Parent Gender Differences in Emotion Socialization Behaviors Vary by Ethnicity and Child Gender

SYNOPSIS

Objective

This study examined ethnicity (African American, European American, and Lumbee American Indian) and child gender as moderators of gender differences in parents’ emotion socialization behaviors.

Design

Mothers and fathers from two samples responded to questionnaires assessing self-expressiveness in the family (N=196) or reactions to children’s negative emotions (N=299).

Results

Differences between mothers and fathers varied as a function of ethnicity. Mothers and fathers showed similar levels of negative expressiveness in European American and African American families, whereas fathers were more negatively expressive than mothers in Lumbee families. Mothers reported more supportive reactions than fathers among European Americans and Lumbees, but African American mothers and fathers reported nearly equal levels of supportive reactions. Parent gender x ethnicity interactions were further moderated by child gender. Mothers were generally more supportive of girls’ negative emotions than fathers across all ethnicities. For boys, however, parent gender differences in supportive reactions to negative emotions varied by ethnicity. Mothers were more supportive than fathers among European American parents of boys, but mothers were less supportive than fathers among African American parents of boys.

Conclusions

Results highlight the contextualized nature of emotion socialization, and the need to consider ethnicity and child gender as influences on mothers’ and fathers’ gender-specific emotion socialization.

INTRODUCTION

Parents’ emotion socialization strategies are central contributors to child development (e.g., Denham, Bassett, & Wyatt, 2007; Eisenberg, Cumberland, & Spinrad, 1998; McDowell, Kim, O’Neil, & Parke, 2002). Two types of strategies particularly noted for their impact are parents’ emotional expressiveness in the family, defined as parents’ positive and negative styles of expression when communicating verbally and nonverbally in their families (Halberstadt, Cassidy, Stifter, Parke, & Fox, 1995), and parents’ reactions to children’s emotions, defined as parents’ supportive and nonsupportive reactions in response to their children’s emotional displays (Fabes, Leonard, Kupinov, & Martin, 2001). These two sets of socialization behaviors are related to children’s emotional experience, expression, understanding, self-regulation, and relationships with peers and family (see reviews by Dunsmore & Halberstadt, 1997; Eisenberg et al., 1998; Halberstadt, Crisp, & Eaton, 1999; Morris, Silk, Steinberg, Myers, & Robinson, 2007).

Receiving less attention, however, have been the myriad factors that shape parents’ emotion socialization behaviors. Cultural- and gender-related factors may be two important influences (Baker & Crnic, 2005; Camras, Kolmodin, & Chen, 2008; Halberstadt & Lozada, 2011; Wong, McElwain, & Halberstadt, 2009). By exploring these factors, both singularly and in conjunction with one another, we hope to reach a greater understanding of the contextual nature of parental emotion socialization and the unique pathways that may differentiate mothers’ and fathers’ treatment of children’s emotion. The present study examined two types of emotion socialization behaviors of mothers and fathers among African American, European American, and Lumbee American Indian families.

Parent Gender Differences in Emotion

Parent gender is widely thought to influence emotion socialization behaviors. In middle-class, European American samples, mothers have been consistently found to be more emotionally expressive than fathers (e.g., Brody, 2000; Dunsmore, Her, Halberstadt, & Perez-Rivera, 2009; Halberstadt et al., 1995; Wong et al., 2009) and more supportive and less nonsupportive of their children’s negative emotions than fathers (Cassano, Perry-Parrish, & Zeman, 2007; Nelson, O’Brien, Blankson, Calkins, & Keane, 2009; Wong et al., 2009). These differences may arise out of fathers’ tendency to mask emotions more than mothers (Dunsmore et al., 2009) or greater discomfort with emotions, which may lead to a pattern in which nonsupport is preferred over the positive acknowledgement of emotions inherent in supportive reactions (Brody, 2000).

Although many lay people and scientists alike have posited that parent gender differences in emotion are innate, there is a growing recognition that emotion experience and expression are to a great degree socially constructed and heavily influenced by cultural context (Brody, 1997; Matsumoto & Hwang, 2012; Shields, 2007). Despite increased attention to the importance of culture on emotion, most studies of emotion socialization continue to rely on primarily middle-class, European American samples without considering how other sociocultural factors (e.g., class, ethnicity) may provide lenses through which emotion socialization is filtered (for exceptions, see Brody, 1997; Fischer, Rodriguez Mosquera, van Vianen, & Manstead, 2004; Leu et al., 2010). As a consequence, the current body of research may exaggerate the universality of emotions, inadvertently giving credence to the trait-like assumption of emotional differences between men and women. In accordance with recent calls to examine contextual factors that may moderate gender-related differences (Shields, 2013), we sought to better assess the degree to which differences in mothers’ and fathers’ emotion-related behaviors vary by ethnicity.

Ethnicity and Gender Differences in Emotion

Current theoretical perspectives argue that socio-cultural norms at least partially account for emotion-related gender differences among European American parents (e.g., Brody, 1997; Durik et al., 2006; Shields, 2007; Wood & Eagly, 2002). Indeed, the more a culture places value on or assumes biological explanations, the more likely gender differences in behavior will exist in that culture (Oyserman & Markus, 1993; Shields, 2013). Thus, it is not surprising that European American women are more emotionally expressive than European American men (Brody, 1997), given the long-standing belief in European American culture that men are more emotionally rational and stoic relative to women, who are considered to be more emotional and nurturing (Shields, 2013). These beliefs may well promote the gender differences in emotional expression (Brody, 1997; Fischer & Manstead, 2000; Shields, 2002; Shields & Warner, 2008; Williams & Best, 1997), but what is unknown is whether these differences are so contextualized that they might vary between sub-cultures.

Because European American adults display more extreme gender stereotypes of emotion compared to African American, Latin American, and Asian American adults (Durik et al., 2006), European American parents may engage in relatively more differentiated emotion socialization practices than parents of other ethnicities. To date no studies directly examine this question, but we note that mothers’ and fathers’ reactions to children’s emotion in a Mexican American sample were very similar (Gamble, Ramakumar, & Diaz, 2007), whereas gender differences predominate in largely European American samples. The present study sought to examine whether gender differences between mothers’ and fathers’ emotion socialization practices were comparable across three different ethnic groups: African American, European American, and Lumbee American Indian.

