Longitudinal Links among Mother and Child Emotion Regulation, Maternal Emotion Socialization, and Child Anxiety

Abstract

Models of transdiagnostic family emotion processes recognize parents’ emotion-related characteristics and behaviors as key contributors to child emotional development and psychological functioning. One such psychological outcome, child anxiety, is prevalent and early emerging, underscoring the importance of identifying early family- and emotion-related mechanisms involved in anxiety risk. We investigated the extent to which mother and child emotion-related traits and behaviors related to child anxiety in a community sample of 175 mother-child dyads. Using three time-points (child ages 2-4 years, assessments 1 year apart), we examined how mothers’ emotion dysregulation predicted their emotion socialization practices (either supportive or non-supportive) and children’s emotion regulation (ER; either attention- or caregiver-focused) over time, in relation to later child anxiety. Models controlled for child inhibited temperament and also tested the role of maternal anxiety in these trajectories. Mothers reported on their emotion dysregulation, emotion socialization, and their own and their child’s anxiety, whereas child ER and inhibited temperament were measured using laboratory observation. In supportive emotion socialization models, maternal emotion dysregulation predicted child anxiety 2 years later. An indirect effect emerged, such that greater maternal emotion dysregulation predicted greater non-supportive emotion socialization, which in turn related to children’s greater caregiver-focused ER. Maternal emotion dysregulation, maternal anxiety, and child inhibited temperament each predicted child anxiety above and beyond other variables, although their shared variance likely accounted for some of the results. Findings lend partial support to current theoretical models of transdiagnostic family emotion processes and child anxiety development, suggesting promising avenues of future research.

Introduction

Foundational models of family emotion processes (e.g., Eisenberg et al., 1998; Morris et al., 2007) posit that parents’ emotion-related practices (e.g., their reactions to children’s emotions, discussions about emotions, modeling of emotion regulation) shape children’s abilities to manage emotions, which in turn engender children’s psychological adjustment. Notably, these models acknowledge that emotion-related parenting practices are influenced by both parents’ and children’s characteristics. Such models are well-suited for research using the developmental psychopathology approach (Sroufe & Rutter, 1984), which emphasizes the examination of normative processes to inform deviations to psychopathology onset. Employing this approach, we assessed the extent to which models of ubiquitous (yet variable) and transdiagnostic family emotion processes may be relevant to anxiety risk in a typically developing sample of children.

Despite its high prevalence (Chavira et al., 2004), relevance to community samples (Muris et al., 1998), and comorbidity with other mental disorders (Tandon et al., 2009), child anxiety remains a disorder whose precedents and mechanistic underpinnings are not entirely understood (Chavira et al., 2004; Zahn-Waxler et al., 2000). Empirical focus on early childhood is critical, given that anxiety symptoms emerge early in development and predict children’s later anxiety outcomes (Mian et al., 2011; Tandon et al., 2009). In the current study, we utilized a longitudinal, multimethod research design to examine how theorized family emotion processes related to anxiety symptoms in a community sample of young children (ages 2-4 years old). Specifically, we assessed the extent to which mothers’ emotion dysregulation distally predicted child anxiety through more proximal emotion-related mechanisms (i.e., maternal emotion socialization, child emotion regulation), while accounting for the established relation between child inhibited temperament and child anxiety. We also tested the role of maternal anxiety, given its hereditary and environmental transmission to child anxiety (Murray et al., 2009) and its relevance to emotion-related parenting (Berg-Nielsen et al., 2002; Morris et al., 2007).

Parent Emotion Dysregulation: A Transdiagnostic Predictor

Emotion dysregulation characterizes individuals’ patterns of experiencing and expressing emotions in ways that are maladaptive (e.g., too intense, long-lasting, context-inappropriate) and interfere with goal-directed activity (Gratz & Roemer, 2004; Thompson, 2019). Given the inherently emotional context of parenting (Hajal & Paley, 2020; Rutherford et al., 2015), we assessed parents’ (i.e., mothers’) emotion dysregulation. Parents are continuously tasked with maintaining a regulated state while also facilitating their young child’s emotion regulation and responding to their needs (Rutherford et al., 2015). Parents who are more emotionally dysregulated may become easily overwhelmed by their own and their child’s emotional experiences and, consequently, be more likely to avoid or mitigate child emotion displays (Buckholdt et al., 2014; Hajal & Paley, 2020). Importantly, researchers postulate that parents’ characteristics and behaviors that transcend diagnoses may ultimately contribute to their children’s anxiety and may be most salient to intervention efforts (Berg-Nielsen et al., 2002). Given its established role as a core feature of various adult psychopathologies (Aldao et al., 2010), emotion dysregulation is likely one such transdiagnostic factor underlying parents’ anxiety symptoms. Parents’ anxiety and anxiety-relevant parenting practices put children at risk for maladaptive outcomes, including children’s own anxiety (Berg-Nielsen et al., 2002), which derives from both hereditary pathways and environmental learning (Murray et al., 2009; Zahn-Waxler et al., 2000). Thus, one may expect parents’ emotion dysregulation and anxiety symptoms to be highly interrelated and dually predictive of children’s later anxiety symptoms.

Considering this precedent, to establish a model of family emotion processes relevant to child anxiety risk, it is necessary to determine the extent to which relations are driven by parents’ transdiagnostic traits (i.e., emotion dysregulation), psychopathology-specific characteristics (i.e., anxiety), and their shared variance. We addressed this need and advanced theory about relations from maternal emotion dysregulation to child psychopathology in three ways. First, we examined two key mechanisms (i.e., maternal emotion socialization, child emotion regulation) of this relation. Second, we applied this model of risk to one particular child outcome, namely, anxiety. Third, we tested processes with and without consideration of maternal anxiety to clarify whether the transdiagnostic factor of emotion dysregulation uniquely accounted for child anxiety risk.

The Mediating Role of Parent Emotion Socialization

Parents’ behaviors (e.g., those that are overly controlling, critical, or negative) towards their children do appear to reinforce avoidance and discourage independence, thereby contributing to children’s greater anxiety symptoms (Berg-Nielsen et al., 2002; Murray et al., 2009). There have been calls for research focused on more specific (e.g., in situation, in behavior; Berg-Nielsen et al., 2002) and proximal (Murray et al., 2009) parenting behaviors. Parent emotion socialization may function as proximal parenting behavior that is specific to children’s emotional experiences, informative given the innately emotional context of parenting (Rutherford et al., 2015), and highly relevant to child anxiety development.

