The cultural stereotype that women are more emotional than men has generated much research on emotion and emotional development (Brody, 1985). Some literature has substantiated that women are more emotionally expressive than men are. For example, women talk more about emotions (Goldschmidt & Weller, 2000) and report being more emotionally expressive (Bronstein, Briones, Brooks, & Cowan, 1996) than do men. The present study looked at one aspect of emotion expression, the verbal expression of emotion, in early adolescents to examine gender differences in their emotion talk with their parents. We also investigated mothers’ and fathers’ emotion talk with their adolescent daughters and sons to examine whether the gender differences in emotion socialization found with young children (e.g., Cervantes & Callanan, 1998; Dunn, Bretherton, & Munn, 1987; Fivush, Brotman, Buckner, & Goodman, 2000) persist into early adolescence.

Beginning when children are young, there are some gender differences in their emotion talk. For example, Cervantes and Callanan (1998) found that 2-year-old girls talked more about emotion than did 2-year-old boys, but there were no differences when children were 3- and 4-year-old. Similarly, girls’ conversation turns concerning emotions were more frequent than those of boys at the age of 24 months (Dunn et al., 1987). Others have found that 4-year-old daughters elaborated more than did same-age sons when talking to mothers (Fivush, Berlin, Sales, Mennuti-Washburn, & Cassidy, 2003). In other words, daughters enhanced the parent–child dialogue by introducing new aspects into the emotional conversation. When discussing their past experiences with their mothers and fathers, 40- to 45-month-old boys mentioned emotion less often than did girls (Fivush et al., 2000).

Why might girls discuss emotions more than do boys? One reason is that girls seem to understand emotions at an earlier age than do boys. Girls score higher on tests that measure how well they understand other people’s emotions (Bosacki & Moore, 2004). Not only do girls understand emotions better than boys do, but they are also better than boys at applying cultural standards of emotion expression in everyday situations. For example, preschool-aged girls are better than boys at hiding negative emotions, such as after having received a disappointing gift (Saarni, 1984, 1989).

A second reason why girls may discuss emotions more than boys is that parents may encourage discussion about emotions more with girls than with boys. As suggested by social cognitive theory (Bandura, 1997), everyday conversations can serve as a context for socializing children into the particular values and beliefs endorsed by their parents and larger cultural community. One such value is the expression of emotion in a gender-consistent manner. Discussing emotions provides children with opportunities for becoming emotionally expressive. Were parents to discuss emotion more with daughters than with sons, daughters would gain relevant knowledge about emotions. In support of this idea, mothers have been found to talk about emotion states more with girls than with boys (Dunn et al., 1987) and to provide more detail regarding emotions in conversations with their 4-year-old daughters than with their sons of the same age (Fivush et al., 2003). The one study that included fathers as well as mothers showed that both mothers and fathers used more emotion expression when discussing sad events with their 40- to 45-month-old daughters than with their same-age sons (Fivush et al., 2000). It was expected that in the present study daughters would use more emotion language in general than would sons.

In addition to the stereotype that women are more emotional than men, there are also stereotypes that associate specific emotions with men and women (Brody, 1985). For instance, fear and sadness may be construed as traditionally feminine-stereotyped emotions, whereas anger and frustration are construed as masculine-stereotyped emotions (Brody, 1996). Anger is a more self-assertive emotion, whereas sadness is considered passive (Leaper, 2002; Shields, 2002). Thus, stereotypes about emotions are consistent with cultural stereotypes about women and men (Anderson, 1998).

There is some support for stereotypes regarding gender and specific emotions. First, approximately twice as many women as men suffer from a depressive disorder each year (12.4 million compared to 6.4 million) (National Institute of Mental Health, 2001). Second, Grossman and Wood (1993) found that women reported more frequent feelings of sadness and reported expressing this emotion more frequently and intensely than did men. Third, women have stated that they express more sadness than men (Allen & Haccoun, 1976). Conversely, men sometimes report more intense anger than women do (see Brody, 1996, for a review).

Findings from studies of adolescents’ emotion expression and experience seem to suggest that emotion expression may be similarly gender-stereotyped. A study of adolescents (12 to 19 years of age) showed a higher frequency of major depressive episodes for girls than for boys (Galambos, Leadbeater, & Barker, 2004). Likewise, middle childhood (Brody, 1984) and adolescent (Stapley & Haviland, 1989) girls report more frequent sadness than do boys. Girls also express more sadness and affection than do boys of the same age (Safyer & Hauser, 1994). Conversely, boys tend to express anger more readily than do girls (Brody & Hall, 1993). In response to vignettes designed to elicit emotion, adolescent girls provided more references to sadness in written responses to both an anger/sad vignette and a fear vignette than did boys of the same age (O’Kearney & Dadds, 2004). As expected, boys produced more references to anger in their written responses to the anger/sad vignette than did girls.