We chose to study African American adults because they appear to exhibit less gender-stereotyped behavior than their European American counterparts. This lack of gender stereotyping among African Americans may be due in part to social and economic histories in which men’s and women’s participation in the work force and responsibilities as parents have been less gender differentiated (Durik et al., 2006; Harris, 1996; Hofferth, 2003). Research suggests that groups historically perceived to be of higher status, such as European Americans, might endorse more traditional stereotypes than historically under-represented groups, such as African Americans, who may be less traditional in their beliefs about gender and emotion (e.g., Rowley, Kurtz-Costes, Mistry, & Feagans, 2007). Moreover, these ethnic differences appear to be reflected in household work, with more equitable divisions of household labor among African American than European American couples (Bianchi, 2000; John & Shelton, 1997). In general, African Americans (and African American women, in particular) tend to be more flexible in their attitudes toward gender, more egalitarian in their gender roles, and more critical of gender inequality than European Americans (see Kane, 2000 for a review). Although speculative, the available evidence suggests that African American mothers and fathers might be more closely aligned in their emotion-related socialization behaviors than European American mothers and fathers.

The opportunity to study the Lumbee American Indian population has not previously been available to gender or emotion researchers. Anecdotal reports from Lumbee informants and a few published reports highlight the deep interconnectedness of the Lumbee to place and people; the Lumbee have lived in the county in which they were studied for at least 350 years and are deeply rooted there (Angell & Jones, 2003; Dial, 1993). Consequently, among the Lumbee, relationships within the community are highly important and families know their histories (Blu, 1980). The American Indian values of self-sufficiency and respect are understood within the context of strong family, church, and community levels of interdependence, and traditional familial structures emphasize fathers as responsible for income and matters outside of the household and mothers for matters in the home, including socialization of children toward kin and church (Angell & Jones, 2003; Bryant, personal communication (May 3, 2012); Bryant & LaFromboise, 2005). These traditional familial structures may parallel the socialization patterns of traditional European American life in which fathers still tend to earn substantially more than mothers, and mothers are still more responsible for their children’s socio-emotional development (Bianchi, 2011; Bird, 1997). Although the dearth of research on emotion socialization among Lumbee parents does not allow for a priori hypotheses, the present study explored whether gendered patterns of emotion socialization within Lumbee families are similar to or different from those of European American parents.

Child Gender and Parent Gender Differences in Emotion

Child gender may also be an important characteristic when considering parents’ emotion socialization practices (Eisenberg et al., 1998). Certainly, European American boys and girls differ in ways similar to adults along dimensions such as emotional expressivity (Brody & Hall, 1993; Chaplin, Cole, & Zahn-Waxler, 2005), and parents possess different desires for boys’ and girls’ emotions (Diener & Lucas, 2004). European American parents, regardless of gender, express more positive emotions with daughters than sons (Garner, Robertson, & Smith, 1997; Halberstadt, 1991) as well as discuss emotions more with daughters than sons (e.g., Adams, Kuebli, Boyle, & Fivush, 1995; Fivush, Brotman, Buckner, & Goodman, 2000). European American fathers may also be more apt to hold children to stereotypical norms than are mothers; fathers are especially likely to differentially encourage and reinforce the expression of gender-stereotyped emotion, particularly negative-submissive emotions, such as sadness and fear (but not necessarily negative-dominant emotions such as anger), in girls versus boys (e.g., Chaplin et al., 2005, Garside & Klimes-Dougan, 2002).

Because past work on child gender and parent gender differences in emotion socialization is almost exclusively seated in a European American context, we followed the suggestions of Brody (1997) and Durik et al. (2006) to examine whether parents’ emotion socialization practices are impacted by the interaction of both child gender and ethnicity. Of particular interest is the potentially differential treatment of boys. In the only published study we are aware of, African American (but not European American) mothers were more nonsupportive of preschool boys’ negative submissive emotions (e.g., nervousness, anxiety) compared to those of girls (Nelson, Leerkes, O’Brien, Calkins, & Marcovitch, 2012). In contrast, recent parenting manuals heavily subscribed to by European American parents have strongly recommended that boys – like girls – should be encouraged to express emotions, such as sadness, nervousness, and anxiety (e.g., Kindlon & Thompson, 2000; Polce-Lynch, 2002). In addition, special attention has been devoted in both academic and popular writings to the unique circumstances of African American boys, including the importance of emotional control in public and the role that parents can play in fostering their success (e.g., Barbarin, 2010; Boyd-Franklin, Franklin, & Toussaint, 2000; Kunjufu, 2007). Given this recent discourse, we sought to assess the differences and similarities in African American, European American, and Lumbee American Indian cultures regarding both mothers’ and fathers’ reactions towards girls’ and boys’ negative emotions. Additionally, because parental treatment of negative-submissive (e.g., sadness, fear) and negative-dominant (e.g., anger, contempt) emotions might vary by ethnicity and child gender as noted above, we distinguished between these two types of negative emotions when possible.

METHOD

Participants

Participants were 495 parents of 4- to- 10-year-olds involved in a larger study (Halberstadt, Dunsmore, Bryant, Parker, Beale, & Thompson, 2013). Parents completed a questionnaire about either their own emotional expressivity in the family (N = 196; 57.7% female; 37.6% African American, 40.2% European American, 20.6% Lumbee American Indian; six respondents did not report ethnicity) or their reactions to their child’s negative emotionality (N = 299; 39.6% female; 29.2% African American, 26.4% European American, and 43.7% Lumbee American Indian; six respondents did not report ethnicity). The African American and European American parents were recruited primarily from a moderate-sized southeastern, metropolitan area. The Lumbee parents were recruited primarily from a small southeastern town comprised of mostly American Indians (82%; US Census, 2000). Only one parent per household participated. The two sub-samples did not differ significantly on any demographic variables.

Mean age was 36.6 years for the sample as a whole, with European American parents being older on average (M = 38.5 years; SD = 7.1 years) than either African American (M = 35.5 years; SD = 9.2 years) or Lumbee (M = 35.8 years; SD = 8.2 years) parents, F(3, 467) = 4.48, p < .01. Education ranged from some high school education to the completion of a graduate degree in all three ethnicities, with 51.1% of the overall sample having completed college. European American participants were more highly educated (76.4 % completed college) than African American (36.7% completed college) and Lumbee (40.4% completed college) participants, F(3, 474)= 30.94, p < .001. The mean child age was 7.1 years (SD = 2.1 years) for the sample as a whole, with no significant difference in child age among the three ethnicities, F(3, 480) = .30, p = .83.

