Etch the Emotional Life: Mother-Child Emotion Socialization from Age 3 to 6 Years

Abstract

Emotion socialization is conceptualized as a relational process, yet children’s role in socializing parents’ emotions is rarely considered. This study explored longitudinal patterns of mother-child emotion socialization from early to middle childhood. Participants were 349 children (51% boys, 49% girls) and their mothers from the southeastern United States. Children were 79.4% White, 12% Black or African American, and 8.3% multi-racial or other; 6.3% were Hispanic/Latino. Mother-child dyads completed an etch-a-sketch task when children were 3, 4, and 6 years old. At each time, mothers’ and children’s expression, coaching, and dismissing of positive and negative emotions were observed. Hierarchical Linear Modeling analyses identified developmental trajectories and within-dyad associations of emotion coaching and dismissing with emotion expression. Over time, expression of positive emotions decreased and expression of negative emotions showed no change for both mothers and children. Mothers decreased in coaching children’s positive emotions and showed no change in coaching children’s negative emotions. Children increased in coaching mothers’ positive emotions and showed no change in coaching mothers’ negative emotions over time. Both mothers and children decreased in dismissing emotions over time. Within dyads, mothers’ increases in coaching and in dismissing related to children’s increased expression. Children’s increases in coaching related to mothers’ increased expression. Findings highlight complexity and dynamics of emotion socialization processes over time.

Etch the Emotional Life: Mother-Child Emotion Socialization from Age 3 to 6 Years

Across the past two decades, a large body of research has accumulated demonstrating the importance of emotion socialization for children’s outcomes (Spinrad et al., 2020). Emotion socialization takes multiple forms, including modeling of emotional expression and behaviors that encourage or discourage children’s emotional expression and shape their attributions about and understanding of emotions (Eisenberg, 2020; Morris et al., 2007). Adjusting responsively to children’s developmental change is important for effective parental socialization in general (Dix, 2000) and effective emotion socialization in particular (Kehoe et al., 2020). However, relatively little research has examined longitudinal change in parental emotion socialization, and most of the research that has done so relied on questionnaire measures rather than observation. Furthermore, although theory and research suggest that children’s behaviors may influence or socialize parents’ emotions (Kuczynski et al., 2016; Saarni et al., 2006), little research has addressed children’s engagement in emotion socialization behaviors. The purpose of the current study was to examine trajectories of emotion socialization behaviors and emotional expression in mother-child dyads from early to middle childhood. The first aim was to examine longitudinal change for both mothers and children. The second aim was to investigate whether within-dyad changes in mothers’ emotion socialization behaviors were associated with changes in children’s emotional expression over time, and conversely whether within-dyad changes in children’s emotion socialization behaviors were related to changes in mothers’ emotional expression over time.

Emotional Development from Early to Middle Childhood and Emotion Socialization

As Saarni and colleagues (2006) review, from early childhood to middle childhood children become increasingly able to identify their experience of mixed emotions, or multiple emotions experienced simultaneously. In middle childhood compared with early childhood, children have more complex understanding of emotional display rules (Saarni et al., 2006) and report greater need to mask or minimize their emotional expression (Zeman et al., 2006). Moreover, children’s emotion vocabulary increases, and their skill at using language to communicate emotions in socially appropriate ways becomes more sophisticated from early childhood to middle childhood (Saarni et al., 2006). In middle childhood, children also begin to use advanced emotion regulation strategies, such as problem solving, reappraisal, and suppression (Saarni, 2008). Relatedly, mothers report better child self-regulation when children are in middle childhood compared with early childhood (Raffaelli et al., 2005). Finally, from early to middle childhood, children become increasingly able to integrate multiple types of cues to discern others’ emotions and calibrate emotional communication to maintain relational harmony (Saarni et al., 2006). Parents engage in emotion socialization behaviors when they detect cues that their child is experiencing an emotion. Indeed, an important component of effective emotion socialization is parents’ identification of subtle or “low intensity” signals of their child’s emotional experience (Gottman et al., 1996; Kehoe et al., 2020). Thus, typical developmental changes in emotional experience, expression, regulation, and understanding that commonly occur in early and middle childhood may be pertinent for emotion socialization. The longitudinal design of the present study includes time points at ages 3, 4, and 6 years – two points in early childhood and one at the beginning of middle childhood – to encompass this period of rapid change in emotional development.

Longitudinal Change in Emotion Socialization

Parents’ behaviors or reactions to children’s emotions as a direct way of socialization are often categorized as supportive or nonsupportive. As emotion socialization is embedded within the sociocultural context, culturally-specific reactions may exist, and whether particular reactions are considered as supportive or nonsupportive may vary according to cultural models of self and self-other relationship as well as familial experiences of racial discrimination or other forms of marginalization (Dunbar et al., 2017; Raval & Walker, 2019). Despite such cultural differences, research with diverse samples demonstrates the benefits of supportive emotion socialization behaviors, which generally show acceptance of children’s emotions and actively guide children’s learning about and management of emotions, and detrimental effects of nonsupportive emotion socialization behaviors, which commonly communicate lack of acceptance or negative judgment of children’s emotions (Spinrad et al., 2020).

In recent years, the field has begun to address stability of emotion socialization over time. Studies with samples ranging from early to middle childhood showed stability of parent-reported supportive and nonsupportive responses to children’s negative emotions across two or three time points (e.g., Fung et al., 2021; Nelson et al., 2016). Ugarte and colleagues (2021) worked with predominantly middle- to upper-middle class European American families and found stability of parent-reported nonsupportive responses to child negative emotions with an early childhood sample followed across four waves of data collection that spanned seven years. Despite this convergence of findings, parents’ report of emotion socialization behaviors may be influenced by social-desirability bias or unintentional adjustment for developmental change, amplifying consistency over time.

Two other studies have addressed change in emotion socialization over time using different methods. Stettler and Katz (2014) conducted semi-structured interviews with European American parents when children were 5, 9, and 11 years old to measure parental meta-emotion philosophy, an emotion socialization mode involving parents’ thoughts, feelings, and beliefs about negative emotions that may affect their emotion socialization practices (Gottman et al., 1996; Katz et al., 2012). An emotion coaching meta-emotion philosophy is characterized by acceptance and validation of emotions and may include supportive emotion socialization behaviors such as talking with children about emotions or helping children strategize how to cope with emotions. An emotion dismissing meta-emotion philosophy is characterized by deprecation or fear of emotions and may include nonsupportive emotion socialization behaviors such as ignoring, criticizing, or attempting to quickly remove children’s emotions (Katz et al., 2012). Stettler and Katz (2014) found curvilinear effects for elements of an emotion coaching philosophy, with decreases from 5 to 9 years of age and increases from 9 to 11 years of age that went above the 5-year-old baseline. We note that the terms “emotion coaching” and “emotion dismissing” are also used in the field for the supportive (or nonsupportive) emotion socialization behaviors that are consistent with the respective meta-emotion philosophies (e.g., Lunkenheimer et al., 2007). Wu and colleagues (2019) observed maternal emotion coaching behaviors for happiness, sadness, anger, and fear during a parent-child reminiscence task when children were 3 years old and again when children were 4 years old in the United States (U.S.). Descriptively, their results are consistent with Stettler and Katz (2014) in showing a decline in coaching of negative emotions across time points. Means for coaching of happiness were consistent across time points. In contrast to the studies using parent report, both Stettler and Katz (2014) and Wu and colleagues (2019) suggest decreases in parents’ coaching of negative emotions across early childhood, calling for studies with different methods to better understand how emotion socialization may or may not change from early to middle childhood.

Behavioral Observation of Parent-Child Emotion-Related Interactions

Behavioral observations are especially advantageous when constructs include nonverbal components (Bakeman & Quera, 2012), which is the case for emotion socialization and emotional expression. Observation of emotion socialization has often been conducted through parent-child reminiscing conversations about emotions (e.g., Hernandez et al., 2018; Lozada et al., 2016; Wang, 2013), which may encourage a focus on emotions that is not necessarily typical for families. In other words, reminiscing conversations about emotions may represent parents’ behavior when intending to discuss emotions with their child, but such intentions may not always be common during families’ everyday activities. Loop and Roskam (2016) randomly assigned Belgian parents to view an emotion coaching video or a video about child development, then had parents and children participate in free play and frustration tasks (i.e., disappointment, puzzle, Lego). Parents who were exposed to the emotion coaching video were more positively expressive and emotionally responsive to their child during free play and provided more sensitive behavioral support to their child during the frustration tasks, compared with parents in the control group.