Parents are emotional role models for their youth, fostering their children’s skills to display, understand, and regulate emotions (Morris et al., 2007; Zeman et al., 2013). Emotion socialization provides a crucial venue through which parents impart their values, beliefs, and practices regarding the acceptability of emotional expressivity to their children (Eisenberg et al., 1998; Zeman et al., 2013). Pertinent to the current study, one method of emotion socialization involves the supportive and non-supportive responses that parents provide to their children’s negative emotion displays (Eisenberg et al., 1998). Supportive responses include encouraging and validating children’s emotions, soothing and comforting them to alleviate distress, or helping them solve their problem (Fabes et al., 2002). In contrast, parents may non-supportively punish (e.g., by scolding them or threatening loss of a privilege) or minimize (e.g., by saying they are overreacting) children’s emotions, or become highly distressed themselves (Fabes et al., 2002; O’Neal & Magai, 2005). Importantly, supportive and non-supportive responses seem to be conceptually distinct, with parents engaging in varying amounts of both (Fabes et al., 2002).

The ways in which parents manage emotions are likely proximal predictors of their emotion-related parenting practices (Hajal & Paley, 2020). For example, because emotionally dysregulated parents may be less aware and in control of their emotions, they may display their own distress and be motivated to punish or dampen children’s emotional experiences to mitigate their own negative feelings. Indeed, theorized linkages between parents’ emotion dysregulation and emotion socialization practices (Morris et al., 2007) have been corroborated by empirical work. Recent findings suggest that greater parent emotion dysregulation predicts decreased supportive (Raval et al., 2018) and heightened non-supportive (Buckholdt et al., 2014; Morelen et al., 2016) responses to school-age children and adolescents. Still, more research is needed to understand these trends in younger samples over time to develop models of anxiety risk and inform early intervention efforts, including those that may be transdiagnostic in nature.

Parents’ supportive and non-supportive emotion socialization responses function as key transdiagnostic targets and contribute to children’s various adaptive and maladaptive outcomes, respectively (Eisenberg et al., 1998). Supportive emotion socialization has been linked to preschoolers’ (Hooper et al., 2018; Mirabile et al., 2016) and adolescents’ (Buckholdt et al., 2014) lower rates of internalizing problems. In contrast, non-supportive emotion socialization relates to greater internalizing symptoms in middle childhood and adolescence (O’Neal & Magai, 2005; Sanders et al., 2015). In fact, Suveg et al. (2005) showed that mothers of clinically anxious school-age children are more prone to discouraging emotions. With some exceptions (e.g., Hooper et al., 2018), the overwhelming literature suggests that parents’ supportive and non-supportive reactions to youths’ emotions are negatively and positively, respectively, related to child anxiety. Thus, we anticipated that emotion socialization would serve as an intermediary mechanism between mothers’ emotion dysregulation and their children’s anxiety. Further, the relation between maternal emotion socialization and child anxiety may itself be mediated (i.e., through child emotion regulation) as part of a serial transmission process.

The Mediating Role of Child Emotion Regulation

Early in life, children begin developing skills to regulate emotions (Cole et al., 2009; Zahn-Waxler et al., 2000) and showing individual differences in basic regulatory abilities (Bridges & Grolnick, 1995; Grolnick et al., 1996). By early toddlerhood, most children have transitioned from near-exclusive dependence on caregivers for emotion management to voluntary, purposeful engagement in their own emotion regulation (ER) strategy use (Calkins, 2007). Young children engage in a variety of ER behaviors (Grolnick et al., 1996), including caregiver-seeking and attention-shifting behaviors. Toddlers employ caregiver-focused behaviors when they either look towards their caregiver for reassurance (e.g., by visually “checking in”) or seek closer contact with them (e.g., by climbing onto their lap) during distressing situations (Diener & Mangelsdorf, 1999). Toddlers also regulate distress by briefly averting their gaze from a distressing stimulus or visually focusing on a source of distraction (e.g., a toy) for longer periods of time (i.e., attention-focused ER; Bridges & Grolnick, 1995; Calkins, 2007). Both caregiver-focused and attention-focused ER behaviors occur contingently with decreases in toddler distress (Buss & Goldsmith, 1998; Diener & Mangelsdorf, 1999), whereas deficits in these abilities relate to children’s greater negative arousal (Calkins, 2007).

Via both environmental and genetic pathways, children’s ER abilities do seem to derive, in part, from parents’ own emotion regulation/dysregulation (Buckholdt et al., 2014; Morelen et al., 2016; Rutherford et al., 2015; Thompson, 2019), with more emotionally dysregulated parents having children with greater difficulties managing emotions. Emotionally dysregulated parents’ expressivity and behaviors may communicate that negative emotions are both uncontrollable and unacceptable, thus precipitating children’s poor ER (Buckholdt et al., 2014). Further, parent emotion socialization and child ER are empirically linked. On one hand, via parents’ supportive responses to their emotions, youth learn adaptive ways to understand and regulate emotions (Cole et al., 2009; Fabes et al., 2002; Hurrell et al., 2015), with this association perhaps holding especially true for children younger than 4 years (Mirabile et al., 2016). On the other hand, non-supportive parental responses have been implicated in prolonging and intensifying children’s emotional states (Fabes et al., 2001), limiting opportunities to learn appropriate coping strategies (Sanders et al., 2015), and contributing to dysregulated affect (Buckholdt et al., 2014; Fabes et al., 2002). Thus, non-supportive responses may hinder ER skill development and teach children to suppress emotions, thereby engendering less adaptive ER (Fabes et al., 2001).

Child ER difficulties are theorized to be transdiagnostic predictors of psychopathology, as well as child anxiety development and maintenance, in particular (Thompson, 2001; Zahn-Waxler et al., 2000). As is also the case for adults, dysregulated emotions may be inherent to anxiety, with limited access to a repertoire of ER strategies and lack of understanding and acceptance of emotions predicting anxiety (Hurrell et al., 2015; Thompson, 2001). Thus, the large corpus of research on the link between child ER and anxiety lends itself to expectations that children’s lower use of ER behaviors would relate to their greater anxiety symptomatology.