Procedure

Individual parents were recruited through direct contact or announcements made at schools, organizations, and small community businesses, flyers posted in places where parents typically congregate, and through radio announcements. An ethnically diverse research team recruited participants and administered study materials individually or in small groups; efforts were made to ensure identical recruitment procedures and relatively equal gender distributions across ethnic groups.

Parents completed the two questionnaires described below as well as others not relevant to the current study. Parents with multiple children were asked to designate a “target child” that fell within the study’s age range for the purposes of the questionnaires. Parents completed the questionnaires individually, either on site or at home, with arrangements made for the timely return of the questionnaires. Parents were paid $15 on completion of the study.

Measures

Self-Expressiveness in the Family Questionnaire (SEFQ; Halberstadt et al., 1995)

The 40-item SEFQ assesses the degree to which parents express positive and negative emotions within the family context. The SEFQ is comprised of a 23-item positive expressiveness subscale (e.g., “telling family members how happy you are”) and a 17-item negative expressiveness subscale (e.g., “expressing anger at someone else’s carelessness”). Parents report how frequently they express themselves within their family on a 9-point scale (1 = not at all frequently to 9 = very frequently). Although our primary goal was to examine these overall scales, the negative expressiveness scale was also separated into negative-dominant (assertive and/or threatening emotional displays, including anger and hostility) and negative-submissive (less assertive emotional displays, including sorrow, sulking, or crying) subscales. Given the potentially important distinction between dominant (e.g., anger) and submissive (e.g., sadness) negative emotions when considering gender differences, these sub-scales were also computed and examined in follow-up analyses. The SEFQ demonstrates good reliability and construct validity for mothers and fathers, and its subscales have been linked to relevant constructs, such as emotional experience and anger expression (e.g., Cassidy, Parke, Butkovsky, & Braungart, 1992; Halberstadt et al., 1995) and children’s emotional expressiveness (e.g., Garner et al., 1997; Halberstadt & Eaton, 2003). Responses were averaged across items for each subscale.

Reliability was similar across parent gender for positive expressiveness (αs = .89 and .91) and negative expressiveness (αs = .88 and .86) for mothers and fathers, respectively. Alphas were also similar across the ethnicities for positive expressiveness (αs = .89, .90, and .94) and negative expressiveness (αs = .90, .84, and .87) for African American, European American, and Lumbee parents, respectively. The correlation between positive and negative expressiveness was modest, r(297) =.19, p = .01. Subscale reliabilities were also similar across parent gender for negative-dominant (αs = .80 and .80) and negative-submissive (αs = .75 and .76) for mothers and fathers, respectively, as well as across the ethnicities (αs = .83, .80, and .75 for negative-dominant; αs = .74, .70, and .86 for negative-submissive) for African American, European American, and Lumbee parents, respectively. The correlation between the two types of negative expressiveness was moderate, r(194) = .52, p < .001. Means and standard deviations are presented in Table 1.

TABLE 1.

Parents’ Mean Scores (and Standard Deviations) for All Measures of Emotion Socialization

	African American		European American		Lumbee American Indian
	Mothers	Fathers	Mothers	Fathers	Mothers	Fathers
Positive expressiveness	7.02 (1.02) n = 39	6.36 (1.34) n = 33	6.86 (1.00) n = 42	6.48 (1.14) n = 34	6.58 (1.29) n = 21	6.86 (1.42) n = 19
Negative expressiveness	4.09 (1.61) n = 39	3.51 (1.07) n = 33	3.98 (1.08) n = 42	3.55 (.90) n = 34	3.72 (.96) n = 21	5.19 (1.23) n = 19
Negative-Submissive	5.23 (1.33) n = 39	4.10 (1.23) n = 33	5.11 (1.33) n = 42	4.35 (.91) n = 34	4.67 (1.41) n = 21	6.19 (1.90) n = 19
Negative-Dominant	3.67 (1.62) n = 39	3.42 (1.10) n = 33	3.61 (1.13) n = 42	3.31 (1.07) n = 34	3.43 (.84) n = 21	4.73 (1.31) n = 19
Supportive reactions	5.48 (1.25) n = 51	5.60 (1.03) n = 35	6.03 (.62) n = 37	5.42 (.80) n = 41	5.92 (.65) n = 58	5.50 (1.03) n = 71
Nonsupportive reactions	2.60 (1.11) n = 51	2.58 (1.22) n = 35	1.79 (.51) n = 37	2.32 (.81) n = 41	2.19 (.98) n = 58	2.45 (.92) n = 71

Note. Means are presented on the first line of each cell. Standard deviations are on the second line in parentheses. Positive and Negative Expressiveness are measured by the SEFQ. Supportive and Nonsupportive reactions are measured by the CCNES.

Coping with Children’s Negative Emotions Scale (CCNES, Fabes et al., 2002)

The CCNES consists of 12 hypothetical vignettes involving children’s negative affect. Parents respond to each vignette by rating the likelihood that they would respond to the child’s emotion with each of six different behaviors on a 7-point scale (1 = very unlikely to 7 = very likely). The six subscales are combined to create two composite factors: supportive and nonsupportive reactions. Supportive reactions consist of problem-focused coping (e.g., “help my child think of constructive things to do”), emotion-focused coping (e.g., “comfort my child and take his/her mind off it”), and encouragement (e.g., “encourage him/her to talk about how it hurts”). Nonsupportive reactions consist of distress (e.g., “feel upset myself”), punitive (e.g., “tell my child to behave or we’ll go home”), and minimizing (e.g., “tell my child not to make a big deal out of it”). The CCNES demonstrates good reliability and validity for mothers and fathers, including links with family characteristics, parental attitudes and behavior, and child outcomes (e.g., Baker, Fenning, & Crnic, 2011; Fabes et al., 2002; Nelson et al., 2009; Wong et al., 2009).