In the current study, we used a challenging etch-a-sketch task that requires parents and children to cooperate in order to draw various shapes. Emotional expressions and responses are commonly elicited without directly asking parents to talk about and engage with emotions, thus enabling the assessment of unprompted engagement in emotion socialization. The task also enabled us to assess socialization and expression of positive emotions (e.g., excitement, pride) as well as negative emotions (e.g., frustration, anger). Brown and Fredrickson (2021) described the role of shared positive emotions in developing social competence, and mothers’ reports of nonsupportive responses to positive emotions were related to children’s problem behaviors in middle childhood in predominantly European American families (Yi et al., 2016). Including socialization and expression of both positive and negative emotions provides a more comprehensive understanding of emotion socialization processes.

Children’s Engagement in Emotion Socialization

Despite recognition that emotions are relational or transactional phenomena that influence the organization or climate of relationships and family systems (Morris et al., 2007; Saarni et al., 2006), research on emotion socialization has rarely addressed children’s potential socialization of their parents’ emotions. Within parent-child interactions, children may engage in emotion socialization behaviors towards their parents, whose emotional expression may be elicited by their children’s behavior. For example, when a parent’s smile is mirrored by their child, the parent’s positive emotion is likely to be enhanced and expressed again. When a child turns their back or abruptly looks away when their parent is frowning, the parent may alter expression to re-engage with their child. Given the greater power parents hold compared with children in parent-child relationships, and parents’ longer history of emotional development, children’s emotion socialization behaviors will not hold the same meaning for parents that parental emotion socialization behaviors will for children. Nonetheless, because emotion socialization is a transactional process, children’s emotion socialization behaviors may impact parents’ emotional expression over time. Parental behaviors have a socialization influence on children regardless of parents’ intent. Similarly, children do not need intent for their emotion socialization behaviors to influence their parents. Therefore, the current study innovates by examining children’s potential role as emotion socializers of their parents.

Current Study Aims and Hypotheses

In this longitudinal study we investigate emotion socialization from age 3 to age 6 with a sample of mother-child dyads from the U.S. who participated in a larger study of emotion and cognition across early development (i.e., the CAP study; e.g., Bruce et al., 2023). The observational variables for the current study are entirely unique from previously published works using this data set. We expand the current literature by a) using an emotion-eliciting observational task to assess families’ unprompted engagement in emotion socialization; b) including expression of and response to both positive and negative emotions; and c) examining active roles of both parents and children in emotion socialization.

The current study has two aims. Aim 1, referred to as the Development Aim, was to examine trajectories over time of mothers’ and children’s emotional expression, emotion coaching, and emotion dismissing, each examined separately. Chaplin and colleagues (2017) observed children completing frustrating tasks with mothers with an economically strained, mostly White sample and found that from age 3 to 5 years, children showed marginally significant increases in expression of happiness, no significant changes in expression of anger, and significant decreases in expression of sadness. Parents’ expressiveness in the family tends to relate to children’s expressiveness (Halberstadt & Eaton, 2003). Thus, we hypothesized increases in expression of positive emotion and decreases in expression of negative emotion over time for both mothers and children. Based on Stettler and Katz (2014), we expected maternal coaching of negative emotions to decrease and maternal dismissing of negative emotions to increase from early to middle childhood. Based on Wu et al. (2019), we hypothesized that maternal coaching of positive emotions would decrease or show no change over time. We also expected that maternal dismissing of positive emotions, parallel to dismissing of negative emotions, would increase over time. Although previous research has not examined children’s socialization of parents’ emotions, based on gains in emotional communication skills from early to middle childhood (Saarni et al., 2006), we expected that children would increasingly engage with parents’ emotions in various forms, and thus increase in both coaching and dismissing over time for both positive and negative emotions.

Aim 2, referred to as the Within-Dyad Association Aim, was to examine at the individual level (or in this study, dyad level) rather than at the group level how within-dyad changes in mothers’ emotion socialization behaviors related to children’s emotion expression over time, and how within-dyad changes in children’s emotion socialization behaviors associated with mothers’ emotion expression over time. Parents’ emotion coaching meta-emotion philosophy is related to children’s emotional competence, including emotion regulation and socially appropriate emotional expression (Katz et al., 2012). Because the challenging task in our study provided a context where both positive and negative emotional expression would be situationally appropriate, we expected that increases over time in mothers’ emotion coaching behaviors would relate to increases over time in children’s expression of the valence of emotion that was coached. In other words, increased coaching of positive emotions would be related to increased expression of positive emotions, and likewise for negative emotions. Although maternal emotion dismissing communicates that children’s emotions should not be expressed, it also fails to support children’s development of emotion regulation and expression management skills (Gottman et al., 1997). Thus, we hypothesized that increases over time in mothers’ emotion dismissing might also relate to increases over time in children’s emotional expression, especially expression of negative emotions. Conversely, though parenting commonly elicits intense positive and negative emotions, mothers have already developed adult emotion regulation and expression management skills and effective parents are motivated to modulate their emotional expressions to match their priorities for their child in the moment (Dix, 2000). Within the context of the challenging cooperative task, we expected that the relation of children’s emotion socialization behaviors to mothers’ emotional expressions would be consistent with behavioral principles because mothers’ responsiveness to children’s positive reinforcement and positive punishment of their emotions would most efficiently return focus to their shared task. Thus, we expected increases over time in children’s emotion coaching to relate to increases over time in mother’s emotional expression and increases over time in children’s emotion dismissing to relate to decreases over time in mothers’ emotional expression, with effects specific to the valence of emotion. We summarize all hypotheses in Table 1.

Table 1.

Summary of Aims, Hypotheses, and Results

Aim 1: Development Aim
	Hypotheses	Results
Over time from early to middle childhood:	• Maternal expression of positive emotions will increase	No; decrease found
	• Maternal expression of negative emotions will decrease	No; no change found
	• Maternal coaching of positive emotions will decrease or show no change	Yes; decrease found
	• Maternal coaching of negative emotions will decrease	No; no change found
	• Maternal dismissing of positive emotions will increase	No; decrease found
	• Maternal dismissing of negative emotions will increase	No; decrease found
	• Children’s expression of positive emotions will increase	No; decrease found
	• Children’s expression of negative emotions will decrease	No; no change found
	• Children’s coaching of positive emotions will increase	Yes
	• Children’s coaching of negative emotions will increase	No; no change found
	• Children’s dismissing of positive emotions will increase	No; decrease found
	• Children’s dismissing of negative emotions will increase	n/aa
Aim 2: Within-Dyad Association Aim
	Hypotheses	Results
Over time from early to middle childhood:	• Increases in maternal coaching of positive emotions will relate to increases in children’s expression of positive emotions	Yes
	• Increases in maternal coaching of negative emotions will relate to increases in children’s expression of negative emotions	Yes
	• Increases in maternal dismissing of positive emotions will relate to increases in children’s expression of positive emotions	Yes
	• Increases in maternal dismissing of negative emotions will relate to increases in children’s expression of negative emotions	Yes
	• Increases in children’s coaching of positive emotions will relate to increases in mothers’ expression of positive emotions	Yes
	• Increases in children’s coaching of negative emotions will relate to increases in mothers’ expression of negative emotions	Yes
	• Increases in children’s dismissing of positive emotions will relate to decreases in mothers’ expression of positive emotions	No; no effect found
	• Increases in children’s dismissing of negative emotions will relate to decreases in mothers’ expression of negative emotions	n/aa

Note:

^aModel was not run because variable Children’s Dismissing of Negative Emotions was not included in the analyses due to floor effect at all three time points.

For both aims, we explored differences according to child gender due to mixed findings in previous research. Katz and Windecker-Nelson (2004) found that mothers of sons endorsed emotion coaching more than mothers of daughters. Cunningham and colleagues (2009) found no child gender differences in parents’ endorsement of an emotion coaching philosophy but did find child gender differences in associations of parents’ emotion coaching philosophy with child emotional skills and adjustment outcomes. No child gender differences were observed in parents’ coaching and dismissing of positive and negative emotions during reminiscing tasks with their child in early childhood (Hernandez et al., 2018) and middle childhood (Lozada et al., 2016). In brief, research using the meta-emotion interview has sometimes shown child gender differences in emotion coaching philosophy or its effects, whereas research using observation of emotion coaching behaviors during reminiscing tasks has not. Given the novelty of the task we used to assess emotion socialization, we explored child gender differences with no directional hypotheses specified.