Mother–child conversations with preschool aged children also conform to the gender stereotypy of specific emotions, such as anger and sadness. For instance, Fivush (1991) reported that mothers, while speaking with their 32- to 35-month-old children, discussed anger more with their sons than with their daughters, whereas they discussed sadness more with their daughters than with their sons. In another study, mothers talked about sadness less frequently with their 2-year-old sons and never spoke about anger with their daughters of the same age (Fivush, 1989).

Although many studies have confirmed the gender stereotypes about emotion, other research concerning discourse behavior does not indicate differences between women and men in either the amount of emotion or the specific emotion verbally expressed (Shimanoff, 1985). Instead, conversational topic plays a larger role in the explicit expression of emotions than does gender (Anderson & Leaper, 1998). Fischer (1993) suggested that emotion attributions may be biased by gender stereotypes that profess women’s greater emotionality. Gender stereotypes may be responsible for how individuals perceive others’ emotions, such that people choose to interpret others’ experiences in a gender consistent manner (Fischer, 1993; Shields, 1987). Thus, findings regarding sex differences in emotion expression may not be so straightforward. Similar processes may operate in how individuals encode their own emotions. In support of this view, self-reported differences in anger, fear, and sadness are often greater than observed differences in emotion expression within people’s conversations (LaFrance & Banaji, 1992).

Given the contradictory findings, in the present study we sought to examine whether there were gender-related patterns in conversations between early adolescents and their parents. This study is different from previous research on gender differences in the gender socialization of emotion with its specific focus on early adolescents. The majority of previous research on emotion expression in everyday conversations has focused on children between the ages of 2 and 4 years with their parents or between adult peers. Whether parents continue to discuss emotions in gender-consistent patterns remains an empirical question. Although early adolescents may be capable of expressing and discussing a wide range of emotions, they may choose not to do so when it is discouraged within their family context (Bronstein et al., 1996). However, emotion expression has been linked with positive socio-emotional adjustment (Gross & John, 2003) and thus they would benefit from learning to express themselves.

Apart from the lack of research on emotion expression in adolescents, children from this age range (10- to 13-year-old) were selected based on findings that suggest some flexibility in gender-related beliefs during this age period. There is a curvilinear pattern of gender flexibility in which children are inflexible about gender-role transgressions at age 5, become flexible and remain flexible until age 13, at which point early adolescents become inflexible again (Stoddart & Turiel, 1985). Research has also supported the view that at 13, gender norms become more rigid (Hill & Lynch, 1983). In their interviews with children in the third through eighth grades, Carver, Yunger, and Perry (2003) found that older children felt less pressure to conform to gender expectations than did younger children. Some have linked the decrease in flexibility in early adolescence to social life transitions, such as the transition into middle school (Alfieri, Ruble, & Higgins, 1996), whereas others have suggested that children’s parents may play a role in how children view gender stereotypes during this period of time. In particular, Crouter, Manke, and McHale (1995) have suggested that parental beliefs toward appropriate activities for girls and boys may become stronger as children age. By examining early adolescent children’s emotional discourse with their mothers and fathers, we explored whether early adolescent and their parents would adhere to gender-stereotyped patterns.

Of course studying parent–child dyads—especially when the child is an early adolescent—raises the difficulty of teasing apart each member’s contribution to the conversation. As proposed by social cognitive theory, parents and children influence each other bi-directionally (Bandura, 1989). Notwithstanding, each member’s contribution was examined separately by calculating which member of the dyad was the first to mention each of the three specific emotions. Because both members were considered equal participants, exploratory analyses were conducted on who was the first member of the dyad to introduce each specific emotion term first. We did not advance specific directional hypotheses.