Cronbach’s alphas were similar across parent gender for supportive reactions (αs = .91 and .92) and nonsupportive reactions (αs = .89 and .90) for mothers and fathers, respectively. Alphas were also similar within each ethnicity for both supportive reactions (.91, .92, and .92) and nonsupportive reactions (.91, .89, and .90) for African American, European American, and Lumbee parents, respectively. The correlation between supportive and nonsupportive reactions was moderate, r(297) = −.42, p < .001. Means and standard deviations are presented in Table 1.

RESULTS

Analyses were conducted in two steps. First, measurement invariance across parent gender and ethnicity was established for the primary variables of interest. Second, hierarchical regression analyses were conducted to examine gender differences between mothers and fathers on each variable as well as the moderating roles of ethnicity and child gender.

Measurement Invariance

Preliminary analyses established measurement invariance among mothers and fathers for both the SEFQ and CCNES measures as well as invariance across the three ethnicities on each measure. A series of Multiple Group Confirmatory Factor Analyses was conducted using MPLUS v. 7.0 (Muthen & Muthen, 2012) following established techniques outlined in Schmitt (2008) and Vandenberg and Lance (2000) for evaluating measurement invariance. The assumption of homogeneity across parent gender and ethnicity was tested by examining increasingly restrictive models in which the form (configural invariance), factor loadings (metric invariance), and intercepts (scalar invariance) were constrained to equality across the groups (mothers and fathers; African American, European American, and Lumbee). To the extent that these models fit the observed data well, there is evidence of measurement invariance.

SEFQ

Invariance analyses were conducted using item parceling, as outlined in Little, Cunningham, Shahar, and Widaman (2002), with parcels (mean scores of three items) chosen at random from within the positive and the negative scales of the SEFQ to serve as observed indicators of latent positive and negative emotional expressiveness factors, respectively. These latent factors were allowed to covary. Parceling provides some distinct psychometric advantages and is particularly useful for specifying simplified models when testing invariance among measures with many items in a modest sample (Little et al., 2002; Meade & Kroustalis, 2006).

When examining invariance by parent gender, the models testing configural invariance, χ² (16) = 15.76, p = .47, and metric invariance, χ²(20) = 20.13, p = .39, fit the data well, with no significant decrease in model fit between the configural and metric invariance models, Δχ² (4) = 4.37, p = .36. Although the model testing scalar invariance showed a significant decrease in model fit, Δχ² (4) = 15.26, p < .01, this model still fit the data well, χ²(24) = 36.39, p = .05; CFI = .98; RMSEA = .07, and only one of the six parcel intercepts varied across parent gender. A partial strong invariance model with only one freed intercept resulted in a model with good fit, χ² (23) = 25.60, p = .32; CFI = 1.00; RMSEA = .03, that was no worse than the metric invariance model, Δχ² (3) = 4.47, p = .22.

When testing invariance by ethnicity, both the configural invariance model, χ² (24) = 23.65, p = 48; CFI = 1.00; RMSEA = .00, and metric invariance model, χ² (32) = 37.79, p = .22; CFI = .99; RMSEA = .05, fit the data well, with no significant decrement in model fit between the two models, Δχ² (8) = 14.14, p = .08. The same model demonstrating partial strong invariance by parent gender (one parcel intercept freed) also fit the data well when testing invariance by ethnicity, χ² (24) = 23.65, p = .48; CFI = 1.000; RMSEA = .00. This model was also not significantly worse than the metric invariance model, Δχ² (6) = 8.08, p = .23. Thus, the SEFQ was deemed sufficiently invariant across parent gender and ethnicity, and analyses proceeded with the positive and negative expressiveness scales.

CCNES

CFA models examining invariance by parent gender on the CCNES also tested the equivalence of a two-factor model for mothers versus fathers (supportive and nonsupportive reactions) with three observed indicators per factor. The observed variables were the computed means for each of the subscales that load onto supportive (i.e., emotion-focused, problem-focused, and encouragement) and nonsupportive (i.e., distress, punishment, and minimizing) reactions. Preliminary confirmatory factor analyses examined each subscale separately, with the individual items of those subscales as observed indicators. These preliminary analyses indicated relatively poor model fit; parents’ responses to half of the CCNES scenarios showed a consistent pattern of low and often non-significant loadings onto the subscales to which they belong. Responses to these scenarios (CCNES items #1, 2, 3, 8, 10, 11) were omitted from subsequent analyses, and tests of invariance proceeded using scale means with these items deleted (αs for the shortened scales ranged from .67 – 85, with a mean α of .80). Although this strategy resulted in the elimination of numerous CCNES items, analyses testing for invariance by ethnicity in different populations have yielded somewhat similar results (Halberstadt, Rogers, Castro, & Garrett-Peters, submitted; Leerkes, Supple, Su, & Cavanaugh, 2013). The remaining items reflect parents’ (supportive and nonsupportive) reactions to children’s negative-submissive emotions, such as nervousness, anxiety, and embarrassment.

Thus, models testing invariance by parent gender and ethnicity consisted of two latent factors, with the shortened (6 item) subscale means of emotion-focused, problem-focused, and encouragement as indicators of the latent supportive reactions factor, and the shortened subscale means of distress, punishment, and minimizing as observed indicators of the latent nonsupportive reactions factor. These two latent factors were allowed to covary. Models testing for invariance by parent gender with the shortened CCNES scales indicated good model fit for configural invariance, χ²(24) = 38.97, p = .01; CFI = .98; RMSEA = .08, and metric invariance, χ²(20) = 34.14, p = .03; CFI = .99; RMSEA = .07, with a non-significant decrease in model fit between the two models, Δχ²(4) = 4.83, p = .31. The model testing scalar invariance by parent gender also fit the data well, χ²(16) = 37.36, p = .04; CFI = .99; RMSEA = .06, with no significant decrease in model fit between the metric and scalar invariance models, Δχ²(4) = 3.22, p = .52.

