Child Development, March/April 2018, Volume 89, Number 2, Pages 620–637

Gender Development in Transgender Preschool Children

Anne A. Fast, and Kristina R. Olson
University of Washington

An increasing number of transgender children—those who express a gender identity that is “opposite” their
natal sex—are socially transitioning, or presenting as their gender identity in everyday life. This study asks
whether these children differ from gender-typical peers on basic gender development tasks. Three- to 5-year-
old socially transitioned transgender children (n = 36) did not differ from controls matched on age and
expressed gender (n = 36), or siblings of transgender and gender nonconforming children (n = 24) on gender
preference, behavior, and belief measures. However, transgender children were less likely than both control
groups to believe that their gender at birth matches their current gender, whereas both transgender children
and siblings were less likely than controls to believe that other people’s gender is stable.

Gender is perhaps the central way in which called a social transition. In the current study, we
children and adults carve the social world into cate- ask whether children like Jazz show patterns of
gories (Maccoby, 1998; Ruble, Martin, & Beren- gender development within the early preschool
baum, 2006). Therefore, it may be unsurprising that years that are similar to or different from gender-
gender is likely the earliest identity and social cate- typical children of the same age.
gory to emerge in development (Lewis & Brooks- Despite developmental psychology’s long and
Gunn, 1979), and that acquiring gender knowledge rich history of studying gender development, chil-
is considered a critical component of early child- dren like Jazz—socially transitioned transgender
hood development (Ruble et al., 2007). A pervasive, children—have largely been absent from these inves-
albeit an often implicit, assumption in society and tigations. This is in part because social transitions
in psychological research is that one’s gender (one’s early in development are relatively new (Ehrensaft,
sense of identity as a boy or girl) aligns with one’s 2011; Hidalgo et al., 2013). However, the unique
sex (determined by one’s anatomy and chromo- developmental experiences of transgender children,
somes at birth). This belief is clearly grounded in especially those who “switch” their gender presenta-
data—for most people, their gender identity aligns tions early in life, may contribute in interesting ways
with their sex. However, it is not always the case; to discussions about how gender and sex function as
rather there are people, termed transgender, whose organizing principles in young children’s lives.
gender identity and sex at birth do not align. One Although their experiences are rare (estimates of
example is reality star, Jazz Jennings, who transgender identities are difficult to find, but one
expressed a female identity as soon as she could recent study of New Zealander high school students
communicate that information to others despite suggested a rate of approximately 1.2% of people
being born a natal male (Goldberg & Adriano, identifying as transgender, Clark et al., 2014; and
2007). When she was 5 years old, her parents likely even fewer have socially transitioned to live as
allowed her to begin living as a girl in everyday life the “other” gender), given that socially transitioned
(meaning that they used the pronoun “she” and a transgender children do exist, it is important to
new female name “Jazz,” but no medical or hor- include their experiences in the study of gender
monal intervention occurred at that age)—a process development. Thus, consistent with arguments con-
cerning the importance of increasing diversity in
This research was supported by NSF Grant 1523632 and the empirical psychology (e.g., Kang & Bodenhausen,
Arcus Foundation. Any opinions, findings, or conclusions 2015; Shelton, 2000), the inclusion of transgender
expressed herein are those of the authors and do not necessarily
reflect the views of the National Science Foundation or the Arcus children will further our understanding of the range
Foundation. The authors wish to thank Madeleine DeMeules,
Lily Durwood, Elizabeth Ake, Sarah Colombo, and Allison Skin-
ner for assistance with data collection.
Correspondence concerning this article should be addressed to © 2017 The Authors
Anne A. Fast, Department of Psychology, University of Washing- Child Development © 2017 Society for Research in Child Development, Inc.
ton, Seattle, WA 98195. Electronic mail may be sent to afast@ All rights reserved. 0009-3920/2018/8902-0020
uw.edu. DOI: 10.1111/cdev.12758
 Transgender Preschool Children 621

of ways in which gender emerges and develops some work with older socially transitioned trans-
while also offering possible contributions to theoreti- gender children suggests that they give gender-typi-
cal discussions of gender development (Dunham & cal (but not sex-typical) responses on measures of
Olson, 2016). In the current study, we aim to do so gender development, such as gendered preferences
by investigating preschool-age socially transitioned (Olson, Key, & Eaton, 2015). Evidence that trans-
transgender children’s gendered preferences, behav- gender children show strongly gendered prefer-
iors, and beliefs. We discuss how these data can add ences (perhaps as strong as controls) paired with a
to our understanding of gender development, inform lack of gender stability beliefs (at least as it has tra-
theories of gender development, and give rise to ditionally been tested) could suggest that the
new research questions concerning the development “boost” from stability beliefs is not needed to show
of gender cognition. the high levels of gendered preferences observed by
gender-typical children.
In addition, this study is likely to spawn further
Theoretical Contribution
theoretical work on questions about gender devel-
Beyond conducting an exploratory analysis of opment that may not come about until we know
the basic gender development of socially transi- how socially transitioned transgender children
tioned transgender children and comparing it to the respond compared to their gender-typical peers. For
development of gender-typical children of the same example, if transgender children do not differ from
age, a secondary goal of the current study was to gender-typical children on some or all measures of
provide data that can begin to speak to broader gender development, we would have some prelimi-
theoretical discussions about gender development nary evidence that gender of rearing in the first few
and transgender children. As one example, some years may not be a large contributor to those partic-
previous studies has claimed that understanding ular aspects of gender development—a hypothesis
gender constancy (that gender is a stable and con- we may then be able to test with future data collec-
sistent attribute) allows for greater organization tion with transgender children or children with
and motivation of strong same-gender preferences other diverse early experiences, such as intersex chil-
and behaviors (Kohlberg, 1966; Slaby & Frey, 1975). dren who were reared as one gender but later iden-
More recent theorizing, however, suggests that tified as the “opposite” gender. Thus, we see this
although full gender constancy knowledge may not study as a catalyst for the establishment of future
be responsible for enhancing gendered preferences research and theories of gender development.
and behavior, understanding gender stability in par-
ticular is a central factor in motivating strong gen-
Gender-Typical Development
der preferences (Ruble et al., 2007). For example, a
very young girl might already display an affinity A large body of research suggests that gender-
toward pink, dolls, and dresses, but once she typical children, or those whose gender identity
understands the stability of her gender, she will aligns with their sex at birth, are attuned to cues
have even more extreme gendered preferences. about gender early in development and begin per-
Socially transitioned transgender children present ceiving gender categories at a young age. Infants
an interesting case to test this idea because unlike display an ability to discriminate male and female
other children, they might not have a belief that faces by 6 months of age (Quinn, Yahr, Kuhn, Sla-
their gender is stable. Anecdotally, many transgen- ter, & Pascalis, 2002), and they can accurately
der children and their families discuss how they match male and female voices to male and female
“used to be” one gender but are another gender faces by their first birthday (Poulin-Dubois, Serbin,
now, after their social transition. Although this con- Kenyon, & Derbyshire, 1994). Around age 2, when
versation can be interpreted as the child having they begin to acquire knowledge of gender labels
been assumed to be one gender and now being rec- (Fenson et al., 1994; Stennes, Burch, Sen, & Bauer,
ognized as a member of the “other” gender, chil- 2005), infants display preferences for objects and
dren may not have this nuanced understanding. people associated with their own gender (Serbin,
Furthermore, as discussed in more detail below, Poulin-Dubois, Colburne, Sen, & Eichstedt, 2001;
there is some initial evidence suggesting that gen- Zosuls et al., 2009) and show rudimentary gender
der nonconforming children (a group that would stereotyping (Levy & Haaf, 1994; Serbin, Poulin-
include transgender children) are less likely than Dubois, & Eichstedt, 2002).
gender-typical children to say that gender is stable The current study focuses on the preschool per-
over time (Zucker et al., 1999). At the same time, iod, as it is a crucial time for gender development.
622 Fast and Olson