Method

Participants

Participants were 349 mother-child dyads (179 boys, 170 girls) recruited in the southeastern U.S. for a study of emotion and cognition across early development (i.e., the CAP study). As part of this longitudinal study, 410 families were recruited in three cohorts when children were infants, of whom 386 children participated in research lab visits with their mother at ages 3, 4, and 6 years. Cohort 1, representing 25% of the original sample, participated at ages 3 and 4 years and not at age 6 years due to the funding schedule. Cohorts 2 and 3 participated at all three time points. An additional 48 children were newly recruited to join the ongoing longitudinal study as part of Cohort 2 for the age 6 visit, 47 of whom completed the mother-child tasks. The sample for the current study included all mother-child dyads who completed the interaction task that we coded during at least one of the three time points. The final sample size of 349 dyads included n = 142 with one time point (including 47 children newly recruited at age 6), n = 44 with two time points, and n = 163 with three time points. Most children were identified by mothers as White (79.4%), 12% were identified as Black or African American, 0.3% were identified as Asian, and 8.3% were identified as multi-racial or other; 6.3% of children were identified as Hispanic or Latino. Mothers reported their highest level of education completed and their age at child’s birth when families were initially recruited. Most mothers completed college education (41.8%), 1.4% had some high school education, 21.2% completed high school, 8.3% completed technical school or had some college education, 25.2% completed graduate school, and 2.0% did not report their educational levels. Maternal age at child’s birth ranged from 14 to 42 years (M = 29.92, SD = 5.55). For each time point, we tested whether there were socio-demographic differences (maternal age, maternal education, child gender, child race/ethnicity) in dyads who participated in the lab visit compared with those who were missing. We found three significant differences. At age 3 and age 4, mothers with higher education were more likely to be missing (χ² = 17.61, df = 4, p = .001 for age 3; χ² = 13.17, df = 4, p = .010 for age 4). At age 3, White children were more likely to be missing (χ² = 14.51, df = 3, p = .002).

Procedure

As part of the larger longitudinal study, children completed developmentally appropriate self-regulation tasks focused on cognition and emotion and mother-child interaction tasks. For the current study, we used only the etch-a-sketch task described below. This interaction task lasted about three to seven minutes and was administered in the second half of a two-hour lab session at each time point. The task was video recorded. Mothers reported demographics as part of a packet of questionnaires completed prior to the lab visit. Upon lab visit completion at each time point, mothers were paid $50. Children were given a small gift at ages 3 and 4 years and were paid $10 at age 6 years. The Institutional Review Board at Virginia Tech approved all procedures. At each time point, informed consent was obtained with all parents and informed assent was obtained with all children.

Measures

Etch-A-Sketch Task

Mother-child dyads completed the etch-a-sketch task (Atzaba-Poria et al., 2014) at ages 3, 4, and 6 years. Each dyad was instructed to be seated at the table in the lab and to work together to draw assigned shapes on an etch-a-sketch with two knobs. Mothers were asked to only control one knob and children only the other throughout the task. Assigned shapes at each time point always included a square (an easier shape) then a house and/or circle (a harder shape). On average, the task lasted 3.25 minutes at age 3, 4.00 minutes at age 4, and 6.75 minutes at age 6.

Behavioral Coding of Emotion Expression, Coaching, and Dismissing

An existing coding scheme (Dunsmore et al., 2013) used with parent-child discourse was adapted for the etch-a-sketch task. Construct validity for the coding scheme is supported by relations of parents’ coaching of their preschool-age child’s positive and negative emotions with their use of positive and negative emotion explanations, respectively. Also, parents’ greater dismissing of child positive emotion and greater coaching of child negative emotion, along with greater elaboration about events, characterized a pattern of parent emotion-related discourse that was related to children’s greater internalizing behavior (Hernandez et al., 2018). In a sample of families seeking treatment for their child’s oppositional behavior, parents’ coaching of positive emotions was related to children’s expression of positive emotions, and parents’ coaching of negative emotions was related to children’s expression of both positive and negative emotions (Dunsmore et al., 2013). For the etch-a-sketch task, in addition to adapting exemplars of emotional expression and emotion socialization, coding took place at 30-second intervals rather than overall codes for the whole interaction. Training emphasized recognizing developmentally typical behavior at each time point (e.g., mothers extending children’s words into full phrases at 3 years to foster language development) and accounting for families’ baseline styles that may reflect culture-relevant values or norms.

Expression of emotions was scored on a 3-point scale: 0 = no emotion, 1 = use of one modality (e.g., verbal words, nonverbal gestures, facial expressions) or mild emotional intensity of expression, and 2 = use of more than one modality or strong emotional intensity (verbal or nonverbal) of expression. Expressions of positive and negative emotions were coded separately, yielding two scores for expression at each interval for both mother and child.

The coding scheme for emotion socialization behaviors was designed to capture the extent to which emotion coaching and emotion dismissing were demonstrated during the 30-second interval. Emotion coaching was scored on a 4-point scale: 0 = no coaching, 1 = acknowledgement of the other’s perspective or behavior with respect to the task, which showed an open stance toward awareness of the other’s experience, 2 = acceptance of the other’s emotion, and 3 = full-blown coaching, which might include actively showing empathy and validation, labeling emotions, and teaching emotion management. Emotion dismissing was scored on a 5-point scale: 0 = no dismissing, 1 = discouragement of the other’s perspective or contribution with respect to the task, which showed a closed stance toward awareness of the other’s experience, 2 = dismissal of the other’s emotion, which conveyed that the emotion was wrong or unimportant, 3 = overriding of the other’s emotion, which not only dismissed the emotion but also attempted to remove it or replace it with another emotion, and 4 = display of contempt, which dismissed not only the emotion but the person. As with expression, coaching and dismissing of positive and negative emotions were coded separately, so the mother and child both received two scores for each emotion socialization behavior at each interval.

Across time points, 37.9% of video recordings were coded for reliability by independent coders. Group consensus meetings were held at least bi-weekly to resolve discrepancies and maintain reliability. Observer drift was checked monthly. Interrater reliability was indexed by intraclass correlation coefficients (ICCs; form: two-way mixed, consistency). Interrater reliability was in the moderate or good range for each coded variable (Koo & Li, 2016): expression of positive emotions (average ICC_weighted =.761 for parents, .759 for children), expression of negative emotions (average ICC_weighted = .680 for parents, .721 for children), coaching of positive emotions (average ICC_weighted = .798 for parents, .708 for children), coaching of negative emotions (average ICC_weighted = .590 for parents, .637 for children), dismissing of positive emotions (average ICC_weighted = .706 for parents, .768 for children), and dismissing of negative emotions (average ICC_weighted = .562 for parents, .550 for children). Cicchetti (1994) describes ICCs ranging from 0.40 to 0.59 as having fair clinical significance; from 0.60 to 0.74 as having good clinical significance; and from 0.75 and 1.00 as having excellent clinical significance. For the three codes with average ICC_weighted values lower than 0.60, less than 3% of coded intervals had non-zero scores (mothers’ coaching of children’s negative emotions: 2.1% of coded intervals had non-zero scores; mothers’ dismissing of children’s negative emotions: 1.8% of coded intervals had non-zero scores; children’s dismissing of mothers’ negative emotions: 0.2% of coded intervals had non-zero scores). Because variability in participants’ scores affects the extent to which coder discrepancies reduce ICCs (Lee et al., 2012), we considered interrater reliability to be acceptable.

At each time point, we averaged codes across intervals to form single scores for mothers’ and children’s expression, coaching, and dismissing of positive and negative emotions that accounted for number of coded intervals per dyad. Thus, at each time point, mothers and children each received six scores for the entire etch-a-sketch task: expression of positive emotions (possible range 0–2; Expression Positive), expression of negative emotions (possible range 0–2; Expression Negative), coaching of positive emotions (possible range 0–3; Coaching Positive), coaching of negative emotions (possible range 0–3; Coaching Negative), dismissing of positive emotions (possible range 0–4; Dismissing Positive), and dismissing of negative emotions (possible range 0–4; Dismissing Negative).