Tests of invariance across ethnicity also indicated good-fitting models for configural invariance, χ²(24) = 38.32, p = .03; CFI = .99; RMSEA = .08, and metric invariance, χ²(32) = 49.69, p = .02; CFI = .98; RMSEA = .08, with no significant decrease in fit between the two models, Δχ² (8) = 11.37, p = .18. The scalar invariance model also showed good absolute fit to the data, χ²(40) = 34.14, p = .03; CFI = .99; RMSEA = .08. Although the reduction in model fit from the metric invariance model was significant based on the change in the chi-square statistic, Δχ² (8) = 17.84, p = .02, alternative fit indices may be more sensitive than chi-square tests for evaluating measurement invariance, with consensus emerging around ΔCFI (< .01) and ΔRMSEA (< .01) as preferred indices for establishing invariance (e.g., Chen, 2007; Cheung & Rensvold, 2002; Meade, Johnson, & Braddy, 2008; Rutkowski & Svetina, 2014). Notably, the change from the weak factorial to the scalar model was not significant based on these criteria (ΔCFI = .009; ΔRMSEA = .008). These results provide sufficient evidence of measurement invariance by parent gender and ethnicity. Thus, supportive and nonsupportive scores were computed by calculating the mean of the three shortened supportive (emotion-focused, problem-focused, and encouragement) and nonsupportive (distress, punishment, minimizing) scales.

Hierarchical Regression Analyses Predicting Emotion Socialization Behaviors

Separate regression analyses were conducted for each of the emotion socialization variables. In all models, parent gender was dummy-coded and entered onto a single block of a regression equation that also included the following covariates: ethnicity, marital status (single-parent vs. two-parent household), child gender, child age, and family occupational status as measured by the family’s highest score on the traditional Hollingshead (1975) inventory. This step tested whether emotion socialization strategies differed as a function of parent gender, after controlling for relevant covariates.

To determine whether parent gender differences varied significantly among the three ethnic groups, a second step including Parent gender x Ethnicity interaction terms was added to each of the regression equations. Ethnicity was dummy-coded with two variables, such that European American families were the reference group (i.e., African American vs. European American and Lumbee vs. European American). Interaction terms were created by computing the product of gender and each of the ethnicity codes; these interaction terms (Parent gender x African American and Parent gender x Lumbee) were added to the second step of each equation.

Child gender was included as an additional moderator only in the equations examining parents’ reactions to children’s negative emotions because emotional expressiveness is not measured as a function of a specific target child. To explore child gender as a further moderator of the parent gender x ethnicity interactions, three-way interaction terms (Parent gender x Ethnicity x Child gender) were computed and entered on the third step of separate regression equations predicting supportive and nonsupportive reactions to children’s negative emotion. In addition to the Child gender x Ethnicity interaction terms described above, Child gender x Parent gender interaction terms were also included on the second step of equations predicting supportive and nonsupportive responses.

Emotional expressiveness

Results of regression analyses predicting positive expressiveness are presented in Table 2. As seen on the first step, mothers reported significantly greater levels of positive expressiveness than fathers even after controlling for all relevant covariates. The second step containing both Parent gender x Ethnicity interaction terms was non-significant, with neither of these interaction terms approaching significance, suggesting that the effect of parent gender on positive expressiveness was not moderated by ethnicity.

TABLE 2.

Summary of Hierarchical Regression Analysis Predicting Emotional Expressiveness

	SEFQ Positive Expressiveness						SEFQ Negative Expressiveness
Variable	Model 1			Model 2			Model 1			Model 2
	B	SE B	β	B	SE B	β	B	SE B	β	B	SE B	β
Covariates
SES	−.01	.04	−.02	.00	.04	−.01	−.03	.04	− .07	−.02	.04	−.05
Marital status	.17	.22	.07	.20	.22	.08	−.21	.22	−.08	−.14	.22	−.06
Child age	.08	.04	.15	.08	.05	.15	−.09	.05	−.15	−.09	.05	−.16*
Child gender	−.16	.18	−.07	−.14	.18	−.06	−.18	.18	−.08	−.11	.18	−.05
Ethnicity
AA vs. EA	.03	.21	.01	−.10	.29	−.04	−.18	.21	−.07	−.07	.29	−.03
LA vs. EA	−.21	.25	−.07	−.08	.61	.03	.04	.26	.02	1.49	.61	.51*
Parent Gender	.55	.19	.24**	.46	.26	.20	.11	.20	.05	.33	.26	.14
Ethnicity × Parent gender
AA x Parent gender				.28	.40	.10				−.10	.40	−.03
LA x Parent gender				−.10	.66	−.03				−1.71	.67	−.56*
R2		.07			.07			.06			.10
F for change in R2		1.72			0.30			1.39			3.36*

Note. AA =African American, EA = European American, LA = Lumbee American Indian; EA = 0 (reference group) in all analyses; Gender coding: 0=boy, 1=girl; 0=father, 1=mother.

^*p < .05.

^**p < .01.

Analyses predicting negative expressiveness are also presented in Table 2. The first step of this equation revealed no significant main effect of parent gender on parents’ expression of negative emotion. However, the second step including the Ethnicity x Gender interaction terms was significant. This was driven by a significant Parent gender x Lumbee American Indian interaction term, indicating that the gender difference between mothers and fathers in the Lumbee sample differed from that of European Americans. To further explicate this interaction, separate regression equations were examined for the subsample of Lumbee parents and the subsample of European American parents. These analyses did not include ethnicity (because each equation was conducted for only one ethnicity), but were otherwise identical to Model 1 presented in the Negative Expressiveness equation found in Table 2. Results indicated that, after controlling for all covariates, parent gender was not a significant predictor of negative expressiveness among European American parents, β = .16, t(73) = 1.30, p = .20. In contrast, parent gender was a significant predictor of negative expressiveness among Lumbee parents, β = −.47, t(38) = −2.28, p = .03, such that Lumbee fathers were more likely to express negative emotion than were Lumbee mothers. Results of these analyses are depicted graphically in Figure 1.

Figure 1.

Negative expressiveness by parent gender and ethnicity.

Note. * denotes a significant difference (p < .05) between mothers and fathers within ethnicity. Error bars represent standard errors.