Throughout these years (3–5 years of age), gender did not differ from gender-typical control children
is highly salient and a powerful motivator of chil- (who were matched on age and expressed gender)
dren’s preferences and behaviors. For example, pre- and gender-typical siblings when considered
school-age children use gender to guide their own according to their gender. When analyzed as a
outfit choices (Halim et al., 2014) and toy choices function of gender assigned at birth (i.e., according
(Eaton, Von Bargen, & Keats, 1981), such that they to natal sex), the transgender children differed from
express interest in objects that are associated with their controls and siblings on every measure. Thus,
their own gender rather than those linked to the from Olson et al. (2015) we can conclude that, by
other gender. Similarly, by age 3 and throughout the elementary years, socially transitioned transgen-
the preschool years, children display a strong pref- der children show gender-typical responding on
erence for same-gender people (Martin & Fabes, many measures of gender development.
2001; Martin, Fabes, Evans, & Wyman, 1999; Shutts, In addition to this more recent work on socially
Pemberton, & Spelke, 2013). Preschool-age children transitioned transgender children, there is a longer
also use gender to guide their expectations of tradition of studying gender nonconforming chil-
others’ appearances and activities (Miller, Lurye, dren—those who defy cultural gender expectations
Zosuls, & Ruble, 2009), with gender stereotype for children of their sex—in the clinical psychology
knowledge developing rapidly during this age and psychiatry literature. Although typically focused
range (see Signorella, Bigler, & Liben, 1993). Finally, on clinical outcomes (e.g., Cohen-Kettenis, Owen,
the preschool years are also thought to be a critical Kaijser, Bradley, & Zucker, 2003), these research
time for developing knowledge of gender constancy teams have occasionally reported on basic gender
(Kohlberg, 1966; Slaby & Frey, 1975). Specifically, development, much as we do in the current study
children at this age are thought to master an under- with socially transitioned transgender children. For
standing that gender is stable from infancy to adult- example, one study found that while siblings of gen-
hood (not until after preschool are they thought to der nonconforming children preferred to play with
understand that gender is consistent across changes toys manufactured for children of their sex, gender
in appearance, Ruble et al., 2007). Moreover, this nonconforming children (Mage = 7.6 years) did not
gender stability knowledge is considered to be a —they equally preferred toys manufactured for their
central factor in enhancing preschool children’s own sex and the other sex (Zucker, Bradley, Doer-
same-gender preferences and behavior (e.g., Ruble ing, & Lozinski, 1985). In contrast, for games, gender
et al., 2007). nonconforming children actually expressed a prefer-
ence for games manufactured for the other sex, but
their siblings did not. Thus, it appears that gender
Gender Nonconforming Children
nonconforming children’s preferences consistently
Despite the large body of work on gender devel- differ from their gender conforming peers (the defi-
opment, most of that work has been conducted nition of gender nonconformity), but only sometimes
with gender-typical children. The current study is was this difference in the direction opposite their sex
an exploratory investigation into whether socially at birth. Furthermore, Zucker et al. (1999) found that
transitioned transgender children show the same a group of 3- to 10-year-old gender nonconforming
patterns of gender development during the pre- children showed an atypical understanding of gen-
school years. To date, there has been no work on der constancy. That is, the gender diverse group of
this question, though there has been one study children was less likely than gender-typical children
reporting on gender development in elementary to believe that their own gender was stable across
school-age children and a few studies reporting on time (gender stability) or across changes in appear-
gender development in a broader range of gender ance (gender consistency) compared to others’ gen-
diverse children. der.
Olson et al. (2015) investigated a similar question Importantly, although these previous findings
about gender development milestones in elemen- suggest that gender nonconforming children have
tary school-age transgender children (Mage = response patterns differing from their same-sex
9 years, 1 month). They found that across several peers, these studies were completed with children
measures—preferences for same-gender peers and who differ from those in the current study in two
objects endorsed by those peers, as well as in the key ways. First, the children in this past work (aside
degree to which they displayed an implicit or expli- from Olson et al., 2015) were not socially transi-
cit gender identity, and implicit gender-based pref- tioned. That is, these children presented in public
erences—socially transitioned transgender children (e.g., attended school) as the gender that aligns with
 Transgender Preschool Children 623

their sex at birth. In contrast, the current study et al., 1999), we expected that young transgender
focuses on children who have socially transitioned, children may be less likely than gender-typical chil-
and thus present a gender that differs from the one dren to see their gender as stable over time. How-
they were assumed to have at birth. Second, previ- ever, drawing upon previous work on older
ous studies with gender nonconforming children socially transitioned transgender children (Olson
likely drew a broader range of children than those et al., 2015), we expected young transgender chil-
included in the current study. That is, there are dren to have just as strong same-gender preferences
many gender nonconforming children who do not and behavior as their gender-typical peers. If trans-
actually believe themselves to be the “other” gender; gender and gender-typical children show similar
instead, they tend to have preferences that align levels of preference but different patterns of stabil-
with the “other” gender while maintaining that they ity, this might suggest that stability is not playing a
identify as a member of the gender group aligning causal role, so much as it is a developmentally co-
with their sex at birth (see Olson, 2016 for more on occurring phenomenon.
this issue). Children in the current study—those Because previous work on gender development
expressing that they are a member of the “other” has included questions about gender identity, we do
gender group and living publically as this “other” so in the current study as well; however, the mea-
gender—perhaps could be thought of as the most sures we use are slightly different from those used
extreme subset of gender nonconforming youth. in past work. The explicit gender identity measure
Therefore, it may be reasonable to assume that these used in the current study asks children what they
socially transitioned transgender children are the feel like they are on the inside (i.e., in their “mind,
most likely to show effects in the direction opposite thoughts, and feelings”). This identity measure was
their sex at birth, leading to our hypothesis that designed to make it clear to children that we were
these children would show patterns of gender asking about their gender identity and not their bio-
responding remarkably similar to children who logical sex. We also investigated gender identity by
share their expressed gender but who differ in their asking children how similar they feel to boys and
sex (as Olson et al., 2015 found with older children). how similar they feel to girls. Finally, we explored
gender stereotyping (vs. flexibility). Because there is
no previous work on transgender or gender noncon-
The Current Study
forming children’s tendency to endorse gender
The present study is an exploratory investigation stereotypes, this was an exploratory investigation.
comparing the gender preferences, behavior and In the current study, we include two control
beliefs of young socially transitioned transgender groups, serving two different purposes. The first is a
children to those of their gender typical peers. To group of gender-typical children that are matched to
this end, we had socially transitioned transgender the transgender children on age and gender identity
children and gender-typical children complete a ser- (henceforth, controls). This group allows us to exam-
ies of basic gender development tasks, including ine whether our transgender sample is showing
measures assessing: (a) gender constancy under- responses typical for their age and gender. The sec-
standing, (b) gender preferences, (c) gendered ond control group is a group of gender-typical chil-
behavior (stereotypicality of outfit worn to the dren who are siblings of transgender or gender
appointment), (d) explicit gender identity, (e) per- nonconforming children (henceforth, siblings). The
ceived similarity to boys and girls, and (f) gender inclusion of the sibling control group allows us to
stereotyping. separate the impact of the lived experience of gender
As described in the Theoretical Contribution sec- diversity from mere knowledge of the existence of
tion earlier, assessing how socially transitioned gender diversity. In the case that transgender chil-
transgender children respond compared to gender- dren respond differently from controls and siblings,
typical peers on measures of gender constancy we might assume personal experience of being trans-
understanding—more specifically, gender stability gender is playing a critical role in this difference. On
understanding—paired with their responding to the other hand, a finding that siblings respond simi-
measures of gender preferences may shed some larly to transgender children but different from con-
light on whether seeing one’s gender as stable over trols would suggest that knowledge of gender
time is always a major contributor to enhancing diversity (or perhaps factors unique to the kinds of
same-gender preferences and behavior at this age. families that have gender nonconforming children in
Based on both anecdotal knowledge and previous them) plays a contributing role rather than just per-
work with gender nonconforming children (Zucker sonal experience as a transgender person.
624 Fast and Olson

We specifically focused on socially transitioned transitioned transgender children, currently ages 3–

transgender children in the preschool years (ages 3–5) 14. This study focused on the 3- to 5-year-old chil-
because it is the age at which—according to pre- dren in that project. Participants in this study
vious work with gender-typical children (Eaton belong to three different groups: (a) socially transi-
et al., 1981; Halim et al., 2014; Shutts et al., 2013; tioned binary (meaning they identify as male or
Signorella et al., 1993)—gender begins to strongly female) transgender children (henceforth, transgen-
motivate children’s preferences and behavior as well der), (b) gender-typical siblings of transgender and
as their beliefs about gender roles. Additionally, the gender nonconforming children, and (c) age- and
preschool years are when an understanding of gen- gender-matched unrelated gender-typical control
der stability is mastered (Ruble et al., 2007; Slaby & children. Because transgender children are rare, the
Frey, 1975), which is particularly interesting with research team traveled extensively to recruit this
regard to a young transgender person’s view of sample. Over the course of 9 months (March 2015
their gender identity. In addition, this is the age at to November 2015), the researchers flew and drove
which gender nonconforming children tend to throughout the United States to meet with families
express their gender-atypical identity and behavior from 17 U.S. states (see Table 1 for list of states) at
(Green, 1976; Zucker, Bradley, & Sanikhani, 1997), a series of conferences and camps for gender
and it is the age at which the first transgender chil- diverse children, at support group meetings for
dren have socially transitioned (to our knowledge). families with transgender children, at our research
laboratory (for area families), or at families’ own
homes to recruit this sample of transgender chil-
Method dren and siblings. Additionally, control participants
were run in a child development laboratory in the
Participants
Pacific Northwestern United States. Despite recruit-
This study is part of a larger longitudinal project ing from different geographic areas, our groups
on gender development, examining the longitudinal had similar other demographics as can be seen in
development of a larger sample of socially Table 1.