Although the majority of past researchers either have used unstructured naturalistic dialogues or mother–child conversations about the child’s past emotional experiences, in the present study we employed a semi-structured discourse design. Specifically, early adolescents and their parents were asked to attribute emotion to protagonists in vignettes designed to educe the three emotions being studied, namely anger, sadness, and frustration. The present study taps emotional attributions in a semi-structured way. For instance, prior research involved the experimenter reading the vignette to a child followed by forced-choice questions pertaining to specific emotions and emotional intensity (e.g., Brody, 1984). Parents and early adolescents read the story themselves and answered open-ended questions, which allowed for further discussion and development of the protagonist’s feelings. We selected this task based on findings that children are more likely to report the experience of an undesirable emotion when they are attributing that emotion to a character rather than to themselves, especially when the three emotions of interest are dysphoric (Brody & Carter, 1982). By presenting the early adolescents with materials that did not explicitly state the emotions appropriate for the protagonist, further exploration of gender differences in an ambiguous emotional context was allowed. Although the absolute amount of emotion expression itself may be limited in an ambiguous context when compared with that of a structured conversation regarding past emotional experiences, we believed that the supposed gender differences under investigation would continue to be perceptible. Underlying this rationale is the suggestion that greater gender differences in children’s emotional expression may be found when children are presented with material or stories that are somewhat indistinct from the emotions under investigation (Brody & Carter, 1982).

Finally, fathers were specifically included, given that mothers and fathers may socialize emotions differently (Garside & Klimes-Dougan, 2002). With the exception of the Fivush et al. (2000) study of preschool children, fathers have not been examined in this literature. Children may be more or less emotionally expressive with fathers or mothers. Thus, early adolescent girls and boys discussed two stories designed to elicit sadness, anger, and frustration with their mothers as well as with their fathers. Based on previous research, it was hypothesized that girls would use more emotion terms than would boys. In addition, it was predicted that girls would use a higher proportion of references to sadness than would boys, and boys would use a higher proportion of references to anger and frustration than would girls. Similarly, it was expected that parents of daughters would use a higher proportion of references to sadness with their children than would parents of sons, whereas parents of sons would use a higher proportion of references to anger and frustration with children than would parents of daughters. Specifically, there were four main hypotheses. First, daughters would use more emotion terms than would sons. Second, daughters would use a higher proportion of references to sadness and a lower proportion of references to anger and frustration in their discussions with both mothers and fathers than would sons. Third, mothers of daughters would use a higher proportion of references to sadness and a lower proportion of references to anger and frustration with their children than would mothers of sons. Fourth, fathers of daughters would use a higher proportion of references to sadness and a lower proportion of references to anger and frustration with their children than would fathers of sons.

Method

Participants

Early adolescents

There were 29 girls (M age = 12.50, SD = 11.85 months) and 27 boys (M age = 12.40, SD = 10.19 months) in the sixth and eighth grades and their mothers and fathers. Families lived in the San Francisco and central coast areas of California. They were recruited from public schools, summer camps, and after-school activities. There was no significant age difference between girls and boys in either age group, F (1, 55) < 1. There were no significant gender differences in birth order.

Mothers

The mean age of the 56 mothers was 43.1 years (SD = 4.8), and their ages ranged from 32 to 53 years. Eighty-two percent of mothers identified themselves as European American, 13% identified themselves as Latina (six of Mexican descent and one of Central American descent), and 5% identified themselves as being of Asian descent. The majority of the mothers were in the paid labor force (76.8%) and these 46 worked a mean of 32.4 h (SD = 11.2) per week for pay. Of the 52 mothers who reported an occupation, 48% were classified as business managers, executives, and other professional posts, 19% were classified as administrative personnel, 14% were classified as housewives, 6% were classified as working in clerical or sales, 4% were classified as skilled manual employees, 4% were classified as unskilled employees, 4% were classified as students, and 2% were classified as disabled based on the Hollingshead Occupational Index. Mothers’ educational background ranged from having graduated from high school (3.6%) to having earned graduate or professional school (32.1%). On average, mothers had a bachelor’s degree.

Fathers

The mean age of the 56 fathers was 46 years (SD = 4.8), and their ages ranged from 34 to 59 years. Seventy-five percent of fathers identified themselves as European American, 14% identified themselves as Latino (seven of Mexican and one of Caribbean descent), 7% identified themselves as being of Asian descent, and 4% identified themselves as African American. Ninety-three percent of fathers were in the paid labor force and these 49 reported working a mean of 45.9 h (SD = 9.4) per week for pay. Of the 52 fathers who reported an occupation, 50% of fathers’ occupations were classified primarily as business managers, higher executives, and other professional posts, 23% as administrative personnel, 14% as skilled manual employees, and 8% were classified as working in clerical or sales based on the Hollingshead Occupational Index. The fathers’ educational background ranged from having completed the 11th grade of high school (1.8%) to having earned a graduate or professional degree (36.4%). On average, fathers had a bachelor’s degree.