Data Analytic Plan

Study variables were the single scores mother-child dyads received at each time point for emotion coaching, emotion dismissing, and emotional expression, as well as mother-reported child gender (0 = boy, 1 = girl). To assess change over time, the three time points were coded with years as the unit of measurement (age 3 as Time 1 = 0, age 4 as Time 2 = 1, age 6 as Time 3 = 3). We used SPSS 27 to conduct preliminary analyses to examine descriptive statistics, child gender differences in other study variables, relations of study variables with families’ socio-demographic characteristics to inform potential covariates, and bivariate correlations. We conducted hierarchical linear modeling (HLM) analyses to examine our two aims. HLM 7.28 was used to run HLM analyses with full maximum likelihood method of estimation. We specified the existence of missing data at level-1 and adopted listwise deletion of missing cases (or in our study, missing time points nested within dyads) when running the analysis option to handle missing data as illustrated by Raudenbush et al. (2011). There were no missing data at level-2. HLM is appropriate for testing hypotheses because it examines change of one variable or associations over time at both within- and between-individual levels (Raudenbush & Bryk, 2002). HLM allows for unbalanced designs and accounts for missing data in longitudinal datasets that vary in terms of participants per time point so dyads with missing data at one or two time points are still included in the analyses (Singer & Willet, 2003).

We tested Aim 1 (Development Aim) by running unconditional linear growth HLM models where coded time was the predictor at level-1. Mothers’ and children’s expression, coaching, and dismissing of positive and negative emotions were separately included as outcomes.

Aim 2 (Within-Dyad Association Aim) addressed whether within-dyad changes in mothers’ (children’s) coaching or dismissing were associated with within-dyad changes in children’s (mothers’) expression, while accounting for between-dyad relations of the predictor and outcome variables. Thus, we conducted HLM models with mothers’ time-varying, within-individual mean scores for coaching and dismissing of positive and negative emotions (each tested separately) and coded time at level-1 predicting children’s expression (of the same-valence emotion) as the outcome. Mothers’ overall across-time mean (time-invariant and between-individual) score for coaching or dismissing was controlled at level-2 initial status and level-2 time slope. Because hypotheses are about within-dyad relations, the effect of interest was mothers’ time-varying score for coaching or dismissing. Then, parallel HLM models were conducted for children’s coaching and dismissing of positive and negative emotions (each tested separately) relating to mothers’ expression (of the same-valence emotion) as the outcome. Again, the effect of interest was children’s time-varying score for coaching or dismissing. Below is one tested model exemplar where t indicates the time interval and i indicates an individual mother-child dyad.

Level-1 model:

$$ChildExpressionPositiv{e}_{ti}={\pi }_{0i}+{\pi }_{1i}\ast \left(Time\right)+{\pi }_{2i}\ast \left(MaternalCoachingPositiv{e}_{time-varying}\right)+{\epsilon }_{ti}$$

Level-2 model:

$${\pi }_{0i}={\beta }_{00}+{\beta }_{01}\ast \left(MaternalCoachingPositiv{e}_{time-invariant}\right)+{\gamma }_{0i}$$

$${\pi }_{1i}={\beta }_{10}+{\beta }_{11}\ast \left(MaternalCoachingPositiv{e}_{time-invariant}\right)+{\gamma }_{1i}$$

$${\pi }_{2i}={\beta }_{20}+{\gamma }_{2i}$$

After running each model under both aims, we added child gender at level-2 to explore whether any of the tested effects varied according to child gender over time. The Bonferroni adjustment was used to control for inflation in Type I error caused by multiple hypothesis tests in each aim (Chen et al., 2017).

Results

Preliminary Analyses

Descriptive Statistics and Data Transformations

Little’s MCAR test result indicated that the pattern of missing dyads at all three time points in our current study did not significantly deviate from a random pattern (χ² = 14.07, df = 9, p = .120). We examined descriptive statistics of all 12 coded variables at each time point. Three variables (Child Expression Positive, Mother Expression Positive, Mother Coaching Positive) showed normal distributions at each time point and thus were kept in their initial values. For Child Dismissing Negative, a floor effect was found, as at each time point more than 97% of dyads had scores of 0. We excluded this variable from subsequent analyses. The other eight variables at each time point were winsorized (1–12 data points per variable) to reduce outliers. Thus, at each time point 11 continuous variables were included in analyses (Child Expression Positive, Child Expression Negative, Mother Expression Positive, Mother Expression Negative, Mother Coaching Positive, Mother Coaching Negative, Mother Dismissing Positive, Mother Dismissing Negative, Child Coaching Positive, Child Coaching Negative, and Child Dismissing Positive).

Child Gender Differences, Potential Covariates, and Bivariate Correlations

We ran t-tests to test child gender differences of all variables. Four significant differences were found. Mothers of daughters coached their child’s negative emotions more at age 3 and dismissed their child’s negative emotions more at age 4 compared with mothers of sons (Mother Coaching Negative, age 3, t(196.86) = −2.03, p = .044; Mother Dismissing Negative, age 4, t(213.14) = −2.52, p = .013). At age 6, girls expressed negative emotions more than boys (t(204.71) = −3.94, p = .000). At age 4, boys dismissed their mothers’ positive emotions more than girls (t(246.68) = 2.19, p = .030). To test for potential socio-demographic covariates, we conducted one-way ANOVA tests for differences in study variables according to child race/ethnicity and maternal educational level, and we ran bivariate correlations between maternal age and study variables. For maternal education, three significant differences were found. Mothers with higher education were higher in Mother Coaching Positive at age 3 (F (4, 211) = 4.78, p = .001) and age 6 (F (4, 225) = 5.49, p = .000), and their children were higher in Child Coaching Negative at age 3 (F (4, 211) = 2.48, p = .045). For child race/ethnicity, White children were lower in Child Dismissing Positive at age 4 (F (2, 258) = 4.25, p = .015) and higher in Child Expression Negative at age 6 (F (3, 233) = 3.17, p = .025), and their mothers were higher in Mother Coaching Positive at age 6 (F (3, 233) = 10.46, p = .000). Mothers who were older at the child’s birth were higher in Maternal Expression Negative at age 6 (r = .16, p = .016) and their children were lower in Child Dismissing Positive at age 3 (r = −.16, p = .020). Given the above socio-demographic differences, we controlled for child race/ethnicity, maternal age, and maternal educational level in subsequent analyses as covariates.

Table 2 shows descriptive statistics and bivariate correlations among study variables at all time points. Each coded variable showed significant positive correlations of earlier time points with later time points. At all three time points, mothers’ coaching was related to children’s expression of the same-valence emotion, and children’s coaching was related to mothers’ expression of the same-valence emotion. At ages 4 and 6, mothers’ dismissing of positive emotions was positively related to children’s positive emotion expression. At all three time points, mothers’ dismissing of negative emotions was positively related to children’s negative emotion expression. At age 3 only, children’s dismissing of positive emotions was negatively related to mothers’ positive emotion expression.

Table 2.

Means, Standard Deviations (SDs), and Correlations Among Study Variables at Three Time Points

