Child Development, xxxx 2012, Volume 00, Number 0, Pages 1–16

1 2 Rigidity in -Typed Behaviors in Early Childhood: A Longitudinal 3 4 Study of Ethnic Minority Children 5 6 May L. Halim Diane Ruble, Catherine Tamis-LeMonda, 7 California State University, Long Beach and Patrick E. Shrout 8 New York University 9 10 11 12 A key prediction of cognitive theories of gender development concerns developmental trajectories in the

13 relative strength or rigidity of gender typing. To examine these trajectories in early childhood, 229 children CE: Deepa R 14 (African American, Mexican, and Dominican) were followed annually from age 3 to 5 years and gender- 15 stereotypical appearance, dress-up play, toy play, and sex segregation were examined. High gender-typing 16 was found across ethnic group, and most behaviors increased in rigidity, especially from age 3 to 4 years. In addressing controversy surrounding the stability and structure of gender-typing it was found that from year 17 to year, most behaviors showed moderately stable individual differences. Behaviors were uncorrelated within 18 age, but showed more concordance in change across time, suggesting that aspects of gender-typing are multi- 19 dimensional, but still show coherence. 20 21

22 Dispatch: 14.12.12Author Received: Journal: No. of pages: 16 PE: Karpagavalli 23 A glance around a playground, a classroom, or a children espouse gender early on could set in 24 toy store will instantly reveal that early childhood motion life-long individual differences in displays B 25 is an important period in gender development for of masculinity or femininity (Martin & Ruble, 2009). 26 American children. Gender is displayed in highly Although the existence of obvious differences in 27 visible, explicit ways. Flocks of girls, especially at gender-typed behaviors among young American 28 ages 3 and 4 years, go about their days swathed in boys and girls is clear, fundamental questions 29 pink, glitter, hearts, and ribbons (Halim et al., 2012; remain unanswered. First, we know little of 30 Ruble, Lurye, & Zosuls, 2007). Boys, too, exhibit whether different aspects of gender-typing emerge 31 their own kind of “gender rigidity” (defined in this at the same time and connect to each other, or 32 article as a strong adherence to gender-stereotypical instead, exhibit specificity in expression. Second, 12057 33 behaviors; Parsons & Howe, 2006). Often, one can questions remain about whether gender-typed 34 see boys kicking the air, flexing their muscles, and behaviors show a predictable pattern of normative 35 running around imagining themselves as Spider- developmental change and stable individual differ- 36 . This period when children display their gen- ences over time or are a function of unpredictable 37 der so openly and ostentatiously provides a situational variation. Third, it remains to be seen 38 fascinating window into the early stages of gender whether the above portrait of gender-typing can be Journal Name Manuscript No.

39 development in the life span. The degree to which found across different ethnic groups, or whether the CDEV 40 focus on White, middle-class samples has led to 41 This work was conducted at New York University’s Center for distorted conclusions. Despite the explicit gender- 42 Research on Culture, Development, and Education (CRCDE), typing seen in young children, there has been little 43 within the Department of Applied Psychology at New York Uni- empirical investigation of it—over time, across 44 versity’s Steinhardt School of Culture, Education, and Human Development. This research was supported by the National different behaviors, and across different populations. 45 Science Foundation grants BCS 021859 and IRADS 0721383 to C. Some may explain these behaviors as just a Disney- 46 S. Tamis-LeMonda. Preparation of this paper was also supported provoked middle-class display. However, we argue 47 in part by a National Institute of Child Health and Human fl Development Research Grant (R01 HD04994 and ARRA Supple- that these behaviors re ect a manifestation of a 48 ment) to D. N. Ruble. We thank our colleagues and staff at the fundamental developmental phenomenon that marks 49 CRCDE, particularly Cristina Hunter, Eva Liang, Yana Kuchirko, the early phases of a life course of a gender divide. 50 Julia Raufman, Emerald Shee, Irene Sze, and Irene Wu, as well as the mothers and children who participated in our research. 51 We also thank David Amodio and Kay Deaux for their feedback 52 on earlier versions of this article. © 2012 The Authors 53 Correspondence concerning this article should be addressed to Child Development © 2012 Society for Research in Child Development, Inc. May L. Halim, California State University, Long Beach, LA. Elec- All rights reserved. 0009-3920/2012/xxxx-xxxx 54 tronic mail may be sent to [email protected]. 1 DOI: 10.1111/cdev.12057 2 Halim, Ruble, Tamis-LeMonda, and Shrout

1 Thus, investigating these questions is critical in explicitly developmental perspective, emphasizing 2 adding to existing literature on gender develop- not just high levels, but change over time, upon 3 ment. which we base this study. 4 These questions are unanswered because the In this study we employed a longitudinal design 5 literature to date is primarily composed of cross- to investigate developmental growth within the 6 sectional studies of middle-class White children. same individuals. We asked whether gender-typed 7 Exciting new insights can be obtained by broaden- behaviors increase in rigidity during ages 3 to 8 ing the study of gender development to include 5 years, as this period is a time of increasing gender 9 samples of ethnically diverse children in a longitu- knowledge, rather than just whether rigidity is 10 dinal design. Insofar as children raised in the high. We also explored whether this rigidity begins 11 United States experience the same global culture to level off during these ages; that is, do some gen- 12 during the developmentally important preschool der-typed behaviors show the curvilinear pattern 13 years, and insofar as increasing rigidity in gender- predicted by cognitive theories of gender develop- 14 typed behaviors reflect a fundamental, generaliz- ment? 15 able developmental theory, they should show the This question is important, because most studies 16 same patterns of gender development. This study on the developmental trajectory of gender-typing 17 aims to address these questions by examining the focus primarily on gender-related beliefs and cogni- 18 change and stability (consistency of a behavior tions. Several studies have shown that children’s 19 across time) of different facets of gender-typing in gender stereotyping increases in rigidity from age 20 an ethnically diverse sample of 3- to 5-year-old 3 to 5 years, which is then generally followed by 21 children. flexibility (Signorella, Bigler, & Liben, 1993; Traut- 22 ner et al., 2005). Other research has found similar 23 increases in rigidity from age 3 to 4 years in Trajectories of Gender Development in Early Childhood 24 attitudes about children who violate gender norms 25 According to cognitive theories of gender devel- (Ruble et al., 2007). But how do manifestations of 26 opment, as children learn about gender categories, rigidity in gender-related behavior develop across 27 they are predicted to be highly motivated to strictly early childhood? Do different types of gendered 28 adhere to gender , and this adherence behaviors show a similar course of increasing rigid- 29 may increase in rigidity as information about gen- ity as seen in gender-related cognitions? Past 30 der categories is constructed and consolidated research examining developmental trajectories in 31 (Martin, Ruble, & Szkrybalo, 2002; Ruble, 1994). gender-typed behavior and preferences has largely 32 These theories view children as active “gender been limited to either peer or toy preferences. This 33 detectives” (Martin & Ruble, 2004), who earnestly research, largely cross-sectional with a few longitu- 34 seek out information about gender, then attempt dinal studies, has found that children increasingly 35 to draw inferences and clear conclusions about prefer to play with same-sex children from pre- 36 gender, and finally apply such conclusions rigidly school through the elementary school years (e.g., 37 to themselves. Subsequently, cognitive theories of Maccoby & Jacklin, 1974; Martin, Fabes, Evans, & 2 38 gender development predict that children will inte- Wyman, 1999; Serbin, Powlishta, & Gulko, 1993) 39 grate their new with other identities and display increased rigidity in gender-typed toy 40 and become more flexible in the application of the preferences (e.g., Servin, Bohlin, & Berlin, 1999). 41 conclusions they drew about gender. These theories In summary, there is little research on the trajec- 42 are, of course, not the only perspectives on gender tory of multiple aspects of gender-typed behavior 43 development. Socialization approaches posit that during the preschool years, as well as little attempt 44 experiences at home, school, and one’s neighbor- to pull together the different facets of gender-typing 45 hood influence children’s gender-typing through in a single study. This study addresses this gap in 46 processes such as social learning and encourage- the literature by looking at four different behaviors 47 ment from others to behave in gender-typed ways. within one study over time. We were interested in 48 Biological approaches emphasize the role of genes, how children adorn themselves as it is relatively an 49 hormones, and the brain in physical and psycholog- unexplored element of gender development and 50 ical sexual differentiation (Ruble, Martin, & seems both prevalent among children and 51 Berenbaum, 2006). These perspectives on gender important to them. Thus, we observed children’s 52 development would predict a high level of gender- gender-typed appearances and their involvement in 53 typed behaviors in early childhood. However, only dress-up play. Gender researchers more commonly 54 cognitive theories of gender development take an study activities and peer choice, so we included Rigidity in Gender-Typed Behaviors 3