The average income of the families was $60,000–74,999, and incomes ranged from $15,000 to over $100,000. Four pairs of parents were divorced. In these cases, parents shared custody of their children. Finally, there was no significant difference between the occupational prestige of parents of girls and that of parents of boys.

Procedure

The second author and a research assistant visited families in their homes on two separate days. Mothers and fathers were visited separately, and the order of the parent visits was counterbalanced. The two visits took place within a week. Thus, children completed tasks with each parent on separate days. Upon arrival, the parent and early adolescent were told that the researchers were interested in how “parents contribute to children’s learning in everyday situations.” After signing a consent form, dyads were told that they would be asked to complete four tasks. Three tasks involved science activities and will not be discussed further. The fourth task was the moral dilemma task. The dyad was given 10 min to complete each task for a total of 40 min of completed tasks per day. After explaining a task to the families, the second author and an assistant turned on the video recorder and left the room so that the parent–child dyad was alone to complete that particular task. The researchers returned at the end of each task to instruct families on the next task. These tasks were counterbalanced using a Latin square design. For one-quarter of the families, the moral dilemmas came before the science tasks. Children and parents spent an additional 20 min completing questionnaires, which are not reported here. For more information about the science tasks or the questionnaires, see Tenenbaum and Leaper (2003).

Interpersonal reasoning tasks

These tasks consisted of a set of cards with printed dilemmas and questions. Parents and children read about two different interpersonal dilemmas, which were adapted from Selman, Beardslee, Schultz, Krupa, and Podorefsky (1986). The dyad had 10 min to complete this task. They were given two story cards (one on top of the other) simultaneously and asked to answer the questions about both cards. They were asked to discuss the second one as soon as they finished discussing the first one. Dyads were requested to participate equally in the discussion. At the end of the 10 min, if the dyad had not finished, the researcher asked them to move on to the next task. Dyads took the entire 10 min to finish the task. A few families did not finish in 10 min; however, videotaping was halted at the 10-min mark.

Each set of dilemmas included one conflict between two child peers and another conflict between a child and an authority figure. Each parent–child dyad received different dilemmas so that the child never read the same dilemmas twice. The order of which type of dilemma the participants read first was counterbalanced across families. Father–child and mother–child dyads received the cards in the same order as each other (i.e., either peer followed by authority conflict, or authority followed by peer conflict), but read different dilemmas. For example:

The main character is Patty. Patty is looking forward to recess because she and her friends are going to practice for the school competition in soccer. They have a game later in the week. During class, Patty’s teacher suggests that Patty stay in at recess to get help in her math. Patty is behind in math and there is a test later in the week.

The second dilemma in each set concerned a conflict between peers, such as:

The main character is Mary. Mary and Sue are friends. They have been assigned to work together on a science project in school and only have two days to finish the project. They meet after school and Mary says she wants to start working on the project right away, but Sue wants to play softball first.

The gendered names of the protagonists in the dilemma stories matched the gender of the child participants. Parents and children were instructed to answer a series of questions after reading the dilemmas, e.g., “What is the problem here?” (see Appendix for complete question list and the other two interpersonal dilemmas). The question of interest to the present study was “How does the main character feel? Why?”

Coding

Message units

Verbatim transcripts were created from the videotapes. Utterances were segmented into message units. A message unit is an individual speech act with a single thought unit that was bound by its intonation. For reliability, the second author and a research assistant separately segmented 24 transcripts (three sixth grade girl–mother and father transcripts, three eighth grade girl–mother and father transcripts, three sixth grade boy–mother and father transcripts, three eighth grade boy–mother and father transcripts). This constitutes 20% of the data set. The coders reached an agreement of 88%. After attaining reliability, the research assistant finished segmenting the transcripts.

Emotion words

The present authors coded any instances that parents and children mentioned anger, sadness, frustration, stress, dislike, and happiness. These were the emotions that occurred with any frequency. Examples of anger included references to being “mad”, “irritated”, or “angry”. Examples of sadness included “sad,” “upset,” “unhappy”, or “miserable”. An example of frustration would be if one mentioned the word “frustrated” (see Tables 1 and 2 for examples of conversations). Stress included references to “fear”, “conflicted”, or “anxiety”, happy included “pleased”, and dislike included “not liking” and “hate”. The codes did not distinguish between emotion state and behavior, or between adjectives, nouns, and verbs.

