stronger gender identification and
(all P < 0.001; see Table 1 for descriptive statistics, see SI Appendix stereotypical clothing preferences.
for further information). This meant, for example, transgender
boys (assigned females), on average, identified as boys, favored Is the Extremity of a Transgender Child’s Gender Identity and
stereotypically masculine toys and clothes, and preferred to be Expression Related to the Amount of Time in Which the Child Has
friends with boys. Been Treated as Their Current Gender? We assessed whether the
amount of time that has passed since a transgender child socially
How Does Transgender Children’s Gender Development Compare to transitioned is associated with the strength of their identification
Their Cisgender Peers’ Development? We found few differences and gender-typed preferences. That is, is it the case that the
between transgender and cisgender participants (see Table 1 for longer a transgender child has presented as their current gender,
descriptive statistics relevant to each measure; see SI Appendix, the more extreme their identification and preferences related to
Figs. S2 and S3 for more details including equivalence tests). On the current gender? Because we found that participant age was
the explicit gender identity measure, which asked participants if correlated with the amount of time (M time since transition =
they were a boy, a girl, or something else, transgender participants, 1.14 y, SD = 1.13 y) that had passed since participants’ transition,
cisgender controls, and cisgender siblings were all more likely to r(317) = 0.29, P < 0.001, we controlled for current age in a series
respond with their current gender (transgender: 84%; cisgender of partial correlations between time since transition and each of
controls: 83%; cisgender siblings: 87%) than with the other gen- the measures described above. After controlling for age, we
der, or as “something else”; these percentages did not statistically found that longer time since transition predicted less stereotyped
differ by participant group, P = 0.782, V = 0.03 Participant groups
also did not differ in terms of their predicted future gender
identity, P = 0.514, V = 0.04, or in their perceptions of how similar ‡
Whether correlations were small-to-moderate or strong (SI Appendix) depended on how
items were scored, such that measures were more highly correlated when coded by
gender (lower scores indicated masculine responses and higher scores indicated feminine
responses) than when coded as a function of a participant’s current gender (lower scores
†
Due to word limits, in this report we include only part of the registered analyses and indicated more “other gender” responding and higher scores indicated more “own
relevant findings. For the full set of analyses, please see SI Appendix. gender” responding).

24482 | www.pnas.org/cgi/doi/10.1073/pnas.1909367116 Gülgöz et al.

 Table 1. Descriptive statistics* of participants’ scores for each measure
Task Control Transgender Sibling

Toy preferences (0–100) 68.42 (20.18) 67.64 (21.63) 70.92 (19.94)

Clothing preferences (0–100) 82.74 (17.67) 87.97 (15.43) 81.63 (18.36)
Peer preferences (0–100) 80.88 (21.67) 79.92 (22.39) 78.34 (24.39)
Similarity to own gender (1 to 5) 4.11 (0.75) 4.20 (0.84) 4.14 (0.91)
Similarity to other gender (1 to 5) 2.12 (0.81) 2.08 (0.88) 2.01 (0.88)
Implicit gender identity (∼ −2 to ∼ +2)† .39 (.47) .26 (.45) .38 (.43)
Current gender identity, % 83 84 87
Future gender identity, % 79 80 85
Outfit at appointment (1–5) 4.10 (0.55) 4.17 (0.55) 4.07 (0.55)

Higher scores on all measures indicate greater alignment with current gender identity.
*Means (SDs) reported for all measures, except the current and future gender identity measures, which report
the percentage of participants who responded with their current gender.
†
Technically, implicit gender identity scores could range above or below ± 2; however, in reality they seldom do.

clothing preferences, r(277) = −0.12, P = 0.04. We observed no socialization. That is, a 10-y-old transgender girl who was labeled a
other significant associations between time since transition and any boy at birth and raised for 9 y as a boy, a 10-y-old transgender girl
of the measures (for all correlations, −0.12 < r < 0.06 and P > who was labeled a boy at birth and raised for 5 y as a boy, and a 10-y-
0.100), indicating no evidence that children who transitioned longer old cisgender girl (sibling or control) who was labeled a girl at birth
ago showed stronger or weaker identities or preferences than