1 these elements in this study to provide reference children’s gender attitudes; Ruble et al., 1998). 2 points for comparison. On the basis of the afore- Looking at cohesion both within-time and across- 3 mentioned studies and in line with proposals that time may shed light on this controversy. On the 4 gender-typing is multidimensional (Huston, 1983; basis of past research (Martin & Fabes, 2001), we 5 Ruble et al., 2006), we anticipated different patterns expected that gender-typed play and sex segrega- 6 of normative change for different types of behav- tion might correlate with one another both within- 7 iors. Physical appearances are salient to children at and across-time, as these two elements feed into 8 a very early age (Ruble, Lurye, et al., 2007), and each other bidirectionally. 9 thus the process of construction and consolidation 10 might “take off” between ages 3 and 4 years Stability in Gender-Typing 11 (Ruble, 1994). Consistent with this idea, in a study 12 of White, middle-class girls, appearances were the Another vital and controversial question is 13 most gender-typed at ages 3 and 4 years and lower whether or not individual differences in gender- 14 at ages 5 and 6 years (Halim et al., 2012). For these typing are stable over time even in the context of 15 reasons, we predicted that children’s actual gender- normative changes of increasing gender-typed 16 typed appearances and their engagement in dress- behaviors (Martin & Ruble, 2009). Does a girl who 17 up play would be characterized by a curvilinear is highly gender-typed compared to her peers at 18 pattern across ages 3–5 years with gender-stereo- age 3 years remain so at age 4 years? Are certain 19 typicality increasing to a peak and then leveling off aspects of gender-typing (e.g., play activities) more 20 or declining. In contrast, participation in gender- stable than others (e.g., sex segregation) across the 21 typed play activities and sex segregation were ages of 3, 4 and 5 years? These questions are 22 expected to continue to increase from ages 3 to important for understanding the processes involved 23 5 years. in gender development over the life span and could 24 have implications for self-perceptions and life 25 choices, such as the valuing of academic domains Multidimensional Nature of Gender-Typing 26 and occupational aspirations, in ways that would 27 The gender development field has conceptual- be stronger than mere fleeting gender proclivities. 28 ized gender-typing as a multidimensional matrix Yet, surprisingly, little empirical data exist on this 29 (Huston, 1983; Ruble et al., 2006). Is there unity in topic, and the question of the stability of individual 30 gender-typing? Or does specificity better describe differences in gender-typing is an area of conten- 31 the way children express their gender? The longitu- tion. 32 dinal design of this study allows us to look at the Prominent gender development theories would 33 coherence of gender-typing both within and across predict stability in gender-typing, if even for a short 34 time. On one hand, there has been a prevalent time, as hypothesized predictors of gender-typing, 35 implicit assumption of coherence in the literature such as hormones or parent modeling of gender- 36 (e.g., if a girl likes to play with dolls, she probably typed behavior, also tend to be stable in childhood 37 loves pink as well). Indeed, the very construct of (see Golombok, 2008). However, some gender 5 38 gender-typing assumes that there are factors that researchers have concluded that, actually, the 39 lead children to become gender-typed as they limited data available show little stability in gen- 40 acquire a set of behaviors and interests that are der-typing (Maccoby, 2002). Instead, Maccoby 41 more typical of their own gender than of the other (2002) suggested that ample variation in situations 42 gender (Maccoby, 2002). Thus, if one is to study surrounding a child affects how gender-typed a 43 gender-typing, there needs to be something coher- child may appear. Furthermore, any stability that is 44 ent to study, even if coherence occurs only across a found may reflect group-level associations (e.g., 45 few attributes, a view evident in many prominent classroom peer influences) rather than individual- 46 3 traditions (e.g., Martin & Halverson, 1981). On the level ones. Other psychologists have challenged this 47 other hand, specificity in gender manifestations is a viewpoint (Martin & Ruble, 2009). Some studies 48 possibility; there have been many failures to find indicate moderate stability in sex segregation in 49 relations among all types of gender-related vari- preschool to elementary school-aged children under 50 4 ables (Ruble et al., 1998). There are also studies that some conditions (e.g., Martin & Fabes, 2001; Serbin 51 find no correlation between gender-typed prefer- et al., 1993), and other studies indicate moderate 52 ences or behaviors and gender- knowl- stability in gender-typed play and toy prefere- 53 edge. Unity may be more evident in smaller subsets nces among preschoolers and kindergarteners 54 of gender-typing variables (e.g., only within (Golombok et al., 2008; Maccoby & Jacklin, 1974; 4 Halim, Ruble, Tamis-LeMonda, and Shrout

1 Martin & Fabes, 2001). In this study, we sought to sparse literature, we explored whether Latino chil- 2 address this controversy by examining several dren might exhibit higher levels of gender-typed 3 behaviors in a single study to determine whether or behavior and show peak gender rigidity earlier 4 not certain aspects of gender-typing show more sta- compared to African American children. 5 bility than others. 6 7 Study Overview Rigidity in Gender-Typed Behaviors Across Ethnic 8 This study assessed four gender-typed behaviors Groups 9 in African American, Mexican, and Dominican chil- 10 In this study we also aimed to test the robust- dren from a large urban Northeastern city. At ages 11 ness of the phenomenon of increasing gender-typed 3, 4, and 5 years, coders rated how gender-typed 12 behaviors across ethnic groups, as the vast majority the child’s appearance was and gathered informa- 13 of gender development research has only sampled tion from mothers on children’s participation in 14 White middle- to upper-middle class children. dress-up play, participation in gender-typed play 15 Because of this bias, only tentative statements as to activities, and the number of girls and boys that 16 what is normative in gender development can be were children’s friends. We examined three types of 17 made. We examined African American and Latino hypotheses: (a) specific theory-driven hypotheses, 18 immigrant children because these ethnic groups (b) those addressing unresolved controversies, and 19 widely differ on factors that could affect gender- (c) exploratory hypotheses. For the first type, on the 20 typing, which would allow us the most stringent basis of cognitive theories of gender development, 21 test of the generalizability. In addition, these groups we predicted that across gender and ethnic groups 22 have an important presence in the United States— children would show high levels and curvilinear 23 African Americans make up a substantial minority, trajectories of increasing rigidity in gender-typed 24 and the population of Latinos is increasing (U.S. behaviors across ages 3–5 years. Furthermore, we 25 Census Bureau, 2011). Because we consider increas- hypothesized that appearance-related behaviors 26 ing gender-typed behaviors to be a fundamental would peak earlier in rigidity and show some level- 27 developmental phenomenon driven largely by ing off compared to other behaviors. For the second 28 changes in cognition, we expected curvilinear trajec- type, we investigated whether these different 29 tories across all ethnic groups. However, we behaviors would be connected to each other in a 30 explored ethnic variation in overall levels or in the constellation of gender-typing and whether they 31 timing of those trajectories based on tentative paralleled one another in how they changed. We 32 evidence showing differences in gender roles and also investigated whether, in the middle of norma- 33 gender socialization by ethnic group. For example, tive changes, there would be stable individual dif- 34 scholars have argued that African American ferences in gender-typing. Finally, for the third 35 women’s ample participation in the workforce has type, we explored whether the levels or trajectories 36 brought about more egalitarian gender roles and of these behaviors would show variation by ethnic 37 less sharply differentiated gender stereotypes in group. 38 African American families (e.g., Stanback, 1985), 39 which has received some empirical support (Jarret, 40 Roy, & Burton, 2002). Also, African American 4- to Method 41 6-year-children have shown less gender stereotype Participants and Procedure 42 endorsement than White children (Albert & Porter, 43 1988), although this has not always been replicated Preschool-aged children (N = 229: 108 girls, 121 = = 44 (Liben & Bigler, 2002). In contrast, machismo and boys) and their mothers (Mage 29.06, SD 5.64) 45 marianismo—or, broadly, male dominance and participated: 84 Dominican American (49 boys, 35 46 submissiveness—are key constructs in girls), 77 African American (42 boys, 35 girls), 68 47 Latino culture (DeSouza, Baldwin, Koller, & Mexican American (30 boys, 38 girls). We assessed 48 Narvaz, 2004). Further, Latino children have been preschoolers at three waves: at age 3 years 49 found to endorse occupational gender stereotypes (M = 3.03 years, SD = 0.14), 4 years (M = 4.21 years, 50 more than White American children (Bailey & SD = 0.16), and 5 years (M = 5.16 years, SD = 0.12). 51 Nihlen, 1990). And a higher proportion of Mexican Of the 229 families, 154 families participated at all 52 and Dominican immigrant 2-year-olds were able to 3 waves, 28 families participated at 2 of the 3 waves 53 produce gender labels compared to African Ameri- and 47 families participated at only 1 of the 3 54 can 2-year-olds (Zosuls, 2008). On the basis of this waves. Comparing families with complete versus Rigidity in Gender-Typed Behaviors 5