		1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23	24	25	26	27	28	29	30	31	32	33
1.	C expression pos 3yr	-
2.	C expression pos 4yr	.09	–
3.	C expression pos 6yr	.20**	.26**	–
4.	C expression neg 3yr	−.11	−.14	.02	–
5.	C expression neg 4yr	.10	.01	.11	.12	–
6.	C expression neg 6yr	−.11	−.13	.06	.26**	.02	–
7.	M expression pos 3yr	.17*	.07	.07	−.11	.04	−.09	–
8.	M expression pos 4yr	.07	−.03	−.04	−.02	−.02	.11	.19**	–
9.	M expression pos 6yr	.00	.08	.13*	−.03	.04	−.09	.24**	.13	–
10.	M expression neg 3yr	−.12	.01	.02	.03	−.09	.18*	.04	.02	.14	–
11.	M expression neg 4yr	−.07	−.01	.11	−.12	−.08	.16*	−.04	.07	.09	.13	–
12.	M expression neg 6yr	.03	.00	.05	.01	.03	.09	.08	.11	−.02	.17*	.07	–
13.	M coaching pos 3yr	.31**	.03	.06	−.04	−.07	−.08	.03	.01	.01	.05	.01	.11	–
14.	M coaching pos 4yr	.03	.27**	.12	.01	−.08	.17*	.04	−.01	.13	.11	.09	.13	.09	–
15.	M coaching pos 6yr	.11	−.08	.21**	.02	−.08	.11	.03	.04	.07	.17*	.07	.27**	.29**	.25**	–
16.	M coaching neg 3yr	−.03	.03	.05	.44**	−.02	.15	−.12	−.05	.02	−.01	.05	.01	−.04	.08	−.03	–
17.	M coaching neg 4yr	.03	.02	.10	.15*	.35**	.13	.01	.11	.14	−.02	.08	−.07	.02	.05	.10	.155*	–
18.	M coaching neg 6yr	.07	−.10	−.05	.14	−.05	.44**	.00	−.02	−.05	.15	−.04	.08	.02	.03	.14*	.04	.08	–
19.	M dismissing pos 3yr	.13	−.03	.07	−.04	−.03	−.07	−.06	.04	.05	.03	.13	.07	−.07	−.08	.08	−.01	.01	−.05	–
20.	M dismissing pos 4yr	−.04	.15*	−.02	.04	.00	−.13	.05	−.19**	−.12	.03	−.11	−.02	−.05	−.21**	−.11	−.06	−.07	−.15	.15*	–
21.	M dismissing pos 6yr	−.04	−.02	.23**	−.01	.01	.04	−.01	.05	−.05	.07	.11	.10	−.08	−.04	−.12	−.02	−.13	−.13*	.23**	.24**	–
22.	M dismissing neg 3yr	−.04	−.07	−.17*	.54**	.05	.10	−.05	.03	−.07	.04	−.11	−.01	−.03	.04	.00	.12	−.08	.07	−.03	−.01	−.03	–
23.	M dismissing neg 4yr	.12	−.01	.04	.08	.62**	−.02	.00	−.09	−.04	−.02	−.12	.11	.00	−.02	−.07	.05	.06	−.06	−.06	.06	.11	.12	–
24.	M dismissing neg 6yr	.01	−.04	.01	.27**	.01	.41**	.04	.14	−.05	.23**	−.05	.02	.06	.01	.01	.28**	.06	.08	.01	.03	.09	.17*	.04	–
25.	C coaching pos 3yr	.22**	.08	.04	−.07	−.05	.02	.21**	.08	−.15*	.12	−.07	.06	.04	.08	.12	.04	−.10	−.07	−.03	−.06	−.03	−.03	.04	.18*	–
26.	C coaching pos 4yr	.19**	.14*	.05	.01	.07	.05	.05	.49**	−.01	−.05	.10	.06	−.02	.18**	.09	.07	.18**	−.05	.07	−.11	.01	.12	−.02	.11	.07	–
27.	C coaching pos 6yr	.11	.20**	.27**	.00	.04	−.10	.14	−.06	.55**	.18*	.03	.07	.12	.15	.19**	.09	.08	−.06	.08	−.10	−.04	−.03	−.01	−.01	.08	.02	–
28.	C coaching neg 3yr	−.07	−.02	.00	−.04	−.09	.06	.17*	.01	−.08	.35**	.01	.09	.08	.14	0.1	−.05	−.04	−.01	−.06	−.07	.07	.02	−.06	.18*	.21**	−.01	−.04	–
29.	C coaching neg 4yr	.10	.18**	.20**	−.06	.06	.06	−.02	.04	.05	.10	.45**	.04	−.01	.10	.09	.04	.07	−.10	.13	−.07	.04	−.10	−.01	.04	.17*	.11	.13	.01	–
30.	C coaching neg 6yr	.10	.17*	.25**	−.13	−.02	.10	.19*	.10	.04	.01	.02	.52**	.00	.12	.16*	−.07	−.04	.06	.05	−.08	.11	−.08	−.03	.02	.18*	.06	.14*	.19*	.06	–
31.	C dismissing pos 3yr	−.18**	−.02	.21**	.07	.03	.25**	−.18**	−.03	.04	.05	−.08	.05	−.28**	−.13	.06	.09	−.11	.08	−.08	.00	.15	−.03	.02	.04	−.08	.02	−.05	−.09	−.10	.00	–
32.	C dismissing pos 4yr	.02	.01	.02	−.03	.09	.14	−.01	−.08	−.07	−.07	−.01	−.03	−.04	−.13*	−.01	.01	−.07	.07	−.02	.11	.05	−.05	.12*	.00	−.16*	−.04	−.06	−.11	−.01	−.09	.17*	–
33.	C dismissing pos 6yr	.08	−.01	.03	.05	−.01	.18**	.03	.02	.02	.02	.02	.08	−.04	.08	.05	.02	−.08	.04	.11	−.04	.17**	.12	−.02	.25**	.08	.18*	.03	.01	.13	.07	.09	.26**	–
	Mean	.38	.24	.19	.09	.09	.11	.25	.19	.14	.10	.12	.08	.45	.42	.39	.04	.02	.05	.37	.50	.23	.03	.05	.02	.09	.17	.15	.01	.05	.03	.21	.19	.15
	SD	.28	.22	.17	.17	.14	.12	.23	.17	.14	.16	.13	.10	.27	.25	.25	.11	.07	.09	.31	.39	.19	.12	.12	.05	.16	.22	.16	.07	.11	.06	.24	.22	.16

Note: C = Child, M = Mother, pos = Positive Emotions, neg = Negative Emotions. 3yr = 3-year-old, 4yr = 4-year-old, 6yr = 6-year-old.

^*p < .05;

^**p < .01.

Aim 1: Development of Emotional Expression, Coaching, and Dismissing Over Time

To test Aim 1, we ran a set of unconditional HLM models for mothers and children separately where coded time was included as a within-individual variable at level-1 and covariates were controlled for at the initial status at level-2. We then added child gender as a between-individual variable at level-2 to explore whether each tested change differed for boys versus girls. Model results are presented in Table 3. The intraclass correlation coefficient (ICC) for all models indicated small to large effects of nesting time points within participants (ICC range = .010–.153; LeBreton & Senter, 2008).

Table 3.

HLM Results of Within- and Between-Individual Variation in Expression, Coaching, and Dismissing of Emotion (Development Aim)

Mother Models
	Expression positive	Expression negative	Coaching positive	Dismissing positive	Coaching negative	Dismissing negative
Fixed Effect
Intercept, π0ί
β00 (SE)	.235 (.084)	.002 (.054)	.271 (.124)	.533* (.163)	−.009 (.039)	.048 (.039)
Time slope, π1i
β10 (SE)	−.032* (.005)	−.007 (.004)	−.027* (.007)	−.064* (.007)	.006 (.003)	−.007* (.002)
Variance Component
Level-1, e (SE)	.023* (.002)	.013* (.001)	.050* (.005)	.082* (.008)	.006* (.001)	.007* (.001)
Intercept, r0 (SE)	.018* (.004)	.008* (.002)	.012 (.006)	.037* (.011)	.004* (.001)	.006* (.001)
Time slope, r1 (SE)	.001 (.001)	.001 (.000)	.000 (.002)	.002 (.003)	.001* (.000)	.001 (.000)
ICC (ρ)	0.055	0.048	0.013	0.021	0.153	0.066
Model Deviance (−2LL)	−460.731	−895.637	15.469	372.373	−1370.870	−1259.502
Child Models
	Expression positive	Expression negative	Coaching positive	Dismissing positive	Coaching negative
Fixed Effect
Intercept, π0i
β00 (SE)	.404* (.104)	.080 (.062)	.033 (.064)	.155 (.119)	−.001 (.033)
Time slope, π1i
β10 (SE)	−.052* (.006)	.006 (.004)	.014* (.005)	−.018* (.006)	.003 (.002)
Variance Component
Level-1, e (SE)	.039* (.004)	.016* (.002)	.033* (.003)	.034* (.003)	.007* (.001)
Intercept, r0 (SE)	.026* (.006)	.008* (.002)	.001 (.004)	.017* (.005)	.000 (.001)
Time slope, r1 (SE)	.002 (.001)	.001 (.001)	.000 (.001)	.001 (.001)	.000 (.000)
ICC (ρ)	0.043	0.035	0.011	0.035	0.010
Model Deviance (−2LL)	−114.394	−790.570	−385.655	−235.835	−1437.619

Note: ICC = Intraclass correlation coefficient.

^*Coefficients were significant at α = .05 level upon Bonferroni adjustment (p_adjusted = .00417 for mother models; p_adjusted = .005 for child models). All models controlled for child race/ethnicity, maternal age, and maternal educational level as covariates. No models showed child gender effects at the adjusted significance levels.

Maternal Expression, Coaching, and Dismissing

The Bonferroni-adjusted significant p value was .00417. Mothers showed no individual variation in expressing positive or negative emotions at age 3 years. Sample-wide, mothers decreased in expressing positive emotions from age 3 to 6 years (β₁₀(SE) = −.032(.005), p = .000) and showed no significant change in expressing negative emotions (p = .059).