Table 1 Example of a conversation.
Full size table
Table 2 Example of a conversation.
Full size table

To test for intercoder reliability, the two authors independently coded 24 transcripts (20% of the data set). Coders were not blind to the hypotheses. Reliability was evaluated with kappa coefficients. According to Fleiss (1981), kappa coefficients above 0.75 reflect excellent agreement. An overall kappa of 0.85 was obtained, with the following kappa coefficients for individual codes: anger, κ = 0.91; frustration, κ = 0.100; sadness, κ = 0.88, stress, κ = 0.100; happiness, κ = 0.85; dislike, κ = 0.95. After reliability was attained, coding was completed within approximately 3 weeks. Disagreements were resolved through discussion.

Results

Study variables and data analysis plan

We coded whether parents and children individually mentioned any of the six emotion words. Thus, each participant received a score for each time they mentioned an emotion term. Because families differed in the amount of time that they devoted to each vignette, we collapsed across the vignettes. To control for differences in the base rate of emotion terms, we created proportion scores for each individual for each of the three target emotions (i.e., sadness, frustration, and anger) examined in this paper by dividing each of these three emotion references individually by the number of the six emotion words used by that participant. We first report descriptive statistics and then hypothesis testing. To test the hypotheses, we first conducted an ANOVA to examine whether daughters used more emotion words than did sons. Second, we carried out six ANOVAs with child gender and parent gender as predictor variables for the proportion of sadness, frustration, and anger references used by children and parents. Child gender was a between-subjects factor, and parent gender was a within-subjects factor, in all of the ANOVA models. Significant main and interaction effects are reported. Third, we report which member of the dyad was the first to mention each of the specific emotions of interest. Finally, we also provide eta2 estimates to indicate a measure of the proportion of variance accounted for by a predictor. Eta2s between 0.01 and 0.09 indicate a small effect size, eta2s between 0.09 and 0.24 indicate a medium effect, and eta2s greater than 0.25 indicate a large effect (Cohen, 1988).

Descriptive statistics

Parents and children were given 10 min for the tasks. Fathers directed a mean of 483.63 words (SD = 210.30) to daughters and a mean of 538.46 words (SD = 249.42) to sons. Mothers directed a mean of 601.63 words (SD = 288.32) to daughters and a mean of 545.42 words (SD = 221.15) to sons during the interpersonal conflict conversations. Neither mothers nor fathers directed a different number of words to sons or daughters, F (1, 54) < 1. In addition, mothers and fathers used roughly equal numbers of words, F (1, 54) = 1.66, ns. Message units were also calculated. Fathers spoke a mean of 72.72 utterances (SD = 33.15) to daughters and a mean of 63.48 utterances (SD = 26.11) to sons. Mothers spoke a mean of 72.90 utterances (SD = 28.86) to daughters and a mean of 69.52 utterances (SD = 25.33) to sons. Neither mothers nor fathers spoke different numbers of utterances to sons or daughters, F (1, 54) = 1.32, ns. In addition, overall mothers and fathers spoke roughly equal numbers of utterances, F (1, 54) < 1.

Mothers used a mean of 5.16 emotion terms (SD = 4.69), while fathers used a mean of 5.96 (SD = 5.50) emotion terms. Forty-eight of fifty-six mothers (86%) used at least one emotion term, whereas 50 of 56 fathers (89%) used an emotion term at least once. While speaking with mothers, children used a mean of 6.30 emotion terms (SD = 5.08), and while speaking with fathers, children used a mean of 8.27 emotion terms (SD = 7.16). Fifty of fifty-six children with mothers (89%) used at least one emotion term, while 54 out of 56 children speaking with fathers (96%) used an emotion term at least once. The six emotion terms constitute the number of possible emotion terms.

Preliminary analyses

There were no significant age or age interaction effects on the total number of emotion words used, the proportion of words that referred to sadness, frustration, or anger, for children, mothers, or fathers. To increase power, age is not included in future analyses.

Child speech hypotheses

Hypothesis 1

To test whether girls would use more emotion terms than would boys, a 2 (daughter, son) × 2 (mother, father) ANOVA was conducted on the total number of emotion references made by daughters and sons. Girls made more overall references (M = 16.81, SD = 8.78) to emotion than did boys (M = 12.11, SD = 4.76), F (1, 54) = 6.20, p − 0.02, eta2 = 0.10. There were no parent gender or parent gender × child gender interactions effects, both Fs (1, 54) < 1.