1 incomplete data revealed no significant differences baseball jerseys), and formal wear (e.g., skinny tie 2 in mother’s education, work status, immigrant with a vest). To create a score of gender-typed 3 status, marital status, cohabitation status, dominant appearance we summed elements for girls and boys 4 language, age, physical health, psychological dis- separately. Girls could reach a maximum score of 8 5 tress, household income, or child cognitive or (M = 3.29, SD = 1.66) and boys could reach a maxi- 6 language skills at age 2. However, girls and Mexican mum score of 5 (M = 2.29, SD = 0.99). Girls’ and 7 children were more likely to have complete data. boys’ scores had different possible ranges because 8 Participants were recruited as part of a larger few boys showed trend-conscious styles or jewelry, 9 longitudinal study from three public hospitals in a and most had short hairstyles. For comparison pur- 10 large northeastern urban city shortly after mothers poses, we created z scores for girls’ and boys’ 11 delivered their children. Eligibility to participate in scores, each centered around the grand mean 12 the study required the mothers to (a) be at least (across ages 3, 4, and 5 years) within gender. 13 18 years old, (b) not live in a shelter, (c) have a Gender-typed dress-up play. Interviewers asked 14 healthy, full-term infant (birth weight > 2,500 g), mothers: “Over the past month, how often did your 15 and (d) self-identify as Mexican, Dominican, or child play with dress-up clothes or costumes like 16 U.S.-born African American. Of all mothers princess or Spiderman costumes, pocketbooks, or 17 approached who were eligible, 50% agreed to par- adult-like shoes?” (0 = Never,1= Once or twice a 18 ticipate. Average family annual income was $20,500 month,2= Once a week,3= Several times a week, to 19 (SD = $14,751). Dominican mothers were 77% first 4 = Everyday; M = 1.53, SD = 1.50). Interviewers 20 generation, and 23% second generation, and were also asked mothers to list the type of dress-up 21 primarily from Santo Domingo. Mexican mothers clothes or costumes with which their child played. 22 were 96% first-generation immigrants and were pri- Only gender-typed costumes or clothes were 23 marily from Puebla. At the infants’ births, 73% of counted (excluding five children at age 3 years, two 24 Dominican, 48% of Mexican, and 65% of African boys at age 4 years, and one girl at age 5 years 25 American mothers completed at least a high school who liked to dress-up as Michael Jackson). The 26 education or a GED. Co-residency rates were high- most common themes included dressing up as a 27 est in Mexican families (87%), followed by Domini- princess for girls (especially Disney princesses at 28 can families (63%), and African American families age 4 years) and dressing up as some sort of super- 29 (46%), F(2, 216) = 14.44, p < .001. hero for boys (especially at age 5 years). Other 30 Interviews took place either at children’s homes themes included dressing up in adult-like clothing, 31 (age 3 years) or at a university (ages 4 and 5 years). including the clothing of parents’, siblings’, or rela- 32 We obtained parental consent either in person or tives’ (especially at age 3 years), and dressing up as 33 through the mail via signed returned consent forms. some sort of TV or cartoon character (e.g., Hannah 34 We paid mothers $75.00 for each interview at each Montana or a Transformer). 35 wave. Gender-typed play. Interviewers asked mothers 36 how often over the past month their children 37 played with three feminine items ([a] kitchen sets, Measures 38 tea sets, and/or food sets, [b] dolls, and [c] soft toys 39 Observations of gender-typed appearance. We coded like teddy bears, stuffed animals, or puppets) and 40 children’s gender-typed appearance from video- three masculine items ([a] toy guns or objects used 41 tapes of the interviews at each wave. Researchers as guns and/or swords or objects used as swords, 42 coded whether they observed a set of specific gen- [b] vehicles like cars, trucks, and trains, and [c] 43 der-typed elements (0 = Not present,1= Present). balls; 0 = Never,1= Once or twice a month,2= Once 44 For girls, elements included the following (j’s a week,3= Several times a week, to 4 = Everyday). 45 range = .70–1.00): dresses or skirts, feminine colors These toys were chosen based on many studies 46 (e.g., pink), feminine hair accessories (e.g., bows), using the same stimuli (Ruble & Martin, 1998). 47 feminine patterns or logos (e.g., flowers), feminine Prior research has shown that mother report of chil- 48 fabric or fit (e.g., taffeta), trend-conscious styles dren’s toy play has good reliability across time 49 (e.g., knee-high boots), formal wear (e.g., patent- (Golombok et al., 2008). To make a scale, for girls 50 leather Mary Janes), and jewelry (e.g., bracelets). we reverse coded the three masculine items and 51 For boys, elements included the following: mascu- averaged them together with the three feminine 52 line colors (e.g., red and black together), masculine items (M = 2.74, SD = 0.54). For boys we reverse 53 patterns or logos (e.g., graffiti), masculine fabric or coded the three feminine items and averaged them 54 fit (e.g., baggy pants), sports-themed styles (e.g., together with the three masculine items (M = 2.70, 6 Halim, Ruble, Tamis-LeMonda, and Shrout

1 SD = 0.55). We then combined girls’ and boys’ scores models, but with alterations (e.g., treated time as a 2 = = into a single measure (aage 3 .59, aage 4 .70, factor for post hoc time period comparisons; used 3 = aage 5 .72). effects coding to look at effects across gender or 4 Sex segregation. Interviewers asked mothers: across ethnicity). 5 “Does your child have the opportunity to spend Observations of gender-typed appearance. Overall, 6 time with other children?” If mothers answered, children, especially girls, were very gender-typed in 7 “Yes,” they were asked, “Who are they?” Mothers their appearances. Dominican girls were more gen- 8 could list up to nine of their child’s peers and der-typed in their appearance compared to Mexican 9 reported the peer’s sex, the peer’s age, how the peer girls. Furthermore, as predicted, gender-typed 10 was related to the child, and if the peer was actu- appearance followed a curvilinear pattern across 11 ally the child’s friend. To be consistent with prior time for the sample as a whole. Trajectories varied 12 studies on sex segregation, we excluded peers that across ethnic groups. See Figure 1a and Table 1. 13 were either over age 10 years or siblings and only Specifically, the mean number of feminine 14 included peers who were counted as friends. At appearance elements at age 4 years was 3.73 15 each age we calculated the proportion of same- (SD = 1.97). Although a mean of 3.73 of a possible 16 sex peers of total peers named (M = 63.01%, eight may seem low, qualitatively, it indicates that 17 SD = 30.00%). on average, girls exhibited between three and four 18 gender-typed elements, such as wearing a pink 19 dress with hearts and an additional big bow in her 20 hair, creating a very feminine appearance. Boys Results 21 exhibited between two and three elements of five 22 We first discuss findings pertaining to normative overall at each age, indicating overall masculine 23 changes in gender-typed behaviors from age 3 to 24 5 years and report any cultural or gender variation (a) 25 in the level and trajectory of these behaviors. Next 26 we report whether these different behaviors are 27 connected to each other in a constellation of gen- 28 der-typing by examining their associations both (a) 29 within time and (b) across time as the behaviors 30 change. Finally, we report the stability of gender- 31 typing across consecutive years. 32 33 34 Change in Gender-Typed Behavior Over Time and

35 Group Differences Colour online, B&W in print 36 Analytic strategy. We examined change over time 37 of gender-typed behavior by conducting multilevel 38 models for each behavior, which included children 39 with at least two waves of data. For each model, 40 fi we rst calculated Type III sums of squares to indi- (b) 41 cate overall omnibus fixed effects for time (centered 42 at age 4 years), gender (female reference group), 43 ethnicity (Dominican reference group), and inter- 44 actions among the three variables and a random par- 45 ticipant intercept effect in predicting each behavior. 46 For ease of communication we present the results 47 using fitted means from the mixed models. For 48 detailed model coefficients and formulas see Halim 49 (2012). We treated gender and ethnicity as factors 50 and time as a covariate, and we assumed that the 51 residuals were uncorrelated and homoscedastic 52 over time. Next, we repeated the model with the 53 Figure 1. (a: left panel) Means of observed gender-typed appear- linear and quadratic components of time separated. ance over time by ethnic group. (b: right panel) Means of gen- 54 For follow-up analyses, we repeated the mixed der-typed dress-up play frequency over time by ethnic group. Rigidity in Gender-Typed Behaviors 7