PSYCHOLOGICAL AND
COGNITIVE SCIENCES
and was raised for 10 y as a girl did not significantly differ in their
children who transitioned more recently (SI Appendix, Table S9). identification and preferences on the assessed measures. These
Discussion findings therefore provide preliminary evidence that neither sex
Transgender children showed a clear pattern of gender develop-
ment associated with their current gender and not their sex
assigned at birth. This pattern was consistent across several mea-
sures of gender identity and multiple, distinct measures of gender
typing. As predicted by self-socialization perspectives, children
showed strong coherence among the identity, preference, and
behavioral measures, irrespective of whether they are transgender
or cisgender, indicating that both within and across groups, chil-
Downloaded from https://www.pnas.org by 47.61.168.101 on April 13, 2023 from IP address 47.61.168.101.

dren showed individual differences consistent across facets of their

gender development. This coherence suggests a powerful role of
individuals’ views of themselves and of self-socialization processes
on the development of gender typing. The similarity between the
transgender group and the cisgender comparison groups was ap-
parent not only in the means but in the distributions, with all
groups displaying variability in responses. For example, while on
average girls in all groups showed feminine responses, all groups
included some girls who showed especially feminine responses and
some girls who showed more masculine responses.
Observing few significant differences between groups was
notable because the transgender children in this sample spent
years being treated as a member of a different gender group than
they currently live as, while the cisgender children have always
been thought of and treated as members of their current gender
group. In fact, even among transgender children, findings did not
vary by whether they had just recently transitioned or have been
living as their current gender for several years (except for clothing
preferences, which was a small effect). This finding further aligns
with a recent discovery that transgender children’s gender devel-
opment before socially transitioning did not significantly differ
from that of a group of age-matched transgender children tested
after socially transitioning (52). Together, these 2 studies suggest
the possibility that children in this early-identifying transgender
cohort may not show major changes in identification, or in their
gender-typed preferences, during the early years, including the
years immediately preceding and following transition. Of course,
this study is cross-sectional and captures a specific moment in
participants’ gender development and our broader culture; these
Fig. 1. Density plots depicting distributions of girls’ and boys’ responses on
findings do not speak to children’s future identities or preferences. each child measure across the 3 participant groups. Higher scores on the x
These findings have implications that may be useful to con- axis indicate greater identification/association with girls; lower scores on the
sider outside of this cohort of children. Our results demonstrated x axis indicate greater identification/association with boys. Scores for boys
that transgender children’s gender development does not appear on these graphs are reverse coded from those in Table 1. Vertical dotted
to show lingering impact of early sex-assignment or sex-specific lines in graphs represent means.

Gülgöz et al. PNAS | December 3, 2019 | vol. 116 | no. 49 | 24483

 assignment at birth nor direct or indirect sex-specific socialization and are not associated with the aspects of gender development assessed
expectations (e.g., rewarding masculine things and punishing feminine here (SI Appendix), but future studies with larger ranges of de-
ones for assigned males) in alignment with early assignment neces- mographic characteristics might better answer these questions.
sarily define how a child later identifies or expresses their gender. As highlighted early in the paper, the current study is unique
Further, the lack of differences between siblings and controls, in that it captures gender development in a newly emerging
observed across measures, suggests that there likely is not a group of transgender children, just as this group of young
unique gender socialization experience in the homes of trans- people are becoming more visible. It is impossible to predict
gender children that is changing gender development. how transitioning at an early age or at this specific historic time
Our findings do not imply that gender socialization is un- may impact later features of children’s development, or how
important in early development. To the contrary, early devel- development might be impacted if the rates of social transition
opment appears to be the time when, for example, children learn continue to shift across time such that these children’s experiences
which toys or clothing or activities are stereotyped as masculine become more common. At the same time, these children are all
or feminine in their culture (62); the children in our study seem living in what is increasingly called a WEIRD (Western, Edu-
to have learned this information by ages 3 and 4, as even our cated, Industrialized, Rich, Democratic; ref. 64) culture. There are
youngest transgender participants showed clearly gendered many other cultures that recognize groups with diverse gender
preferences. Therefore, the transgender children in our sample identities (1–5); whether the observed patterns would be seen in
are showing signs of broader knowledge about gender likely “transgender” children in these other cultures is, as yet, unknown.
gained through living in their society; they just do not appear to These findings illustrate that children develop a sense of identity
show an impact of early direct socialization geared toward the at an early age, that this identity is not necessarily determined by
gender they were assumed to be at birth, at least on the measures sex assigned at birth, and that children may hold on to this identity
assessed here. Instead, in these cases, it appears that transgender even when it conflicts with others’ expectations. Why children
children’s motivation to self-socialize (31) is likely occurring, and develop such strong feelings of identity and how identity is main-
the focus is on learning about characteristics of the other gender tained or changed when in conflict with other socialization forces
rather than learning about characteristics of the gender assigned are seldom addressed in theories of gender development. Our
to them at birth. It is possible that there are children for whom findings also demonstrate that once a child is living in line with an
early direct socialization is more influential in shaping their identity different from the one they were assumed to have at birth
gender identities, and future work might help uncover factors and were initially socialized to have, they are likely to show the
that lead to such developmental trajectories. same patterns of gender development as a child who develops a
One interesting note for future work that came out of these gender aligning with their assigned sex and socialization. Although
findings is the importance and implications of how we define questions remain about whether young children’s identities and
participant groups. In the current work, we constructed our preferences will remain consistent over situations and stable over
transgender and cisgender participant groups by using the align- time, these data suggest the power of one’s own internal sense of
ment of one’s pronouns and one’s assigned sex. However, when we identity (however formed initially) and the consequences of self-
asked children how they identify, a few transgender and cisgender socialization on how one thinks about and expresses one’s gender.
Downloaded from https://www.pnas.org by 47.61.168.101 on April 13, 2023 from IP address 47.61.168.101.