In sum, there is support for both parent emotion dysregulation and emotion socialization predicting child ER, as well as the role of these variables in child anxiety risk. Thus, we propose that emotion socialization and child ER may operate as serial mediators in the association between maternal emotion dysregulation and child anxiety. An integrated model is supported by theory (Eisenberg et al., 1998; Morris et al., 2007) and research assessing relations between some, but not all, of these variables (Buckholdt et al., 2014; Han & Shaffer, 2013; Morelen et al., 2016). Our model addresses the need to identify mechanistic relations in a younger sample, across time, and with consideration of established child psychopathology predictors.

Accounting for Child Inhibited Temperament

Capturing the extent to which a child shows wariness, fear, and withdrawal in response to unfamiliar individuals or contexts (Kagan et al., 1989), child inhibited temperament has emerged as a stable, robust predictor of anxiety (Rapee & Coplan, 2010). That is, children who display, or are perceived to display, higher reticence and wariness towards novel situations are more likely to develop anxiety than children who do not (Murray et al., 2009; Rapee & Coplan, 2010). Given this established relation, and because inhibited temperament may account for part of children’s heritable, biologically-based risk for anxiety (Rapee & Coplan, 2010; Zahn-Waxler et al., 2000), models of processes involved in anxiety risk should consider child inhibited temperament. Thus, in the present study, we tested whether transdiagnostic models of family emotion processes predicted child anxiety above and beyond the effects of child inhibited temperament on anxiety.

The Current Study

Foundational theory suggests that maternal emotion dysregulation and emotion socialization, as well as children’s early ER attempts, predict child psychopathology. Recent studies (e.g., Buckholdt et al., 2014; Morelen et al., 2016) have begun to address these gaps, though they have been conducted using cross-sectional designs, in relatively small samples, and in later developmental periods. Relations have not yet been tested in integrated models using a multimethod, longitudinal research design and toddler sample. Models of emotion processes have high potential to inform our understanding of child anxiety risk, particularly if said models consider other key anxiety predictors (e.g., child inhibited temperament, maternal anxiety). As such, we tested anxiety-relevant models of family emotion processes and risk factors in a community sample of mothers and children who were assessed at child age 2 (Time 1 [T1]), 3 (Time 2 [T2]), and 4 years (Time 3 [T3]). Our primary aim was to examine how maternal emotion dysregulation related to mothers’ emotion socialization (either supportive or non-supportive), children’s ER (either attention- or caregiver-focused), and subsequent child anxiety (see Figure S1 in Online Resource). We expected that greater T1 maternal emotion dysregulation would predict less supportive and more non-supportive T2 emotion socialization, fewer T2 child ER behaviors, and greater T3 child anxiety. We hypothesized that T1 maternal emotion dysregulation would serially predict T3 child anxiety through T2 emotion socialization and T2 child ER. Effects were expected to exist above and beyond T1-T2 child inhibited temperament.

Two additional sets of models were run to inform directionality of relations and determine the extent to which transdiagnostic family emotion processes predicted child anxiety above and beyond maternal anxiety. First, preliminary analyses assessed associations between mothers’ emotion dysregulation and anxiety (across T1 and T2) as they predicted later child anxiety. It was expected that T2 maternal emotion dysregulation would uniquely predict T3 child anxiety and account for the relation between T1 maternal anxiety and T3 child anxiety. Second, we retested primary models while controlling for T1 maternal anxiety. We expected that hypothesized effects in the primary models would persist, accounting for T1 maternal anxiety.

Method

Participants

The sample included 175 mothers and their toddlers (42.9% [75] girls), recruited in the midwestern United States as part of a larger study. Mothers and toddlers, respectively, were 92%/89.7% non-Hispanic and 1.7%/3.4% Hispanic, and regarding race were 88.6%/80% European American, 2.9%/1.1% Asian or Pacific Islander, 0.6%/1.7% African American, 0.6%/1.1% Latinx, 0.6%/0% Native American, and 0.6%/9.1% had multiple racial identities. Some mothers declined to respond about themselves (6.3%) or their toddlers (6.9%). Families were diverse in socioeconomic status (SES), with 32.2% of families reporting an annual family income >$40k (M = $51-60k, Range = <$15k to >$100k). Mothers averaged 15.47 years of education (Range = 9^th grade to PhD); 30.8% of mothers reported a high school degree or less.

Mothers were recruited from local birth announcements and community programs/offices (e.g., WIC program, farmer’s markets, doctor offices). Families enrolled in the study at any of the three time-points. Families came in for T1 when children were 24-30 months old (M = 26.83, SD = 1.96), T2 when 36-42 months old (M = 39.20, SD = 2.88), and T3 when 48-54 months old (M = 51.67, SD = 3.72). At T1, T2, and T3, mothers’ average ages were 32.18 (SD = 5.29), 33.79 (SD = 5.09), and 34.60 (SD = 5.15) years. We used a rolling recruitment method: 145 families enrolled at T1, 18 enrolled at T2, and 12 enrolled at T3. Of the sample, 82.3% of families had data for T1, 72.6% had data for T2, and 64.0% had data for T3. Further, 44.6% of families had data for three time-points, 29.7% had two time-points, and 25.7% had one time-point.

Procedure

Miami University institutional research board approval was obtained for the larger study, and mothers provided written informed consent at all time-points. Mothers were mailed consent forms and questionnaires, which they then returned in-person at their laboratory visit (1.5-3 hours, depending on time-point). At each time-point, mothers were debriefed, receiving compensation of $50 for their time, and their children received a small gift (worth <$5).

At T1 (or first entry time-point for families starting at T2 or T3), the questionnaire packet assessed demographic information, maternal emotion dysregulation, child inhibited temperament, and maternal anxiety. At their lab visit, mothers and toddlers engaged in activities from the Laboratory Temperament Assessment Battery (Lab-TAB; Buss & Goldsmith, 2000) and past studies (Buss, 2011). Relevant to this study, the “Risk Room” episode was video-recorded to provide an observation measure of child inhibited temperament. A research assistant (RA) led the mother-toddler dyad into a room with toys (i.e., tunnel, trampoline, balance beam, box, lion mask), instructing the toddler to play with the toys and the mother to limit her involvement. The RA came back after 3 min and prompted the child to interact with each toy.