1 Table 1 2 Means and Standard Deviations of Gender-Typed Behaviors 3 4 Age 3 years Age 4 years Age 5 years Across time 5 M SD M SD M SD M SD 6 7 Observed appearance rigidity 8 Boys overall 2.52 0.94 2.36 1.05 1.99 0.91 2.28 0.99 a,b a b 9 Girls overall 3.32 1.32 3.73 1.97 2.80 1.51 3.28 1.66 f À a,d b,d,e À a,d,e À d 10 African American children 0.24 0.68 0.23 1.04 0.24 0.93 0.07 0.93 Mexican childrenf 0.11a,e 0.96 À0.15a,b,d 1.05 À0.52b,d 0.86 À 0.18d 0.99 11 Dominican childrenf 0.43a,e 0.85 0.37a,e 1.23 À0.16b,e 0.93 0.20e 1.05 12 All children overallf 0.12a 0.88 0.16a 1.13 À0.30b 0.91 À 0.01 1.00 13 Gender-typed dress-up play 14 Boys overall 0.71a,d 1.13 1.29b,d 1.35 1.22a,b,d 1.38 1.08d 1.31 15 Girls overall 1.89e 1.6 2.15e 1.57 1.96e 1.50 2.00e 1.55 16 African American children 1.02a 1.45 1.63a,b 1.54 1.96b,d 1.49 1.56 1.54 17 Mexican children 1.51a,b 1.55 1.76a 1.60 1.08b,e 1.29 1.46 1.50 18 Dominican children 1.31 1.47 1.71 1.43 1.63d,e 1.53 1.55 1.48 a b a,b 19 All children overall 1.29 1.49 1.70 1.52 1.57 1.48 1.52 1.50 Gender-typed play 20 a a,b b 21 Boys overall 2.60 0.56 2.73 0.62 2.78 0.46 2.70 0.55 Girls overall 2.64a 0.53 2.74a,b 0.54 2.81b 0.50 2.73 0.53 22 African American children 2.52a 0.61 2.76b,d,e 0.59 2.68a,b,d 0.52 2.66 0.58 23 Mexican children 2.64a 0.55 2.61a,b,d 0.65 2.80b,d,e 0.52 2.66 0.58 24 Dominican children 2.68a 0.48 2.82a,b,e 0.50 2.89b,e 0.38 2.80 0.46 25 All children overall 2.62a 0.55 2.73b 0.59 2.80b 0.48 2.72 0.54 26 Sex segregation 27 Boys overall 0.48a 0.29 0.71a,b 0.26 0.73b 0.28 0.65 0.30 28 Girls overall 0.49a 0.28 0.68a,b 0.29 0.75b 0.25 0.64 0.30 29 African American children 0.55 0.28 0.66 0.27 0.71 0.30 0.65 0.29 a b b 30 Mexican children 0.44 0.28 0.70 0.32 0.78 0.25 0.64 0.32 Dominican children 0.46a 0.29 0.72b 0.25 0.73b 0.25 0.64 0.29 31 a b b 32 All children overall 0.48 0.29 0.70 0.28 0.74 0.27 0.64 0.30 33 Note. Means not sharing the same (a) a, b, c superscripts within row, and (b) the same d, e superscripts within columns and sections 34 are significantly different at p < .05. Superscripts are only given where findings were significant. fMeans given for boys and girls sepa- 35 rately are in original scale for ease of interpretability. Means when girls and boys are combined are the z-scored means. 36 37 appearances as well, such as wearing dark, baggy The level of gender-typed appearance was not 38 clothes with a graffiti pattern. different from age 3 to 4 years (p = .646). Age 5 39 In exploring ethnic differences we found that gender-typed appearance was lower than age 3 and 40 across time, the model revealed a main effect of 4 levels (p’s < .001). We also found an interaction 41 ethnicity, F(2, 485.4) = 5.49, p = .004, qualified by a between time and ethnicity, F(2, 333.8) = 2.83, 42 gender by ethnicity interaction, F(2, 485.4) = 6.79, p = .025. Both Dominican (time linear p = .001, time 43 p = .001, such that the effect was concentrated in quadratic: p = .041) and African American children 44 the girls. Dominican girls (M = 4.05, SD = 1.74) showed significant curvilinear patterns (time linear: 45 were significantly more gender-typed in their ns, time quadratic: p = .004). Mexican children 46 appearances compared to Mexican girls (M = 2.82, showed a significant negative linear trend over 47 SD = 1.58), p = .015, but not compared to African time (p < .001). See Figure 1a. The interaction 48 American girls (M = 3.04, SD = 1.37), p = .165. between time and gender was not significant, F(2, 49 Among boys, there were no differences in gender- 333.9) = 1.98, ns. 50 typed appearance among the three ethnic groups. Gender-typed dress-up play. As predicted, girls 51 The average level of gender-typed appearance played dress-up more frequently than boys at all 52 changed over time, and the overall effect was curvi- ages, F(1, 175.2) = 35.36, p < .001. We explored 53 linear as expected, time linear: F(1, 337.1) = 16.40, whether Latino children played dress-up more 54 p < .001; time quadratic: F(1, 330.9) = 8.65, p = .004. frequently than African American children, but there 8 Halim, Ruble, Tamis-LeMonda, and Shrout

1 = = = ’ were no differences among the three ethnic groups SDage 4 .89; Mage 5 .97, SDage 5 .82). Girls 2 (F(2, 175.3) = .91, ns), nor any gender by ethnicity female-typed and boys’ male-typed play was high 3 interaction, F(2, 175.3) = 1.21, ns. Furthermore, as at all time points and did not change. 4 expected, dress-up play followed a curvilinear pat- Sex segregation. At age 3 years, 48% of children’s 5 tern across time and we again found cultural varia- friends were of the same gender, at age 4 years, 70% 6 tion in these trajectories. Specifically, the overall were of the same gender, and at age 5 years, 74% 7 dress-up play mean of 2.00 for girls can be inter- were of the same gender. As predicted, sex segrega- 8 preted as girls playing dress-up about once a week tion increased over time in a curvilinear fashion, and 9 on average, whereas the boys’ mean of 1.08 can be there was no evident variation by ethnicity, F(2, 10 interpreted as playing dress-up about once or twice 164.8) = .01, ns, or gender, F(1, 164.8) = .05, ns,or 11 a month on average during the preschool years. interaction between the two, F(2, 164.8) = 1.12, ns. 12 Examined another way, 56% of girls and 31% of Specifically, the model revealed a main effect 13 boys played dress-up anywhere from once a week for time, F(2, 310.9) = 39.04, p < .001, linear: F(1, 14 to every day across the preschool years. 312.2) = 69.02, p < .001; quadratic: F(1, 310.0) = 10.10, 15 As for developmental change, we found a signifi- p = .002. Sex segregation increased from ages 3 to 16 cant main effect for time, F(2, 325.3) = 5.26, 5 years with the sharpest increase from age 17 p = .006, with both linear, F(1, 327.0) = 4.58, 3 to 4 years (from 3 to 4 years, p < .001; from 4 to 18 p = .033, and quadratic, F(1, 323.7) = 5.99, p = .015, 5 years, p = .094). 19 components. Dress-up play increased in frequency Summary. As expected, children exhibited rigid 20 from ages 3 to 4 years (p = .002), then leveled off in male- and female-typed behaviors during the early 21 frequency from ages 4 to 5 years, but this small childhood years. They dressed and played in gen- 22 decrease was not significant (p = .246). The model der-typed ways and mainly were friends with 23 also revealed an interaction between time and same-gender peers. We found few ethnic differ- 24 ethnicity, F(4, 325.3) = 4.72, p = .001. African ences in overall levels of gender-typing except that 25 American children increased linearly over time in Dominican girls dressed in more stereotypical ways 26 dress-up play frequency (p < .001). Mexican than did Mexican girls. 27 children increased then decreased in dress-up play Our expectations for the developmental pattern 28 frequency, peaking at age 4 years(quadratic: of increased rigidity in gender-typed behaviors 29 p = .011). Dominican children did not significantly were supported as well. From age 3 to 5 years chil- 30 change over time. See Figure 1b. The interaction dren’s friends were increasingly same-gender peers. 31 between time and gender was not significant, F(2, Also, children increasingly avoided cross-gender- 32 325.3) = .76, ns. typed play. Showing a slightly different pattern, 33 Gender-typed toy play. We first conducted mixed children increasingly played dress-up from age 3 to 34 models on the full six-item gender-typed play scale 4 years then remained steady. In contrast, children’s 35 followed by exploration of a breakdown of the gender-typed appearances showed some movement 36 scale: three items of cross-gender-typed play and toward flexibility, remaining highly stereotypical at 37 three items of same-gender-typed play. Both boys ages 3 and 4 years, then becoming less so from age 38 and girls showed high levels of same-gender-typed 4 to 5 years. Our exploration of ethnic variation in 39 play (approximately once a week to several times a trajectories showed ethnic differences for two of the 40 week) and low levels of cross-gender-typed play four behaviors (appearance and dress-up play). 41 (approximately one to two times a month), and 42 boys and girls did not significantly differ in their Coherence in Gender-Typed Behaviors 43 levels, F(1, 175.9) = .27, ns. An exploration of differ- 44 ences by ethnic group, F(2, 176.0) = 2.56, ns, or by The second goal was to examine the multidimen- 45 gender and ethnic group, F(2, 176.0) = 2.33, ns, sional nature of gender-typing and speak to the 46 revealed no significant effects. controversy surrounding whether or not different 47 Gender-typed toy play (six-item scale) linearly aspects of gender-typing develop in concert. We 48 increased over time across ethnicity and gender, first tested whether exhibiting gender rigidity in 49 F(2, 328.1) = 6.67, p = .001. Children engaged in one behavior tended to be associated with rigidity 50 more gender-typed play at ages 4 and 5 years in other behaviors at each age by examining corre- 51 compared to age 3 years, p = .010, p < .001, respec- lations for the sample as a whole, by gender, and 52 tively. This pattern was driven by linearly decreas- by ethnicity, similar to what most of the little exist- 53 ing levels of cross-gender-typed play, p < .001 ing research has done (Martin & Ruble, 2009). We 54 = = = (Mage 3 1.40, SDage 3 .84; Mage 4 1.22, next expanded upon this literature and tested Rigidity in Gender-Typed Behaviors 9