Table 1
Participant Family Demographics

Controls Siblings Transgender Statistic p Value

Child’s agea 5.03 (8.52) 4.93 (8.24) 4.99 (7.82) F(2, 93) = 0.15 .858
Child’s genderb
Male 22% 50% 22% v2(2) = 6.72 .035
Female 78% 50% 78%
Child’s racec
White 78% 75% 61% v2(2) = 1.87 .392
Monoracial, non-White 8% 4% 6%
More than one 14% 21% 33%
Household annual incomed
Less than $25,000 3% 0% 3% F(2, 93) = 0.46 .632
$25,001 to $50,000 0% 17% 5%
$50,001 to $75,000 19% 0% 8%
$75,001 to $125,000 36% 50% 31%
More than $125,000 42% 33% 53%
Parental political ideologye 2.35 (1.11) 1.54 (0.66) 1.64 (0.76) F(2, 93) = 8.07 .001
U.S. state of residencef WA AZ, CA, CO, IL, IN, AZ, CA, CO, DC, IL,
MA, MD, MO, NJ, IN, MA, MD, NM,
OH, OR, VT, WA, WI OR, PA, WA

Note aMean age (years) with standard deviation (months) of child participants. One-way analysis of variance was conducted on child’s
age in months. bPercentage of male-identified and female-identified children. cPercentage of child in each racial category. Chi-square
analysis was conducted with two categories (White and non-White). dPercentage of families in each income bracket. One-way analysis
of variance was conducted on a 5-point scale variable that maps onto the five options provided. eMean and standard deviation of par-
ents’ political ideology on a scale ranging from 1 (very liberal) to 7 (very conservative). fCurrent U.S. states of residence for participants in
each group.
 Transgender Preschool Children 625

transgender participants. These controls were

Transgender Children
matched on age, such that controls’ ages at test
Thirty-six 3- to 5-year-old transgender children were within 4 months of the transgender children’s
who had socially transitioned (Mage = 4.99 years, age at test, and matched on expressed gender (such
SD = 7.82 months) participated, including 28 trans- that a transgender girl—a natal male who lives as a
gender girls (natal males) and 8 transgender boys girl—was matched to a gender-typical girl), which
(natal females). Perhaps not surprisingly, as social is the same matching approach as utilized by Olson
transitions often occur later in development, our et al. (2015). Gender-typical matched controls were
sample skewed toward the older age of this range, recruited through a university database of families
with two 3-year-olds, thirteen 4-year-olds, and interested in participating in child development
twenty-one 5-year-olds participating. The transgen- research, and families were informed this was part
der children were socially transitioned at the time of a study of children with diverse gender identities
of participation, meaning they were all living as the and expressions.
gender “opposite” of their natal sex. Using the cri-
teria for full transitions from Steensma, McGuire,
Procedure and Materials
Kreukels, Beekman, and Cohen-Kettenis (2013), par-
ticipants had to be using the pronoun, clothing, and After obtaining parental consent and verbal
hairstyles associated with the “other” gender to assent, experimenters administered the following
count as socially transitioned. Every socially transi- measures. Parents simultaneously provided demo-
tioned transgender 3- to 5-year-old child who was graphic and other information. The procedures took
run during the recruitment period is reported in approximately 30 min to complete. All participants,
this article. including parents, received $10, and children addi-
tionally received a small toy prize for participation.
Siblings
Gender Constancy
In order to recruit as large of a sample of siblings
as possible, we recruited a group of children who Participants were asked questions about their
were siblings of transgender children, irrespective own and others’ gender stability and consistency to
of whether the transgender sibling was in the 3- to examine their gender constancy understanding. We
5-year-old range (only two of the current siblings chose to separate first-party and third-party gender
had one in this sample), or who were siblings of constancy because previous work on gender con-
children who were gender nonconforming (i.e., chil- stancy understanding of gender-atypical children
dren who had behaviors and preferences counter to examined constancy knowledge in this way (Zucker
gender stereotypes and who had not yet transi- et al., 1999). Furthermore, it was plausible that
tioned). Twenty-four 3- to 5-year-old gender-typical transgender children would view the constancy of
siblings of transgender or gender nonconforming their own gender differently compared to the gen-
children (Mage = 4.93 years, SD = 8.24 months; 12 der constancy of others—after all, unlike children
natal males, 12 natal females) participated during themselves, most people the children know have
the same time period as the transgender children. had a stable gender.
These children were recruited and run while attend- First-party stability. Two questions about partici-
ing the same support group meetings, conferences, pants’ own gender stability were taken from a pre-
and camps as our transgender sample and were vious work (Slaby & Frey, 1975). Participants were
recruited via the same recruitment techniques, as asked about their gender in the past (“When you
we always stated that we were interested in sibling were a little baby, were you a little boy or a little
participants as well. All siblings age 3–5 years old girl?”) and in the future (“When you grow up, will
who were run during the recruitment period are you be a dad or a mom?”). Because we were inter-
included in this article. ested in the degree to which transgender children
responded that their gender was stable from the
past compared to their responses about the stability
Controls
of their gender going into the future, we separately
Additionally, thirty-six 3- to 5-year-old gender- analyzed this measure by item. To examine
typical children (Mage = 5.03 years, SD = 8.52 whether transgender children differ from siblings
months; 8 natal males, 28 natal females) were and controls in their pattern of responding to the
recruited to participate as matched controls of the two first-party stability questions, we coded
626 Fast and Olson

whether participants responded to each question the gender corresponding to their expressed gender
with their expressed gender or with the “opposite” and 0 if they gave any other answer (e.g., opposite
gender (i.e., the gender “opposite” of natal sex for gender, both genders). These two variables (Cron-
gender-typical controls and siblings, and the gender bach’s a = .79) were then summed to create a first-
that aligns with natal sex for transgender partici- party consistency total score for each participant
pants). Participants who did not provide a response (with a possible range from 0 to 2, see Table 3 for
to an item were excluded from analyses of that means). Participants had to respond to both first-
item, leading to the exclusion of one control partici- party consistency items to be included in analyses
pant, one sibling, and two transgender participants of this measure, which resulted in the exclusion of
on the past stability item as well as one control par- one control participant and three transgender par-
ticipant and four transgender participants on the ticipants.
future stability item. Additionally, for these consistency items, chil-
Third-party stability. To measure participants’ dren were asked to provide a justification of why
understanding of others’ gender stability, four ques- they gave each response. The field is split about
tions were adapted from previously validated mea- whether or not to code justifications given by 3- to
sures (Ruble et al., 2007; Szkrybalo & Ruble, 1999). 5-year-old children—some previous researchers
Participants were shown pictures of four different code justifications to consistency questions (Arthur,
targets (a boy, a girl, a woman, and a man) one at Bigler, & Ruble, 2009; Ruble et al., 2007), but many
a time and answered one question about each tar- do not (Bussey & Bandura, 1984; Frey & Ruble,
get. When participants were shown a boy or a girl, 1992; Lobel & Menashri, 1993; Marcus & Overton,
they were asked, “When this kid was a little baby, 1978; Slaby & Frey, 1975; Warin, 2000). Because our
was this kid a boy or a girl?” and when partici- participants tended to give nonsense, “I don’t
pants were shown a man or a woman, they were know,” or no justifications (40% did so at least once
asked, “When this grown-up was little, was this during the first-party consistency task, leaving very
grown-up a boy or a girl?” Participants’ responses few responses that could be coded), and because
to the four third-party stability questions were 3- to 5-year-olds in our study and in past work gen-
coded as 1 if they responded with the gender-con- erally do not pass gender consistency measures
stant response (e.g., saying that a boy will be a anyway, we did not use these justification
man) and 0 if they gave any other answer (e.g., responses for first-party or third-party consistency
opposite gender, both genders). These four vari- measures; doing so would necessarily mean chil-
ables (Cronbach’s a = .90) were summed to create a dren would perform even worse on these items.
third-party stability total score (with a possible range Third-party consistency. To measure participants’
from 0 to 4, see Table 3 for means). Participants understanding of others’ gender consistency, four
had to respond to all four third-party stability items questions were adapted from previously validated
to be included in analyses of this measure, resulting measures (Ruble et al., 2007; Szkrybalo & Ruble,
in the exclusion of one control participant and two 1999). Participants were shown four new targets (a
transgender participants. boy, a girl, a woman, and a man) and were asked a
First-party consistency. Two questions were taken question about each target. When participants saw
from previous work (Slaby & Frey, 1975) to assess a boy or a girl, they were asked, “If this kid wore
participants’ own gender consistency. Participants [opposite gender’s] clothes, would this kid be a boy
were asked two questions: “If you wore [opposite or a girl?” and when participants saw a man or a
gender’s] clothes, would you be a boy or a girl?” woman, they were asked, “If this grown-up did the
and “If you played [opposite gender’s] games, work that [opposite gender] do, would this grown-
would you be a boy or a girl?” Critically, when up be a man or a woman?” Participants’ responses
presenting the consistency questions, we asked to the four third-party consistency questions were
about the gender “opposite” children’s expressed coded as 1 if they responded with the gender-con-
identity (e.g., a gender-typical or transgender girl stant response (e.g., saying that a boy will be a
was asked about boys’ clothes or games). In this man) and 0 if they gave any other answer (e.g.,
way, all children were asked about clothing that opposite gender, both genders). Then, these vari-
would have been less common for them to wear in ables (Cronbach’s a = .87) were summed to create a
their current everyday life, as the measure was third-party consistency total score (with a possible
originally designed to function in that way. Partici- range of 0–4, see Table 3 for means). Participants
pants’ responses to the two first-party consistency were required to respond to all four third-party
questions were coded as 1 if they responded with consistency items in order to be included in
 Transgender Preschool Children 627