Follow-up analyses were also conducted to determine whether these effects held for the negative-submissive and/or negative-dominant subscales of the SEFQ. These two regression equations were identical to the one examining the full SEFQ negative expressiveness scale, but substituted the negative-submissive and negative-dominant SEFQ subscales as dependent variables; results are presented in Table 3. For negative-submissive expressiveness, a significant effect of parent gender on the first step indicated that mothers as a whole were more likely than fathers to express negative-submissive emotions. However, as in analyses examining the entire negative expressiveness scale, this effect was moderated by ethnicity. In particular, the Parent gender x Lumbee interaction term on the second step was significant, β = −.54, t(160) = −2.54, p = .01. For negative-dominant expressiveness, parent gender was not a significant predictor, although the interaction between Parent gender and Ethnicity approached significance, β = −.41, t(160) = −1.89, p = .06, and was in the same direction as the interaction term predicting negative-submissive expressiveness. Although this interaction was significant only for the negative-submissive (and not the negative-dominant) subscale, the similar magnitude and direction of these effects suggest that these two subscales were not markedly different from one another.

TABLE 3.

Summary of Hierarchical Regression Analysis Predicting SEFQ Negative Expressiveness Subscales

	SEFQ Negative-Submissive						SEFQ Negative-Dominant
Variable	Model 1			Model 2			Model 1			Model 2
	B	SE B	β	B	SE B	β	B	SE B	β	B	SE B	β
Covariates
SES	−.02	.04	−.05	−.01	.04	−.01	−.02	.04	− .05	−.02	.04	−.04
Marital status	−.41	.26	−.14	−.31	.26	−.10	−.05	.22	−.02	−.01	.22	−.01
Child age	.02	.05	.02	.01	.05	.02	−.10	.05	−.17*	−.10	.05	−.18*
Child gender	−.14	.21	−.05	−.03	.21	−.01	−.14	.18	−.06	−.10	.18	−.04
Ethnicity
AA vs. EA	−.29	.25	−.10	−.32	.34	−.11	−.10	.21	−.04	.07	.29	.03
LA vs. EA	−.11	.30	−.03	1.62	.72	.46*	.08	.26	.03	1.13	.62	.39
Parent gender	.61	.23	.22**	.75	.31	.27*	−.06	.20	−.03	.18	.27	.08
Ethnicity × Parent gender
AA x Parent gender				.21	.48	.06				−.28	.41	−.10
LA x Parent gender				−1.99	.79	−.54*				−1.27	0.68	−.41
R2		.09			.13			.04			.06
F for change in R2		2.05*			3.72*			0.89			1.83

Note. AA =African American, EA = European American, LA = Lumbee American Indian; EA = 0 (reference group) in all analyses; Gender coding: 0=boy, 1=girl; 0=father, 1=mother.

^*p < .05.

^**p < .01.

Reactions to children’s negative emotions

Results of hierarchical regression analyses predicting parents’ supportive reactions to children’s negative emotion are presented in Table 4. As seen in the first step of the equation, parent gender was a significant predictor of supportive reactions, such that mothers were significantly more supportive than fathers. There was also a significant Parent gender x Ethnicity interaction term on the second step, with the contrast comparing African American with European American parents being significant. When conducting separate regression equations for each ethnicity (identical to Model 1 of Table 4, without including ethnicity), European American mothers were more supportive than European American fathers, β = .41, t(76) = 3.66, p < .001, and Lumbee mothers were more supportive than fathers, β = .20, t((118) = 2.20, p = .03. In contrast, African American mothers and fathers reported similar levels of supportive reactions, β = −.02, t(78) = −.17, p = .87. These results are presented graphically in Figure 2.

TABLE 4.

Summary of Hierarchical Regression Analysis Predicting CCNES Supportive Reactions

Variable	Model 1			Model 2			Model 3
	B	SE B	β	B	SE B	β	B	SE B	β
Covariates
SES	.04	.03	.10	.04	.03	.11	.04	.03	.09
Marital status	.26	.15	.12	.24	.15	.11	.31	.15	.15*
Child age	.01	.03	.02	.01	.03	.01	.01	.03	.02
Child gender	.19	.12	.10	.08	.18	.04	.29	.18	.15
Ethnicity
AA vs. EA	−.07	.17	−.03	.36	.27	.17	.73	.29	.34*
LA vs. EA	.09	.15	.05	−.06	.29	−.03	.06	.29	.03
Parent gender	.33	.13	.17**	.49	.26	.24	1.00	.29	.50**
Ethnicity X Parent gender
AA x Parent gender				−.65	.32	−.25*	−1.70	.43	−.65**
LA x Parent gender				−.21	.31	−.06	−.81	.40	−.25
Parent gender x Child gender				.25	.25	.11	−.73	.37	−.32
Ethnicity X Child gender
AA x Child gender				−.15	.27	−.06	−.94	.38	−.35*
LA x Child gender				.13	.11	.16	.07	.11	.08
Ethnicity X Parent gender X Child gender
AA x Parent gender x Child gender							2.03	.57	.64**
LA x Parent gender x Child gender							1.14	.50	.27*
R2		.06			.09			.14
F for change in R2		2.52*			1.45			1.39

Note. AA =African American, EA = European American, LA = Lumbee American Indian; EA = 0 (reference group) in all analyses. Gender coding: 0=boy, 1=girl; 0=father, 1=mother.

^*p < .05.

^**p < .01.

Figure 2.

Supportive reactions to children’s submissive negative emotion by parent gender and ethnicity.

Note. * denotes a significant difference (p < .05) between mothers and fathers within ethnicity. Error bars represent standard errors.

However, this two-way interaction was further moderated by child gender. In particular, a significant three-way interaction on the third step indicated that the Parent gender x African American x Child gender interaction was a significant predictor of supportive reactions. Furthermore, the Parent gender x Lumbee x Child gender interaction was also significant on the third step. Results of these three-way interactions are presented graphically in Figures 3 and 4 and were explicated by analyzing separate regression equations for girls and boys. Each of these equations was identical to Model 2 in Table 4, except they did not contain child gender or any interactions including child gender. For girls, non-significant Parent gender x Ethnicity terms for African American vs. European American, β = .12, t(133) = .65, p = .52, and Lumbee vs. European American, β = .10, t(133) = .66, p = .51, contrasts on the second step indicated that differences between mothers’ and fathers’ supportive reactions did not vary by ethnicity. As seen in Figure 3, African American and Lumbee mothers of girls were more supportive than fathers of girls, after controlling for all covariates. Although this parent gender difference only approached significance for European American parents, the overall results suggest a consistent pattern of mothers’ more supportive reactions to girls’ negative emotions than fathers in all three ethnicities.