1 Age 3 Age 4 Age 5 2 3 e1 e1 4 5 .14† .22* Appearance 6 Appearance Appearance .10 7 .06 .05 .03 8 .01 e2 e2 -.08 -.06 9 .03 .49*** -.03 .40*** 10 Dress-up Play Dress-up Play Dress-up Play -.06 11 12 .09 -.14 e3 e3 13 -.12 -.04 -.02 14 .38*** .53*** .05 15 Toy Play Toy Play Toy Play 16 .05

17 -.04 e4 e4 18 19 .10 .24* Sex Segregation 20 Sex Segregation Sex Segregation 21 22 23 24 25 Figure 2. Structural equation model showing within-age correlations and stability estimates among gender-typed behaviors. 26 Note. Standardized coefficients are depicted. Coefficients not in boldface are nonsignificant. † < * < *** < 27 p .10. p .05. p .001. 28 whether a change in one gender-typed behavior gender-typing within measures, we simultaneously 29 from 1 year to the next was related to change in tested whether gender-typing scores at one age 30 other gender-typed behaviors across the same time predicted gender-typing scores at the consecutive 31 period. The focus on change has received new age, which will be discussed later in this section. 32 attention in recent years (McArdle, 2009). On the Results indicated satisfactory model fit, v2(40) = 33 basis of the assumption that a coherent develop- 55.63, CFI = .887, RMSEA = .052. Unexpectedly no 34 mental process underlies the changes, we hypothe- predicted associations were found between any mea- 35 sized that the various behaviors would be sures at any age (Figure 2). Furthermore, multiple 36 moderately associated with each other, within and groups analyses indicated no significant improve- 37 across age. Given the number of associations we ment in model fit when separating the model by 38 wanted to investigate, and the possibility of making child gender, v2(8) = 12.57, ns, or ethnicity, 39 Type I errors, to estimate within-age associations v2(16) = 24.02, ns. 40 we conducted a structural equation models that Change-to-change associations among gender-typed 41 simultaneously included all associations. We used behaviors. We calculated difference scores for the 42 Mplus to estimate the model and used full informa- four gender-typed behaviors (e.g., dress-up play at 43 tion maximum likelihood estimation, which age 3 years from dress-up play at age 4 years) and 44 allowed participants with incomplete data to be assessed associations among these difference scores. 45 included. We also estimated the across-age correla- Three of the 12 correlations were significant 46 tions in the context of Mplus because it allowed the (Table 2). An increase in sex segregation was posi- 47 matrix of correlations to be estimated simulta- tively associated with an increase in dress-up play 48 neously and also allowed the handling of missing from age 3 to 4 years, r(176) = .17, p = .024, and with 49 data under the assumption of a data missing at an increase in gender-typed play from 4 to 5 years, 50 random pattern. r(176) = .16, p = .034. Unexpectedly, an increase in 51 Within-age correlations among gender-typed behav- sex segregation from 4 to 5 years was associated 52 iors. In the model, we tested associations among dif- with a decrease in gender-typed appearance from 53 ferent gender-typed behaviors within each age 4 to 5 years, r(176) = À .19, p = .011. There were no 54 (N = 145; Figure 2). To investigate stability of significant differences by ethnicity. However, in 10 Halim, Ruble, Tamis-LeMonda, and Shrout

1 Table 2 Discussion 2 6 Correlations Among Change Scores of Gender Rigidity Measures 3 Cognitive theories of gender development posit that 4 Measure 1 2 3 4 as young children learn about gender categories and 5 that they belong to one of these categories, they will Change between be highly motivated to conform to gender stereo- 6 consecutive ages types (Martin et al., 2002). As children actively con- 7 1. Gender-typed — .05 À .01 À .19* 8 appearance struct and consolidate information about gender, 9 2. Dress-up play .12 — À .06 À .12 their strict adherence may increase in rigidity (Ruble, 10 3. Play with .00 À .06 — .16* 1994). Do gender-typed behaviors of young children 11 gender-typed toys follow this predicted pattern of increasing rigidity? À * À 12 4. Proportion of .01 .17 .10 — The current results lend support to this conceptuali- 13 same-sex peers zation of gender development, and indicate its robustness across children from different ethnic 14 Note. N = 175. Correlations below the diagonal refer to changes 15 in variables from age 3 to 4 years. Correlations above the diago- backgrounds. Each of the four gender-typed behav- 16 nal refer to changes in variables from age 4 to 5 years. See Halim iors followed a course of change from age 3 to 17 (2012) for further details. 5 years, with nearly all behaviors moving toward fi ’ 18 greater rigidity. Speci cally, children s dress-up play, gender-typed play, and sex segregation increased in 19 examining correlations by gender we found that for rigidity from age 3 to 4 years. Children’s observed 20 boys only, increases in gender-typed appearance gender-typed appearance was already high at age 3, 21 were positively associated with increases in gender- but did not increase in level. Findings point to the 22 typed play at the same time interval, r (93) = .31, 3–4 transition from age 3 to 4 years as an especially 23 p = .002; r (93) = .20, p = .055. Surprisingly, for 4–5 important time in children’s gender development for 24 girls these relations trended in the negative direction, the active consolidation of gender information. This 25 r (83) = À.21, p = .055; r (93) = À.19, p = .087, 3–4 4–5 consolidation manifests in multiple aspects—from 26 and differed from boys’ correlations (p < .001, gendered activities to the peers with whom children 27 p = .004, respectively). 28 play. The period from age 4 to 5 years showed more 29 divergence. Sex segregation trended toward increas- 30 Stability of Gender-Typing Within Measures ing further, corroborating past, largely cross-sec- tional, research on middle-class White children 31 Stability in children’s gender-typing was tested (Ruble et al., 2006). In contrast, dress-up play and 32 in the structural equation model as above (Figure 2) gender-typed play remained unchanged from age 4 33 with gender-typing scores at one age predicting to 5 years, although cross-gender-typed play 34 gender-typing scores at the consecutive age. Nota- decreased. Gender-typed appearance decreased in 35 bly, both gender-typed dress-up play and gender- rigidity, suggesting growing flexibility. 36 typed play showed considerable stability across In light of the behavioral rigidity we observed, 37 consecutive ages (dress-up: b = .49, p < .001, age 3–4 an interesting juxtaposition is that during the same 38 b = .40, p < .001; play: b = .38, p < .001, age 4–5 age 3–4 developmental period children show flexibility in 39 b = .53, p < .001). Gender-typed appearance age 4–5 some gender-related judgments. For example, 40 (b = .22, p = .010) and sex segregation (b = .24, research has shown that a majority of preschoolers 41 p = .012) were only significantly related from age viewed exclusion from gender-stereotyped activities 42 4 to 5 years. The estimates from age 3 to 4 years based on gender (e.g., girls excluding a boy from 43 were smaller than those from age 4 to 5 years and playing dolls with them because he is a boy) as 44 were only a trend for appearance (b = .14, unfair and wrong (Killen, Pisacane, Lee-Kim, & 45 p = .083), but not significant for sex segregation Ardila-Rey, 2001; Theimer, Killen, & Stangor, 2001). 46 (b = .10, p = .237). Multiple group analyses indi- Furthermore, the judgment that exclusion based on 47 cated no significant improvement in overall model gender is wrong becomes more pronounced, com- 48 fit when examining separate models for each gen- plex, and more based on moral reasoning rather 49 der or for each ethnicity as reported above. How- than on social convention as children develop 50 ever, examination of the data revealed that the 51 through elementary school (Killen, 2007). In addi- stability of gender-typed appearance was evident ’ 52 = tion, around age 5 or 6 years, children s gender- for girls, even from age 3 to 4 years, (bage 3–4 .31, fl 53 = = = stereotype beliefs start to grow in their exibility p .001, bage 4–5 .32, p .002) but not for boys ’ 54 = = = = (Trautner et al., 2006). Thus, at times, children s 7 (bage 3–4 .02, p .894, bage 4–5 .13, p .364). Rigidity in Gender-Typed Behaviors 11