analyses of this measure, which resulted in the three of the four items in each category, resulting in
exclusion of two control participants, two sibling the inclusion of all participants (three children
participants, and two transgender participants. skipped one toy item—one control, one sibling, one
transgender—and two transgender children skipped
one clothing item, but due to the averaging
Preferences
approach, these participants could nonetheless be
Participants were asked about their peer, toy, included in analyses).
and clothing preferences. For analysis purposes, these three measures were
Peer preference. Participants saw eight separate combined to create a preferences composite score.
pairs of children and were asked to point to the Because the measures were on two different scales,
child they would like to be friends with the most the peer preference (ranging from 0 to 6), toy pref-
(Olson et al., 2015). In six of the trials, the pair erence (1 to 5), and clothing preference (1 to 5)
included a male child and a female child, matched scores were first standardized into percent of maxi-
on perceived age and attractiveness, while two filler mum possibility (POMP) scores (Cohen, Cohen,
trials included two apparently male children or two Aiken, & West, 1999). The POMP scores were calcu-
apparently female children. A peer preference score lated by first subtracting the minimum possible
was calculated for each participant, representing score on the scale from the observed scores. That
the number of times on mixed-gender pair trials difference was then divided by the difference
(0–6) the participants chose peers who were the between the maximum and minimum possible
gender that matched their own expressed gender scores on the scale, which was then multiplied by
(e.g., number of times a gender-typical girl or trans- 100. Once POMP scores were calculated for each
gender girl picked girls). Participants who did not preference score (peer, toy, and clothing), they were
provide a response on every trial of the task (with averaged to create the preferences composite score.
the exception of the two filler trials) were excluded Although we do report the means on the original
from analyses of this item, which resulted in the scales in the table below, the analyses are con-
exclusion of one control (missed five of six items) ducted with the preferences composite score.
and one transgender participant (missed four of six
items).
Stereotype Flexibility
Toy and clothing preferences. Participants saw four
sets of five toys and four sets of five outfits and A task was adapted from Liben and Bigler (2002)
were asked to point to the toy they would like to to assess the degree to which participants endorsed
play with the most or which outfit they liked the flexibility about gender activity stereotypes. Partici-
best. For example, one group of toy items included pants were told that they would hear a list of activ-
an orange tool set, a red barbecue set, a board game, ities that people can do (e.g., gymnastics and video
a purple stove set, and a pink kitchen set. An exam- games; see Supporting Information for full list), and
ple set of clothing items included a pair of plaid to say who they think should do each activity: boys,
cargo shorts with athletic T-shirt, a pair of gray girls, or both boys and girls. Responses were coded
jeans with blue button-down shirt, a pair of blue into a stereotype flexibility score, which is the number
jeans with green T-shirt, a pair of blue jeans with of times each participant responded that “both boys
pink tank-top, and a purple dress with sparkles. and girls” should do an activity that was previ-
Thus, each set of five toys or outfits could be ously deemed either stereotypically male or stereo-
arranged from 1 (most stereotypically masculine) to 5 typically female. Because 5 of the 15 items were
(most stereotypically feminine). These items had previ- intended to be gender-neutral activities, those items
ously been pilot tested with a group of gender-typi- were excluded from analyses. Thus, only the 10
cal children to determine how stereotypically girl- items about gendered activities were included in
like or boy-like they were, and in the current study, participants’ stereotype flexibility scores, which rep-
the items were found to be highly reliable (toy resent the number of times participants responded
items: Cronbach’s a = .74; clothing items: Cron- with the “both” option on gendered items (ranging
bach’s a = .92). Responses were averaged to create a from 0 to 10). Some participants did not even begin
toy preference score and a clothing preference score. For the measure (one control participant, two siblings,
boys, scores were recoded such that higher numbers and two transgender participants), and thus are
represent more gender-consistent preferences (the excluded from analyses on this measure. Of the
scale was already ordered that way for girls). Analy- participants who did start this stereotype measure,
ses included participants who responded to at least five did not respond to all of the questions,
628 Fast and Olson

resulting in the additional exclusion of four trans- We calculated three scores for this similarity
gender participants and one sibling on this mea- measure. First, we created a similarity to my gender
sure. score, which is an average of the five items about
other kids with the same gender as the participant
(Cronbach’s a = .70). For example, for transgender
Explicit Gender Identity
girls and gender-typical girls, the similarity to my
Participants reported their gender identities gender score is the average of the items asking
using the explicit gender identity measure that about similarity to other girls. Next, we calculated a
Olson et al. (2015) used with elementary school- similarity to other gender score, which is an average
aged transgender children. Before answering gender of the five about other kids with the “opposite”
identity questions, participants were told that gender as the participant (Cronbach’s a = .71). For
everybody has an outside part (physical body) and transgender girls and gender-typical girls, the simi-
an inside part (mind, thought, and feelings) of larity to other gender score would be the average
them. Participants were further told that for some of the items asking about similarity to boys, for
people the outside and inside parts are the same, example. We also created a similarity difference score
and for other people they are different. For exam- by calculating the difference between the similarity
ple, a person could be a boy on the outside and feel to my gender score and the similarity to other gen-
like a boy on the inside or could be a boy on the der score. This difference score was always calcu-
outside and feel like a girl on the inside. Addition- lated in the direction of the participant’s expressed
ally, participants were told that some people feel gender (e.g., scores for transgender and gender-
like they are both, neither, or that it changes over typical girls were calculated by subtracting the
time. Finally, participants reported (a) what they “similarity to boys” from the “similarity to girls”
feel like on the inside right now, and (b) what they average).
think they will feel like on the inside when they Participants had to answer all questions included
grow up: a boy or man, a girl or woman, neither, in each composite score to be included in analyses
both, it changes over time, or I don’t know. Partici- of that score. Because this similarity measure was
pants who did not provide a response to either always the last task in the procedure, a number of
explicit identity item were excluded from analyses participants did not even begin the measure (3 con-
of that particular item (first (now) item: two trans- trol participants, 7 siblings, and 10 transgender par-
gender participants and two siblings; second ticipants), and thus are excluded from analyses on
(grown up) item: one control participant, two trans- this measure. Of the participants who did start this
gender participants, and two siblings). similarity measure, three missed at least one ques-
tion contributing to each composite score, resulting
in the additional exclusion of one control partici-
Similarity
pant, one sibling, and one transgender participant
Participants completed a task developed by Mar- from analyses of all scores for this measure. Finally,
tin, Andrews, England, Zosuls, and Ruble (2016), one additional transgender participant was
measuring how similar children think they are to excluded from analyses of only the similarity to other
boys and girls. Participants answered 10 questions gender and similarity difference scores for not answer-
(5 for similarity to boys and 5 for similarity to girls) ing one of the questions about similarity to children
and responded on a scale ranging from 0 (very dif- of the other gender.
ferent) to 4 (very similar). More specifically, partici-
pants were asked, “How similar do you feel to
Outfit at Appointment
boys[girls]?” “How much do you act like boys[-
girls]?” “How much do you look like boys[girls]?” To measure participants’ gender expression in
“How much do you like to do the same thing as everyday life, without telling parents or children in
boys[girls]?” and “How much do you like to spend advance, two experimenters independently rated
time with boys[girls]?” The response scale included the outfit worn by each participant at the testing
a visual representation of each option to help par- session on a scale ranging from 1 to 5 (allowing for
ticipants understand the possible responses. This half-point ratings) with lower numbers representing
visual representation displayed circles labeled more stereotypical boy outfits and higher numbers
“You” and “Boys”[“Girls”], with the circles varying representing more stereotypical girl outfits (r = .94,
in the degree of overlap or separation, mapping on p < .001). However, in some cases (n = 13) only one
to the degree of (dis)similarity to other kids. experimenter was able to provide an outfit rating,
 Transgender Preschool Children 629

and in those cases we just used the one experi- participants in the three groups differed signifi-
menter’s rating (unfortunately, due to experimenter cantly in their tendency to say their expressed gen-
error, for three participants—one transgender par- der in response to the question about their past
ticipant and two siblings—the experimenter did not gender, v2(2) = 57.32, p < .001, φ = .789; however,
indicate a rating, thus those three were excluded participants in the three groups were no different in
from analyses for this measure). Experimenters their tendency to say their expressed gender in
were told that the most masculine outfits consisted response to the question about their future gender,
of clothing items such as male-stereotypic sports v2(2) = .081, p = .960, φ = .030. In response to the
attire, superhero costumes, and men’s formal wear, question about their gender as a baby, only 21% of
whereas the most feminine outfits consisted of frilly transgender participants said their expressed gen-
dresses or skirts, princess costumes, and sparkly der, whereas 97% of controls and 96% of siblings
accessories. Experimenters also considered the col- said their expressed gender. On the other hand,
ors (e.g., pink) and style (e.g., fitted vs. baggy shirt) when asked about their gender as an adult, 97% of
when determining outfit ratings. transgender participants, 97% of controls, and 96%
of siblings replied with their expressed gender (see
Table 2). To best understand this result, imagine a
Other Measures
child like Jazz from the Introduction—a natal boy
Importantly, and in the spirit of transparency who identifies as a girl. If she was the modal partic-
(Simmons, Nelson, & Simonsohn, 2011), we note ipant in our study, she would have said she was a
that these measures were given as part of a larger boy as a baby, but will be a woman as a grown-up.
study about gender development and mental health
among gender diverse children. Therefore, during
First-Party Consistency
this time period we did collect data on two mea-
sures not reported here. First, 72% of the children A one-way ANOVA on first-party consistency total
in this study completed a measure of gender essen- scores indicated that participants in the three
tialism. However, that measure was intended for a groups did not differ in the degree to which they
article in progress on essentialism, which also believed their gender would remain consistent
includes participants within a larger age range (i.e., across situational changes, F(2, 89) = 0.96, p = .389,
children who are older and are not included in the g2p ¼ 0:02 (see Table 3 for proportion of participants
current article). Second, we added a new measure in each group who gave consistent responses).
—on gender encoding—part way through this Within each group, children’s responses did not dif-
study; however, this measure was only completed fer from chance responding: transgender, t(32) =
by 18% of our participants and as such, will be .96, p = .344, d = 0.17; siblings, t(23) = .46, p = .647,
reported in a separate article. Furthermore, while d = 0.09; controls, t(34) = .90, p = .377, d = 0.15.
the current participants were completing these mea- Thus, consistent with past research examining pre-
sures, their parents completed a variety of measures school-age children, irrespective of whether they
(e.g., mental health), but as those measures were were transgender or not, children did not systemat-
not relevant to the present article (which is focused ically believe gender was consistent across changes
on children’s own behaviors, beliefs, and attitudes), in appearance.
they have been excluded from this article as well.