Figure 3.

Supportive reactions to children’s submissive negative emotion by parent gender and ethnicity for girls only

Note. * denotes a significant difference (p < .05) between mothers and fathers within ethnicity. Error bars represent standard errors.

Figure 4.

Supportive reactions to children’s submissive negative emotion by parent gender and ethnicity for boys only

Note. * denotes a significant difference (p < .05) between mothers and fathers within ethnicity. Error bars represent standard errors.

For boys, however, significant Parent gender x Ethnicity contrasts for both African American vs. European American, β = −.55, t(134) = −3.68, p < .01, and Lumbee vs. European American, β = −.21, t(136) = −1.66, p = .10, ethnicities emerged. These results indicated that parent gender differences differed by ethnicity among parents of boys, with the pattern of findings for European American parents differing significantly from the pattern for African American and Lumbee parents. In particular, after controlling for all covariates, European American mothers of boys were more supportive than European American fathers of boys, whereas African American mothers of boys were less supportive than African American fathers of boys. Lumbee mothers and fathers of boys did not differ in their supportive reactions. These results are depicted graphically in Figure 4.

Results of an identical regression equation with nonsupportive reactions as the dependent variable are presented in Table 5. Parent gender was a significant predictor on the first step of this equation, indicating that fathers reported more nonsupportive reactions to children’s negative emotion than mothers. No significant interactions emerged when predicting parents’ nonsupportive reactions to children’s negative emotions, suggesting that this parent gender difference was not moderated by ethnicity or the combination of ethnicity and child gender.

TABLE 5.

Summary of Hierarchical Regression Analysis Predicting CCNES Nonsupportive Reactions

Variable	Model 1			Model 2			Model 3
	B	SE B	β	B	SE B	β	B	SE B	β
Covariates
SES	−.08	.03	−.20**	−.09	.03	−.22**	−.09	.03	−.21**
Marital status	−.15	.14	−.07	−.13	.14	−.06	−.17	.14	−.08
Child age	.01	.03	.03	.02	.03	.04	.01	.03	.03
Child gender	−.16	.11	−.09	−.03	.17	−.02	−.06	.18	−.03
Ethnicity
AA vs. EA	.32	.16	.15	−.05	.25	−.03	−.07	.28	−.03
LA vs. EA	.07	.15	.04	.16	.28	.08	.10	.28	.05
Parent gender	−.27	.12	−.14*	−.36	.24	−.19	−.55	.28	−.28
Ethnicity X Parent gender
AA x Parent gender				.48	.30	.19	.67	.42	.26
LA x Parent gender				.35	.29	.11	.76	.39	.24
Parent gender x Child gender				−.36	.24	−.16	−.01	.36	−.01
Ethnicity X Child gender
AA x Child gender				.22	.26	.09	.25	.37	.10
LA x Child gender				−.12	.11	−.15	−.09	.11	−.11
Ethnicity X Parent gender X Child gender
AA x Parent gender x Child gender							−.36	.55	−.12
LA x Parent gender x Child gender							−.76	.48	−.18
R2		.11			.1			.14
F for change in R2		4.55**			1.36			1.24

Note. AA =African American, EA = European American, LA = Lumbee American Indian; EA = 0 (reference group) in all analyses; Gender coding: 0=boy, 1=girl; 0=father, 1=mother.

^*p < .05.

^**p < .01.

DISCUSSION

This study is among the first to examine whether the parent gender differences often found in European American populations for two key aspects of parental socialization, emotional expressiveness and reactions to children’s negative emotions, vary across ethnicity and child gender. Overall, mothers and fathers differed in expected ways on both measures of emotion socialization practices. However, some of these differences were qualified by ethnicity and/or child gender, even after controlling for child age, family occupational status, and the number of parents in the home.

Parents’ Emotional Expressiveness

Consistent with prior research with primarily European American samples (e.g., Halberstadt et al., 1995; Hughes & Gullone, 2010; Wong et al., 2009), our findings suggest that mothers are more positively expressive than fathers within the context of their families. In contrast to the highly contextualized results related to negative emotions, these results are consistent with both biological and socio-cultural explanations of gender differences in the expression of positive emotion. However, our results indicated that gender differences in parents’ expression of negative emotions differed markedly across ethnicity, in both the magnitude and direction of the effects. Specifically, European American and African American mothers and fathers were similar in their levels of negative expressiveness, but Lumbee American Indian parents showed a strikingly different pattern, with fathers reporting significantly greater levels of negative emotional expression at home than mothers. This effect may be driven by the belief among Lumbee parents that men should be encouraged to openly express emotion (Parker et al., 2012), including displays of submissive negative emotions. Although future research is needed to explore variations in mothers’ and fathers’ expressiveness, these results suggest that gender differences in the expression of negative emotions are embedded within cultural contexts and the degree to which mothers and fathers differ may depend on values and/or roles inculcated in their cultural or ethnic backgrounds.

Parents’ Reactions to Children’s Negative Emotions

Consistent with previous work (Cassano et al., 2007; McElwain et al., 2007; Nelson et al., 2009), fathers reported more nonsupportive reactions than mothers to children’s submissive negative emotions, such as sadness and fear. Fathers’ outwardly harsh or punitive reactions to negative emotion are consistent with biological explanations of parent gender differences, but could also reflect the pervasive social construction of the father as traditional disciplinarian across all three cultures. However, a more nuanced picture emerges when examining parents’ supportive reactions toward children’s negative emotions. Whereas European American and Lumbee American Indian mothers were significantly more supportive of their children’s submissive negative emotions than fathers, African American mothers and fathers were similarly supportive of their children’s submissive negative emotions. This suggests greater gender equity in parents’ support for children’s emotions among African American parents.

Additionally, the parent gender difference in supportive reactions was further contextualized when considering ethnicity and child gender simultaneously. Specifically, gender differences in supportive reactions were relatively consistent across ethnicities among parents of girls, but varied by ethnicity among parents of boys. Among parents of girls, mothers were more supportive than fathers overall, and this pattern was similar across all three ethnicities (albeit slightly stronger among African American and Lumbee parents). The consistency of this gender-differentiated support from parents of all ethnicities may reflect a widespread (and gender-stereotyped) lay theory in which mothers play the primary role in supporting their daughters during times of emotional distress, particularly those involving the outward expression of submissive negative emotions like sadness.