At T2, mothers answered questions about their emotion dysregulation, emotion socialization, and child’s temperament. At the lab, the child again participated in the Risk Room. To provide an observational measure of ER, the child also engaged with a RA dressed up as a clown and a remote-controlled spider. In the 5-min “clown” episode, the clown introduced herself and invited the child to play with toys (i.e., bubbles, beach balls, musical instruments) for 1 min each. The clown asked the child to help clean up before saying goodbye. The 2-min “spider” episode began with the toddler seated in their mother’s lap. A remote-controlled spider approached and retreated from them twice, with 10-sec pauses between movements. A familiar experimenter then entered the room, gave the child three prompts to touch the spider, and showed them that the spider was a toy. Episodes were video-recorded for behavioral coding.

At T3, mothers answered questions about their child’s anxiety symptoms before their laboratory visit. In the current study, only the child anxiety questionnaires were used. Of note, several of the measures (i.e., maternal emotion dysregulation, anxiety, and emotion socialization) were also utilized at other time-points for preliminary and supplementary models.

Measures

Maternal emotion dysregulation (T1).

Mothers completed the 36-item Difficulties in Emotion Regulation Scale (DERS; Gratz & Roemer, 2004) to assess their emotion dysregulation along six dimensions: nonacceptance of emotions, difficulties engaging in goal-directed behavior, impulse control difficulties, limited access to ER strategies, lack of emotional awareness, and lack of emotional clarity (e.g., “When I’m upset, I lose control over my behaviors”). Mothers responded to items on a 1 (almost never) to 5 (almost always) scale. We averaged all items (α = .94) to derive the final emotion dysregulation variable. The DERS has been used with parents in past studies (e.g., Buckholdt et al., 2014; Morelen et al., 2016) and has shown acceptable to high test-retest and internal reliability, as well as construct validity with respect to other measures of ER and clinically-relevant outcomes (Gratz & Roemer, 2004).

Maternal emotion socialization (T2).

Mothers completed the 82-item Coping with Toddlers’ Negative Emotions Scale (CTNES; Spinrad et al., 2004). The CTNES presented 12 hypothetical scenarios (e.g., “If my child loses some prized possession and reacts with tears…”), for which mothers used a 7-point scale (1 = Very Unlikely to 7 = Very Likely) to indicate the likelihood that they would engage in seven different responses to their child’s negative emotion. Both supportive and non-supportive emotion socialization response composites were created. The 36-item supportive responses composite (α = .94) was derived from the averages of the Expressive Encouragement (α = .93; e.g., “I would tell my child it’s okay to cry”), Emotion-Focused Reactions (α = .83; e.g., “I would comfort my child and try to make him/her feel better”), and Problem-Focused Reactions (α = .87; e.g., “I would help my child think of something else to do”) subscales (rs = .54-.74, ps ≤ .001). The 36-item non-supportive responses composite (α = .90) was derived from the averages of the Distress Reactions (α = .83; e.g., “I would feel upset and uncomfortable because of my child’s reactions”), Minimization Reactions (α = .86; e.g., “I would tell my child that he/she is overreacting”), and Punitive Reactions (α = .82; “I would tell my child to behave or we’ll have to go home right away”) subscales (rs = .30-.61, ps = .002-≤.001). CTNES composites were created as consistent with past research (e.g., Gudmundson & Leerkes, 2012; Hurrell et al., 2015; Morelen et al., 2016). The CTNES has shown acceptable to excellent test-retest and internal reliability, and construct validity with respect to other parenting indices (Fabes et al., 2002; Spinrad et al., 2004).

Observed child ER (T2).

Toddler ER behaviors during clown and spider episodes were coded using Lab-TAB coding protocol (Buss & Goldsmith, 2000) and INTERACT (Mangold, 2017). Coders rated behaviors as present/absent continuously throughout the episode (at 25 frames/sec rate). Caregiver-focused ER behaviors denoted toddlers’ looks towards their mother or self-initiated efforts to increase proximity to their mother. Attention-focused ER behaviors involved instances in which toddlers looked away from the stimulus either briefly and without apparent focus or towards a different target for 2+ secs. We derived a frequency score (i.e., summed discrete instances of behavior) and a proportion score (i.e., total time engaged in behavior, relative to episode time, range = 0-1) for each ER type. See Online Resource for observed averages and ranges prior to creating a composite. Frequency and proportion ICCs ranged from .90-.98 (M_ICC = .95) for caregiver-focused ER and from .77-.91 (M_ICC = .85) for attention-focused ER. To account for tradeoffs conferred in using either frequency- or proportion-based scores, both factors were standardized and averaged within episode. The two standardized averages were averaged across episodes per behavior. A separate coder team scored child distress (e.g., negative facial expressions/vocalizations) during these episodes on a 5-point scale (1 = No Distress Shown to 5 = Distress Entire Episode). Episode distress ICCs ranged from .98-.99 (M_ICC = .99). Children’s ER behavior averages were regressed on their distress score to account for their distress in encountering the stimuli (Buss, 2011), providing final ER variables.

Child anxiety symptomatology (T3).

Mothers completed the 7-item Anxiety/Worry subscale (α = .66; e.g., “[my child] seems nervous, tense, or fearful”) and 5-item Inhibition to Novelty subscale (α = .85; e.g., “[my child] takes a while to feel comfortable in new places”) of the 126-item Infant-Toddler Social and Emotional Assessment (ITSEA; Carter et al., 2003). Using a 3-point scale (0 = Not true/Rarely to 2 = Very true/Often), mothers reported how well/often the statements described their child in the past month. The ITSEA is appropriate for children ages 1-4 years old (Carter et al., 2003; Mian et al., 2011). In past studies, it has demonstrated acceptable to excellent test-retest and internal reliability, as well as construct validity with respect to other measures of child functioning and symptoms.

Additionally, mothers completed the 5-item Generalized Anxiety subscale (α = .77; e.g., “[my child] has difficulty stopping him/herself from worrying”) and 6-item Social Anxiety subscale (α = .85; e.g., “[my child] is afraid of meeting or talking to unfamiliar people”) of the 34-item Preschool Anxiety Scale (PAS; Spence et al., 2001). Mothers responded on a 5-point scale (0 = Not true at all to 4 = Very often true) with respect to how well statements described their child. The PAS is appropriate for preschool-age children and has shown good test-retest and internal reliability, as well as construct validity (Spence et al., 2001). For analyses, the two ITSEA and two PAS subscale scores (means of items) were each standardized and averaged together to create a T3 child anxiety symptom score. The four indices of T3 child anxiety symptoms were positively inter-related (rs = .26-.68, ps = .017 to < .001).¹

Child inhibited temperament (T1-T2).