Regarding emotion coaching and dismissing, only mothers’ dismissing of positive emotions showed individual variation at age 3 years. Mothers’ coaching and dismissing of positive emotions both showed sample-wide decreases over time (coaching: β₁₀(SE) = −.027(.007), p = .000); dismissing: β₁₀(SE) = −.064(.007), p = .000). Mothers’ coaching of negative emotions showed no change over time (p = .057), whereas mothers’ dismissing of negative emotions decreased over time (β₁₀(SE) = −.007(.002), p = .004). Across all six maternal models, adding child gender at level-2 showed no child gender effects at the adjusted significant level (ps ≥ .033).

Children’s Expression, Coaching, and Dismissing

The Bonferroni-adjusted significant p value was .005. Children showed individual variation in expressing positive emotions at age 3 years. Children also showed sample-wide decreases in expressing positive emotions (β₁₀(SE) = −.052(.006), p = .000) and no significant change in expressing negative emotions over time (p = .144).

Children’s coaching and dismissing of positive emotions both showed no individual variation at age 3 years. Children’s coaching of positive emotions increased from age 3 to 6 years (β₁₀(SE) = .014(.005), p = .0047), whereas their dismissing of positive emotions decreased over time (β₁₀(SE) = −.018(.006), p = .002). Children’s coaching of negative emotions showed no individual variation at age 3 years and no significant sample-wide change over time (p = .134). Children’s dismissing of negative emotions was not included in analyses due to floor effects. Across all five child models, there were no child gender effects at the adjusted significance level when adding child gender at leval-2 (ps ≥ .010).

Aim 2: Within-Dyad Associations of Emotion Coaching and Dismissing with Expression

We ran a second set of HLM models to test Aim 2. In each model, time and mothers’ (or children’s) time-varying, within-individual coaching/dismissing variables were entered at level-1 to predict children’s (or mothers’) expressions of the same valence emotions that were coached/dismissed. We also controlled the effect of the between-individual coaching/dismissing variable on the initial status and change over time at level-2 in addition to controlling for covariates. We then added child gender as a between-individual variable at level-2 to each model testing whether the association varied across child gender. Model results are presented below in Table 4. Because hypotheses regard within-dyad associations, only results regarding relations of mothers’ time-varying coaching/dismissing to children’s expressions (and vice versa) are described in text below.

Table 4.

HLM Results of Within- and Between-Individual Variation in Dyadic Relations of Coaching and Dismissing to Expression (Within-Dyad Association Aim)

Maternal Emotional Responses → Children’s Emotional Expressions Models
	Coach pos → Express pos	Dismiss pos → Express pos	Coach neg → Express neg	Dismiss neg→ Express neg
Fixed Effect
Intercept, π0i
β00 (SE)	.222 (.130)	.339 (.138)	.118 (.088)	.112 (.077)
Mean ERV, β01 (SE)	.339* (.074)	−.004 (.054)	.625* (.103)	.796* (.116)
Time slope, π1i
β10 (SE)	.032 (.065)	−.024 (.071)	−.014 (.044)	−.035 (.038)
Mean ERV, β11 (SE)	−.061 (.034)	.031 (.029)	.002 (.050)	−.001 (.064)
Time-varying ERV, π2ί
β20 (SE)	.231* (.049)	.120* (.040)	.605* (.078)	.785* (.077)
Variance Component
Level-1, e (SE)	.036* (.004)	.035* (.004)	.013* (.001)	.010* (.001)
Intercept, r0 (SE)	.022* (.006)	.028* (.006)	.007* (.002)	.005* (.001)
Time slope, r1 (SE)	.001 (.001)	.002 (.001)	.001 (.000)	.001 (.000)
Time-varying ERV slope, r2 (SE)	.029* (.038)	.033* (.022)	.066 (.064)	.110* (.059)
Model Deviance (−2LL)	−169.199	−129.424	−925.651	−1058.576
Children’s Emotional Responses → Maternal Emotional Expressions Models
	Coach pos → Express pos	Dismiss pos → Express pos	Coach neg → Express neg
Fixed Effect
Intercept, π0i
β00 (SE)	.335* (.124)	.318* (.115)	−.011 (.097)
Mean ERV, β01 (SE)	.386* (.075)	−.131 (.068)	.641* (.140)
Time slope, π1i
β10 (SE)	−.107 (.046)	−.070 (.047)	−.004 (.038)
Mean ERV, β11 (SE)	−.014 (.034)	.043 (.031)	.009 (.067)
Time-varying ERV, π2i
β20 (SE)	.380* (.045)	−.060 (.043)	.587* (.074)
Variance Component
Level-1, e (SE)	.017* (.002)	.022* (.002)	.010* (.001)
Intercept, r0 (SE)	.019* (.003)	.019* (.004)	.007* (.002)
Time slope, r1 (SE)	.002* (.001)	.001 (.001)	.001 (.000)
Time-varying ERV slope, r2 (SE)	.046* (.031)	.024 (.025)	.085* (.055)
Model Deviance (−2LL)	−588.283	−477.912	−1041.961

Note: ERV=emotional response variable (coaching or dismissing – identified in model column header); “Mean ERV” = average score across time points, or time-invariant score; “Time-varying ERV” = score for the ERV at each time point, mean-centered around the time-invariant score; pos=positive; neg=negative.

^*Coefficients were significant at α = .05 level upon Bonferroni adjustment (p_adjusted = .00625 for mother→child models; p_adjusted = .00833 for child→mother models). All models controlled for child race/ethnicity, maternal age, and maternal educational level as covariates. No models showed child gender effects at the adjusted significance levels.

Mothers’ Coaching and Dismissing Relating to Children’s Emotional Expression

The Bonferroni-adjusted significant p value was .00625. All four models predicting children’s expression of positive and negative emotions showed significant within-dyadic effects of maternal coaching and dismissing, with significant positive slopes for time-varying maternal coaching and dismissing of children’s positive and negative emotions. The more mothers coached positive emotions, the more children expressed positive emotions (β₂₀(SE) = .231(.049), p = .000). Similarly, the more mothers dismissed positive emotions, the more children expressed positive emotions (β₂₀(SE) = .120(.040), p = .003). Likewise, the more mothers coached negative emotions, the more children expressed negative emotions (β₂₀(SE) = .605(.078), p = .000), and the more mothers dismissed negative emotions, the more children expressed negative emotions (β₂₀(SE) = .785(.077), p = .000). Across all four models, adding child gender at level-2 did not yield significant gender effects (ps ≥ .049).

Children’s Coaching and Dismissing Relating to Mothers’ Emotional Expression

The Bonferroni-adjusted significant p value was .00833. Two out of three models predicting mothers’ expression of positive and negative emotions showed significant within-dyadic effects of children’s coaching and dismissing. Significant positive slopes were found for time-varying child coaching of mothers’ positive emotions and time-varying child coaching of mothers’ negative emotions. The more children coached positive emotions, the more their mothers expressed positive emotions (β₂₀(SE) = .380(.045), p = .000). Similarly, the more children coached negative emotions, the more their mothers expressed negative emotions (β₂₀(SE) = .587(.074), p = .000). The slope for time-varying child dismissing of mothers’ positive emotions was not significant (p = .167). Across all three models, when child gender was added at level-2 there were no significant gender effects (ps ≥ .326).

Discussion

A robust body of research has demonstrated concurrent and longitudinal associations between parental emotion coaching or dismissing and children’s psychosocial adjustment (see Katz et al., 2012 for a review). Although change in parental emotion coaching as children develop had been theorized (Gottman et al., 1996; Katz et al., 2012), little research had examined longitudinal change in parental emotion coaching. In the current study we examined trajectories and within-dyad associations for emotion socialization (coaching and dismissing) and emotional expression in mother-child dyads from early childhood to middle childhood. We contribute novel scientific knowledge by observing emotion socialization behaviors during an emotion-eliciting cooperative task, examining socialization and expression of both positive and negative emotions, and investigating children’s emotion socialization behaviors in relation to their mothers’ expressions as well as mothers’ emotion socialization behaviors in relation to their children’s expressions. Results partially support our hypotheses (see Table 1 for an overview). Overall, results demonstrate children’s active engagement in the emotion socialization process from age 3 to 6 years, increases in children’s and decreases in mothers’ coaching of positive emotions over time, and longitudinal change in both mothers’ and children’s emotional expression in relation to the other’s emotion socialization.