children identified as something other than the gender aligning

with their current pronouns. This finding suggests that perhaps Materials and Methods
these experimenter-created groups, based on pronouns, are We collected data from 317 transgender children (208 transgender girls;
somewhat arbitrary. Had we used children’s self-categorizations of Mage = 96.82 mo, SD = 28.37; 68% White; see SI Appendix, Fig. S1 for a map
gender to divide our groups, for example, into boys, girls, and of where transgender participants have been recruited; see SI Appendix for
people who prefer a different term, these categorizations would full demographics of all participants)§.
likely have resulted in groups that cross our boundaries of cis- We recruited 2 comparison groups. The first was the Sibling Group. When
gender and transgender. How this approach might impact findings possible, we recruited the closest in age cisgender sibling (n = 189; 82 girls;
Mage = 96.72, SD = 29.72; 68% White) of each transgender participant who
is unknown and has implications for an ongoing discussion on how
was also 3–12 y of age. We also recruited a Control Group of unrelated
gender identity should be assessed in research (63). cisgender participants to match to each transgender child by gender and age
The present work provides a foundation for additional ques- (n = 316; 207 girls; Mage = 97.23, SD = 28.27; 70% White). Control partici-
tions to be addressed in future research on the transgender ex- pants were recruited from a university database, from 1 major metropolitan
perience in children. For example, our transgender participants’ area in the Pacific Northwest.
families affirmed their child’s gender identity (as evidenced by Participants received a battery of tasks assessing multiple indicators of gender
allowing their child’s early social transition), which was a necessary development, including explicit gender identity, implicit gender identity (52),
condition as the current study. Future work is needed to provide perceived similarity to their own and “opposite” gender (18), toy, peer, clothing
insight into the transition process (e.g., who transitions, what preferences, and gender-typed behavior (outfit at appointment). Subsets of
factors enable successful transitions), how transitioning influences children in this study also completed other tasks (for more on those measures
and why these were kept, see SI Appendix). All 3 groups received the same
gender development (although see ref. 52), and the role support
measures. Parents filled out a questionnaire packet containing questions about
and affirmation may play in gender identification and expression.
demographics, and their participating children’s gender identity and gender
A second future question concerns how these findings would typing (see SI Appendix for more details). Children participated only if they
or would not generalize to samples with different demographic provided oral (ages 3–8 y) or written (ages 9 and up) assent, and their parents
characteristics (i.e., samples with a different range of household provided written consent. The study and materials were approved by the Uni-
income, parental political ideology, race, and level of education; versity of Washington Institutional Review Board (protocol no. 44772).
see SI Appendix). The present study included comparison groups
with similar backgrounds to our transgender sample, which Data Availability. The registration, data, and analysis script for the results
allowed greater confidence that differences in demographics presented in this work can be found online at https://osf.io/q2kuw/ (doi:
could not explain differences between groups, but future work 10.17605/OSF.IO/Q2KUW).
would benefit from asking whether groups with other charac-
teristics (e.g., in other countries, with families from lower so-
cioeconomic backgrounds) differ from the groups presented Although there are some children who use nonbinary pronouns, (e.g., “they,” “ze”) and
§

consider themselves “transgender,” the present work focused on children who had socially
here and, if so, consider the reasons that might explain the transitioned, and exclusively used binary “he” or “she” pronouns that were not associated
differences (e.g., differences in access to supportive schools). with their sex assigned at birth; a criterion determined before the study began. Please see
Our initial data on this issue suggest that demographic characteristics https://osf.io/duy7b/ for our laboratory’s standards for inclusion of participants in papers.