Behaviors indicating inhibited temperament during T1 and T2 Risk Room episodes were coded using INTERACT (Mangold, 2017). Specifically, coders scored the number of seconds it took the child to touch a toy for the first time, the number of toys touched (reversed), the number of seconds spent playing (reversed), and the number of seconds spent within 2 ft of their mother. ICCs ranged from .92-1.00 (M_ICC = .98). The inhibited temperament variable comprised the average of z-scored variables.

Mothers completed the Shyness subscales (5 and 6 items, respectively) of the Early Childhood Behavior Questionnaire - Short Form (ECBQ; Putnam et al., 2006) at T1 and the Children’s Behavior Questionnaire - Short Form (CBQ; Rothbart et al., 2001) at T2. Subscales assessed the extent to which mothers saw their child as slow to approach and uncomfortable/inhibited in new social situations. Mothers responded on a 7-point scale (1 = Never to 7 = Always) to items like “when approaching unfamiliar children playing, how often did your child seem uncomfortable?” (ECBQ) and “[my child] acts shy around new people” (CBQ). Both measures have shown adequate internal consistency in validation studies and good internal consistency presently (αs = .82, .85). We z-scored the two observed and two mother report measures, then averaged them together to create a T1-T2 child inhibited temperament composite.

Maternal anxiety symptoms (T1).

We measured maternal anxiety symptoms by creating a composite of the 16-item Penn State Worry Questionnaire (PSWQ; Meyer et al., 1990) and the 20-item Social Interaction Anxiety Scale (SIAS; Brown et al., 1997). PSWQ items were scored on a 5-point scale (1 = Not at all typical to 5 = Very typical) and provided a measure of trait worry (α = .94; e.g., “I am always worrying about something”). SIAS items were scored on a 5-point scale (0 = Not at all characteristic or true of me to 4 = Extremely characteristic or true of me) and provided a social interaction anxiety measure (α = .94; e.g., “When mixing socially, I am uncomfortable”). Both measures showed acceptable internal consistency and construct validity in past studies. The PSWQ and SIAS means were strongly and positively related (r = .55 p < .001). We standardized each, then averaged them to create a maternal anxiety composite.²

Data Analytic Strategy

We first examined patterns of data missingness and descriptive statistics, before assessing bivariate correlations among primary variables, as well as correlations between primary variables and demographic variables (e.g., child sex, family SES³) to determine covariates. For all model testing, we analyzed path models through Mplus v.7.3 (Muthén & Muthén, 2012) using full information maximum likelihood (FIML) estimation (Enders & Bandalos, 2001; Graham, 2009). Path analysis models simultaneously assessed interrelations and indirect effects among variables.

For primary analyses, we ran four path analysis models (see Figure S1 in Online Resource) that varied on T2 emotion socialization (supportive or non-supportive), T2 child ER (caregiver- or attention-focused), and demographic covariates. T1 maternal emotion dysregulation predicted T2 emotion socialization, T2 child ER, and T3 child anxiety. T2 emotion socialization was associated with T2 child ER and T3 child anxiety. T2 child ER and T1-T2 child inhibited temperament predicted T3 child anxiety. We tested a serial mediation in which T1 maternal emotion dysregulation predicted T3 child anxiety through T2 emotion socialization and T2 child ER, as well as simple indirect effects along the serial mediation path.

We also ran two sets of models to determine whether maternal anxiety accounted for (a) the relation between T1 maternal emotion dysregulation and T3 child anxiety and (b) relations among variables in the full primary models. First, we tested a cross-lagged path model (see Figure S2 in Online Resource) wherein paths were estimated for construct stability, within-time-point correlations, and cross-construct prediction across time for mothers’ emotion dysregulation and anxiety from both T1 and T2. Maternal emotion dysregulation and anxiety were modeled to predict T3 child anxiety. Second, we examined relations when T1 maternal anxiety was added into primary models (see Figures S3-S4 in Online Resource). Due to observed bivariate associations, we allowed mothers’ T1 maternal anxiety to correlate with their T1 emotion dysregulation, predict T2 non-supportive emotion socialization (in 2 of 4 models), and predict T3 child anxiety. Indirect effects were again tested in these models. Minimally acceptable model fit was indicated by non-significant χ² values, RMSEA and SRMR values ≤.08, CFI values ≥.90, and TLI values ≥.95. We reported AIC and sample-size adjusted BIC (SABIC). We estimated confidence intervals (CIs) for indirect effects using 10,000 bootstrap samples. For brevity, only statistically significant pathways (p ≤.05) and indirect effects (CIs not containing 0) were reported in text. See Figures 1-2 (main document) and Figures S2-S4 (Online Resource) for model fit indices, unstandardized path coefficients, and standard errors.

Figure 1.

Path analysis model with supportive emotion socialization and caregiver-focused (Panel A) and attention-focused (Panel B) emotion regulation. Child emotion regulation represents the variable residualized on lab-observed distress. Covariates (i.e., SES out of caregiver-focused emotion regulation [b = 0.14, SE = 0.08, p = .075]) are not shown for ease of presentation. The Panel A model had mostly close fit χ² (4) = 4.11, p = .392; AIC = 1790.73, SABIC = 1790.69; RMSEA = 0.012 (90% CI [ 0.00, 0.12]); SRMR = 0.036; CFI = 1.00, TLI = 0.99. The Panel B model’s fit ranged from acceptable to close (with only TLI indicating mediocre fit), χ² (2) = 3.40, p = .183; AIC = 1327.92, SABIC = 1327.89; RMSEA = 0.063 (90% CI [ 0.00, 0.18]); SRMR = 0.044; CFI = 0.97, TLI = 0.85. Unstandardized path coefficients are shown with standard error estimates in parentheses. Gray lines represent non-significant paths, dashed lines represent marginally significant paths, and solid black lines represent statistically significant paths. T1 = Time 1, T2 = Time 2, T3 = Time 3. ^tp ≤ .10, *p ≤ .05, **p ≤ .01, ***p ≤ .001.

Figure 2.

