Gendered Peer Involvement in Girls with Congenital Adrenal Hyperplasia: Effects of Prenatal Androgens, Gendered Activities, and Gender Cognitions

Abstract

A key question in understanding gender development concerns the origins of sex segregation. Children’s tendencies to interact with same-sex others have been hypothesized to result from gender identity and cognitions, behavioral compatibility, and personal characteristics. We examined whether prenatal androgen exposure was related to time spent with boys and girls, and how that gendered peer involvement was related to sex-typed activities and gender identity and cognitions. We studied 54 girls with congenital adrenal hyperplasia (CAH) aged 10–13 years varying in degree of prenatal androgen exposure: 40 girls with classical CAH (C-CAH) exposed to high prenatal androgens and 14 girls with non-classical CAH (NC-CAH) exposed to low, female-typical, prenatal androgens. Home interviews and questionnaires provided assessments of gendered activity interests and participation, gender identity, and gender cognitions. Daily phone calls over seven days assessed time spent in gendered activities and with peers. Girls with both C-CAH and NC-CAH interacted more with girls than with boys, with no significant group differences. The groups did not differ significantly in gender identity or gender cognitions, but girls with C-CAH spent more time in male-typed activities and less time in female-typed activities than did girls with NC-CAH. Time spent with girls reflected direct effects of gender identity/cognitions and gender-typed activities, and an indirect effect of prenatal androgens (CAH type) through gender-typed activities. Our results extend findings that prenatal androgens differentially affect gendered characteristics and that gendered peer interactions reflect combined effects of behavioral compatibility and feelings and cognitions about gender. The study also shows the value of natural experiments for testing hypotheses about gender development.

INTRODUCTION

Why do children segregate by sex? Children’s affinity for interaction with same-sex others has been well-documented (e.g., Fabes, Shepard, Guthrie, & Martin, 1997; Maccoby & Jacklin, 1987), and the consequences of sex segregation for psychological development (e.g., for social skills, interests) have been widely discussed (e.g., Leaper, 1994; Maccoby, 1998).

Explanations of Sex Segregation

There is considerable interest in the causes of sex segregation, with focus on two primary (but not mutually exclusive) contributors: behavioral compatibility and gender cognitions.

The Role of Behavioral Compatibility

The most common explanation for sex segregation involves behavioral compatibility; particular attention has been given to the role of gendered play style and activities, such as toy play, with additional consideration of gendered personal characteristics, such as aggression (e.g., Maccoby & Jacklin, 1987; Martin & Fabes, 2001; Mehta & Strough, 2009). It can be difficult to disentangle gendered behavior as a cause versus a consequence of sex segregation, but shared behaviors have been seen to both result from and drive peer selection. On the one hand, frequency of play with same-sex partners longitudinally predicted increases in gendered behaviors, including personal characteristics (e.g., aggression) and gendered (rough-and-tumble) play (Martin & Fabes, 2001). On the other hand, children selected partners partly on the basis of gender-typed activities (Martin et al., 2013).

Peer selection might also be facilitated by personal characteristics that prompt play with one sex over another, particularly those that reflect the different play styles of boys’ and girls’ groups. Compared to girls’ groups, boys’ groups tend to be larger, less structured, more peer than adult-directed, less verbal, more physically rough and active, competitive rather than cooperative, more likely to involve dominance hierarchies, and less likely to be concerned with group harmony (reviewed in Berenbaum, Martin, Hanish, Briggs, & Fabes, 2008; Leaper, 1994). Thus, sex differences in personal characteristics (e.g., aggression, dominance, activity level, empathy, and verbal skills) and desire for compatible playmates may lead children to segregate by sex. Sex differences in play styles of boys’ and girls’ groups may also influence children’s peer choice beyond personal compatibility. For example, assertive girls may play with other girls to avoid boys’ dominance attempts (Powlishta & Maccoby, 1990). Studies linking children’s gendered characteristics to their time with same- and other-sex peers are limited, and most focus on characteristics directly involved in play, such as activity level and rough play. We know little about the role of global gendered personal characteristics, such as sensitivity or empathy (for a discussion of the issues and data, see Martin, Fabes, Hanish, Leonard, & Dinella, 2011).

The Role of Gender Cognitions

Moving beyond explanations of sex segregation based on behavioral compatibility, Martin et al. (2011) focused on the importance of gender cognitions, that is, gender identity and subsequent perceptions about similarity to one’s sex. They proposed a comprehensive model of sex segregation that incorporates both experienced similarity (behavior) and expected similarity (gender cognitions), and that fits within a broader framework of gender schema and other constructivist theories. Their model enables explicit tests of those theories regarding the role of gender cognitions in behavior. Martin et al. then tested their model in a study of a predominantly non-Hispanic white sample in the United States, examining the relative contributions to observed sex-typed peer interactions of behavior (rough play, aggression, activity level) and gender cognitions (perceived similarity to boys and to girls). Observed peer interactions were correlated as expected with cognitions and with some aspects of behavior. Direct comparison of contributors via regression analyses highlighted the importance of gender cognitions: Observed frequency of interaction with boys relative to girls was predicted by perceived similarity to boys relative to girls beyond effects of rough play; the latter was a significant predictor of peer interaction frequency only for boys, perhaps because of low rates of rough play in girls.

Subsequent work validated the importance of gender cognitions for peer interactions, extending the evidence to an additional aspect of social cognition, ingroup bias, and to another country (Italy) (Gasparini, Sette, Baumgartner, Martin, & Fabes, 2015). Cognitions were linked to the frequency of dyadic play in girls but not boys (although both sexes showed gendered dyadic play and expected biases for same-sex others): Global liking for girls was associated positively with frequency of dyadic play with girls; both global liking and positive attributions for girls were associated negatively with dyadic play with boys.

Congenital Adrenal Hyperplasia as a Natural Experiment

A unique opportunity to further our understanding of contributors to sex segregation is provided by girls with classic congenital adrenal hyperplasia (C-CAH). Because of a genetic mutation leading to an enzymatic deficiency, these girls are exposed to excess androgens during prenatal development. They are reared as girls and the overwhelming majority identify as girls, but they show some behavioral masculinization (discussed below, and reviewed in Berenbaum & Beltz, 2016). They provide information about prenatal androgen effects on peer involvement, and about the behavioral paths by which androgens might act to influence that involvement.

Girls with C-CAH are usually compared to typical girls (often their unaffected female relatives), but another valuable comparison is girls with a mild form of CAH, non-classical CAH (NC-CAH), which results in exposure to high levels of androgens only after birth and is not associated with behavioral masculinization (Hall et al., 2004; Nordenström, Servin, Bohlin, Larsson, & Wedell, 2002).¹ These girls thus provide a comparison for other factors related to CAH (e.g., having a chronic disease).