24484 | www.pnas.org/cgi/doi/10.1073/pnas.1909367116 Gülgöz et al.

 ACKNOWLEDGMENTS. We thank all participating families, as well as those 1715068 (to S.G.), BCS-1523632, and SMA-1837857 (to K.R.O.); National Insti-
who have helped with data collection and data entry, listed in SI Appendix. tute of Child Health and Human Development Grant HD092347 (to K.R.O.);
This research was funded by National Science Foundation Grants BCS- and the Arcus Foundation.

1. R. Astuti, “‘It’s a boy! It’s a girl!’ Reflections on sex and gender in Madagascar and 33. Y. Dunham, K. R. Olson, Beyond discrete categories: Studying multiracial, intersex,
beyond” in Bodies and Persons: Comparative Perspectives from Africa and Melanesia, and transgender children will strengthen basic developmental science. J. Cogn. Dev.
M. Lambek, A. Strathern, Eds. (Cambridge University Press, Cambridge, UK, 1998). 17, 642–665 (2016).
2. E. Blackwood, Sexuality and gender in Native American tribes: The case of cross- 34. S. Coates, E. S. Person, Extreme boyhood femininity: Isolated behavior or pervasive
gender females. Signs J. Women Culture Soc. 10, 27–42 (1984). disorder? J. Am. Acad. Child Psychiatry 24, 702–709 (1985).
3. C. Callender et al., The North American berdache. Curr. Anthropol. 24, 443–470 35. G. A. Rekers, O. I. Lovaas, Behavioral treatment of deviant sex-role behaviors in a
(1983). male child. J. Appl. Behav. Anal. 7, 173–190 (1974).
4. J. M. Mageo, Male transvestism and cultural change in Samoa. Am. Ethnol. 19, 443– 36. G. A. Rekers, O. I. Lovaas, B. Low, The behavioral treatment of a “transsexual” pre-
459 (1992). adolescent boy. J. Abnorm. Child Psychol. 2, 99–116 (1974).
5. P. L. Vasey, N. H. Bartlett, What can the Samoan “Fa’afafine” teach us about the 37. P. McHugh, Transgender surgery isn’t the solution. Wall Street Journal. https://
Western concept of gender identity disorder in childhood? Perspect. Biol. Med. 50, www.wsj.com/articles/paul-mchugh-transgender-surgery-isnt-the-solution-1402615120.
481–490 (2007). Accessed 13 May 2014.
6. A. H. Devor, Witnessing and mirroring: A fourteen stage model of transsexual identity 38. K. J. Zucker et al., Gender constancy judgments in children with gender identity
formation. J. Gay Lesbian Psychother. 8, 41–67 (2004). disorder: Evidence for a developmental lag. Arch. Sex. Behav. 28, 475–502 (1999).
7. D. B. Hill, E. Menvielle, K. M. Sica, A. Johnson, An affirmative intervention for families 39. M. Barreto, R. Spears, N. Ellemers, K. Shahinper, Who wants to know? The effect of
with gender variant children: Parental ratings of child mental health and gender. J. audience on identity expression among minority group members. Br. J. Soc. Psychol.
Sex Marital Ther. 36, 6–23 (2010). 42, 299–318 (2003).
8. M. A. Hidalgo et al., The gender affirmative model: What we know and what we aim 40. S. Cheryan, J. Cameron, Z. Katagiri, B. Monin, Manning up: Threatened men com-
to learn. Hum. Development 56, 285–290 (2013). pensate by disavowing feminine preferences and embracing masculine attributes.
9. J. Malpas, Between pink and blue: A multi-dimensional family approach to gender Soc. Psychol. 46, 218–227 (2015).
nonconforming children and their families. Fam. Process 50, 453–470 (2011). 41. K. Bussey, A. Bandura, Self-regulatory mechanisms governing gender development.
10. A. Campbell, L. Shirley, L. Caygill, Sex-typed preferences in three domains: Do two- Child Dev. 63, 1236–1250 (1992).
year-olds need cognitive variables? Br. J. Psychol. 93, 203–217 (2002). 42. K. Bussey, A. Bandura, Social cognitive theory of gender development and differen-