Path analysis model with non-supportive emotion socialization and caregiver-focused (Panel A) and attention-focused (Panel B) emotion regulation. Child emotion regulation represents the variable residualized on lab-observed distress. Covariates (i.e., SES out of non-supportive emotion socialization [bs = −0.16 to −0.17, SEs = 0.08-0.09, ps = .053-.057], SES out of caregiver-focused emotion regulation [b = 0.17, SE = 0.08, p = .024]) are not shown for ease of presentation. The Panel A model had very close fit, χ² (3) = 1.42, p = .702; AIC = 1741.89, SABIC = 1741.85; RMSEA = 0.00 (90% CI [ 0.00, 0.10]); SRMR = 0.022; CFI = 1.00, TLI = 1.00. The Panel B model also had very close fit, χ² (4) = 3.19, p = .527; AIC = 1740.38, SABIC = 1740.34; RMSEA = 0.00 (90% CI [ 0.00, 0.10]); SRMR = 0.034; CFI = 1.00, TLI = 1.00. Unstandardized path coefficients are shown with standard error estimates in parentheses. Gray lines represent non-significant paths, dashed lines represent marginally significant paths, and solid black lines represent statistically significant paths. T1 = Time 1, T2 = Time 2, T3 = Time 3. ^tp ≤ .10, *p ≤ .05, **p ≤ .01, ***p ≤ .001.

Results

Preliminary Analyses

Missing data.

Taken together, 28.14% of final primary variable values were missing, 8.57% of which were due to families starting at T2 or T3. The study design shared characteristics with planned missing and accelerated longitudinal designs (Graham, 2009) in which participants do not complete all measures due to rolling recruitment, but larger sample sizes are retained. FIML was an appropriate method for handling missing data (Enders & Bandalos, 2001).

Missing values existed for T1-T2 child inhibited temperament (n = 12, 6.9%), T1 maternal anxiety (n = 37, 21.1%), T1 DERS (n = 38; 21.7%), the T2 CTNES (n = 68; 38.9%), T2 observed child ER or distress coding (n = 54; 30.9%), the T3 child anxiety composite (n = 63, 36.0%), mostly due to beginning the study at a later time-point, attrition (mostly due to moving out of the area), mothers filling out partial or no packets, or there were recording technical difficulties. These families did not differ on family income or any other primary or contextual study variables. Little’s missing completely at random (MCAR) test indicated that the pattern of missingness did not deviate from a MCAR pattern (χ²[85] = 89.34, p = .353).

Descriptive statistics and bivariate associations.

Variable descriptives and bivariate correlations are presented in Table 1. All variables showed adherence to a normal distribution (skew < |2.00|, kurtosis < |4.00|), with one exception. Child generalized anxiety (PAS) had a kurtosis of 6.43 and was thus square-root transformed (kurtosis = 3.30). See the Online Resource for correlations between child ER behaviors within and across episodes, as well as correlations between maternal emotion dysregulation and aspects of non-supportive emotion socialization. SES related to non-supportive emotion socialization and caregiver-focused ER, so it was included as a covariate in models containing those variables. Child biological sex differences did not exist for any primary construct (all ps > .05), so it was not considered further in analyses.

Table 1.

Descriptive Statistics and Correlations among Primary Variables

Variable	Mean (SD)	Range	1	2	3	4	5	6	7	8	9	10	11
1. Child biological sex	—	—	—	.13t	.13	.08	.10	.002	.03	−.03	−.07	.01	.05
2. Family socioeconomic status	−0.01 (0.92)	−2.04 - 1.74		—	−.21*	.16	−.22*	−.10	.19*	−.05	−.07	−.15t	−.21*
3. T1 maternal emotion dysreg.	1.80 (0.50)	1.00 - 3.39			—	−.04	.38***	−.02	−.15	.22*	.32**	.17*	.71***
4. T2 maternal supportive ES	5.46 (0.88)	1.93 - 6.97				—	.02	.16	.06	.008	.08	.11	.05
5. T2 maternal nonsupportive ES	2.60 (0.78)	1.22 - 4.50					—	−.03	.16	.16t	.14	−.004	.52***
6. T2 child episode distress	1.56 (0.69)	1.00 - 3.50						—	.12	−.009	.18t	.26**	.09
7. T2 child caregiver-focused ER	−0.04 (0.71)	−1.07 - 2.59							—	.006	−.09	−.02	−.08
8. T2 child attention-focused ER	−0.02 (0.69)	−0.99 - 2.07								—	−.11	.16t	.20*
9. T3 child anxiety	−0.01 (0.79)	−1.30 - 3.33									—	.57***	.29**
10. T1-T2 child inhibited temp.	0.01 (0.73)	−2.33 - 2.05										—	.21*
11. T1 maternal anxiety	0.00 (0.88)	−1.43 - 1.96											—

Note. Dysreg = dysregulation, ES = emotion socialization, ER = emotion regulation, Temp = temperament.

Means, standard deviations, ranges, and correlations were computed after variable transformations with available data but prior to handling missing data. Family socioeconomic status, child caregiver-focused and attention-focused regulation, child anxiety, child inhibited temperament, and maternal anxiety were all variables comprised of multiple measures that were standardized and then averaged together, as described in text. Ns for correlations ranged from 82 to 171. T1 = Time 1, T2 = Time 2, T3 = Time 3

^tp < .10,

^*p < .05,

^**p < .01,

^***p < .001.

Preliminary model of maternal emotion dysregulation and anxiety, in relation to child anxiety.

The model (see Figure S2 in Online Resource) had very close fit. Maternal emotion dysregulation and anxiety were positively correlated at T1 (p <.001) and T2 (p <.004). Emotion dysregulation and anxiety demonstrated stability (ps <.001). Neither construct predicted change in the other; T2 maternal anxiety did not predict T3 child anxiety. T2 maternal emotion dysregulation positively predicted T3 child anxiety (p = .023) while accounting for T2 maternal anxiety, supporting our continued investigation of maternal emotion dysregulation as a uniquely important predictor of focus in subsequent models.

Primary Models

Across all models, T1-T2 child inhibited temperament robustly and positively predicted T3 child anxiety (ps <.001). Additionally, in all models, T1 maternal emotion dysregulation was positively correlated with T1-T2 child inhibited temperament (ps =.020-.026)

Supportive emotion socialization and caregiver-focused ER.