Peer Preferences in Girls with Congenital Adrenal Hyperplasia

Previous studies of peer preferences in girls with C-CAH suggest that they are more likely than typical girls to prefer boy playmates, but there are several factors to consider in interpreting the results. First, the difference between girls with and without C-CAH is not large, and many prefer girl playmates (Berenbaum & Snyder, 1995; Ehrhardt & Baker, 1974; Hines & Kaufman, 1994; Pasterski et al., 2011; Servin, Nordenström, Larsson, & Bohlin, 2003); this contrasts with the very large sex difference (d = 3–4), and no overlap between typical boys and girls (Berenbaum et al., 2008; Martin & Fabes, 2001). Second, measures in several previous studies consisted of a single general item. This includes an item asked of parents by an interviewer not blind to the child’s condition (Ehrhardt & Baker, 1974); a question asked of children about the names of their most frequent playmates (with follow-up to clarify the sex of the playmates) with the possibility that playmates could be adults or relatives (Hines & Kaufman, 1994); and a single question about best friends, with the sex recorded (Servin et al., 2003). Third, no previous study directly captured everyday peer interactions; questions were about preferences (Berenbaum & Snyder, 1995; Pasterski et al., 2011) or global reports of peer experiences (Ehrhardt & Baker, 1974; Hines & Kaufman, 1994; Servin et al., 2003).

Gender Development in Girls with Congenital Adrenal Hyperplasia

Studies of other aspects of early gender development in girls with C-CAH show that early androgens have differential effects on psychological characteristics hypothesized to contribute to peer interactions. Girls with C-CAH have gender-atypical activity interests and participation; compared to girls without C-CAH (including unaffected sisters and typical girls in the population), they show greater interest and participation in boy-typed activities, and lower interest and participation in girl-typed activities (e.g., Berenbaum & Hines, 1992; Berenbaum & Snyder, 1995; Pasterski et al., 2005; Servin et al., 2003). The differences between girls with and without C-CAH are sometimes as large as the sex differences, with little overlap between girls with severe C-CAH and control girls. Girls with C-CAH are gender-atypical, on average, on some personal characteristics; for example, girls with C-CAH were reported by their mothers to be more aggressive and active than their unaffected sisters (Pasterski et al., 2007). In contrast, the vast majority of girls with C-CAH identify as girls, although they may have reduced degree of identification compared to unaffected girls (e.g., Berenbaum & Bailey, 2003; Dessens, Slijper, & Drop, 2005; Pasterski et al., 2015). There has been little study of their gender-related cognitions, however, and results are inconsistent: In one study, girls with C-CAH (aged 10–13) had typical gender attitudes (Endendijk, Beltz, McHale, Bryk, & Berenbaum, 2016); in another study, girls with C-CAH (aged 4–11) differed from unaffected female relatives in reduced verbal and behavioral preferences for objects labelled to be for girls (Hines et al., 2016).

Understanding Sex Segregation by Studying Girls with Congenital Adrenal Hyperplasia

The differential effects of prenatal androgens revealed in studies of girls with C-CAH create an opportunity to examine both androgen influences on peer interactions and behavioral contributors to them, particularly the relative role of activities (which are masculinized to a considerable degree in girls with C-CAH) versus gender identity and cognitions (which are relatively female-typical in girls with C-CAH). If peer selection is driven by behavioral compatibility for activities, then girls with C-CAH should interact more with boys than do typical girls; if peer selection is driven by gender identity and cognitions, then they should interact primarily with girls.

In a study providing a partial test of these possibilities, Pasterski et al. (2011) used the Playmate and Play Style Preferences Structured Interview (Alexander & Hines, 1994; Fridell, Owen-Anderson, Johnson, Bradley, & Zucker, 2006) to examine the relative importance to playmate choice of gendered play style (toy or activity) and playmate sex (called gender label) in children aged 3 to 10 years with and without C-CAH. Children were shown pictures that contained a stick figure representing a person who was male, female, or gender neutral (denoted by appearance and clothing), above a toy or activity that was male-typed or female-typed. In the nonconflict condition (as determined by the match of playmate sex and play style), unaffected boys and girls preferred own-sex playmates and sex-typical play styles, but girls with C-CAH preferred boy playmates and male-typed play styles. In the conflict condition (own-sex playmate paired with other-sex activity or other-sex playmate paired with own-sex activity), boys were most likely to choose playmates based on activity rather than apparent sex, unaffected girls chose playmates using playmate sex and play style about equally, and girls with C-CAH chose playmates with male-typed play styles. The general conclusion was that androgens contribute to sex differences in playmate selection, through preferences for play styles. The results are intriguing, but there are several reasons that results might not reflect what happens in real-world social interactions: Test stimuli did not depict playmates engaging in activities or toy play (they were shown below the stick figures); playmate sex (a stick figure) was less salient than play style (realistic picture of the activity or toy); and observational data show that children spend the majority of their time with same-sex others, regardless of the activity with which they are engaged (Berenbaum et al., 2008; Martin & Fabes, 2001), in contrast to their preference for play style over playmate sex in this study. Furthermore, the contribution of gender identity and cognitions to playmate selection was not tested.

Current Study

We studied the nature and correlates of gendered peer interactions in girls with C-CAH and NC-CAH, using data from daily phone dairies of girls’ time use and from global self reports of gendered characteristics collected during home interviews and in questionnaires. Our study is novel in methods and participants. We used multiple methods, including a cued recall approach to collecting daily diary reports of time spent with peers in an effort to reduce memory demands and self-concept biases that can characterize global reports of daily activities (e.g., McHale, Crouter, & Tucker, 1999). We also compared girls with CAH who differed in amount and timing of androgen exposure.

We had two study goals. First, we examined how early androgens influence time spent with peers of both sexes, by testing whether girls with C-CAH and NC-CAH differed in time spent with boys and with girls. Because girls with C-CAH are exposed to excess levels of prenatal androgens, and girls with NC-CAH are exposed to female-typical levels of prenatal androgens, differences between these groups of girls reflect effects of prenatal androgens. Previous work suggests several possibilities. If peer involvement is based on behavioral compatibility related to gendered activities, then girls with C-CAH should spend more time with boys than should girls with NC-CAH, as a result of their strong interest and engagement in boy-typical activities (as discussed above). If peer involvement is based on identity and cognitions related to being a girl, then girls with C-CAH should spend time primarily with girls, consistent with their female-typical identity, and typical gender-role attitudes (as discussed above). If peer involvement is based on a combination of behavior and gender cognitions, then the peer interactions of girls with C-CAH will depend on the degree to which behavioral compatibility versus identity and cognitions are important. They might have fewer interactions with both boys and girls than would girls with NC-CAH because their identity and cognitions might initially lead them to spend time with other girls, but lack of shared interests might ultimately limit their time together. Alternatively, they might interact more with both boys and girls, seeking out boys whose activity interests are compatible, and girls for their similar gender identity and cognitions. Or, they might have increased interactions with boys and decreased interactions with girls, but to a lesser extent than if behavioral compatibility was the sole determinant of peer interactions.