1 gender-typed behaviors might misalign with their occur between Dominican and African American 2 gender-related beliefs, perhaps due to feeling pres- groups because some Dominicans are perceived 3 sure from peers to conform to gender norms in and/or identify as Black and also because of the 4 their behavior. This is an interesting area for future influence of African American hip hop culture on 5 investigation and highlights the importance of Dominican youth (Duany, 2008; Torres-Saillant & 6 examining in future research whether the develop- Hernandez, 1998). The uniqueness of Mexican chil- 7 mental shifts we observed are linked to children’s dren’s trajectories may have been due to less accul- 8 growing understanding of gender, as cognitive turation. Although both Dominican and Mexican 9 theories would predict, as it is possible that other families were largely first generation, Dominicans 10 perspectives (e.g., socialization, biological) could resided in the United States for more years than 11 also explain rigidity in gender-typed behaviors. Mexicans, and a small proportion of them (15%) 12 Children’s physical appearance may be the first were second generation. 13 to move toward flexibility because it may peak in 14 rigidity earlier as well, as physical appearance is Group Similarities and Differences in Levels of 15 one of the most salient traits to children in early Gender-Typing 16 8 childhood (Ruble & Dweck, 1995). Indeed, research 17 on the “pink, frilly dress” phenomenon and girls’ A strength of this study was our inclusion of 18 extremely intense interest in playing dress-up sug- multiple measures of gender-typed behavior, as 19 gests that this appearance rigidity could start as stressed by the matrix of gender-typing (Ruble 20 early as age 2 (DeLoache, Simcock, & Macari, 2007; et al., 2006). This strength became evident when we 21 Halim et al., 2012). Gender-related cognitions, espe- examined group differences in levels of gender-typ- 22 cially gender stereotypes, tend to increase in rigid- ing. If we had relied on only one measure, we 23 ity then grow more flexible (e.g., Trautner et al., would have erroneously concluded that one group 24 2005), and interestingly, in this same study, chil- was more gender-typed than the other, when, in 25 dren who showed earlier rigidity showed earlier fact, each group manifested gender rigidity in 26 flexibility. Findings here suggest that the trajectory different behaviors. 27 of gender-typed appearance is similar to that of Gender similarities and differences. Notably, we 28 gender-related cognitions. found differences in the ways that boys and girls 29 Patterns in the timing and shape of trajectories expressed their gender. Girls engaged in dress-up 30 for the four behaviors were robust across gender. play more than did boys. And although we could 31 That no gender differences in the timing or shape not formally test differences in gender-typed 32 of any trajectory for any of the four behaviors were appearance, girls appeared to find a larger number 33 seen is somewhat surprising as some research on of ways to dress in feminine ways compared to 34 White, middle-class children would suggest that boys. Thus in two of the four measures, girls’ mani- 35 girls sometimes precede boys in gender develop- fested their gender rigidity in self-adornment rather 36 9 ment (La Freniere, Strayer, & Gauthier, 1984; Signo- than in what they play with or with whom they 37 rella et al., 1993; Zosuls et al., 2009). Perhaps in play. Prior research has suggested that boys some- 38 Latino and African American cultures, masculinity times show stronger gender-typed behavior com- 39 (i.e., machismo) is strongly emphasized so that gen- pared to girls, but these studies focused largely on 40 der development for boys occurs as early as it does play (Huston, 1983; Perry, White, & Perry, 1984; 41 for girls (e.g., DeSouza et al., 2004; Rowan, Pernell, Turner, Gervais, & Hinde, 1993). Our results sug- 42 & Akers, 1996). gest that girls’ adherence to maintaining a feminine 43 In terms of ethnic variation, Dominican and appearance and in pretending to be princesses 44 African American children were similar in their should also be examined as important gendered 45 developmental trajectories. Mexican children differ behaviors. Future research should examine the 46 the most. For some measures (gender-typed appear- broader societal implications of this early emphasis 47 ance, dress-up play), Mexican children’s trajectories on physical appearance for girls. If girls tie their 48 suggest that their peaks in rigidity and subsequent self-worth to one’s physical appearance this could 49 flexibility in gender-typed behavior occurred earlier. potentially put girls at risk for psychological dis- 50 The concordance in the trajectories of African tress later in life (Crocker & Wolfe, 2001; Nolen- 51 American and Dominican children may be due to Hoeksema, 2001). 52 possessing similar demographics (e.g., mother’s Consistent with past research, boys and girls did 53 education, father cohabitation, urban background). not differ in levels of sex segregation (Fabes, 54 Alternatively, more cross-cultural influence may Martin, & Hanish, 2003; Howes, 1988). They also 12 Halim, Ruble, Tamis-LeMonda, and Shrout

1 did not differ in levels of gender-typed play Coherence of Gender-Typing 2 (including cross-gender-typed play), which may be 3 surprising. Studies have shown that White, middle- Whether or not different dimensions of gender- 4 class grade-school boys compared to girls tend to typing show coherence within a single individual 5 reject cross-gender-typed behaviors and preferences has long been debated (Huston, 1983). We found no 6 more (Pickering & Repacholi, 2001; Raskin & Israel, significant correlations among the different behav- 7 1981). The lack of a gender difference in our sample iors within each age even though at the group level, 8 could possibly be due to the young age or to the 3 of the 4 behaviors increased in rigidity from ages 3 9 different ethnic background and class of our sam- to 4 years. Our findings support studies suggesting 10 ple. This is an area for future research. that gender-typing dimensions are distinct, at least 11 Cultural similarities and differences. Another when assessed within age. When we examined 12 strength of this study was the diversity of the change across successive years, a different picture 13 sample. We explored whether Latino children emerged. Increases in sex segregation were positively 14 would have higher levels of gender-typed behav- correlated with increases in dress-up play and gen- 15 iors compared to African American children based der-typed play at certain time intervals consistent 16 on literature suggesting egalitarian gender roles in with past research on the bidirectional link between 17 African American families (e.g., Hill, 1971). Results gender-typed play and sex segregation (Martin & 18 largely indicated no differences in levels. Indeed, it Fabes, 2001). It was surprising that changes in 19 was remarkable that African American and Latino gender-typed appearance showed no connections to 20 children often looked the same. Both African changes in dress-up play and further surprising that 21 American and Latina girls exhibited similar levels changes in gender-typed appearance showed no 22 of dress-up play, gender-typed play, and sex seg- positive connections to any other behavior for girls. 23 regation. And boys, as we mentioned earlier, In fact, increases in gender-typed appearance were 24 showed no ethnic group differences. These results related to decreases in gender-typed play for girls 25 indicate that rigidity in gendered behaviors is and decreases in sex segregation for both boys and 26 robust and generalizable to children of multiple girls from age 4 to 5 years. Perhaps these negative or 27 ethnic backgrounds. null relations are due to the disparate trajectories of 28 One exception was that Dominican girls dressed each behavior with gender-typed appearance show- 29 in more feminine ways than did Mexican girls. ing an earlier period of rigidity. Alternatively per- 30 Qualitative interviews with Caribbean American haps peers are rejecting children who look highly 31 communities have revealed persistent traditionalism gender-stereotypical at later ages. This is a question 32 in gender roles concerning household responsibili- for future research. 33 ties and liberties (Lopez, 2002). These values could The different conclusions drawn from within- 34 translate into the encouragement toward Dominican time versus across-time associations between differ- 35 girls to maintain a highly feminine appearance. Of ent gender-typed behaviors may shed light on the 36 course, in interpreting differences, it is important to conundrum of gender-typing coherence (Martin & 37 recognize that although groups in our sample were Ruble, 2009). On one hand, the within-time data 38 similar on a number of demographic variables and suggest that at one point in time, children may 39 representative of the ethnic groups in New York express their gender in a variety of ways. A child 40 City, a limitation of our study was that not all who dresses in an extremely feminine way does not 41 demographic variables were identical. Future necessarily play with only dolls and kitchen sets. 42 research should attempt to equalize other demo- The lack of associations may also indicate that 43 graphic variables to determine the cause of these different behaviors emerge at different times in 44 ethnic group differences. response to different antecedents (e.g., cognitive 45 Why we found no ethnic differences in boys is a development or social pressures). On the other 46 question for future research. Perhaps both Latino hand, the across-time data suggest that gender-typ- 47 and African American cultures value masculinity in ing shows some unity with some gender-typed 48 boys to such a high degree (DeSouza et al., 2004; behaviors changing in parallel to others for certain 49 Rowan et al., 1996) that it would mask any cultural groups and at certain age intervals. Overall then, 50 variation. Indeed, the boys often wore baggy cloth- gender-typed behaviors seem to exhibit specificity 51 ing with contrasting red and black and graffiti in one sense (within-time), but more cohesion in 52 designs suggesting an early “toughness” that may another (across-time). These data stress the impor- 53 symbolize masculinity as young as age 3 years in tance of looking at associations between variables 54 an urban setting. across time in future research. Rigidity in Gender-Typed Behaviors 13