Hypothesis 2

To test whether daughters would use a higher proportion of references to sadness in their discussions with both mothers and fathers than would sons, a 2 (daughter, son) × 2 (mother, father) ANOVA was conducted on the proportion of references to sadness children used. Neither a child gender nor a parent gender main effect was significant, both Fs (1, 54) < 1. However, a parent × child gender interaction effect was significant, F (1, 54) = 9.54, p = 0.003, eta2 = 0.16. Follow-up ANOVAs revealed that when talking to mothers, daughters (M = 0.22, SD = 0.29) and sons (M = 0.10, SD = 0.15) used an equal proportion of sadness terms, F (1, 54) = 3.49, p = 0.07, eta2 = 0.06. In contrast, when speaking with fathers, sons used a higher proportion of references to sadness (M = 0.24, SD = 0.33) than did daughters (M = 0.10, SD = 0.14), F (1, 54) = 4.17, p < 0.05, eta2 = 0.07.

To test whether sons would use a higher proportion of references to anger in their discussions with both mothers and fathers than would daughters, a 2 (daughter, son) × 2 (mother, father) ANOVA was conducted on the proportion of references to anger children used. There were no significant effects of child gender, parent, or parent × child gender interaction, all Fs (1, 54) < 1.

To test whether sons would use a higher proportion of references to frustration in their discussions with both mothers and fathers than would daughters, a 2 (daughter, son) × 2 (mother, father) ANOVA was conducted on the proportion of references to frustration children used. Daughters (M = 0.66, SD = 1.32) used a higher proportion of references to frustration than did sons (M = 0.11, SD = 0.42), F (1, 54) = 4.20, p < 0.05, eta2 = 0.07. There were no significant effects of parent or parent × child gender interaction effects, both Fs (1, 54) < 1.

Parent speech hypotheses: Mothers and fathers

Hypotheses 3 and 4

To test whether mothers and fathers of daughters would use a higher proportion of references to sadness with their children than would mothers and fathers of sons, a 2 (daughter, son) × 2 (mother, father) ANOVA was conducted on the proportion of references to sadness mothers and fathers used. There were no significant effects of child gender, parent gender. Finally, there was not a significant parent × child gender interaction effect, all Fs (1, 54) < 1.

To test whether mothers and fathers of sons would use a higher proportion of references to anger with their children than would mothers and fathers of daughters, a 2 (daughter, son) × 2 (mother, father) ANOVA was conducted on the proportion of references to anger mothers and fathers used. There were no significant effects of child gender, parent gender, nor a parent × child gender interaction effect, all Fs (1, 54) < 1.

To test whether mothers and fathers of sons would use a higher proportion of references to frustration with their children than would mothers and fathers of daughters, a 2 (daughter, son) × 2 (mother, father) ANOVA was conducted on the proportion of references to frustration mothers and fathers used. Parents used a higher proportion of references to frustration when speaking with daughters (M = 0.69, SD = 1.07) than with sons (M = 0.15, SD = 0.36), F (1, 54) = 4.20, p < 0.05, eta2 = 0.07. There were no significant effects of parent gender, or parent × child gender interaction effects, all Fs (1, 54) < 1.

First mentions of specific emotions

Finally, to try to tease apart the direction of effects, we examined which member of the dyad was the first to mention sadness, anger, and frustration. The proportion of times that the parent initiated the specific emotion was divided by the total number of conversations in which the specific emotion term was used.

Mothers

During mother–child discussions, 17% of the mothers mentioned sadness before their sons did, whereas 47% of the mothers mentioned sadness before their daughters did. However, a Fisher’s Exact Test (df = 1) did not reveal significant differences in whether mothers of daughters or sons were more likely to initiate discussions of sadness. When discussing anger, 9% of the mothers mentioned the emotion before their sons did, whereas 18% of the mothers mentioned the emotion before their daughters did. Once again, however, the difference was nonsignificant, Fisher’s Exact Test (df = 1). In addition, a Fisher’s Exact Test (df = 1) revealed no significant difference between mother and child when initiating discussions of frustration. Eighty percent of the mothers mentioned frustration before their sons did, whereas 71% of the mothers mentioned frustration before their daughters did.

Fathers

During father–child conversations, 29% of fathers mentioned sadness before their sons did, whereas 43% of fathers mentioned sadness before their daughters did. Analyses, again however, failed to reveal any significant differences in who initiated the use of sadness, χ 2 (1, N = 56) = 0.73, ns. Conversely, a Fisher’s Exact Test (df = 1) revealed a significant difference in the initiation of anger (p = 0.04). In these conversations, 90% of daughters mentioned anger before their fathers did, whereas 57% of sons mentioned anger before their fathers did. Talking with fathers, thus, daughters were more likely than sons to initiate anger. Within the father–daughters discussions regarding frustration, 75% of girls mentioned frustration before their fathers did. Chi-square analyses could not be conducted because father–son dyads never mentioned frustration.