Second, we examined specifically how time spent with girls and boys reflected the relative roles of behavior and gender-related cognitions, and whether those influences were indirect effects of prenatal androgens. With respect to behavioral compatibility, we focused on activities because they show large effects of both sex and prenatal androgens (reviewed in Blakemore, Berenbaum, & Liben, 2009) and have been invoked as drivers of sex segregation (e.g., Martin & Fabes, 2001; Martin et al., 2013; Mehta & Strough, 2009). We tested whether, across girls with C- and NC-CAH, time spent with boys versus girls was associated with interest and engagement in gender-atypical activities, and with nonstereotypical gender identity and cognitions, as well as whether those characteristics mediated effects of prenatal androgen exposure (CAH type) on time with peers. In exploratory analyses, we also considered whether gendered peer involvement was associated with gendered characteristics beyond activities, specifically the gendered personality qualities of expressivity and instrumentality.

METHOD

Participants

Details of recruitment and sample characteristics are provided elsewhere (Endendijk et al., 2016).² In brief, girls with CAH were recruited through family support groups and pediatric endocrinologists. Deidentified information provided by the support group suggests that about 80–85% of eligible families participated in the study.

Participants were told that the goal was to learn about the development of characteristics related to being male or female (gender development), that we planned to study girls with CAH aged 10–13 to learn about the girls’ interests and activities, thoughts and feelings about being female, and family relationships, and that the study would help us to understand how gender development is shaped by hormones and family relationships. They received a description of the study methods. Informed consent was provided by both parents and girls.

The sample consisted of 54 girls with CAH due to 21-hydroxylase deficiency, aged 10 to 13 years (M = 11.90, SD = 1.16). Most girls came from middle-class families. Racial background was primarily European American (N = 50), with a small representation of other races (1 girl was African American, 1 was Asian, 2 were multiracial); ethnic background was primarily non-Hispanic, but two girls were Hispanic. Type of CAH (classical or non-classical) was determined using mother report or genotype when mother report was not available (88% agreement between mother report and genotype); most girls with C-CAH had the severe, salt-losing, form of CAH. Girls with C-CAH (N = 40) and NC-CAH (N = 14) were not significantly different in age, parental education, or family income (ps > .10).

Measures

During home interviews lasting approximately three hours, we collected girls’ reports of their gendered activity interests and participation, gendered personal characteristics, gender identity, and gender role attitudes. At the end of the home interview, girls received questionnaires to complete and return by mail; these questionnaires provided additional data on gender cognitions. Some of these data have been reported previously with respect to group comparisons and links among global reports of activities, identity, and attitudes (Endendijk et al., 2016); these measures were used here as predictors of time spent with male and female peers.

Over the two-to-four week period following the home interviews, seven nightly phone interviews were conducted (five weekday and two weekend nights) about a half hour before bedtime. During each phone call, girls were guided through a list (provided during the home interview) of 93 non-school activities, and asked which activities they had engaged in that day, from the time they woke up that morning until the time of the phone call. For each activity they engaged in, girls reported who else was present and the duration (in minutes) of the activity. Activity companions were distinguished by sex and roles/relationships (e.g., father, sister, male peer, unrelated female adult). We then aggregated reports of minutes spent in individual activities and in particular social contexts, across all activities and all seven calls, to create the measures of interest described below. We used this cued recall procedure, focused on specific activities (e.g., “How many times did you wash dishes today?”) rather than global activities (e.g., “How many times did you do chores today?”) because it is the optimal means of promoting memory (e.g., McHale et al., 1999; McHale, Kim, Whiteman, & Crouter, 2004a; McHale, Shanahan, Updegraff, Crouter, & Booth, 2004b).

Measures used for this study are summarized in Table 1. They are described below.

Table 1.

Measures

Construct & Measure	Source of Data	References
Peer Involvement
% time spent with boys	Daily phone calls	McHale et al. (1999, 2004a, 2004b)
% time spent with girls	Daily phone calls	McHale et al. (1999, 2004a, 2004b)
Gendered Activities
% time in male-typed activities	Daily phone calls	McHale et al. (1999, 2004a, 2004b)
% time in female-typed activities	Daily phone calls	McHale et al. (1999, 2004a, 2004b)
Interest, male-typed activities	Interview questionnaire	Endendijk et al. (2016)
Interest, female-typed activities	Interview questionnaire	Endendijk et al. (2016)
Participation, male-typed activities	Interview questionnaire	Endendijk et al. (2016)
Participation, female-typed activities	Interview questionnaire	Endendijk et al. (2016)
Gender Identity/Cognitions
Core gender identity	Interview questionnaire	Berenbaum & Bailey (2003)
Gender typicality	Interview questionnaire	Egan & Perry (2001)
Gender contentedness	Interview questionnaire	Egan & Perry (2001)
Intergroup bias	Interview questionnaire	Egan & Perry (2001)
Felt pressure for gender conformity	Interview questionnaire	Egan & Perry (2001)
Gender role attitudes	Interview questionnaire	Antill et al. (1996)
Gender centrality	Interview questionnaire	Ruble et al. (2007)
Gender evaluation	Interview questionnaire	Ruble et al. (2007)
Gendered Personal Characteristics
Instrumentality	Interview questionnaire	Antill et al. (1993)
Expressivity	Interview questionnaire	Antill et al. (1993)

Peer Involvement

Peer involvement was indexed by the proportion of time spent with male peers only and with female peers only (no mixed-sex peer groups, adults or child relatives such as siblings were present) relative to total time spent in activities with social partners. That is, we calculated the number of minutes reported across the seven phone calls that girls spent with boys (no one else present) and with girls (no one else present), and divided each sum by total time spent in activities with others. We used proportional time scores because they were less skewed than absolute time scores, but the proportional and absolute time scores were highly correlated (rs > .90), and results were similar with the two types of scores. In line with the work of others (Martin et al., 2011), we also calculated the difference between the proportion of time spent with boys and the proportion of time with girls to assess tendency to spend time with opposite- versus same-sex peers.