1 may already reflect their preferences for feminine Stability in Gender-Typing 2 clothing, whereas boys’ appearances may not reflect 3 Did children show stability in specific gender- their own preferences until later ages. 4 typed behaviors across time? Or is gender-typing These findings argue for continued research on 5 primarily a group-level rather than an individual- individual-level differences in gender-typing, not 6 level phenomenon (Maccoby, 2002)? We found just group-level ones. They suggest that even 7 stability in dress-up play and gender-typed play though gendered behaviors do not necessarily align 8 across successive ages for the whole sample. Girls in a constellation of gender-typing within a certain 9 also showed stability in their gender-typed appear- age, when just one behavior is examined, that 10 ance. In addition, children’s sex segregation was behavior shows stability within an individual girl 11 stable between the ages of 4 and 5 years. R’s ran- or boy over time. Future research should examine 12 ged from .18 to .47. These associations further sup- whether these individual differences during ages 13 port our claim that children may adopt 3–5 years have downstream influences on children’s 14 individualized ways of displaying gender through identity, behavior, preferences, and values later in 15 play, appearance, and dress-up play that continue life. For example, it would be interesting to see 16 over time. Moreover, these stability coefficients whether girls who are low on gender-typing tend 17 were robust and generalizable across different to become tomboys in middle childhood or whether 18 ethnic groups. girls who are high on gender-typing tend to dis- 19 The lack of stability for sex segregation from age engage with mathematics and science in adoles- 20 3 to 4 years is consistent with some past research cence. These future directions would further speak 21 (Maccoby & Jacklin, 1974), but not with other to possible antecedents of gender-typing and allow 22 research (Martin & Fabes, 2001; Serbin et al., 1993). us to further examine stability within the context of 23 Perhaps this is not surprising given that our normative increasing rigidity followed by flexibility 24 research design was most consistent with Maccoby later in development. 25 and Jacklin’s (1987) study which examined children 26 over a 2-year period from age 4½ to 6½ years. The Limitations 27 lack of sex segregation at age 3 years may partly 28 explain this lack of stability—the mean percent of It is important to note that mothers provided 29 same-gender friends was not significantly different data for several of the behaviors we examined, with 30 from 50%. This suggests that sex segregation in the exception of observed gender-typed appearance. 31 low-income African American and Latino popula- We stress the importance of continued research on 32 tions emerges later than what is found in White, these behaviors using more observational tech- 33 middle-class samples (around 27 months for girls, niques to replicate our results. In addition, although 34 36 months for boys; Ruble et al., 2006). On another the developmental trajectories we observed support 35 note, the lack of stability in sex segregation from the hypotheses of cognitive theories of gender 36 age 3 to 4 years, and the lower level of stability development, it is important to recognize other per- 37 from age 4 to 5 years, compared to the other three spectives on gender development, such as biological 38 behaviors, hints at the possibility that sex segrega- and socialization perspectives (Ruble et al., 2006). 39 tion may reflect different underlying processes. Parents’ own adherence to gender norms may influ- 40 Perhaps sex segregation is less of a reflection of the ence the importance they place on children’s gender 41 internally motivated, child-directed consolidation of conformity, and families make decisions that influ- 42 gender, and more of a reflection of socialization ence what their children wear, what they play with, 43 and of children’s experiences in playing with the and with whom they play. Nevertheless, some 44 other gender, such as girls’ dislike for boys’ tenden- studies have shown that parent influence on chil- 45 cies toward rough-and-tumble play (Pellegrini & dren’s appearances is minimal across ages 3–6 years 46 Smith, 1998). (Halim et al., 2012), and children are likely to have 47 The lack of stability in appearance for boys is considerable independence from parents in play 48 intriguing as well and opens up directions for future and peer choices in preschool contexts. However, 49 research. Perhaps appearances are not as central to exploring other influences is an important area of 50 boys in expressing their gender compared to other future research. 51 behaviors such as play (Miller, Lurye, Zosuls, & In addition, another limitation of our study was 52 Ruble, 2009). It is also possible that girls get to our inability to examine acculturation’s influence on 53 choose their clothing at earlier ages than boys do. gender-typed behavior. In our sample, ethnicity 54 Thus, starting from age 3 years, girls’ appearances was confounded with immigration status as all 14 Halim, Ruble, Tamis-LeMonda, and Shrout

1 African American children were third generation or References 2 later, and all Latino children were first or second Albert, A. A., & Porter, J. R. (1988). Children’s gender- 3 generation. Even within the Latino group, the vast 4 role stereotypes: A sociological investigation of psycho- majority of children had foreign-born parents, so logical models. Sociological Forum, 3(2), 184–210. 5 fi we were unable to compare rst versus second gen- doi:10.1007/BF01115290 6 eration children. It would be interesting for future Bailey, B. A., & Nihlen, A. S. (1990). Effect of experience 7 research to examine whether acculturation plays a with nontraditional workers on psychological and 8 role in the level and timing of gender-typing. For social dimensions of occupational sex-role stereotyping 9 example, it is possible that since American culture by elementary school children. Psychological Reports, 66, 10 values equality, including gender equality (Tamis- 1273–1282. doi:10.2466/PR0.66.4.1273-1282 11 Lemonda & McFadden, 2009), later generation chil- Chuang, S. S., & Tamis-LeMonda, C. S. (2009). Gender 12 dren would show lower levels of gender-typing roles in immigrant families: Parenting views, practices, and child development. Sex Roles, 60, 451–455. 13 and later peaks in gender-typing. At the same time doi:10.1007/s11199-009-9601-0 10 14 it is important to note that all of the children in our 15 Collins, P. H. (1990). Black feminist thought: Knowledge, con- sample were part of a large urban context, and sciousness, and the politics of empowerment. Boston: 16 indeed, even part of a larger United States context. Unwin Hymen. 11 17 The levels and patterns of gender-typed behavior Crocker, J., & Wolfe, C. T. (2001). Contingencies of self- 18 we observed may reflect this larger context as well. worth. Psychological Review, 3, 593–625. doi:10.1037/ 19 0033-295X.108.3.593 20 DeLoache, J. S., Simcock, G., & Macari, S. (2007). Planes, 21 trains, automobiles–and tea sets: Extremely intense 22 Conclusion interests in very young children. Developmental Psychol- 23 ogy, 43, 1579–1586. doi:10.1037/0012-1649.43.6.1579 A child can express gender in multiple behaviors— 24 DeSouza, E. R., Baldwin, J., Koller, S. H., & Narvaz, M. through what they wear, through the costumes they 25 (2004). Latin American perspective on the study of gen- don, through the objects they play and do not play der. In M. A. Paludi (Ed.), The psychology of gender (pp. 26 41–68). Westport, CT: Praeger. 27 with, and through the peers that they befriend. On Duany, J. (2008). Quisqueya on the Hudson: The transna- 28 a group level, most of these behaviors generally showed similar trajectories from age 3 to 4 years, tional identity of Dominicans in Washington Heights. New 29 York: CUNY Dominican Studies Institute. 30 with children increasingly acting in gender-stereo- typical ways, consistent with predictions based on Fabes, R. A., Martin, C. L., & Hanish, L. D. (2003). Young 31 children’s play qualities in same-, other-, and mixed-sex cognitive theories of development. From age 4 to 32 peer groups. Child Development, 74, 921–932. 5 years, the behaviors diverged. The timing and 33 doi:10.1111/1467-8624.00586 levels of behavior sometimes varied by ethnic 34 Golombok, S., Rust, J., Zervoulis, K., Croudace, T., Gol- 35 group, highlighting the importance of studying gen- ding, J., & Hines, M. (2008). Developmental trajectories 36 der development in multiple, diverse groups. of sex-typed behavior in boys and girls: A longitudinal 37 Despite group-level parallels, at the individual general population study of children aged 2.5-8 years. Child Development, 79, 1583–1593. doi:10.1111/j.1467- 38 level, children who were highly gender-typed in one way (e.g., wearing a lot of pink dresses) were 8624.2008.01207.x 39 not necessarily gender-typed in another way (e.g., Halim, M. L. (2012). Gender-typed behavior in ethnically diverse 40 playing frequently with dolls). Nevertheless, there young children: Developmental trajectories, cognitive under- 41 was some coherence across pairs of gender-typed pinnings, and consequences for gender identity and attitudes 42 (Doctoral dissertation). New York: New York University. behaviors in change across time, as well as moder- 43 Halim, M. L., Ruble, D. N., Lurye, L., Greulich, F., Zosuls, ate stability in individual gendered behaviors from 44 fi K., & Tamis-LeMonda, C. S. (2012). The case of the 45 year to year. In short, our ndings contribute to the pink frilly dress and the avoidance of all things “girly”: ’ ’ 46 ongoing conversation about developmental trajecto- Girls and boys appearance rigidity and cognitive theo- ries of gender-typing, multidimensional nature, and 47 ries of gender development. Under review. stability of individual differences in an ethnically Hill, R. B. (1971). The strengths of Black families. New York: 48 diverse sample through a longitudinal design. The Emerson Hall. 49 examination of young children’s gender-typed Howes, C. (1988). Relations between child care and 50 behaviors provided a snapshot into a dynamic, schooling. Developmental Psychology, 24,53–57. 51 doi:10.1037/0012-1649.24.1.53 52 exciting period of gender development, involving both stability and change that could have implica- Huston, A. C. (1983). Sex-typing. In E. M. Hetherington 53 (Ed.), Handbook of child psychology: Socialization, personality, 54 tions for gender throughout the life course. Rigidity in Gender-Typed Behaviors 15