Table 2
Proportion of Participants Responding to First-Party Stability Items
Results With Expressed Gender
Effect sizes for analyses of variance (ANOVAs) and Proportion N v2(2) p Value
post hoc comparisons were calculated using a
spreadsheet created by Lakens (2013). First-party stability past
Controls .97 35 57.32 < .001
Siblings .96 23
Gender Constancy Transgender .21 34
First-party stability future
First-Party Stability Controls .97 35 0.81 .960
Siblings .96 24
Chi-square analyses on responses to the first- Transgender .97 32
party stability questions indicated that the
630 Fast and Olson

Table 3 F(2, 87) = 2.69, p = .073, g2p ¼ 0:06. However, if

Means and Proportion of Participants “Passing” Constancy Items anything, the mean scores were higher (indicating
M SD N Statistic p Value
greater belief in consistency) among the sibling and
transgender groups compared to the control group
Third-party stability (see Table 3 for proportion of participants in each
Controls 3.89 0.47 35 F(2, 90) = 5.36 .006 group who gave consistency-relevant responses).
Siblings 3.17 1.49 24 Responses from control participants did not differ
Transgender 2.94 1.56 34 from chance responding, t(33) = .76, p = .454,
First-party consistency d = 0.13; however, transgender participants were
Controls 0.86 0.94 35 F(2, 89) = 0.96 .389
marginally more likely than chance to say that
Siblings 1.08 0.88 24
others’ gender is consistent, t(33) = 2.00, p = .054,
Transgender 1.15 0.91 33
Third-party consistency
d = 0.38, as well as siblings, t(21) = 2.06, p = .052,
Controls 1.76 1.81 34 F(2, 87) = 2.69 .073 d = 0.44. In sum, transgender children and their sib-
Siblings 2.68 1.56 22 lings, but not control participants, trended toward
Transgender 2.53 1.54 34 believing that another person’s gender was likely to
be consistent across changes in appearance.
Note Third-party stability total scores and third-party consistency
total scores have a possible range of 0–4. Possible range for first-
party consistency total scores is 0–2. Preferences
We conducted a one-way ANOVA on the prefer-
ences composite score to test whether transgender,
Third-Party Stability
sibling, and control participants differed in the
A one-way ANOVA on third-party stability total degree to which they prefer same-gender peers,
scores revealed that the three groups differed in the toys, and clothing (see Table 4 for means on origi-
degree to which they believed other people’s gen- nal scales and preferences composite). Participants in
der would remain stable over time, F(2, 90) = 5.36, the three groups did not differ in the degree to
p = .006, g2p ¼ 0:11. Tukey’s honestly significant dif- which they prefer same-gender peers and items,
ferent tests indicate that transgender participants F(2, 91) = 2.21, p = .116, g2p ¼ 0:05. In all groups,
were significantly less likely to say that others’ gen- children were significantly more likely than chance
der is stable over time compared to control partici- to prefer peers and items in the direction of their
pants, p = .006, d = 0.83, but they were not own gender: transgender, t(34) = 15.53, p < .001,
different from siblings, p = .775, d = 0.15. Siblings d = 2.62; siblings, t(23) = 4.68, p < .001, d = 0.96;
were marginally different from control participants controls, t(34) = 10.43, p < .001, d = 1.76.
in the degree to which they endorse gender stability
in others, p = .079, d = 0.71 (see Table 3 for the pro-
Stereotyping
portion of participants in each group who gave
stable responses). Overall, all groups were signifi- A one-way ANOVA on gender stereotype flexibil-
cantly more likely than chance to believe that gen- ity scores (ranging from 0 to 10, see Table 4 for
der would be stable: transgender, t(33) = 3.53, means) suggests that the transgender, sibling, and
p = .001, d = 0.60; siblings, t(23) = 3.83, p = .001, control participant groups did not differ on how
d = 0.78; controls, t(34) = 23.69, p < .001, d = 4.0. much they endorse flexibility of gender stereotypes,
Thus, while all groups generally believed that other F(2, 83) = 0.77, p = .464, g2p ¼ 0:02. In all groups,
people’s gender was typically stable across time, participants were significantly more likely than
transgender children and to a lesser extent, the sib- chance (3.33) to say that both boys and girls should
lings of transgender and gender nonconforming do the gendered activities: transgender, t(29) = 5.52,
children, responded that occasionally another p < .001, d = 1.0; siblings, t(20) = 3.86, p = .001,
child’s gender could change across their life span. d = 0.84; controls, t(34) = 4.75, p < .001, d = 0.80.
Figure 1 displays responses broken down by partic-
ipant group and type of stereotype item.
Third-Party Consistency
A one-way ANOVA on third-party consistency Gender Identity
total scores revealed a marginal difference between
groups in participants’ tendency to say that others’ In terms of gender identity, we categorically
gender is consistent across situational changes, coded participants’ responses, such that participants
 Transgender Preschool Children 631

Table 4 it changes over time, or I don’t know). Chi-square

Preference, Stereotyping, Similarity, and Outfit Means by Group analyses revealed that transgender, sibling, and
M SD N
control participants did not differ in their likelihood
of responding with their expressed identity when
Peer preferencea asked about both their current gender identity,
Controls 4.94 1.63 35 v2(4) = 3.19, p = .526 φ = .186, and future gender
Siblings 4.29 1.90 24 identity, v2(4) = 3.77, p = .438, φ = .204 (see Table 5
Transgender 5.09 1.36 35 for a summary of participant responses to both the
Toy preferenceb current and future gender identity questions).
Controls 3.96 0.66 36
Siblings 3.67 0.94 24
Transgender 3.55 0.85 36 Similarity
Clothing preferenceb
Controls 4.49 0.88 36
The transgender, sibling, and control groups also
Siblings 4.14 1.10 24 not differ on how similar they felt to children of
Transgender 4.65 0.59 36 their same gender, F(2, 70) = 0.15, p = .861,
Stereotype flexibilityc g2p ¼ 0:004, one-way ANOVA, and how similar
Controls 5.54 2.76 35 they felt to children of the other gender, F(2, 69) =
Siblings 6.29 3.51 21 0.26, p = .768, g2p ¼ 0:007, one-way ANOVA (see
Transgender 6.43 3.08 30 Table 4 for means). The three groups also did
Similarity to my genderd not differ on how similar they feel to their own
Controls 3.08 0.87 32 gender versus the other gender, F(2, 69) = 0.04,
Siblings 3.10 0.87 16
p = .966, g2p ¼ 0:001, one-way ANOVA. All groups
Transgender 2.97 0.94 25
tended to see themselves as similar to their own
Similarity to other genderd
Controls 1.13 0.96 32
gender more than to the other gender: transgender,
Siblings 1.20 1.11 16 t(23) = 6.60, p < .001, d = 1.35; siblings, t(15) = 5.70,
Transgender 0.99 0.81 24 p < .001, d = 1.43; controls, t(31) = 8.06, p < .001,
Similarity differencee d = 1.42.
Controls 1.95 1.37 32
Siblings 1.90 1.33 16
Transgender 2.02 1.50 24
Outfit at Appointment
Outfit at appointmentf A one-way ANOVA on outfit ratings (ranging
Controls 4.14 0.59 36 from 1 to 5, see Table 4 for means) indicated that
Siblings 4.13 0.66 22
transgender, sibling, and control participants did
Transgender 4.25 0.63 35
not differ in the degree to which they wore outfits
Note aPeer preference scores range from 0 to 6, with higher num- stereotypically consistent with their expressed gen-
bers representing greater preference other children of one’s own der to the appointment, F(2, 90) = 0.39, p = .680,
gender. bToy and clothing preference scores range from 1 to 5, g2p ¼ 0:009. Children from all groups, on average,
with higher numbers representing more gender-typed prefer-
ences. cStereotype flexibility scores range from 0 to 10, indicating wore clothing that is stereotypically associated with
the number of times children said “both” in response to the their gender: transgender, t(34) = 11.74, p < .001,
stereotypic activities, such that higher numbers representing d = 1.98; siblings, t(21) = 8.02, p < .001, d = 1.71;
greater flexibility. dSimilarity to my gender and similarity to
other gender scores range from 1 to 5, representing the degree to controls, t(35) = 11.64, p < .001, d = 1.94.
which participants think they are similar to children of the same
gender as them or the other gender, respectively. eSimilarity dif-
ference scores represent the degree to which participants think
they are similar to children of their own gender compared to Discussion
children of the other gender (similarity to my gendersimilarity
to other gender). fOutfit at appointment scores range from 1 to 5, Across all measures of preference, behavior, stereo-
with higher numbers representing more gender-typed outfits.
typing, and identity, if coded according to chil-
dren’s expressed gender, preschool-age socially
transitioned transgender children never significantly
could either respond with (a) their expressed gen- differed from their gender-matched peers (age- and
der, (b) the “opposite” of their expressed gender gender-matched controls and preschool-age siblings
(opposite of natal sex for controls and siblings and of transgender or gender nonconforming children),
same as natal sex for transgender participants), or mirroring a previous finding about preferences and
(c) one of the additional options (i.e., neither, both, identity with older socially transitioned transgender
632 Fast and Olson

100%
90%
Mean Percentage of Responses 80%
70%
60%
50% Girl
40% Boy
30% Both
20%
10%
0%
Control Sibling Trans Control Sibling Trans Control Sibling Trans
Boy Stereotypes Girl Stereotypes Neutral

Figure 1. Mean percentage of responses to stereotype items by participant group and item type.