Gender-typed Activities

We used several measures of gendered activities, capturing interest and engagement in a set of activities that show consistent sex differences and that were assessed with both questionnaires and daily phone calls. These activities are listed in the Appendix, classified as male-typed and female-typed, as determined from previous data (references below).

Time in gendered activities

As noted, during each phone call, girls reported on their time spent in a variety of activities, including leisure activities, household tasks, and homework (e.g., McHale et al., 1999, 2004a, 2004b). We focused on 30 activities that were also assessed in global self-report measures of activity interests and participation (see Appendix).

Time in gendered activities was indexed by the proportion of time spent in male-typed and in female-typed activities relative to time in all activities (gendered and nongendered). We calculated the number of minutes across seven days spent in male-typed and in female-typed activities, and divided each by total time in all activities. Again, proportional time scores were less skewed than absolute time scores, but the two types scores were highly correlated and produced similar results.

Global reports of gender-typed activity interests and participation

Reports of activity interests and participation were obtained during the home interview. Girls rated their level of interest (using a 1 to 4 scale) in each of 50 household and leisure activities, as well as their frequency of participation (0 = never, 1 = sometimes, 2 = always) in 78 activities derived from other measures (Berenbaum, 1999; McHale et al., 1999; Rekers, Sanders, Rasbury, Strauss, & Morey, 1989). We focused on 23 interest items and 30 participation items (see Appendix) that show consistent sex differences in typical children and were endorsed by at least 10% of girls in this sample. Scores on male- and female-typed interests and participation scales were created using previous data from typical children and recent work on gender-typed activities (Berenbaum, 1999; Egan & Perry, 2001; Martin & Dinella, 2012; McHale et al., 1999; McHale, Kim, Dotterer, Crouter, & Booth, 2009; Ohannessian, 2009) and corrected for overall response set (tendency to endorse items) by subtracting a participant’s average response (mean on all items) from the mean gender-typed score. Internal consistency reliabilities (alpha coefficients) ranged from .64 to .76. Results of group comparisons were described elsewhere (Endendijk et al., 2016); these data were used here as predictors of time spent with peers.

Gender Identity and Cognitions

Core gender identity

Gender identity was assessed during the home interview with 10 questions about comfort with (versus distress over) being a girl (Berenbaum & Bailey, 2003). Each item was scored for typicality on a scale from 0 (female-typical identity) to 2 (male-typical identity) and scores were summed to produce a measure of gender identity that ranged from 0 (complete female-typical identity) to 20 (complete male-typical identity). Internal consistency reliability was .67; the higher reliability of .76 in the Berenbaum and Bailey (2003) study likely reflects the additional variation provided by the group of tomboys in that study. Some results on this measure were reported elsewhere (Endendijk et al., 2016); the measure was used here as a predictor of time spent with peers.

Gender identity

Typicality, contentedness, bias, and pressure. During the home interview, participants answered 30 questions assessing four dimensions of gender identity (Egan & Perry, 2001). Girls rated each item as to whether it was true or not true of them and then their degree of endorsement (sort of or very); responses were coded from 1 (not very true) to 4 (very true), and high scores indicated high levels of the characteristic. Gender typicality reflects a girl’s sense of herself as a typical girl. Gender contentedness measures satisfaction with gender assignment. Intergroup bias assesses attribution of positive versus negative qualities to girls versus boys as groups. Felt pressure reflects the feeling of pressure for gender conformity. Internal consistencies ranged from .55 to .85. These four subscales were used to compare girls with C-CAH and NC-CAH, and as predictors of gendered peer involvement. One girl with C-CAH did not complete this measure.

Gender role attitudes

Attitudes were assessed with a modified version of the Children’s Attitudes Toward Women Scale (Antill, Cotton, Russell, & Goodnow, 1996; Crouter, Whiteman, McHale, & Osgood, 2007). Girls rated each of 19 statements on a scale from 1 to 4 (really untrue, sort of untrue, sort of true, really true); high mean scores reflect traditional gender role attitudes, with reliability of .81. A description of the measure and some results were reported elsewhere (Endendijk et al., 2016); the measure was used here as a predictor of time spent with peers.

Gender centrality and evaluation

This measure assessed aspects of identification with one’s gender group, that is, centrality of gender and personal evaluation of gender; it was developed by Ruble et al. (2007) based on other research on social identity (e.g., Luhtanen & Crocker, 1992). In a questionnaire completed at home after the interview, girls rated how well each of 12 statements described them, using a scale from 1 (does not describe you) to 5 (describes you exactly); high scores indicate high levels of centrality and evaluation. Internal consistency reliabilities were .91 for evaluation and .79 for centrality. Data were missing for 9 girls (8 C-CAH, 1 NC-CAH) who did not return the questionnaire. Data on the evaluation subscale have been published (Endendijk et al., 2016); the measure was used here as a predictor of time spent with peers. Data on the centrality subscale were used in comparisons of girls with C-CAH and NC-CAH, and as a predictor of time spent with peers.

Gendered personal characteristics

Expressivity and Instrumentality. The 12-item Antill Trait Questionnaire (Antill, Russell, Goodnow, & Cotton, 1993) was used to measure the female-typed trait of expressivity (e.g., sensitivity, kindness), and the male-typed trait of instrumentality (e.g., independence, competitiveness). For each item, girls rated how often they exhibited particular characteristics, using a scale from 1 (almost never) to 5 (almost always). Scores for each scale reflect averages of six items, with high scores signifying high levels of expressivity and instrumentality; reliabilities were .73 and 50. Data on both scales were used in exploratory analyses to compare girls with C-CAH and NC-CAH, and as predictors of gendered peer involvement.

Composite Scores of Activities and Identity/Cognitions

In order to increase measurement sensitivity and minimize statistical errors (Type I and II), we constructed composite scores for activities and identity/cognitions from the measures described above. The composites were created by averaging standardized scores for individual measures shown in Table 1, adjusting the signs as needed so that all measures were scored in the same direction. For activities, we constructed composites of male-typed activities (three items, α = .70), female-typed activities (three items, α = .61), and the difference between them to reflect involvement in and preference for gender-atypical activities (six items, α = .73), paralleling other work using difference scores to assess stereotypy of activities (McHale et al., 1999). For identity/cognitions, we considered using separate composites of identity and cognitions, but several scales appear to measure both; for example, the multidimensional measure of identity (Egan & Perry, 2001) contains subscales considered by others to reflect cognitions (intergroup bias), and the centrality scale might be considered to reflect identity. The justification for this eight-item scale as a measure of a single construct is provided by its internal consistency reliability (α = .70); supplementary analyses (item correlations, minimal changes to alpha when measures are omitted, and factor analyses) confirmed that all scales were similarly good measures of the construct.