This model (see Figure 1) had mostly close fit. T1 maternal emotion dysregulation positively predicted T3 child anxiety (p = .045), above and beyond effects of T2 supportive emotion socialization, T2 child caregiver-focused ER, and T1-T2 child inhibited temperament. No relations emerged among T1 maternal emotion dysregulation, T2 supportive emotion socialization, and T2 caregiver-focused ER.

Supportive emotion socialization and attention-focused ER.

The second model (see Figure 1) had fit that mostly ranged from acceptable to close fit. In this model, the same unique relation from T1 maternal emotion dysregulation to T3 child anxiety present in the first model emerged (p = .029). There were no other relations among primary variables.

Non-supportive emotion socialization and caregiver-focused ER.

Fit indices indicated that this model (see Figure 2) had very close fit. Greater T1 maternal emotion dysregulation predicted more T2 non-supportive emotion socialization (p = .001) but no longer predicted T3 child anxiety. T2 maternal non-supportive emotion socialization related to T2 child caregiver-focused ER (p = .025), such that when mothers engaged in more non-supportive emotion socialization, they had toddlers with more caregiver-focused ER attempts. Moreover, there was a significant indirect effect from T1 maternal emotion dysregulation to T2 child caregiver-focused ER through T2 non-supportive emotion socialization (ab = 0.14, SE = 0.07, 95% CI = [0.026, 0.316]). Specifically, greater maternal emotion dysregulation predicted more non-supportive emotion socialization practices, which in turn contributed to more child caregiver-focused ER.

Non-supportive emotion socialization and attention-focused ER.

This model (see Figure 2) had very close fit, χ² (4) = 3.19, p = .527; AIC = 1740.38, SABIC = 1740.34; RMSEA = 0.00 (90% CI [ 0.00, 0.10]); SRMR = 0.034; CFI = 1.00, TLI = 1.00. In this model, similar effects were observed with T1 maternal emotion dysregulation relating to T2 non-supportive emotion socialization (p = .001). No emotion processes predicted child anxiety.

Models Accounting for Maternal Anxiety

We ran four additional models (see Figures S3-S4 in Online Resource) to determine the extent to which maternal anxiety explained primary model results. Models fit the data closely (see Online Resource). In all models, T1 maternal emotion dysregulation and T1 maternal anxiety were robustly, positively correlated (ps <.001), with neither emerging as a predictor of T3 child anxiety. Instead, only T1-T2 child inhibited temperament predicted T3 child anxiety (ps <.001). T1 maternal anxiety, rather than T1 maternal emotion dysregulation, predicted T2 non-supportive emotion socialization (ps <.001). Additionally, in models with non-supportive emotion socialization, mothers who were more emotionally dysregulated at T1 (p = .046) or engaged in less non-supportive emotion socialization at T2 (p = .047) had children with fewer caregiver-focused ER behaviors. T1-T2 child inhibited temperament was positively associated with T1 maternal anxiety (ps = .033-.040) and, similarly to primary models, positively correlated with T1 maternal emotion dysregulation (ps = .033-040). No indirect effects emerged.

Supplementary Analyses

We ran several supplementary analyses to complement the main aim of this paper (see Online Resource), including a series of alternative models in which emotion socialization was measured at T1 or T3 (instead of T2). Models revealed that maternal emotion socialization predicted child ER but not vice versa, supporting our theorized directionality of constructs.

Discussion

We examined integrated models in which maternal emotion dysregulation, emotion socialization, and child ER collectively predicted child anxiety symptoms. This research built on prior work by (a) assessing associations among these factors and child anxiety, (b) focusing on early childhood, (c) examining linkages across time, and (d) incorporating consideration of maternal anxiety and child inhibited temperament. Models fit the data well, with individual path coefficients differing throughout models. Our findings characterize family emotion processes involved in child anxiety risk that likely have broader transdiagnostic significance.

Several themes emerged in primary models. In supportive emotion socialization models, maternal emotion dysregulation predicted child anxiety, as hypothesized. This link existed over and above effects of supportive emotion socialization, child ER, and child inhibited temperament. This finding is among the first to signal a direct linkage between maternal emotion dysregulation and child anxiety in early childhood, and it supports targeting transdiagnostic parent-level factors like emotion dysregulation in clinical interventions for child anxiety. Perhaps the presence of this relation points to an underlying, biologically-based shared vulnerability between mothers and their children. In conjunction, children may be at risk for emotional difficulties and anxiety through their general exposure to their mothers’ emotion dysregulation and related modeled behaviors that are distinct from emotion responses. Notably, germane theoretical models of family emotion processes (e.g., Eisenberg et al., 1998; Morris et al., 2007) have highlighted how parents’ characteristics may affect their behaviors, as well as the ways in which parents’ behaviors engender child ER difficulties that precede negative adjustment. Theory has not, however, posited a direct relation between parent characteristics and child psychopathology, as observed here. Thus, our findings highlight the possibility that mechanisms outside of parent emotion socialization and child ER may predict child anxiety risk. Of note, this effect emerged in models of supportive, but not non-supportive, emotion socialization. Perhaps in these latter models, direct relations between maternal emotion dysregulation and child anxiety were partially obstructed after accounting for the shared variance among maternal emotion dysregulation, non-supportive emotion socialization, and child emotion regulation.

There was greater evidence for hypothesized transmission of family emotion processes in models involving non-supportive emotion socialization. Mothers reporting greater emotion dysregulation reported more non-supportive emotion responses, though they were not more or less supportive to child emotions. Perhaps mothers’ maladaptive ER more readily engenders non-supportive, but not less supportive, parenting practices. There is some precedent for this, although the extant literature offers equivocal findings (Morelen et al., 2016; Raval et al., 2018). Still, present results do evidence the importance of maternal emotion dysregulation as a proximal predictor of non-supportive emotion responses, consistent with emotion-related parenting theory (Hajal & Paley, 2020). It remains unclear whether supportive or non-supportive responses are more important for child anxiety risk, as emotion socialization did not predict child anxiety.