Table 5 However, in terms of children’s responses to

Explicit Gender Identity Response Percentages by Group the gender constancy measures, the results were
fairly mixed. Transgender children differed from
Controls Siblings Transgender
the other children in that they tended to say that
(%) (%) (%)
they were a different gender as an infant than
Current gender identity their current gender in everyday life. However,
Expressed gender 67 55 73 they were just as likely as both control groups to
“Opposite” of expressed 8 5 6 say that their gender in adulthood would be con-
gender gruent with their current gender. This pattern of
Anything else 25 40 21 responding to the gender stability items may
Future gender identity reflect the way that transgender children’s families
Expressed gender 66 50 68
often talk about their gender—that everyone
“Opposite” of expressed 6 5 0
believed them to be one gender (the one associ-
gender
Anything else 28 45 32
ated with their sex), but now and in the future
they have a different gender. Whether the family
Note The “anything else” response category includes the discussion reflects, leads to, or merely co-occurs
responses “neither,” “both,” “it changes over time,” and “I don’t with this responding is currently unclear; however,
know.”
this pattern of responding is notably different from
the way many binary-identified (male or female)
transgender adults often discuss their gender—as
children (Olson et al., 2015). That is, young trans- always existing in this one way. One possible
gender children were just as likely as gender-typical explanation for this difference is that children are
children to (a) show preferences for peers, toys, and interpreting the statement about their identity as
clothing culturally associated with their expressed an infant as being a question about sex, rather
gender, (b) dress in a stereotypically gendered out- than gender, or that they are responding by con-
fit, (c) endorse flexibility in gender stereotypes, and sidering how other people treated them rather
(d) say they are more similar to children of their than how they felt. If older transgender children,
gender than to children of the other gender. Trans- particularly those approaching puberty, are better
gender children were also just as likely as controls able to interpret these gender stability items due
and siblings to say that they identify with their to stronger awareness of the distinction between
expressed gender, both now and in the future, sex and gender or better separating their own
when given multiple other choices. These findings beliefs about their gender from the beliefs others
suggest that, in many ways, the basic gender devel- had about their gender, perhaps they would be
opment of socially transgender children is quite more likely than younger transgender children to
similar to that of other children. say that their gender has been stable across their
 Transgender Preschool Children 633

whole life. Future work can investigate this ques- behaviors, preferences, stereotypes, and real-life
tion, as well as assess whether preschool-age trans- clothing choices to that of gender-typical children.
gender children change the way they think about The largest difference in responses from the partici-
their gender as a young child based on how the pant groups was in the domain of constancy, where
question is presented (i.e., framed in terms of gen- transgender children were less likely than the gen-
der or sex). Transgender children did not differ der-typical groups to believe that their own gender
from the control groups in thinking about the con- will remain stable from infancy to adulthood. The
sistency of their gender identity across superficial fact that transgender children had atypical
changes (e.g., clothing), though none of the groups responses on gender constancy measures, but typi-
showed anywhere near ceiling level performance cal responding on measures of gender preference,
on these items. behavior, and stereotyping, sheds some doubt on
Interestingly, transgender children were less Kohlberg’s (1966) claim that full gender constancy
likely to see other people’s gender as stable over time understanding enhances gendered beliefs, prefer-
compared to gender-typical controls. Although this ences, and behaviors, as well as more recent claims
finding could at first be seen as support for the that gender stability understanding in particular can
claim that transgender children have quite a differ- boost same-gender preferences and behavior gender
ent understanding of gender than their gender-typi- typing (Halim et al., 2014; Ruble et al., 2007).
cal peers, the fact that transgender children did not Although the development of gender stability
differ from siblings on their third-party stability understanding seems to co-occur with increases in
responding suggests instead that this effect may be gender typing for gender-typical children, the cur-
the result of knowledge that gender is not stable rent work suggests that this knowledge is not nec-
over time for some people. Importantly, children in essary, nor does it appear to be a central
all three groups generally believed that gender contributor to strong same-gender preferences and
would be stable across development for most peo- behavior (for more discussion and evidence on this
ple, meaning that even the transgender children point, see the associated Supporting Information).
and siblings made this assumption. The difference Of course the best test of this causal question
was that the transgender and sibling groups would be to conduct a longitudinal study of even
seemed to assume that occasionally there is an indi- younger children, a test we hope researchers will
vidual for whom this is not the case. Future conduct in the future.
research could examine whether knowledge of Kohlberg (1966) and more recent theorists (Halim
transgender people is causally related to this pat- et al., 2014; Ruble et al., 2007) did not distinguish
tern of responding by teaching a group of gender- between past and future identity when discussing
typical children about transgender children and children’s knowledge of gender stability. However,
then later assessing their third-party gender stabil- nearly all children in our transgender sample
ity. With regard to third-party consistency, we believed their current expressed gender will be their
found that transgender children were actually more gender in the future, responded with gendered
likely than the control participants to respond that preferences, endorsed gender stereotypes, and
other people’s gender was consistent in identity chose gendered clothing to the same degree as gen-
across clothing and hairstyle changes; though the der-typical children. Therefore, perhaps it is reason-
difference was not quite significant, transgender able to consider an amendment to this cognitive
children differed from chance responding, while theory of gender development, such that children’s
controls did not. This finding could reflect transgen- sense of gender stability from the current moment
der children’s knowledge of the fact that gender into the future (with the removed assumption that
identity can exist irrespective of what one wears gender is necessarily consistent with a child’s sex at
since the children personally experienced a time birth) is what motivates or contributes to especially
during which they wore clothes of a gender that strong gendered preferences and behavior, rather
did not match their gender identity. However, than the belief that past gender must be stable.
given the fact that this difference between groups Again, a longitudinal study would best answer the
was not significant and that the sample size was causal component of this question. In addition,
small, we hesitate to draw a particularly strong given the current debate about the degree to which
conclusion on this point. gender identity is stable in transgender children
Taken together, the current work suggests that (Olson, 2016; Soh, 2016; Steensma, Biemond, de
preschool-age transgender children display similar Boer, & Cohen-Kettenis, 2011; Steensma et al., 2013;
patterns of gendered responding in terms of their Vilain & Bailey, 2015; Zucker & Bradley, 1995), it
634 Fast and Olson