PSYCHOLOGICAL AND
COGNITIVE SCIENCES
11. B. I. Fagot, M. D. Leinbach, R. Hagan, Gender labeling and the adoption of sex-typed tiation. Psychol. Rev. 106, 676–713 (1999).
behaviors. Dev. Psychol. 22, 440–443 (1986). 43. E. E. Maccoby, Parenting and its effects on children: On reading and misreading be-
12. D. N. Ruble, C. L. Martin, “Gender development” in Handbook of Child Psychology: havior genetics. Annu. Rev. Psychol. 51, 1–27 (2000).
Social, Emotional, and Personality Development, W. Damon, N. Eisenberg, Eds. (Wi- 44. L. Frisén et al., Gender role behavior, sexuality, and psychosocial adaptation in
ley, 1998), pp. 933–1016.
women with congenital adrenal hyperplasia due to CYP21A2 deficiency. J. Clin.
13. K. M. Zosuls et al., The acquisition of gender labels in infancy: Implications for gender-
Endocrinol. Metab. 94, 3432–3439 (2009).
typed play. Dev. Psychol. 45, 688–701 (2009).
45. H. F. L. Meyer-Bahlburg et al., Prenatal androgenization affects gender-related be-
14. A. A. Fast, K. R. Olson, Gender development in transgender preschool children. Child
havior but not gender identity in 5-12-year-old girls with congenital adrenal hyper-
Dev. 89, 620–637 (2018).
plasia. Arch. Sex. Behav. 33, 97–104 (2004).
15. D. N. Ruble et al., The role of gender constancy in early gender development. Child
46. C. L. Martin, L. M. Dinella, Congruence between gender stereotypes and activity
Dev. 78, 1121–1136 (2007).
preference in self-identified tomboys and non-tomboys. Arch. Sex. Behav. 41, 599–
16. R. G. Slaby, K. S. Frey, Development of gender constancy and selective attention to
610 (2012).
same-sex models. Child Dev. 46, 849–856 (1975).
47. J. M. Bailey, K. T. Bechtold, S. A. Berenbaum, Who are tomboys and why should we
17. S. K. Egan, D. G. Perry, Gender identity: A multidimensional analysis with implications
study them? Arch. Sex. Behav. 31, 333–341 (2002).
Downloaded from https://www.pnas.org by 47.61.168.101 on April 13, 2023 from IP address 47.61.168.101.

for psychosocial adjustment. Dev. Psychol. 37, 451–463 (2001).