Analysis Plan

Analyses were structured to meet the goals of the study (and parallel in part previous work in typical children that examined predictors of gendered peer interaction (Martin et al., 2011)): assessing effects of early androgens on gendered peer involvement, and activities and identity/cognitions as predictors of gendered peer involvement. First, to assess early androgen effects, we compared girls with C-CAH and NC-CAH on gendered peer involvement, and on hypothesized predictors (activities, identity/cognitions, gendered personal characteristics). We used two-tailed t-tests for peer involvement (as discussed earlier, the groups could reasonably be expected to differ in either direction, and previous studies showing greater interest in boys as playmates were based on preference and not time spent with peers). We used one-tailed t-tests for hypothesized predictors of peer involvement (girls with C-CAH were expected to have scores on all characteristics that were more male-typed or less traditional than those of girls with NC-CAH, although differences on identity have been small and differences on cognition not always found).

Second, to examine contributors to gendered peer involvement, we used correlation and regression analyses. In an initial step, we examined correlations across the groups (pooling girls with C-CAH and NC-CAH³) between the criterion measures (proportional time spent with boys, girls, and the difference between them) and the hypothesized predictors (male-typed and female-typed activities, and the difference between them; identity/cognitions). The primary analysis was a parallel multiple mediation model to predict gendered peer involvement from degree of prenatal androgen exposure, reflected in CAH type (predictor) and from composite scores of activities and identity/cognitions (parallel mediators). This model considers the influence of all variables of interest, testing whether degree of prenatal androgen exposure (CAH type) directly influences gendered peer involvement, and whether effects on gendered peer involvement of activities and identity/cognitions are direct (independent of prenatal androgens), indirect (a function of prenatal androgens), or both. We also tested reverse mediation models because the study was not longitudinal (although the outcome was assessed a few weeks later than most measures of the mediators), examining whether gender peer involvement could instead be a mediator of the link between CAH type and activities or identity/cognitions.

All models used ordinary least squares regression to estimate the direct effects of a predictor and multiple mediators on an outcome, as well as bias-corrected confidence intervals to determine the indirect effects of the predictor on the outcome through the mediators. The models were estimated using the PROCESS macro in SPSS (Hayes, 2013) with 90% bias-corrected confidence intervals with 5,000 bootstrap estimates, and Type 1 error of .05, with one-tailed tests for directional hypotheses concerning mediators. This type of resampling method is the current standard for mediation analyses; it is superior to traditional approaches because it reduces Type II errors and provides parameter estimates of the indirect effect (MacKinnon & Fairchild, 2009).

Preliminary analyses informed the main analyses. We correlated age with gendered peer involvement, using two-tailed tests because we had no hypotheses about age effects in this narrow range; any significant age effects would need to be considered in regression (mediation) analyses. We also examined correlations between predictors (composites of activities and identity/cognitions) to aid in interpretation of the mediation analyses.

RESULTS

Comparisons of Girls with C-CAH and NC-CAH

Our first question concerned androgen effects on gendered peer involvement and on characteristics that might contribute to that involvement. There were no significant differences between girls with C-CAH and NC-CAH in time spent with boys or girls, as shown in Table 2. Regardless of androgen exposure, girls spent most of their time with other girls and little time with boys. (Although we conducted two-tailed tests as planned, we note that differences were also nonsignificant with one-tailed tests.)

Table 2.

Descriptive Statistics and Group Comparisons for Gendered Peer Involvement, Activities, Identity, and Cognitions

Measure (scale or absolute range in parentheses) a	C-CAH N = 32–40 b M (SD)	NC-CAH N = 13–14 b M (SD)	d	t(df)
Peer Involvement
Proportion of time spent with
Boys only (0–1)	.01 (.04)	.04 (.12)	−.34	−1.39 (52)
Girls only (0–1)	.10 (.13)	.14 (.14)	−.30	−1.00 (52)
Boys – Girls	−.08 (.14)	−.09 (.21)	.06	.25 (52)
Gendered Activities
Proportion of time spent in
Male-typed activities (0–1)	.14 (.09)	.08 (.08)	.70	2.23 (52) *
Female-typed activities (0–1)	.10 (.08)	.15 (.09)	−.59	−1.83 (52) *
Activities composites
Male-typed activities (z-score)	.18 (.80)	−.51 (.52)	1.02	3.01 (52) **
Female-typed activities (z-score)	−.12 (.73)	.34 (.72)	−.63	−2.00 (52) *
Male-typed – female- typed activities	.30 (1.31)	−.85 (.90)	1.02	3.02 (52) **
Gender Identity/Cognitions
Gender typicality (1–4)	2.63 (.84)	2.60 (1.04)	.03	.13 (51)
Gender contentedness (1–4)	2.74 (.66)	2.54 (.63)	.31	.98 (51)
Felt pressure (1–4)	1.72 (.44)	1.69 (.43)	.07	.16 (51)
Ingroup bias (1–4)	2.98 (.49)	2.73 (.57)	.47	1.55 (51)
Gender centrality (1–5)	3.48 (.93)	3.45 (.75)	.03	.11 (43)
Gender id/cognitions composite (z-score)	.02 (.59)	−.09 (.43)	.21	.62 (42)
Gendered Personal Characteristics
Instrumentality (1–5)	3.65 (.62)	3.60 (.51)	.09	.28 (52)
Expressivity (1–5)	3.55 (.60)	3.88 (.60)	−.55	−1.80 (52) *

^aHigh scores on individual measures and activity composites reflect high level of characteristic stated. High scores on composite of gender identity/cognitions reflect female-typicality/gender-traditionality.

^bMissing data for measures of gender identity (1 girl with C-CAH), and gender centrality and evaluation (8 girls with C-CAH, 1 girl with NC-CAH)

Group difference significant, one-tailed t-test,

^*p < .05,

^**p < .01

Comparisons between girls with C-CAH and NC-CAH on gendered activities confirmed expectations. Compared to girls with NC-CAH, girls with C-CAH were less gendered in their activities, as shown in Table 2: they spent a significantly greater proportion of time in male-typed activities, and a significantly lower proportion of time in female-typed activities. These findings extend previously described androgen effects on self-reported activity interests and participation in this sample (Endendijk et al., 2016), by demonstrating androgen effects on daily reports of time use. Not surprisingly, analyses of the composite measures of gendered activities, based on new time use data and published self-report data, revealed substantial and significant differences between girls with C-CAH and NC-CAH, as shown in Table 2.