Development Aim: Change in Emotional Expression and Emotion Socialization Over Time

We had expected that expression of positive emotions would increase and expression of negative emotions would decrease from age 3 years to age 6 years, based on previous research on emotional expression during frustrating tasks during early childhood (Chaplin et al., 2017). Instead, we found that expression of positive emotions declined for both mothers and children whereas expression of negative emotions showed no change over time for both mothers and children. In retrospect, the frustrating parent-child task used by Chaplin and colleagues (2017; i.e., waiting for several minutes before the child could open an appealing gift) may have had situational demands that promoted positive social engagement to distract from the gift. Successfully carrying out instructions for the etch-a-sketch task used in the current study, however, involved active coordination between mothers and children to jointly turn knobs to create a desired picture. The etch-a-sketch task, therefore, may have had situational demands that diminished positive social engagement between mothers and children as they focused on completing the picture. This may have been especially likely as children became increasingly competent in engaging in the task from 3 years to 6 years of age and completion of the picture became more co-directed rather than solely mother-directed. As children develop more autonomy, mothers’ positive emotion expression may decrease due to their perceptions of children’s self-sufficiency in emotion regulation (Klimes-Dougan & Zeman, 2007).

The absence of sample-wide change in children’s or mothers’ expression of negative emotion from 3 to 6 years may reflect appropriate calibration of the etch-a-sketch task to be challenging for mother-child dyads at each time point, as instructions varied to include a square, house with triangle roof, and circle. Furthermore, though children’s understanding of emotional display rules increases from early childhood to middle childhood (Saarni et al., 2006), perceived need to mask emotional expressions is lower with parents than with peers (Zeman et al., 2006). Thus, situational demands for children’s and mothers’ expression of negative emotions during a challenging cooperative task may have remained relatively stable across time points. In families’ daily lives, many activities require coordination of parents’ and children’s goals, such as family meals, homework completion, morning and evening routines. Our results suggest that when tasks are challenging and require cooperation to create a product, both mothers’ and children’s expression of positive emotion may decrease from early childhood to middle childhood as less maternal expression of positive emotion and elicitation of the child’s positive emotion is needed to maintain joint prioritization of completing the product (Dix, 2000).

Based on prior research (Stettler & Katz, 2014), we hypothesized that maternal coaching of negative emotions would decrease whereas maternal dismissing of negative emotions would increase from early to middle childhood. Inconsistent with expectations, coaching of negative emotions stayed stable and dismissing of negative emotions decreased over time. Methodology may account for these differences in findings. Whereas Stettler and Katz (2014) used the meta-emotion interview to assess parents’ coaching and dismissing philosophies, we relied on behavioral observation during a task that focused mothers’ and children’s attention on cooperating to create a picture rather than on emotions. Parents’ emotion socialization behaviors may not always match their beliefs. Because the task was challenging, elicitation of children’s negative emotions may have been expected by mothers, therefore mothers may have stayed constant across time points with supportive socialization behaviors towards their child’s negative emotions. As children’s own emotion regulation skills develop from early to middle childhood (Saarni, 2008), mothers may have increasingly perceived it as unnecessary to actively dismiss children’s negative emotions in order to maintain engagement with the task.

Little extant research had examined longitudinal change in maternal socialization of positive emotions, so we tentatively hypothesized decreases or no change for maternal coaching of positive emotions (based on Wu et al., 2019) and increases for maternal dismissing of positive emotions, parallel to dismissing of negative emotions. Findings for mothers’ coaching and dismissing of positive emotions were partially consistent with our tentative expectations. Both coaching and dismissing of positive emotions decreased from early to middle childhood. As discussed above regarding expression of positive emotions, this pattern may reflect mothers’ perception of children’s increasing autonomy and skill at handling their own emotions (Klimes-Dougan & Zeman, 2007; Saarni, 2008), resulting in less active engagement with children’s positive emotions – both coaching and dismissing – to further foster that autonomy.

No extant research had examined children’s coaching and dismissing of parents’ emotions, so hypothesized increases in children’s emotion coaching and dismissing over time were based on general gains in children’s emotional communication skills from early to middle childhood (Saarni et al., 2006). Children’s coaching of mothers’ positive emotions increased and dismissing of mothers’ positive emotions decreased over time. There was no change in children’s coaching of mothers’ negative emotions over time. Children’s dismissing of mothers’ negative emotions was too rare to include in analyses. Children’s longitudinal growth in coaching mothers’ positive emotions and decline in dismissing mothers’ positive emotions may reflect their increasing engagement in and initiation of more positive, mutual interactions with their mothers as they develop (Sperling & Repetti, 2018). Taken together, these trajectories suggest children’s active role in the emotion socialization process, especially for positive emotions. In general, findings for both mothers and children suggest different sample-wide trajectories for coaching positive versus negative emotions. Perhaps, as children develop from early to middle childhood, mothers can take a more passive role in responding to children’s positive emotions as children increasingly share initiation of mutual positive affect. For negative emotions, which may still be challenging for children to handle on their own, mothers’ active coaching and reduced dismissing may build on and support children’s own developing regulation skills.

Within-Dyad Association Aim: Within-Dyad Relations of Emotion Socialization to Expression

Hypotheses for relations of mothers’ emotion coaching and dismissing with children’s expression of emotion were based on meta-emotion theory (Gottman et al., 1997). We expected that mothers’ increased coaching and dismissing of positive emotions over time would be related to children’s increased expression of positive emotions, and likewise for negative emotions. Analyses controlled between-individual effects for precision in addressing within-dyad relations of maternal coaching and dismissing to child expression. Results were consistent with hypotheses. Within dyads, mothers’ time-varying coaching of emotions, whether positive or negative, was positively related to children’s expression of same-valence emotion over time. The same pattern held for maternal dismissing of positive and negative emotions.

From a behavioral perspective, mothers’ attention to children’s emotions – whether coaching or dismissing – may reinforce children’s emotional expression. Sperling and Repetti (2018) observed emotion socialization during naturalistic parent-child interactions and found that parent ignoring of children’s emotions increased child positive or neutral emotional reactions. They suggest that ignoring may sometimes give children opportunities to practice and develop socio-emotional skills. Active ignoring of emotions is a form of emotion dismissing. Using a person-centered approach, Miller and colleagues (2015) found that a high involvement pattern of parental emotion socialization, characterized by high endorsement of both emotion coaching and emotion dismissing as well as positive and negative expressivity, was related to children’s greater growth in effortful control in early childhood. They propose that the high involvement pattern reflects parents’ active engagement in emotion socialization, which may sensitively apply multiple strategies to meet children’s needs in the moment – coaching when children need assistance to identify and manage emotions, dismissing when children need assistance focusing attention elsewhere.

According to meta-emotion theory, there may be different underlying mechanisms through which maternal coaching and dismissing relate to children’s increased emotion expression over time. Parental emotion coaching was longitudinally related to children’s emotion regulation, suggesting that children’s increased emotion expression following maternal emotion coaching may reflect children’s perception that mothers invite their emotion expression as part of maintaining a close relationship (Gottman et al., 1996). When compared with families in a wait-list control condition, parents participating in an emotion coaching intervention showed decreases in emotion dismissing as well as increases in emotion coaching, and their preschool-age children showed increases in emotion knowledge (Havighurst et al., 2010). If emotion dismissing hinders children’s development of emotion knowledge and emotion regulation, children’s greater emotion expression may reflect attempts to elicit mothers’ help, lack of understanding of display rules, or difficulty managing emotional experience and expression. Future research including broader measurement of mothers’ emotion socialization beliefs and skills and children’s emotion-related behaviors, competencies, and outcomes will be needed to distinguish between these varied interpretations of results.

For children’s emotion coaching and dismissing, we hypothesized that increases over time in coaching would relate to increases over time in mother’s emotional expression and increases over time in dismissing would relate to decreases over time in mothers’ emotional expression, again with effects specific to the valence of emotion. Results partially supported hypotheses. Within dyads, children’s time-varying coaching of emotion, whether positive or negative, was positively related to mothers’ expression of valence-matching emotion over time. Within-dyad patterns of mothers’ expressions following children’s coaching responses align with social expectations for mothers. Mothers, who are often children’s main caregivers, are expected to be emotionally responsive to children’s needs and reactions in order to be a good mother (Elliott et al., 2015). This may induce them to express more emotions when children coach their emotions, particularly during a task that requires cooperative action. Children’s time-varying dismissing of mothers’ positive emotions was not associated with mothers’ display of positive emotions over time. Again, we were unable to test the longitudinal association of children’s dismissing of negative emotions with mothers’ negative emotion expression due to rare occurrence of children’s dismissing mothers’ negative emotions. Though we are cautious about interpreting a null finding, we note that mothers as adults hold power in parent-child relationships, and their emotions and expressions may be influenced by their culturally embedded socialization goals and strategies as well as their child’s dismissive behavior (Dix, 2000; Raval & Walker, 2019).