In contrast, differences between mothers and fathers varied significantly by ethnicity among parents of boys. European Americans showed the greatest parent gender discrepancy, with mothers significantly more supportive of boys’ submissive negative emotions than fathers. Lumbee mothers and fathers were relatively equal in their levels of support for boys’ submissive negative emotions. Strikingly, however, African American mothers were actually less supportive of boys’ submissive negative emotions than were African American fathers. These results are consistent with another finding of African American mothers being less supportive of boys’ negative emotions than those of girls relative to their European American counterparts who are more supportive of boys (Nelson et al., 2012). Indeed, in Nelson et al. (2012) African American mothers of boys perceived the most negative consequences for the display of their children’s negative emotions compared to European American mothers, perhaps leading them to engage in emotion socialization practices aimed at discouraging signs of weakness in male children. Alternatively, it may be that African American mothers are more likely to ignore their boys’ emotional reactions rather than respond in any explicit way, or that they devote more of their effort and energy to socializing girls than boys (Mandara, Varner, & Richman, 2010).

Although African American fathers are often faced with the tension of balancing traditional parenting role expectations with the relatively egalitarian attitudes toward work and family roles in African American families (e.g., Hunter & Davis, 1992), results from this study support previous work suggesting that African American fathers’ treatment of girls versus boys may be less gender stereotyped than European American fathers (Hossain & Roopnarine, 1993). In general, African American and European American boys have different socialization experiences that likely reflect different societal expectations for, and perceptions of, this generation of males. Future research should continue to examine the possibility that the beliefs and attitudes governing the treatment of boys’ negative emotions may be both culture-and gender-specific.

These findings as a whole suggest that submissive negative emotions are particularly likely to generate differences between mothers’ and fathers’ socialization as well as contextual differences as a function of ethnicity and child gender. Parents’ reactions in this study, as measured by the CCNES, were essentially responses to children’s nervousness, fearfulness, and embarrassment, and the gender differences in parents’ emotional expressiveness were slightly stronger when expressing sorrow, disappointment, and distress. Modern views of emotion and gender roles may very well allow for a great deal of flexibility and variability in parents’ expression of – and reactions to – these sorts of negative emotions. It is, of course, unclear at this time whether the patterns discussed in the current study hold for more dominant negative emotions, such as anger, contempt, and hostility. Those emotions may be accompanied by a different set of socialization processes reflected in unique patterns of gender and cultural differences.

Summary

The examination of emotion socialization practices of mothers and fathers from three ethnicities contributes to a more nuanced and contextualized understanding of emotion, culture, gender roles, and the interplay among them. Gender differences in some aspects of emotion socialization were consistent across ethnicity and child gender. However, our findings demonstrate that the ways in which mothers and fathers differ in their emotion socialization behaviors are often qualified by ethnicity and occasionally by child gender.

Limitations and Future Directions

Although the present study contributes to our understanding of gender-specific patterns of emotion socialization, there are a number of limitations to address in future research. First, although there is a wide net of construct validity for the questionnaires used in this study, all data were collected via parent reports. Reports from both children and parents, as well as observational methods over time, may elucidate the underlying processes at work in emotion socialization. It may also be useful to examine parents’ reactions to children’s emotion in public versus private settings, and whether those contexts alter the pattern of results observed here.

Although we were able to include ethnic minorities in the United States who are under-represented in research, the operational definition of “culture” was still constrained. While acknowledging this limitation, our findings suggest that at least three sub-cultures within the United States have markedly different gender roles and gender-related expectations for parents’ and children’s emotional lives. These findings emerged even after controlling for child age, number of parents in the home, and family occupational status, suggesting that the differences we found are more likely to be cultural rather than economic. These differences are particularly noteworthy given that the three ethnicities live in close geographical proximity to one another, and may be exposed to the other ethnicities’ cultural norms and expectations.

To ensure independent responses, we collected data from parents in different families. However, given the importance of family relationships for individuals’ parenting behavior, and the complex dynamics of the co-parenting relationship (e.g., McHale, 1995), examining mothers and fathers from the same families might provide a more complete picture of the ways in which parents divide family labor in general and emotion socialization tasks in particular, and whether those tasks are divided along gender lines. Moreover, the ethnicity of the partners of participating parents was not available. Future work examining emotion socialization behaviors among partners from different ethnic backgrounds may also provide useful information regarding the influence of culture on gender differences in parenting.

Despite these limitations, this work adds to our understanding of individual differences in emotion socialization by assessing both fathers and mothers, and including parents from three ethnic groups, one of which (Lumbee American Indians) has been vastly underrepresented in parenting research. In doing so, this study contributes to the literatures on emotion socialization and parental gender roles by identifying ethnicity and child gender as contextual factors that may play a role in shaping both mothers’ and fathers’ emotion expression and reactions to children’s submissive negative emotion. We hope that future research exploring similarities or differences among mothers and fathers will be sensitive to the ethnic, cultural, socioeconomic, and gender characteristics of the samples being explored. Such an approach may well lead to an increased appreciation of the development of gender-specific emotion socialization practices, and influences on those practices that are both universal and culturally unique.

IMPLICATIONS FOR PRACTICE, APPLICATION, THEORY, AND POLICY

These results speak to the importance of examining the cultural contexts of parent gender differences in emotion socialization. In contrast to models that focus on innate explanations for gender differences, the ethnic differences observed in this study support the notion that gender differences in emotion-related aspects of parenting are socially constructed, and might be partially attributed to one’s cultural background (e.g., Brody, 1997; Leu et al., 2010). Gender roles do not develop in isolation, and the collection of beliefs and values shared within ethnicity – and the status and history of that ethnicity within the broader society-at-large – may shape the unique attitudes and behaviors of mothers and fathers. Thus, it is important for parents, practitioners, and teachers working with parents to recognize that parental goals and behaviors may be embedded in the scripts that parents believe are appropriate for their gender and ethnicity. Given that these scripts are culturally created, practice aimed at promoting adaptive patterns of parental emotion socialization should be culturally sensitive to variation in emotional expression and reactions to children’s emotion by gender and ethnicity.