would be interesting to connect children’s reasoning “insides” and “outsides” was intended to make it
about gender stability and their actual later life clear that we wanted to know how they felt on the
identity. inside (in their mind), however, some may have
interpreted this as a question about their body (in-
side their clothes), interpreting this question differ-
Limitations and Future Work
ently than intended.
As with all work, but especially exploratory Relatedly, we may have also encountered an
work like the present project, there are considerable issue with interpretation in our gender constancy
limitations in interpretation, and suggestions for measure. Unlike the gender identity measure, we
improvements to make in further work. First, with did not attempt to specify whether we were refer-
regard to our measures, we have some concerns ring to gender and sex when asking about past ver-
with the gender identity measure, developed by sus future identity (i.e., stability) and identity upon
Olson et al. (2015), for use with preschool-age chil- situational changes (i.e., consistency). Thus, our
dren. Rather than utilize traditional questions about results on these measures could reflect the fact that
gender identity (e.g., are you a boy or a girl?), we young children may not understand sex and gender
opted to provide children with a range of possible as distinct, and further, that we did not provide
responses (both, neither, it changes over time, I adequate explanation for this distinction. For exam-
don’t know) that do not conform to a binary view ple, the difference between transgender and gender-
of gender identity. Unexpectedly, a very large num- typical children’s responses on the gender stability
ber of siblings and controls responded with one of measure might have occurred because in order to
the nonbinary options. One interpretation of this answer “correctly,” transgender children would
result is that children have more diverse identities have to understand the distinction between sex and
than we typically assume them to have. We are gender, but gender-typical children would not. That
skeptical of this interpretation because (a) we doubt said, it is also worth noting that some transgender
that such a large number of children, especially in children at this age experience considerable body
the control groups, have nonbinary identities given dysphoria. This experience might suggest that
that older children do not express these identities young transgender children actually have a repre-
as much (Olson et al., 2015), and (b) anecdotally, sentation of the distinction between sex and gender
sometimes the very same children who gave nonbi- (i.e., they know that more female-identified people
nary answers on these items clearly identified them- do not have penises, hence their dislike of their
selves as a member of a binary gender group penises). Furthermore, these findings may suggest
(always the one they lived in during everyday life) that transgender children, and to a lesser extent
at other times in the visit (something we unfortu- their siblings, have a deeper or more complex
nately did not regularly assess). Furthermore, these understanding of sex and gender than other chil-
findings are at odds with the results of the similar- dren. A more systematic examination of transgen-
ity measure which showed that children generally der children’s representation of this distinction
identified with their gender group. Thus, another should be conducted in future work in order to
possibility is that these younger children—who begin to disentangle these issues.
could not read the response options—forgot the These findings open up questions regarding not
options that were listed first (boy, girl) and only how transgender children interpret the items
responded with the options that were listed later for these particular measures or how they represent
(the nonbinary ones) or simply liked the idea of sex versus gender, but also more broadly whether
providing a nontraditional answer. Perhaps future children at this age understand the distinction
work could focus on just three options (e.g., boy, between gender and sex. Although there have been
girl, something else) to allow for nonbinary selec- some research programs aimed at addressing these
tions, but maintain a lower number of options to questions (Bem, 1989; Volbert, 2000), future work
reduce demand. Finally, this gender identity mea- might capitalize on the existence of transgender
sure was designed to inquire about children’s gen- children to validate these measures, for example.
der identity rather than their sex. Thus, the items Despite these limitations to some of the measures
were worded such that children were asked about utilized in the present study, in general, most mea-
what they are on the “inside” following an experi- sures appeared to be interpretable and consistent
menter’s statement that the “inside” is our mind, with past research for control participants, giving
thoughts, and feelings, as opposed to the “outside,” us greater confidence in their interpretation. The
which is our body. This information about considerable similarity across our three samples on
 Transgender Preschool Children 635

many of these basic gender development measures representing more stereotypic responses for child’s
also suggests that the earliest years of rearing—the gender) and because the sibling group never dif-
time during which transgender children were raised fered from both of the other groups, it is unlikely
as a gender “opposite” that of the one they cur- that the gender breakdown of groups influenced
rently live as—may have minimal impact on many responding on the measures.
of our measures, such as preferences, stereotypes, A third, more general limitation of the current
and behavior. Instead, perhaps these constructs work is that the children included in this study are
develop for the first time during the preschool quite unique, and therefore, generalizing from these
years, a time at which the transgender children in results must be met with caution. Within the
this study have already socially transitioned. Exam- greater population, very few children are transgen-
ining children who transition later would be incred- der, fewer socially transition (by the preschool
ibly interesting. Alternatively, perhaps these years), and fewer still sign up for research studies.
constructs develop even earlier and are based on Therefore, it remains an open question how widely
some of the same underlying mechanisms that lead these results will generalize. Moreover, the small
transgender children to identify as the “opposite” sample size of the current work is another general
gender in the first place. Future empirical and con- concern that limits our ability to observe significant
ceptual work in this area is needed. differences between groups. One the other hand,
A second limitation of the current work is that, although most measures of basic gender develop-
although the gender- and age-matched controls ment revealed no differences between transgender
were not different from the transgender children on and other children, considering the number of tests
many participant and family demographic mea- conducted, it is important to take caution when
sures, the groups did differ on parent political ide- interpreting the significant effects. Conducting
ology. More specifically, the parents of the many statistical tests increased the possibility of
transgender and sibling children reported being Type I errors, making replication especially impor-
more politically liberal than the (also very liberal) tant.
parents of the control children. On one hand, it is
possible that more liberal parents are more likely to
Conclusions
(a) have transgender or gender nonconforming chil-
dren, (b) allow their children to socially transition, As a society and as psychological researchers, we
or (c) sign up to participate in research. Alterna- are increasingly aware that there are individuals
tively, having a transgender or gender nonconform- who identify, early in development, as a gender
ing child may make a parent more liberal (in fact, other than the one aligned with their sex at birth.
we have had parents anecdotally report this while Therefore, such children should be included in
filling out the political orientation item). However, work on basic gender development, as in the cur-
despite differences between our groups on parental rent work, in order to both expand our knowledge
political orientation, we found remarkably few dif- of gender developmental experiences and
ferences between groups on our other measures, strengthen our (widely debated) theories of gender
suggesting that while political orientation could development. In that way, as psychologists, we can
influence social transitioning or identifying a child lead not only the field, but also broader societal cul-
as transgender, it is unlikely (at least within the lib- ture, in understanding that gender and sex identi-
eral range) to hugely influence the measures ties are much broader than have traditionally been
assessed here. Our participant groups differed sig- studied.
nificantly on one additional demographic variable
—the gender breakdown of each group. Critically,
our control participants were matched to transgen- References
der participants based on gender. For example, a Arthur, A. E., Bigler, R. S., & Ruble, D. N. (2009). An
transgender girl (natal male) was matched with a experimental test of the effects of gender constancy on
gender-typical girl. Thus, importantly, there was no sex typing. Journal of Experimental Child Psychology, 104,
difference in gender of children in our control and 427–446. doi:10.1016/j.jecp.2009.08.002
transgender participant groups, which were made Bem, S. (1989). Genital knowledge and gender constancy
up of 80% girls and 20% boys. However, the sibling in preschool children. Child Development, 60, 649–662.
group was 48% girls and 52% boys. Importantly, doi:10.2307/1130730
because our variables were coded with respect to Bussey, K., & Bandura, A. (1984). Influence of gender
each child’s gender (i.e., higher numbers constancy and social power on sex-linked modeling.
636 Fast and Olson

Journal of Personality and Social Psychology, 47, 1292– what we aim to learn. Human Development, 56, 285–290.
1302. doi:10.1037/0022-3514.47.6.1292 doi:10.1159/000355235
Clark, T. C., Lucassen, M. F., Bullen, P., Denny, S. J., Kang, S. K., & Bodenhausen, G. V. (2015). Multiple identi-
Fleming, T. M., Robinson, E. M., & Rossen, F. V. (2014). ties in social perception and interaction: Challenges and
The health and well-being of transgender high school opportunities. Annual Review of Psychology, 66, 547–574.
students: Results from the New Zealand adolescent doi:10.1146/annurev-psych-010814-015025
health survey (Youth’12). Journal of Adolescent Health, Kohlberg, L. (1966). A cognitive-developmental analysis
55, 93–99. doi:10.1016/j.jadohealth.2013.11.008 of children’s sex-role concepts and attitudes. In E. E.
Cohen, P., Cohen, J., Aiken, L. S., & West, S. G. (1999). Maccoby (Ed.), The development of sex differences (pp. 82–
The problem of units and the circumstance for POMP. 173). Stanford, CA: Stanford University Press.
Multivariate Behavioral Research, 34, 315–346. doi:10. Lakens, D. (2013). Calculating and reporting effect sizes
1207/S15327906MBR3403_2 to facilitate cumulative science: A practical primer for
Cohen-Kettenis, P. T., Owen, A., Kaijser, V. G., Bradley, t-tests and ANOVAs. Frontiers in Psychology, 4, 863.
S. J., & Zucker, K. J. (2003). Demographic characteris- doi:10.3389/fpsyg.2013.00863
tics, social competence, and behavior problems in chil- Levy, G. D., & Haaf, R. A. (1994). Detection of gender-
dren with gender identity disorder: A cross-national, related categories by 10-month-old infants. Infant Behav-
cross-clinic comparative analysis. Journal of Abnormal ior and Development, 17, 457–459. doi:10.1016/0163-6383
Child Psychology, 31, 41–53. doi:0091-0627/03/0200- (94)90037-X
0041/0 Lewis, M., & Brooks-Gunn, J. (1979). Social cognition and
Dunham, Y., & Olson, K. R. (2016). Beyond discrete cate- the acquisition of self. New York, NY: Plenum Press.
gories: Studying multiracial, intersex, and transgender Liben, L. S., & Bigler, R. S. (2002). The developmental
children will strengthen basic developmental science. course of gender differentiation: Conceptualizing, mea-
Journal of Cognition and Development, 17, 642–665. suring, and evaluating constructs and pathways. Mono-
doi:10.1080/15248372.2016.1195388 graphs of the Society for Research in Child Development, 67
Eaton, W. O., Von Bargen, D., & Keats, J. G. (1981). Gen- (Serial No. 269), 1–183.
der understanding and dimensions of preschooler toy Lobel, T. E., & Menashri, J. (1993). Relations of concep-
choice: Sex stereotype versus activity level. Canadian tions of gender-role transgressions and gender con-
Journal of Behavioural Science/Revue Canadienne Des stancy to gender-typed toy preferences. Developmental
Sciences Du Comportement, 13, 203–2009. doi:10.1037/ Psychology, 29, 150. doi:10.1037/0012-1649.29.1.150
h0081181 Maccoby, E. (1998). The two sexes: Growing up apart,
Ehrensaft, D. (2011). Gender born, gender made: Raising coming together. Cambridge, MA: Harvard University
healthy gender-nonconforming children. New York, NY: Press.
Workman. Marcus, D. E., & Overton, W. F. (1978). The development
Fenson, L., Dale, P. S., Reznick, J. S., Bates, E., Thal, D. J., of cognitive gender constancy and sex role preferences.
& Pethick, S. J. (1994). Variability in early communica- Child Development, 49, 434–44. doi:10.2307/1128708
tive development. Monographs of the Society for Research Martin, C. L., Andrews, N. C., England, D. E., Zosuls, K.,
in Child Development, 59(Serial No. 242), 1–173. doi:10. & Ruble, D. N. (2016). A dual identity approach for
2307/1166093 conceptualizing and measuring children’s gender iden-
Frey, K. S., & Ruble, D. N. (1992). Gender constancy and tity. Child Development, 88, 167–182. doi:10.1111/cdev.
the “cost” of sex-typed behavior: A test of the conflict 12568
hypothesis. Developmental Psychology, 28, 714–721. Martin, C. L., & Fabes, R. A. (2001). The stability and con-
doi:10.1037/0012-1649.28.4.714 sequences of young children’s same-sex peer interac-
Goldberg, A. B., & Adriano, J. (2007, April 27). “I’m a girl” : tions. Developmental Psychology, 37, 431–446. doi:10.
Understanding transgender children. Retrieved from http:// 1037//0012-1649.37.3.431
abcnews.go.com/2020/story?id=3088298&page=1 Martin, C. L., Fabes, R. A., Evans, S. M., & Wyman, H.
Green, R. (1976). One-hundred ten feminine and mascu- (1999). Social cognition on the playground: Children’s
line boys: Behavioral contrasts and demographic simi- beliefs about playing with girls versus boys and their
larities. Archives of Sexual Behavior, 5, 425–446. doi:10. relations to sex segregated play. Journal of Social and
1007/BF01541335 Personal Relationships, 16, 751–771. doi:10.1177/0265
Halim, M. L., Ruble, D. N., Tamis-LeMonda, C. S., 407599166005
Zosuls, K. M., Lurye, L. E., & Greulich, F. K. (2014). Miller, C. F., Lurye, L. E., Zosuls, K. M., & Ruble, D. N.
Pink frilly dresses and the avoidance of all things (2009). Accessibility of gender stereotype domains:
“girly”: Children’s appearance rigidity and cognitive Developmental and gender differences in children. Sex
theories of gender development. Developmental Psychol- Roles, 60, 870–881. doi:10.1007/s11199-009-9584-x
ogy, 50, 1091–1101. doi:10.1037/a0034906 Olson, K. R. (2016). Prepubescent transgender children:
Hidalgo, M. A., Ehrensaft, D., Tishelman, A. C., Clark, L. What we do and do not know. Journal of the American
F., Garofalo, R., Rosenthal, S. M., . . . Olson, J. (2013). Academy of Child and Adolescent Psychiatry, 55, 155.e3–
The gender affirmative model: What we know and 156.e3. doi:10.1016/j.jaac.2015.11.015
 Transgender Preschool Children 637