Child Gender Differences

Results showed relatively few child gender effects overall and no gender effects in the HLM models. Maternal coaching of negative emotions was higher for daughters than for sons at age 3 years. This contrasts with previous research using the meta-emotion interview that shows greater endorsement of coaching of negative emotions by mothers of sons compared with mothers of daughters (Katz & Windecker-Nelson, 2004). However, this finding is consistent with the reminiscing literature, which shows greater frequency and variety of references to emotions in parents’ conversations about past events with daughters compared with sons in early childhood (Fivush & Zaman, 2015). Girls also dismissed mothers’ positive emotions less than boys did at age 4 years. This may reflect greater socialization emphasis on social bonding and connection for daughters compared with sons (Fivush & Zaman, 2015). Overall, findings suggest that child gender was not a systematic factor in children’s and mothers’ emotion socialization longitudinally, nor in their within-dyad associations with each other’s expression over time.

Strengths, Limitations, and Future Directions

We emphasize key strengths of the current study. First, our sample size of 349 mother-child dyads afforded statistical power for testing our hypotheses with analyses that distinguished sample-wide trajectories, between-individual time-invariant effects, and within-dyad time-varying effects over time. This allowed precision in drawing inferences from findings. Relatedly, including three time points when children were aged 3, 4, and 6 years was a strength. By focusing on development from early to middle childhood, times of important changes in emotional competencies, our study contributes to a better understanding of how emotion socialization processes unfold longitudinally. We addressed the transactional nature of emotion socialization by testing children’s behaviors with mothers’ expressions as well as mothers’ behaviors with children’s expressions. This was particularly developmentally appropriate given children’s typical growth in emotional communication and regulation that might afford greater impact on parents’ emotion socialization behaviors (Saarni et al., 2006; Saarni, 2008). It will be important for future research to address the meaning and functions of children’s emotion coaching and dismissing behaviors, which may be different than parents’ because children’s role does not encompass fostering parents’ development. Including positive as well as negative emotions also strengthened our study, providing a broader examination of emotion socialization and expression over time. Finally, behavioral observation of mother-child interactions during a challenging cooperative task provided a rigorous measurement of multiple components of emotion socialization in a context that may have an important similarity to occasions in families’ lives when responding to each other’s emotions is not the main focus, yet emotion socialization is taking place regardless.

It is also important to recognize some limitations that inspire future directions. Interrater reliability for coaching and dismissing of negative emotions was lower than is desirable. Although concern about this was mitigated by the low occurrence of these codes, this nonetheless added variance to analyses. It will be important for future research to test generalizability of findings in tasks that elicit more responses to negative emotions. Indeed, children’s dismissing of mothers’ negative emotions was so rarely observed that it was excluded from our analyses. Perhaps children’s dismissing of mothers’ negative emotions does not become common until late childhood or adolescence. Future research including a wider developmental period with more time points may inform a more comprehensive picture of how emotion socialization develops and fluctuates over time.

Regarding the sample, even though the missing dyads at all time points are missing at random and HLM analyses retained dyads with missing data at one or two time points (Singer & Willet, 2003), only 46.7% of all the participating dyads had complete data across three time points. Also, dyads who participated in the lab visit differed significantly from those who were absent on maternal education at age 3 and age 4 as well as child race at age 3. This may limit the generalizability of our findings. We included a U.S. community sample with mothers who were willing to participate in lab visits with their child throughout the years, the majority of whom identified as White and had completed college education or higher. Although research demonstrates benefits of emotion coaching in clinical samples (e.g., Dunsmore et al., 2013; Katz & Windecker-Nelson, 2004), findings regarding trajectories across time may not generalize to clinical samples. Similarly, although research demonstrates benefits of emotion coaching for African American youth (Cunningham et al., 2009) and Chinese preschoolers (Qiu & Shum, 2021), future research is needed with culturally, ethnoracially, and socioeconomically diverse populations to determine generalizability of current findings and unravel nuanced, culturally and contextually informed underlying mechanisms. For example, although White, African American, and Lumbee American Indian parents expressed favorable views about children’s displays of happiness, only White parents showed concern about children holding on to emotions and hindering daily life (Parker et al., 2012). Among African American families, parental racial socialization and emotion socialization may intertwine to influence children’s adaptive development, as effective racial socialization to manage and prepare for experiences of discrimination and to foster racial/ethnic pride may require flexible use of emotion socialization strategies and effective emotion socialization requires attention to race-related contextual and environmental demands (Dunbar et al., 2017). With a sample that predominantly included White, African American, and biracial families, parents’ emotion coaching during family conversations was higher in families with higher socioeconomic status, whereas emotion dismissing did not differ according to families’ socioeconomic status (Lunkenheimer et al., 2007). Also, we only included mother-child dyads and did not find systematic differences across child gender. Baker and colleagues (2011) found that fathers’ emotion socialization behaviors were structured differently than mothers’ in relation to children’s social competence. Literature would benefit from studies examining parent-child same-gender and across-gender effects as well as considering other potentially influential sociocultural factors such as families’ residential places and social class as well as race, ethnicity, and culture.

In terms of measurements, we noted above the strength of our use of the etch-a-sketch task to elicit emotion expression and socialization responses. As with all observation tasks, it is unknown whether families’ behavior in the lab reflects their behavior at home. Comparing tasks with varying situational demands that provide analogues for different aspects of families’ daily lives may illuminate within-family flexibility and complexity in emotion socialization processes. With older children, combining observations with video-guided reports will provide insight into the impact of emotion socialization behaviors and children’s active role in emotion socialization processes.

Practical Implications

From early to middle childhood, children advance in emotional skills and parents as key socialization agents continue to foster child emotional development through family interactions (Eisenberg, 2020). In the current study, we found that expression of and responses to emotions were not static but rather changing over time. Children’s active engagement in emotion socialization increased over time and parents demonstrated responsiveness to children’s emotion socialization behaviors over time. Findings suggest the importance of taking both children’s developmental stage and family transactional processes into consideration in intervention and prevention efforts to better serve families and promote child functioning. According to positivity resonance theory, shared positive emotions function to foster relational closeness and overall well-being while expressing inner states (Brown & Fredrickson, 2021). Our study is consistent with this theory as findings indicate children’s increased initiation of mutual positive affect as they develop emotional autonomy, and mothers’ emotional responsiveness to their child’s socialization behaviors. The Tuning into Kids parenting program, which teaches parents to be emotionally aware and coach children’s negative emotions, has shown beneficial effects on children’s emotional competence and reduced behavioral problems in diverse cultural groups (e.g., Havighurst et al., 2010; Qiu & Shum, 2021). Because our findings highlight the role of positive emotions during mother-child emotion socialization, it may be useful for parenting education and intervention programs to include guidance about identifying and facilitating children’s positive emotions in addition to effective socialization of children’s negative emotions. In summary, practical implications of findings for intervention and prevention include consideration of both positive and negative emotions and more thorough application of a systems view of socialization and family interactions recognizing mutual influences of family members on one another.

Conclusions

In conclusion, results indicate different trajectories (both sample-wide and within-dyads) for emotion socialization behaviors for positive versus negative emotions from early to middle childhood. Findings support relations of coaching and dismissing to expression within mother-child dyads. Both maternal coaching and dismissing of emotions related to children’s increased expression. Children’s coaching related to mothers’ increased expression. This demonstrates that children play an active role in the emotion socialization process through recognizing mothers’ emotional cues and initiating mutual interconnections with mothers. Overall, our study sheds light on the complex, interconnected, and dynamic emotional lives of mother-child dyads.

Highlights.

• Children play an active role in emotion socialization transactions.

• Mothers decreased and children increased in coaching of positive emotions over time.

• Maternal emotion coaching and emotion dismissing related to children’s increased expression over time.

• Children’s emotion coaching related to mothers’ increased expression over time.