Discussion

The results confirmed the first hypothesis that daughters would make more total references to emotion than would sons. Contrary to expectation, however, sons were found to use a higher proportion of references to sadness than did daughters when speaking with fathers. Once more, in contrast to the hypothesis, daughters used a higher proportion of references to frustration than did sons. Finally, parents used a higher proportion of references to frustration when speaking with daughters than with sons. Although the first finding supports stereotypical gender patterns, the other three findings call into question stereotypical patterns of affective speech.

Early adolescents followed stereotypical norms in that girls made more frequent references to emotions than did boys. Previous researchers have argued that women are more emotionally expressive than men. For example, women talk more about emotions (Goldschmidt & Weller, 2000) and report being more emotionally expressive (Bronstein et al., 1996) than do men. The findings in the present study confirm these stereotypical patterns in an adolescent population.

By contrast, use of specific emotional terms such as sadness did not follow stereotypical patterns. For instance, daughters and sons made an equal amount of references to sadness with their mothers, which was unexpected. Even more unexpected, however, was that boys spoke of sadness more than did girls when partnered with their fathers. Mothers and fathers were found to refer to sadness with equal frequency with their daughters and sons. The findings suggest that partner gender influences boys’ talk. Although some have suggested that mothers and fathers may socialize emotion expression in different ways and to differing degrees (Garside & Klimes-Dougan, 2002), the lack of research on children’s emotional expression when fathers are their conversational partners inhibits postulations as to why sons spoke more of sadness to their fathers than did daughters. Perhaps by attributing sadness to the protagonists instead of to themselves, fathers and sons were able to express sadness in a comfortable manner.

The context of the vignette itself may have played a role in boys’ greater propensity compared to girls in expressing sadness with fathers. For instance, Kelly and Hutson-Comeaux (1999) suggested that men are more emotionally expressive in achievement domains, whereas women are more emotionally expressive in interpersonal domains. Because the vignettes used in the present study utilized both interpersonal (e.g., friendship, romantic) and achievement (e.g., goal or work-related events) domains, the sons and fathers might have chosen to view all of the vignettes as strictly achievement-related. Hence, the sons may have felt comfortable attributing sadness to the protagonist when paired with their fathers more so than did the daughters because of the dyads’ understanding of the domain of the vignette in question. In contrast, while speaking with mothers, boys may not have not viewed the vignettes as singularly achievement-oriented, which may account for girls’ and boys’ similar references to sadness when speaking with mothers. Why sons and their fathers did not use more references to anger and frustration, however, remains to be seen. Further research on the father’s role in emotion socialization needs to be conducted to explore this supposition.

It was hypothesized that boys and girls would conform to stereotyped expressions of emotion, but the findings differed from expectations regarding the masculine-stereotyped emotions of anger and frustration. Girls were more likely to mention frustration in conversations with their parents than were boys. In addition, girls and boys were found to discuss equal amounts of anger with their mothers and with their fathers. Although anger is masculine stereotyped (Shields, 1984), findings regarding the expression of this emotion are far from conclusive. Brody (1996) argues that gender differences are not always found because emotion expression, including that of anger, are context specific. The findings in the present study regarding the masculine-stereotyped emotions of anger and frustration conflict with the small number of studies that indicate a gender difference in expression. Moreover, girls mentioned anger before their fathers did. Whereas girls and boys made a similar number of references to anger within the present study, girls focused a higher proportion of their talk on frustration than did boys. Thus, girls and boys in the present sometimes study spoke in a nontraditional manner.

Why might the girls and boys in this sample sometimes have spoken in a nontraditional manner in terms of the specific emotions used? Research suggests that, prior to high school, children may be less gender-typed than when they are older. According to Stoddart and Turiel (1985), there is a curvilinear pattern of gender flexibility in which children are inflexible about gender-role transgressions at age 5, become and remain flexible until around age 13, at which point early adolescents become inflexible again. In their beliefs about the self and others, children become more flexible from middle childhood until early adolescence (Katz & Ksansnak, 1994). Past work on emotion expression and socialization during parent–child conversations has largely been confined to younger children. Age may partially explain why the girls acted in a gender nontraditional manner.

Second, the parents and children discussed other people’s emotions. Thus, we cannot answer whether the conversations would have differed if the parents and early adolescents had conversed about their own experiences. Parents and children did, however, make frequent analogies to the children’s own lives in which they had experienced similar situations. Future researchers should examine parent–adolescent conversations about children’s personal experiences to shed light on this matter.