In contrast, girls with C-CAH were not significantly different from girls with NC-CAH in gender identity or cognitions, either on individual measures or on the composite score, as shown in Table 2. These findings extend others using this sample, which showed that girls with C-CAH and NC-CAH have female-typical core identity, gender-role attitudes, and attitudes about being a girl (Endendijk et al., 2016).

Exploratory analyses of personal characteristics, shown in Table 2, revealed that girls with C-CAH had moderately lower scores than did girls with NC-CAH on expressivity, but the groups were not significantly different on scores of instrumentality. These scales were also examined in relation to time spent with peers, as reported below.

Predictors of Time Spent with Peers

Our second question concerned predictors of time spent with boys and with girls. Of particular interest was the relative strength of prediction from gendered activities versus gender identity and cognitions, and whether such effects were direct or indirect due to prenatal androgens; this was tested via a parallel multiple mediation model. As noted, we initially examined correlations of gendered peer involvement with composite scores of activities and identity/cognitions.

Preliminary Analyses

Age was not significantly correlated with time spent with boys or girls, all rs(52) < .20, ps > .10, two-tailed. Therefore, age was not considered in subsequent analyses. Composite scores of activities and identity/cognition were not significantly correlated with each other: Correlation of identity/cognition with male-typed activities, r(42) = −.12; with female-typed activities, r(42) = .06; with the difference between male-typed and female-typed activities, r(42) = −.10.

Correlations of Gendered Peer Involvement with Activities and Identity/Cognitions

Correlations between peer involvement (proportions of time in activities spent with boys and with girls) and both sets of predictor characteristics (gendered activities and identity/cognitions), in the combined sample of girls with C-CAH and NC-CAH, are shown in Table 3.⁴ Time spent with girls was associated with both identity/cognitions and activities, particularly male-typed activities. Time spent with boys was not significantly associated with either set of characteristics, perhaps reflecting the limited time with boys among this group of girls. Exploratory analyses of links between peer involvement and personal characteristics showed that neither expressivity nor instrumentality were significantly associated with time spent with girls. Across all characteristics, correlations between proportion of time spent with boys versus girls (that is, the difference score) were similar to those with time spent with girls, but in the opposite direction due to the scaling of the variables.

Table 3.

Correlations of Gendered Peer Involvement with Gendered Activities, Identity/Cognitions, and Personal Characteristics

	Proportion of time spent with:
	Boys	Girls	Boys – Girls
Activities composites (N=54) a
Male-typed	.02	−.38 **	.33 **
Female-typed	−.05	.13	−.13
Male-typed – female-typed	.04	−.30 *	.28 *
Identity/cognitions composite (N=44) a b	−.07	.32 *	−.29 *
Personal characteristics (N=54) a
Instrumentality	.03	−.07	.08
Expressivity	.11	.20	−.13

^aHigh scores on activity composites and personal characteristics reflect high level of characteristic stated. High scores on composite of gender identity/cognitions reflect female-typicality/gender-traditionality.

^bMissing data for measures of gender identity (1 girl with C-CAH), and gender centrality and evaluation (8 girls with C-CAH, 1 girl with NC-CAH).

Correlation significantly different from 0, one-tailed test,

^*p < .05,

^**p < .01.

Regression Analysis to Predict Gendered Peer Involvement

A parallel multiple mediation model extended correlation results to show unique patterns of direct and indirect effects of activities and identity/cognitions on gendered peer involvement. We originally intended to use as the main outcome measure the relative preference for boys versus girls (that is, the difference in proportion of time spent with each), but the limited variability in time spent with boys led us instead to use the measure of time spent with girls as the outcome; it was standardized to facilitate interpretation.

Results, depicted in Fig. 1, showed that time spent with girls was positively predicted by female-typical identity/cognitions and inversely predicted by the difference between male-and-female-typed activities, and that CAH type indirectly influenced time spent with girls through those activities. These results were evident in a set of three regression analyses and two estimates of indirect effects. In two regression analyses examining direct effects of CAH type on the mediators, results paralleled group comparisons: CAH type predicted activities, F(1, 42) = 7.53, p < .01, R² = .15 (girls with C-CAH girls were more gender-atypical in their activities than girls with NC-CAH), but not identity/cognitions, F(1, 42) = .39, p > .05, R² = .01. In the third regression of combined effects on time spent with girls of both CAH type and the mediators, the overall model was significant, F(3, 40) = 3.59, p < .05, R² = .21, with significant prediction by identity/cognitions and activities, but not CAH type (see Fig. 1). Indirect effects of CAH type on time spent with girls via the mediators were estimated with 90% bias-corrected bootstrap confidence intervals. The confidence interval for activities did not include 0, but the confidence interval for identity/cognitions did, indicating that the former but not the latter is an indirect effect.

Figure 1.

Parallel multiple mediation model showing that proportion of time spent with girls is predicted directly by female-typical identity/cognitions and gender-atypical (male-typed minus female-typed) activities, and indirectly by prenatal androgen exposure (CAH type) through activities. Ordinary least squares regression was used, and 90% bias-corrected confidence intervals with 5,000 bootstraps were estimated for indirect effects (shown in brackets). Unstandardized estimates are shown, but proportion of time spent with girls was standardized priori to the analysis. Solid lines indicate statistically significant links; dotted lines indicate nonsignificant links. *p < .05; **p < .01; one-tailed tests were used for mediators.

There was no support for reverse mediation: Time spent with girls did not mediate effects of CAH type on either activities or identity/cognitions. For activities, the combined effects model was significant, F(3, 40) = 4.09, p < .05, R² = .23; only CAH type was a significant predictor (b = 1.00, p < .05), and time spent with girls was not a significant mediator (CI: [−.03, .55]). For identity/cognitions, the combined effects model was not significant, F(3, 40) = 1.99, p > .05, R² = .13; although time spent with girls was a significant predictor (b =.18, p < .05), it was not a significant mediator (CI: [−.25, .01]). These results increase confidence in the hypothesized model.

DISCUSSION

Our results provide novel evidence about the origins of sex segregation. Using a unique natural experiment that enables separation of the effects of early hormones and rearing, and of different aspects of gender development, we showed that gendered peer involvement was not significantly associated with prenatal androgen exposure, but was related to both gendered activities and identity/cognitions. Furthermore, prenatal androgens indirectly influenced time spent with girls through gendered activities.