48. S. Ahlqvist, M. L. Halim, F. K. Greulich, L. E. Lurye, D. Ruble, The potential benefits and
18. C. L. Martin, N. C. Andrews, D. E. England, K. Zosuls, D. N. Ruble, A dual identity
risks of identifying as a tomboy: A social identity perspective. Self and Identity 12,
approach for conceptualizing and measuring children’s gender identity. Child Dev.
563–568 (2013).
88, 167–182 (2017).
49. K. R. Olson, A. C. Key, N. R. Eaton, Gender cognition in transgender children. Psychol.
19. J. L. Yunger, P. R. Carver, D. G. Perry, Does gender identity influence children’s psy-
Sci. 26, 467–474 (2015).
chological well-being? Dev. Psychol. 40, 572–582 (2004).
50. K. R. Olson, E. A. Enright, Do transgender children (gender) stereotype less than their
20. M. L. Halim, D. Ruble, C. Tamis-LeMonda, P. E. Shrout, Rigidity in gender-typed be-
peers and siblings? Dev. Sci. 21, e12606 (2018).
haviors in early childhood: A longitudinal study of ethnic minority children. Child Dev.
51. J. R. Rae, K. R. Olson, Test-retest reliability and predictive validity of the Implicit As-
84, 1269–1284 (2013).
21. P. La Freniere, F. F. Strayer, R. Gauthier, The emergence of same-sex affiliative pref- sociation Test in children. Dev. Psychol. 54, 308–330 (2018).
erences among preschool peers: A developmental/ethological perspective. Child Dev. 52. J. R. Rae et al., Predicting early childhood gender transition. Psychol. Sci. 30, 669–681
55, 1958–1965 (1984). (2019).
22. E. E. Maccoby, The Two Sexes: Growing up Apart, Coming Together (Belknap/Harvard 53. J. Cohen, Statistical power analysis. Curr. Dir. Psychol. Sci. 1, 98–101 (1992).
University Press, Cambridge, MA, 1998). 54. J. A. C. Sterne, G. D. Smith, Sifting the evidence-what’s wrong with significance tests?
23. E. E. Maccoby, C. N. Jacklin, Gender segregation in childhood. Adv. Child Dev. Behav. Phys. Ther. 81, 1464–1469 (2001).
20, 239–287 (1987). 55. C. L. Martin, C. F. Halverson, A schematic processing model of sex typing and
24. C. L. Martin, R. A. Fabes, The stability and consequences of young children’s same-sex stereotyping in children. Child Dev. 52, 1119–1134 (1981).
peer interactions. Dev. Psychol. 37, 431–446 (2001). 56. K. M. Zosuls, D. N. Ruble, C. S. Tamis-LeMonda, Self-socialization of gender in African
25. C. L. Martin, R. A. Fabes, S. M. Evans, H. Wyman, Social cognition on the playground: American, Dominican immigrant, and Mexican immigrant toddlers. Child Dev. 85,
Children’s beliefs about playing with girls versus boys and their relations to sex seg- 2202–2217 (2014).
regated play. J. Soc. Pers. Relat. 16, 751–771 (1999). 57. B. I. Fagot, M. D. Leinbach, The young child’s gender schema: Environmental input,
26. S. M. McHale, J. Y. Kim, S. Whiteman, A. C. Crouter, Links between sex-typed time use internal organization. Child Dev. 60, 663–672 (1989).
in middle childhood and gender development in early adolescence. Dev. Psychol. 40, 58. C. L. Martin, J. K. Little, The relation of gender understanding to children’s sex-typed
868–881 (2004). preferences and gender stereotypes. Child Dev. 61, 1427–1439 (1990).
27. L. A. Serbin, K. K. Powlishta, J. Gulko, C. L. Martin, M. E. Lockheed, The development 59. M. M. Patterson, Self-perceived gender typicality, gender-typed attributes, and
of sex typing in middle childhood. Monogr. Soc. Res. Child Dev. 58, 1–99 (1993). gender stereotype endorsement in elementary-school-aged children. Sex Roles 67,
28. M. L. Halim et al., Pink frilly dresses and the avoidance of all things “girly”: children’s 422–434 (2012).
appearance rigidity and cognitive theories of gender development. Dev. Psychol. 50, 60. M. Weinraub et al., The development of sex role stereotypes in the third year: Re-
1091–1101 (2014). lationships to gender labeling, gender identity, sex-typed toy preference, and family
29. V. Jadva, M. Hines, S. Golombok, Infants’ preferences for toys, colors, and shapes: Sex characteristics. Child Dev. 55, 1493–1503 (1984).
differences and similarities. Arch. Sex. Behav. 39, 1261–1273 (2010). 61. S. Gülgöz, K. R. Olson, Gender development among transgender and cisgender chil-
30. L. Kohlberg, “A cognitive-developmental analysis of children’s sex-role concepts and dren. Open Science Framework. https://osf.io/q2kuw/. Deposited 31 October 2019.
attitudes” in The Development of Sex Differences, E. E. Maccoby, Ed. (Stanford Uni- 62. C. Leaper, “Gender and social-cognitive development” in Handbook of Child Psy-
versity Press, Stanford, CA, 1966), pp. 82–173. chology and Developmental Science, R. M. Lerner, Ed. (Wiley, 2015), pp. 1–48.
31. C. L. Martin, D. N. Ruble, J. Szkrybalo, Cognitive theories of early gender develop- 63. C. C. Tate, J. N. Ledbetter, C. P. Youssef, A two-question method for assessing gender
ment. Psychol. Bull. 128, 903–933 (2002). categories in the social and medical sciences. J. Sex Res. 50, 767–776 (2013).
32. W. Mischel, “Sex-typing and socialization” in Carmichael’s Manual of Child Psychol- 64. J. Henrich, S. J. Heine, A. Norenzayan, The weirdest people in the world? Behav. Brain
ogy, P. H. Mussen, Ed. (Wiley, 1970), vol. 2, pp. 3–72. Sci. 33, 61–83 (2010).

Gülgöz et al. PNAS | December 3, 2019 | vol. 116 | no. 49 | 24485