Of course, past studies have not shown differences in anger expression (Anderson & Leaper, 1998; Shimanoff, 1985; for reviews, see Fischer, 1993 and Shields, 1987). Gender differences are often larger in self-report measures than in expression or observation of experience (La France & Banaji, 1992). Perhaps, individuals distort their perceptions and memory of emotions to conform to gender stereotypes. Developmental researchers have found that children frequently distort gender-inconsistent information (e.g., a picture of a man in the kitchen) to gender-consistent information (e.g., a picture of a woman in the kitchen) on memory tasks (Bigler & Liben, 1990, 1992; Levy & Fivush, 1993; Martin & Halverson, 1983; Welch-Ross & Schmidt, 1996). Similar processes may contribute to stereotypes about gender and emotionality. Thus, boys may not express frustration and anger more readily than girls do.

Fathers used a higher proportion of emotion talk focused on frustration with their daughters than with their sons. Frustration can be understood as a mitigated form of anger. Although frustration is stereotyped as a masculine emotion, it can be seen as a much less masculine stereotyped emotion than anger is. Therefore, the fathers in present study may have viewed frustration as a more appropriate response for girls in dilemmas that they saw as likely to invoke anger. Nonetheless, frustration is not as passive as sadness. This finding further questions the cultural stereotype that girls endorse more passive emotions than do boys.

That fathers were less likely to uphold gender stereotypes than mothers is noteworthy and unexpected. Prior studies have indicated that gender-differentiated treatment is more likely among fathers than mothers (Siegal, 1987; Tenenbaum & Leaper, 2003). Leaper (2000) has suggested that men tend to be more concerned than women with the adoption of gender-typed behavior. Although both mothers and fathers engaged in gender-differentiated talk with sons and daughters, fathers spoke in a nontraditional manner with their daughters. It is difficult to understand why we found the opposite pattern. Perhaps fathers in the present sample were more gender egalitarian. Unfortunately, we did not assess their gender beliefs, which is a limitation of the present study. Future researchers need to assess whether emotional expression is related to gender stereotypical attitudes and beliefs.

Finally, we did not find age effects, nor have others shown age effects either in how children speak or in how parents speak to children of these ages. For example, research on parents’ science talk to children has not shown differences based on age (e.g., Tenenbaum & Leaper, 2003). By the time children have reached early adolescence, parents and children may rely on patterns of interaction that are relatively stable. Future longitudinal research is needed to test the stability of children’s and parents’ talk at these ages. Longitudinal designs may be better than cross-sectional designs at uncovering subtle age-related changes.

Limitations

Two other limitations of the present study are worth noting. First, we did not fully tease apart which partner was more influential in deciding which emotions were most relevant to the dilemmas. Children actively contribute to their socialization (Bandura, 1997). The conversational context of when an emotion was mentioned could provide valuable information. Second, the conversations were very structured and only about a specific topic. Emotional disclosure varies along with conversational topic (Anderson & Leaper, 1998). To understand emotion in its conversational context, it would be necessary to examine more than one conversational topic.

Conclusions

As suggested by social cognitive theory (Bandura, 1997), parents socialize children into the specific attitudes and values endorsed by their cultural community through daily activities, such as conversations. One such value is the expression of emotion that conforms to culturally prescribed gender roles. Discussing emotion provides children with opportunities to become emotionally expressive. Social cognitive theory proposes that not only do parents influence their children, but that children and their parents also influence each other bi-directionally (Bandura, 1989). Given that mothers were more likely to discuss frustration in their conversations with daughters than with sons, and, in turn, daughters were more likely than sons to mention frustration to their parents, the children and their parents in the present study may have taken cues from each other as to what was an acceptable emotion for daughters in different dilemmas. Apart from the specific emotions studied, the boys and girls in this sample expressed themselves in a gendered-stereotypical manner in that the daughters used more emotion words than did the sons. Although this behavior is congruent with gender stereotypes that women are more emotional than men, the same stereotyped expression of emotion was not found within the adults’ discourse in this sample. Mothers and fathers were not found to differ in the number of emotion words they spoke. The absence of stereotyped expression from the parents in the sample should not, however, be taken as a shortage of stereotypic gender role socialization concerning emotion for these children. Given girls’ use of more emotion words than boys in the present study, it seems that these adolescent boys and girls had acquired stereotypes regarding the appropriateness of feminine emotionality, but from whom is unclear. Future researchers should take into account the multi-faceted role of modeling gender-typed behavior for children in order to address more fully these issues.