Implications for Understanding the Nature and Causes of Gender Development

Our study contributes in several ways to understanding the nature and causes of gender development. This includes information about androgen effects on peer involvement, contributions to peer involvement of gendered activities and gender identity and cognitions, and the differential effects of prenatal androgens on different aspects of gender development. The study also emphasizes the value of natural experiments such as CAH to address broad questions about development, moving beyond ascribing outcomes to hormones versus socialization.

Prenatal Androgen Effects on Gendered Peer Involvement

With time use data derived from a daily, cued recall procedure, we provided evidence that prenatal androgens play little role in time spent with male and female peers. Other data on peer choices in younger girls with C-CAH showed that they were more likely than typical girls to prefer boys as playmates, but that many girls with C-CAH preferred girls as playmates; results were based on preferences in experimental tasks (Berenbaum & Snyder, 1995; Pasterski et al., 2011) or global self reports (Hines & Kaufman, 1994; Servin et al., 2003). It is important to study how peer preferences become enacted–or not–in everyday time with peers, including the effects of age, other characteristics (e.g., gender cognitions, as discussed below), peers themselves, and the broader gendered world. For example, some girls–with or without CAH–may prefer to play with boys because of shared interests, but their female-typical identity and cognitions may lead them to spend time with girls; they may also be rejected by boys. In contrast to preferences, aspects of the larger social context not measured here (e.g., sex-segregated sports teams and clubs, family and school routines) may also play a role in girls’ time use. A fruitful future research topic is the identification of contextual moderators of links between girls’ global reports of gendered preferences and their time use.

Prediction of Gendered Peer Involvement

We showed that gendered peer involvement was associated with both activities and gender identity/cognitions, and that the former but not the latter was an indirect effect of prenatal androgens. These relations were found for time spent with other girls. We did not examine mediation involving time spent with boys, because it was not significantly correlated with predictors, likely reflecting the limited time spent with boys among the girls in this sample. Other studies that documented associations for interactions with both boys and girls were conducted with children within the context of preschool (Gasparini et al., 2015; Martin et al., 2011), so age, opportunity, or choice all may explain why the girls in our sample spent little time with boys, and why our significant effects involved time spent with girls. In another study of girls with CAH (Pasterski et al., 2011), playmate preferences in an experimental task were related to gendered play style preferences rather than playmate sex when the two were in conflict, but there was not a test of their independent contributions or explicit measure of gender cognitions.

We thus confirmed and extended recent studies highlighting the contributions of gender cognitions to gendered peer involvement, including effects of beliefs about others’ approval (Martin, Fabes, Evans, & Wyman, 1999), perceived similarity to one’s gender group (Martin et al., 2011), and ingroup bias (Gasparini et al., 2015). We assessed multiple components of gender thoughts and feelings (identity, centrality, evaluation, bias, attitudes) but given our relatively small sample, used a composite measure of gender identity and cognitions. We thus did not determine whether some aspects were more important than others, and did not examine the separate contributions of identity versus cognitions (although emphasize that all components measured the construct to a similar extent). Furthermore, the identity and cognition measures were more highly related to each other than might be expected on the basis of other data and theory (e.g., Egan & Perry, 2001; Ruble, Martin, & Berenbaum, 2006). This may reflect increased variability in this sample compared to typical girls, increasing the ability to detect correlations. These are topics worth future study.

With respect to the role of gendered activities, time spent with girls was primarily related (negatively) to male-typed, rather than female-typed, activities. The importance of male-typed activities could reflect a number of factors, including the increased variability in these activities provided by girls with classical CAH. It may also be the case that behavioral compatibility involves eschewing activities typical of the other sex rather than engaging in those typical of one’s own sex, given that children spend much of their time in nongendered activities. Further, many female-typed activities (e.g., reading, writing, drawing) can be down alone, whereas some male-typed activities such as sports often include others; this might have contributed to our findings linking time spent with girls to male-typed rather than female-typed activities.

Importantly, findings concerning contributors to peer involvement depend on the selection, measurement, and variability of the predictors, so inferences about specific predictors and the relative power of cognitions versus behavior should consider how results might change with different measures or samples. For example, Martin et al. (2011) found that gendered peer interactions were better predicted by gender cognitions than by behavior, but they did not assess gender-typed activities, which are a key feature of children’s behavior. Similarly, we examined a specific set of characteristics, and cannot address the role of other gendered characteristics or different aspects of gender cognitions.

Differential Effects of Prenatal Androgens

We confirmed and extended previous findings that prenatal androgens have differential effects on different aspects of gender, with large effects on gendered activities and small or no effects on gender identity and cognitions. In doing so, we add support for the multidimensional nature and causes of gender development (Ruble et al., 2006).

Extending earlier work, we showed that large androgen effects on self-reported activity interest and participation are paralleled by effects on time use measured in daily phone calls. Girls with C-CAH differed from girls with NC-CAH in time spent in male-typed activities, confirming their interview reports (Endendijk et al., 2016); these results extend those from other studies showing more self-reported interests and participation in male-typed activities among females with CAH than females without CAH (unaffected sisters and population comparisons) (discussed above and reviewed in Berenbaum & Beltz, 2011, 2016).

We also confirmed that prenatal androgen exposure has minimal effects on gender identity and cognitions, measured in multiple ways. Girls with C-CAH were female-typical on gender typicality, gender contentedness, felt pressure, intergroup bias, and gender centrality: they were not significantly different from girls with NC-CAH on any of these measures, and both groups scored in the range of typical girls in other studies (e.g., Egan & Perry, 2001). These results extend published results from this sample that girls with C-CAH are typical in their binary gender identity and attitudes about gender roles and about being a girl (Endendijk et al., 2016), and findings from other samples that the overwhelming majority of girls and women with C-CAH have female-typical binary gender identity but a reduced degree of identification (discussed above and reviewed in Berenbaum & Beltz, 2011, 2016; Meyer-Bahlburg et al., 2016). But, future work is needed on gender cognitions in girls with CAH, given how little has been done, and to help reconcile our results with others showing girls with CAH to have reduced verbal and behavioral preferences for objects labelled to be for girls (Hines et al., 2016). Inconsistencies may reflect differences in measures: focusing on objects (even gender-neutral ones) might remind girls with C-CAH that their own object preferences are not female-typical, whereas focusing on global attitudes might allow them to consider the ways in which they are like girls.

The Value of CAH for Understanding Development

Our findings emphasize the value of CAH and other natural experiments for addressing broad questions about psychological and behavioral development. The design we used is typically employed to test hormonal influences on gender development, but here was also applied to test constructivist perspectives on gender development (Martin, Ruble, & Szkrybalo, 2002). In particular, the variation in activities exhibited by girls with C-CAH and the differential effects of androgens on activities and identity/cognitions enabled an efficient test of their relative contributions to gendered peer involvement.