1 and social development (Vol. 4, pp. 387–467). New York: Nolen-Hoeksema, S. (2001). Gender differences in depres- 2 Wiley. sion. Current Directions in Psychological Science, 10, 3 Jarret, R., Roy, K., & Burton, L. (2002). Fathers in the 173–176. doi:10.1111/1467-8721.00142 4 “Hood”: Insights from qualitative research on Parsons, A., & Howe, N. (2006). Superhero toys and boys’ 5 low-income African-American men. In C. S. Tamis- physically active and imaginative play. Journal of 6 LeMonda, N. Cabrera (Eds.), Handbook of father Research in Childhood Education, 20, 287–300. 7 involvement: Multidisciplinary perspectives (pp. 211–248). doi:10.1080/02568540609594568 8 Mahwah, NJ: Erlbaum. Pellegrini, A. D., & Smith, P. K. (1998). Physical active Killen, M. (2007). Children’s social and moral reasoning play: The nature and function of a neglected aspect of 9 about exclusion. Current Directions in Psychological Sci- play. Child Development, 69(3), 577–598. doi:10.1111/ 10 ence, 16, 32–36. j.1467-8624.1998.tb06226.x 11 Killen, M., Pisacane, K., Lee-Kim, J., & Ardila-Rey, A. Perry, D. G., White, A. J., & Perry, L. C. (1984). Does 12 (2001). Fairness or stereotypes? Young children’s priori- early sex typing result from children’s attempts to 13 ties when evaluating group exclusion and inclusion. match their behavior to sex role stereotypes? Child 14 Developmental Psychology, 37, 587–596. doi:10.1037/ Development, 55, 2114–2121. doi:10.2307/1129784 15 0012-1649.37.5.587 Pickering, S., & Repacholi, B. (2001). Modifying children’s 16 Liben, L. S., & Bigler, R. S. (2002). The developmental gender-typed musical instrument preferences: The 17 course of gender differentiation. Monographs of the Soci- effects of gender and age. Sex Roles, 45, 623–643. 18 ety for Research in Child Development, 67(2, Serial No. doi:10.1023/A:1014863609014 12 – 19 ???), 1 147. doi:10.1111/1540-5834.00189 Raskin, P. A., & Israel, A. C. (1981). Sex-role imitation in Lopez, N. (2002). Hopeful girls, troubled boys: Race and gen- children: Effects of sex of child, sex of model, and sex- 20 13 der disparity in urban education. ????, NY: Routledge. role appropriateness of modeled behavior. Sex Roles, 7, 21 Maccoby, E. E. (2002). Gender and group process. Current 1067–1077. doi:10.1007/BF00287584 22 Directions in Psychological Science, 11,54–58. Rowan, G. T., Pernell, E. J., & Akers, T. A. (1996). Gender 23 doi:10.1111/1467-8721.00167 role socialization in African American men: A concep- 24 Maccoby, E. E., & Jacklin, C. N. (1974). The psychology of tual framework. Journal of African American Men, 1, 25 sex differences. Stanford, CA: Stanford University Press. 3–22. doi:10.1007/BF02733916 26 Martin, C. L., & Fabes, R. A. (2001). The stability and con- Ruble, D. N. (1994). A phase model of transitions: Cogni- 27 sequences of young children’s same-sex peer interac- tive and motivational consequences. In M. Zanna (Ed.), 28 tions. Developmental Psychology, 37, 431–446. Advances in experimental social psychology (pp. 163–214). 29 doi:10.1037/0012-1649.37.3.431 New York: Academic Press. 30 Martin, C. L., Fabes, R. A., Evans, S. M., & Wyman, H. Ruble, D. N., Lurye, L. E., & Zosuls, K. M. (2007). Pink (1999). Social cognition on the playground: Children’s frilly dresses (PFD) and early gender identity. Princeton 31 beliefs about playing with girls versus boys and their Report on Knowledge (P-ROK), 2(2), ???–???. Retrieved 32 relations to sex segregated play. Journal of Social and ???? ???, ????, from http://www.princeton.edu/prok/ 33 Personal Relationships, 16, 751–771. doi:10.1177/ issues/2-2/pink_frilly.xml 15 34 0265407599166005 Ruble, D. N., & Martin, C. L. (1998). Gender develop- 35 Martin, C. L., & Ruble, D. N. (2004). Children’s search for ment. In W. Damon (Ed.), Handbook of child psychology 36 gender cues: Cognitive perspectives on gender develop- (Vol. 3, 5th ed., pp. 933–1016). New York: Wiley. 37 ment. Current Directions in Psychological Science, 13, Ruble, D. N., Martin, C., & Berenbaum, S. (2006). Gender 38 67–70. doi:10.1111/j.0963-7214.2004.00276.x development. In W. Damon (Ed.), Handbook of child psy- 39 Martin, C. L., & Ruble, D. N. (2009). Patterns of gender chology (Vol. 3, 6th ed., pp. 858–932). Hoboken, NJ: 40 development. Annual Review of Psychology, 61, 353–381. Wiley. 41 doi:10.1146/annurev.psych.093008.100511 Ruble, D. N., Taylor, L. J., Cyphers, L., Greulich, F. K., Martin, C. L., Ruble, D. N., & Szkrybalo, J. (2002). Cogni- Lurye, L. E., & Shrout, P. E. (2007). The role of gender 42 tive theories of early gender development. Psychological constancy in early gender development. Child Develop- 43 Bulletin, 128, 903–933. doi:10.1037/0033-2909.128.6.903 ment, 78, 1121–1136. doi:10.1111/j.1467-8624.2007. 44 McAdoo, J. L. (1988). The roles of Black fathers in the 01056.x 45 socialization of Black children. In H. P. McAdoo (Ed.), Serbin, L. A., Powlishta, K. K., & Gulko, J. (1993). The 46 14 Black families (pp. 257–269). Newbury Park, CA: Sage. development of sex-typing in middle childhood. Mono- 47 McArdle, J. J. (2009). Latent variable modeling of differ- graphs of the Society for Research in Child Development, 48 ences and changes in longitudinal data. Annual Review 58(2, Serial No. 232). 49 of Psychology, 60, 577–605. doi:10.1146/annurev. Servin, A., Bohlin, G., & Berlin, L. (1999). Sex differences ’ 50 psych.60.110707.163612 in 1-, 3-, and 5-year-olds toy-choice in a structured 51 Miller, C. F., Lurye, L. E., Zosuls, K. M., & Ruble, D. N. play-session. Scandinavian Journal of Psychology, 40, – 52 (2009). Accessibility of gender stereotype domains: 43 48. doi:10.1111/1467-9450.00096 Developmental and gender differences in children. Sex Signorella, M. L., Bigler, R. S., & Liben, L. S. (1993). Devel- 53 Roles, 60, 870–881. doi:10.1007/s11199-009-9584-x opmental differences in children’s gender schemata 54 16 Halim, Ruble, Tamis-LeMonda, and Shrout

1 about others: A meta-analytic review. Developmental or differential? Infant and Child Development, 14, 365– 2 Review, 13, 147–183. doi:10.1006/drev.1993.1007 381. doi:10.1002/icd.399 3 Stanback, M. H. (1985). Language and Black ’s Turner, P. J., Gervais, J., & Hinde, R. A. (1993). Gender- 4 place: Evidence from the Black middle class. In P. A. typing in young children: Preferences, behavior and 5 Treichler, C. Kramarae, B. Stafford (Eds.), For alma cultural differences. British Journal of Developmental 6 mater: Theory and practice in feminist scholarship (pp. 177– Psychology, 11, 323–342. doi:10.1111/j.2044-835X.1993. 7 193). Urbana, IL: University of Illinois Press. tb00607.x 8 Tamis-Lemonda, C. S., & McFadden, K. E. (2009). The U.S. Census Bureau. (2011). Statistical abstract of the United States of America. In M. H. Bornstein (Ed.), United States, population. Retrieved April 13, 2011, 9 Handbook of cultural developmental science (pp. 299–322). from http://www.census.gov/compendia/statab/2011 10 16 ????, NY: Psychology Press. /tables/11s0005.pdf. 11 Theimer, C. E., Killen, M., & Stangor, C. (2001). Young Zosuls, K. M. (2008). Gender development in African Ameri- 12 children’s evaluation of exclusion in gender-stereotypic can, Dominican, and Mexican immigrant toddlers: Self 13 peer contexts. Developmental Psychology, 37, 18–27. socialization in three cultural contexts (Doctoral disserta- 14 doi:10.1037/0012-1649.37.1.18 tion). New York: New York University. 15 Torres-Saillant, S., & Hernandez, R. (1998). The Dominican Zosuls, K. M., Ruble, D. N., Tamis-LeMonda, C. S., 16 17 Americans. ?????, CT: Greenwood Press. Shrout, P. E., Bornstein, M. H., & Greulich, F. K. (2009). 17 Trautner, H. M., Ruble, D. N., Cyphers, L., Kirsten, B., The acquisition of gender labels in infancy: Implications fl 18 Behrendt, R., & Hartmann, P. (2005). Rigidity and exi- for sex-typed play. Developmental Psychology, 45, 688– 19 bility of gender stereotypes in children: Developmental 701. doi:10.1037/a0014053 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54