Olson, K. R., Key, A. C., & Eaton, N. R. (2015). Gender Clinical Child Psychology and Psychiatry, 16, 499–516.
cognition in transgender children. Psychological Science, doi:10.1177/1359104510378303
26, 467–474. doi:10.1177/0956797614568156 Steensma, T. D., McGuire, J. K., Kreukels, B. P., Beekman,
Poulin-Dubois, D., Serbin, L. A., Kenyon, B., & Der- A. J., & Cohen-Kettenis, P. T. (2013). Factors associated
byshire, A. (1994). Infants’ intermodal knowledge about with desistence and persistence of childhood gender
gender. Developmental Psychology, 30, 436–442. doi:10. dysphoria: a quantitative follow-up study. Journal of the
1037/0012-1649.30.3.436 American Academy of Child & Adolescent Psychiatry, 52,
Quinn, P. C., Yahr, J., Kuhn, A., Slater, A. M., & Pascalis, 582–590. doi:10.1016/j.jaac.2013.03.016
O. (2002). Representation of the gender of human faces Stennes, L. M., Burch, M. M., Sen, M. G., & Bauer, P. J.
by infants: A preference for female. Perception, 31, (2005). A longitudinal study of gendered vocabulary and
1109–1121. doi:10.1068/p3331 communicative action in young children. Developmental
Ruble, D. N., Martin, C. L., & Berenbaum, S. A. (2006). Psychology, 41, 75–88. doi:10.1037/0012-1649.41.1.75
Gender development. In W. Damon & R. M. Lerner (Ser- Szkrybalo, J., & Ruble, D. N. (1999). “God made me a
ies Eds.) & N. Eisenberg (Vol. Ed.), Handbook of child psy- girl”: Sex-category constancy judgments and explana-
chology. Vol. 3: Social, emotional, and personality tions revisited. Developmental Psychology, 35, 392–402.
development (6th ed., pp. 858–932). New York, NY: Wiley. doi:10.1037/0012-1649.35.2.392
Ruble, D. N., Taylor, L. J., Cyphers, L., Greulich, F. K., Vilain, E., & Bailey, J. M. (2015, May 21). What should
Lurye, L. E., & Shrout, P. E. (2007). The role of gender you do if you son says he’s a girl? Los Angeles Times.
constancy in early gender development. Child Develop- Retrieved from http://www.latimes.com/opinion/op-
ment, 78, 1121–1136. Retrieved from http://www.jstor. ed/la-oe-vilain-transgender-parents-20150521-story.html
org/stable/4620693 Volbert, R. (2000). Sexual knowledge of preschool chil-
Serbin, L. A., Poulin-Dubois, D., Colburne, K. A., Sen, M. dren. Journal of Psychology & Human Sexuality, 12, 5–26.
G., & Eichstedt, J. A. (2001). Gender stereotyping in doi:10.1300/J056v12n01_02
infancy: Visual preferences for and knowledge of gen- Warin, J. (2000). The attainment of self-consistency through
der-stereotyped toys in the second year. International gender in young children. Sex Roles, 42, 209–231. doi:10.
Journal of Behavioral Development, 25, 7–15. doi:10.1080/ 1023/A:1007039222998
01650250042000078 Zosuls, K. M., Ruble, D. N., Tamis-LeMonda, C. S.,
Serbin, L. A., Poulin-Dubois, D., & Eichstedt, J. A. (2002). Shrout, P. E., Bornstein, M. H., & Greulich, F. K. (2009).
Infants’ response to gender-inconsistent events. Infancy, The acquisition of gender labels in infancy: Implications
3, 531–542. doi:10.1207/S15327078IN0304_07 for gender-typed play. Developmental Psychology, 45,
Shelton, J. N. (2000). A reconceptualization of how we 688–701. doi:10.1037/a0014053
study issues of racial prejudice. Personality and Social Zucker, K. J., & Bradley, S. J. (1995). Gender identity disor-
Psychology Review, 4, 374–390. doi:10.1207/ der and psychosexual problems in children and adolescents.
S15327957PSPR0404_6 New York, NY: Guilford.
Shutts, K., Pemberton, C. K., & Spelke, E. S. (2013). Zucker, K. J., Bradley, S. J., Doering, R. W., & Lozinski, J. A.
Children’s use of social categories in thinking about (1985). Sex-typed behavior in cross-gender-identified
people and social relationships. Journal of Cognition and children: Stability and change at a one-year follow-up.
Development, 14, 35–62. doi:10.1080/15248372.2011. Journal of the American Academy of Child Psychiatry, 24,
638686 710–719. doi:10.1016/S0002-7138(10)60114-8
Signorella, M. L., Bigler, R. S., & Liben, L. S. (1993). Zucker, K. J., Bradley, S. J., Kuksis, M., Pecore, K., Birken-
Developmental differences in children’s gender sche- feld-Adams, A., Doering, R. W., . . . Wild, J. (1999).
mata about others: A meta-analytic review. Developmen- Gender constancy judgments in children with gender
tal Review, 13, 147–183. doi:10.1006/drev.1993.1007 identity disorder: Evidence for a developmental lag.
Simmons, J. P., Nelson, L. D., & Simonsohn, U. (2011). Archives of Sexual Behavior, 28, 475–502. doi:10.1023/A:
False-positive psychology: Undisclosed flexibility in 1018713115866
data collection and analysis allows presenting anything Zucker, K. J., Bradley, S. J., & Sanikhani, M. (1997). Sex dif-
as significant. Psychological Science, 22, 1359–1366. ferences in referral rates of children with gender identity
doi:10.1177/0956797611417632 disorder: Some hypotheses. Journal of Abnormal Child Psy-
Slaby, R. G., & Frey, K. S. (1975). Development of gender chology, 25, 217–227. doi:10.1023/A:1025748032640
constancy and selective attention to same-sex models.
Child Development, 46, 849–856. doi:10.2307/1128389
Soh, D. W. (2016, January 4). The transgender battle line: Supporting Information
Childhood. The Wall Street Journal. Retrieved from
http://www.wsj.com/articles/the-transgender-battle- Additional supporting information may be found in
line-childhood-1451952794 the online version of this article at the publisher’s
Steensma, T. D., Biemond, R., de Boer, F., & Cohen-Kette- website:
nis, P. T. (2011). Desisting and persisting gender dys- Data S1. Correlation Table and Stereotyping
phoria after childhood: A qualitative follow-up study. Measure Activities.
This document is a scanned copy of a printed document. No warranty is given about the
accuracy of the copy. Users should refer to the original published version of the material.