Limitations

Several limitations need to be considered in interpreting our results. First, the sample was relatively small, and the groups were unequal in size. Data were missing, particularly on one measure of gender cognitions. Nevertheless, the sample was relatively large among studies of girls with C-CAH, especially those that go beyond global self-report; for example, the number of girls with CAH in five studies that included more than questionnaires ranged from 15 to 43 (Hines et al., 2016; Meyer-Bahlburg et al., 2004; Nordenström et al., 2002; Pasterski et al., 2005, 2011). We had enough power to detect moderate to large effects, and did confirm hypothesized group differences in activities, as well as associations between time spent with peers and both gendered activities and identity/cognitions. The sample size may have made it difficult to determine whether girls with C-CAH spend less time with peers overall, a question worthy of further work. Second, the recruitment strategy prevented determination of the representativeness of the sample. It is important to note, therefore, that these results parallel those obtained in samples of girls with CAH recruited in other ways with respect to masculinized activities but female-typical gender identity (e.g., Berenbaum & Bailey, 2003; Meyer-Bahlburg et al., 2004). Third, the cross-sectional nature of the data limits causal inferences about origins of time spent with peers. But, similar designs have been used to study sex segregation in typical children (e.g., Gasparini et al., 2015; Martin et al., 2011). Fourth, we examined only the amount of time spent with male and female peers, not the nature of peer interactions. Fifth, time use might be constrained by contextual factors not assessed here that prevent children from expressing their preferences.

Design Strengths

Important design features balanced study limitations. First, this is the first study to measure time use in females with CAH–or any sample with atypical hormone exposure. Peer involvement has previously been studied only with global self reports, which are subject to greater social desirability biases and memory distortions than are daily data (McHale, Bartko, Crouter, & Perry-Jenkins, 1990). Although studies of girls with CAH (ages 3–11) have included behavioral observations of play (e.g., Berenbaum & Snyder, 1995; Pasterski et al., 2005), studies of activities in older females (adolescents and adults) have been restricted to global measures (self- and parent-report) from interviews and questionnaires. Second, we used measures widely used in studies of typical children, to facilitate application of our work to gender development in general and comparison of our sample with typical girls. The girls in our sample had scores on these measures within the range of typical girls in other samples (see references in Method for each measure). Third, the sample was homogeneous in age. Most studies of females with CAH involve participants across a broad age range, generally in childhood or in adulthood; we focused on a narrow age range, and one that is not often represented in studies of CAH. Fourth, the sample provided comparison for aspects of hormone exposure. Girls with NC-CAH provided important control for recruitment biases and for aspects of the disease suggested to affect behavior, such as postnatal androgen excess and medical treatment (for discussion, see Blakemore et al., 2009). It would be valuable to include a sample of unaffected sisters of girls with C-CAH, as is often done in studies of girls with CAH, but our focus on girls in a narrow age range prevented us from doing that here.

Summary and Conclusions

The evidence we provided from girls who varied in degree of prenatal androgen exposure advances understanding of the nature and causes of gender development. Girls’ gendered peer involvement does not seem to be strongly influenced by exposure to moderate levels of androgens during prenatal development. Across girls with varying exposure, time with peers was explained instead by both gendered activities and gender identity/cognitions, confirming and extending findings in samples of young, typical children. The effect of activities, but not identity/cognitions, was partly due to prenatal androgens. Extending other work on androgen influences on gender development, we also showed that prenatal androgens have large effects on time spent in male-typed leisure activities during the transition to adolescence, but small or no effects on multiple aspects of gender identity and cognitions. Finally, this study illustrates the value of natural experiments for understanding hormonal and social influences on gender development and links among gender-related characteristics.

APPENDIX

Scale of male-typed gendered activities, measured by the Activity Interest Questionnaire (AIQ), the Adolescent Activity Questionnaire (AAQ), and Daily Diaries

Interest in male- typed activities (AIQ)	Participation in male-typed activities (AAQ)	Time in in male- typed activities (phone diaries)
Baseball	Baseball	Baseball
Football	Football	Football
Basketball	Basketball	Basketball
Hunting and fishing	Hunting or fishing	Hunt or fish
Building models or other things	Build things (e.g., models, towers, or roadways)	Build models or other things
Playing with/collecting trucks, cars, action figures	Play with or collect trucks, cars, or action figures	Play with/collect trucks, cars, or action figures
Watching sports on TV	Watching sports on TV	Watching sports on TV
	Follow sports or a sports team online or in newspapers
Motorbiking, motorcycling, or 4- wheeling	Ride motorcycles, dirt bikes, or 4- wheelers	Motorbiking, motorcycling, or 4- wheeling
Computer and video games	Play computer or other online games	Play computer and video games
	Play videogames (e.g., Playstation, Xbox, Wii)
Home repairs	Use tools
	Yard work or taking out the garbage
Science		Do science homework (e.g., Biology or Astronomy)
Math		Do math homework (e.g., Number problems)
	Work with engines and electronic Maintain vehicles

Scale of female-typed gendered activities, measured by the Activity Interest Questionnaire (AIQ), the Adolescent Activity Questionnaire (AAQ), and Daily Diaries

Interest in female- typed activities (AIQ)	Participation in female-typed activities (AAQ)	Time in in female- typed activities (phone diaries)
Cheerleading	Cheerleading	Cheerleading
Dancing	Dancing	Dance (e.g., ballet, jazz)
Gymnastics	Gymnastics	Gymnastics
Playing with/collecting dolls or figurines	Collect or play with dolls or figurines	Play with/collect dolls or figurines
Writing poetry, short stories, or in diary	Write poetry, short stories, or in a diary	Write poetry, short stories, or in a diary
Reading stories, magazines, or newspapers	Read fiction stories or books	Read stories, magazines, or newspapers
Shopping	Go shopping	Go shopping (not including grocery shopping or car maintenance)
Cooking and baking	Cook or bake	Prepare a meal or snack (e.g., cook or set table)
Drawing, painting, pottery, other arts and crafts	Paint or draw	Draw, paint, pot, other arts and crafts related activities
	Make arts and crafts
	Sew, knit, embroider, or other needlework
	Work with pottery or ceramics
Housework (e.g., laundry, cleaning)	Clean house or do dishes	Do dishes
		Vacuum, dust, or straighten up (not including own things in own room)
	Do laundry	Laundry
Language arts		Do language arts homework (e.g., reading or writing)
	Play with make- up or dress up