Particularly strong paths emerged in the model with non-supportive emotion responses and caregiver-focused ER. Here, greater maternal emotion dysregulation predicted greater non-supportive emotion socialization, which in turn related to children’s greater caregiver-focused ER. Mothers’ emotion dysregulation and non-supportive responses were strongly related in the hypothesized direction; however, mothers’ more frequent non-supportive responses were associated with children’s greater use of caregiver-focused ER. Though this result lends support to the indirect transmission of ER, it also suggests that more non-supportive parenting is associated with a greater reliance on one’s caregiver. It may be that when learning from an emotionally dysregulated mother who responds non-supportively to emotions, children struggle to develop early independent ER. As such, they may seek out and rely on external ER from their mother (e.g., via caregiver-focused behaviors) at a higher rate. There is some evidence that children who show more ER attempts elicit more non-supportive responses (Premo & Kiel, 2014). Alternatively, as emotion socialization and child ER were measured at the same time-point, it may be that children who elicit more attention from their mothers naturally receive more non-supportive (and supportive) responses, merely because of children’s elicitation. Importantly, however, our supplementary analyses assessing directionality found that emotion socialization predicted child ER, but not vice versa, across time. Thus, perhaps a child-elicited-effects explanation is less likely.

Notably, the association between non-supportive emotion responses and children’s caregiver-focused ER was present only in full models (but not preliminary correlations), perhaps due to statistical suppression or inconsistent mediation. Aspects of non-supportive emotion responses independent from maternal emotion dysregulation may be controlled and warm, potentially discouraging child negative affect and reinforcing adaptive ER. This may be, in part, what has been captured in some studies showing that parents’ non-supportive responses are not necessarily detrimental to child psychological adjustment (e.g., Hooper et al., 2018; Raval et al., 2018). More broadly, perhaps the observed, positively-valenced, indirect relation reflects that mothers who are more emotionally dysregulated and prone to reacting non-supportively have children who are more concerned with and oriented to their mothers’ emotional reactions.⁴

Overall, serial mediation effects did not emerge. The 1-year time frame between study time-points may have attenuated interrelations, perhaps due to competing processes among variables in or outside the model. Further, the multimethod nature of this study likely decreased shared method variance. Indeed, most associations to and from observed child ER were relatively weak. Additionally, these ER behaviors may have been too context- or age-specific to be strongly related to mothers’ earlier behaviors and children’s later anxiety. Perhaps more global measures or profiles of children’s ER behaviors across contexts would provide a better representation of child ER. It should also be noted that an adolescent sample and cross-sectional design were utilized in one of the only other studies to test these theorized family emotion processes as predictors of child psychopathology (i.e., Buckholdt et al., 2014). Specifically, Buckholdt et al. (2014) found cross-sectional evidence for intergenerational transmission of emotion dysregulation from parent to adolescent, through parents’ non-supportive emotion responses and in relation to adolescent internalizing and externalizing symptoms. Our findings expand upon these results by using a longitudinal design and applying theorized models’ (i.e., Eisenberg et al., 1998; Morris et al., 2007) to early childhood processes. Further research should examine replicability of these results in other toddler samples.

We additionally tested models that accounted for maternal anxiety, given its status as a strong predictor of child anxiety (Murray et al., 2009). Theory emphasizes the centrality of emotion dysregulation to psychopathology (Aldao et al., 2010), and indeed, we found robust positive relations between mothers’ emotion dysregulation and anxiety symptoms. In preliminary models, maternal emotion dysregulation uniquely predicted child anxiety across time, above and beyond maternal anxiety. However, in the larger models of family emotion processes accounting for maternal anxiety, neither mothers’ emotion dysregulation nor their anxiety uniquely predicted child anxiety. This suggests that maternal anxiety may not be a primary or unique contributor to child anxiety when considering transdiagnostic emotion-related characteristics and behaviors. Substantial shared variance between maternal emotion dysregulation and anxiety, coupled with both constructs’ correlations with child inhibited temperament (which did predict child anxiety), likely prevented either from uniquely predicting child anxiety. Interestingly, mothers’ anxiety, controlling for their emotion dysregulation, predicted their non-supportive responses. Perhaps mothers’ anxious traits, rather than their dysregulated affect, more readily drive their non-supportive responses. Maternal and child characteristics likely interact in complex ways that unfold over time, the full understanding of which is beyond the scope of this paper. Still, our results do provide evidence that maternal emotion dysregulation’s relevance to models of family emotion processes predicting child anxiety is not entirely accounted for by maternal anxiety.

Limitations and Future Directions

The study was strengthened by its use of multiple measures, consideration of multiple transdiagnostic emotion processes and child anxiety predictors, and longitudinal design. Still, limitations exist. Like many multi-wave studies with families and young children, missing data across time-points was a concern. We utilized FIML as a gold-standard method for handling missing data, but examination of larger family samples with fewer missing time-points would bolster confidence in our conclusions. Relatedly, models did not account for baseline emotion socialization and child anxiety, and doing so may provide clarity into their change over time. Because we assessed a community sample, the extent to which (a) children in our sample met clinical threshold for anxiety and (b) these processes would replicate in a clinical sample remains unclear. We did not investigate how parents manage their emotions within the family context, which is specifically crucial for children’s emotional development and recommended for the field (e.g., Rutherford et al., 2015). Also, although our measures of emotion socialization collapsed across discrete emotions, parental strategies vary by type of emotion expressed (e.g., O’Neal & Magai, 2005). Responses to child worry may be most salient to anxiety risk. We did not find evidence that child ER predicted mothers’ responses to emotions, but child ER does likely involve bidirectional processes with parents (e.g., Premo & Kiel, 2014). Future work may test feedback loops in which child ER affects parents’ emotion responses. Finally, other contextual factors likely influence trajectories of emotion processes and child anxiety over time. Our sample was diverse in SES but mostly racially and ethnically homogenous. It is critical to understand how emotion socialization operates across families of diverse backgrounds and contexts, for example, by studying how parents respond to children’s emotion displays consequent to discrimination (Dunbar et al., 2017). Assessment of these racialized emotion socialization practices may offer unique insight into how ER norms are taught to and internalized in children who face greater adversity/threat and show different opportunities for resilience.

Conclusions and Implications

This study offers a unique contribution as a multi-wave, multi-method model of mothers’ and children’s emotion-related characteristics and behaviors in relation to child anxiety. Results partially support theorized roles of family emotion processes with transdiagnostic significance, as well as those specific to child anxiety risk, over and above effects of child inhibited temperament and maternal anxiety on child anxiety. Future work should further delineate sequences of family emotion processes and their interactions with other child anxiety risk factors, as well as explore mothers’ emotion dysregulation in early child anxiety interventions.

Data Availability:

The authors are not making a dataset publicly available as consent was not specifically sought from participants to do so. However, de-identified data will be distributed upon reasonable request.