Secure Attachment in Infancy Predicts Context-Dependent Emotion Expression in Middle Childhood

Abstract

Attachment security has been linked to healthy socioemotional development, but less is known about how secure attachment in infancy relates to emotional functioning in middle childhood, particularly across multiple contexts. The present study examined associations between secure attachment in infancy and children’s context-dependent emotion expression during a parent-child interaction at age nine (N = 78) among families with Child Protective Services involvement (i.e., children at risk for emotion dysregulation). Results indicated that children classified as securely attached in infancy exhibited less task-incongruent affect (i.e., less positive affect during a distressing discussion, less negative affect during a positive discussion) and a greater decrease in negative affect from a distressing discussion to a positive discussion than children classified as insecurely attached. In addition, secure children were rated as more appropriate in their emotion expression than insecure children. The present study highlights attachment as a promising intervention target for children at risk for emotion dysregulation.

Children first develop patterns of emotion expression and regulation in the context of the parent-child relationship (Ainsworth et al., 1978; Bowlby, 1969). In attachment theory, attachment classifications are hypothesized to correspond to specific patterns of emotion regulation (Cassidy, 1994). Empirical evidence has begun to support these theoretical propositions, particularly for infants and young children (Zimmer-Gembeck et al., 2017). In general, children who are securely attached to their parents have been found to exhibit more effective emotion regulation and to use more optimal coping strategies than children who are insecurely attached (Brumariu, 2015; Zimmer-Gembeck et al., 2017). Developmentally, children in middle childhood are expected to exhibit task-congruent affect (i.e., negative affect in a distressing task) in age-appropriate ways, minimize task-incongruent affect (i.e., positive affect in a distressing task), and adjust to changing task demands. However, although theory contends that infant attachment quality should predict appropriate and flexible emotion expression across development, most studies fail to measure emotion across varied affective contexts and longitudinally beyond early childhood. Further, associations between early attachment and emotional development are particularly important to examine among children most at risk for later problems regulating emotions, such as maltreated children (Cicchetti & Olsen, 2012). The present study examined secure (versus insecure) attachment in infancy as a predictor of children’s emotion expression in a two-part parent-child discussion in middle childhood among children at risk for maltreatment. Specifically, we investigated associations between infant attachment quality and children’s abilities to 1) openly and appropriately express task-congruent affect, and 2) adjust their emotion expression as task demands change.

Emotion Expression in Context

To further our understanding of the development of emotion, it is critical to study children’s emotion expression across different contexts. First, it allows us to understand whether children are able to exhibit task-congruent affect and minimize task-incongruent affect, and to express their emotion in age- and socially appropriate ways. For example, it is normative for children in middle childhood to signal clearly to their parents when they are distressed and in need of support by expressing their emotion with situationally and age-appropriate facial expressions, body language, and verbal statements. Expressing distress with parents would be indicative of emotion understanding and emotion flexibility – recognizing one’s own emotional experience and adaptively expressing negative affect in specific contexts to elicit support as needed (Cooke et al., 2016, 2019). Alternatively, they may suppress their emotion and avoid outwardly displaying their internal experience, or they may become behaviorally dysregulated and maximize the expression of their emotion. By measuring child emotion expression in a task specifically designed to elicit negative affect in the presence of the parent, we can interpret children’s task-congruent and -incongruent emotion expression in developmental context.

Importantly, children in middle childhood are also expected to recover from distress after receiving regulatory support or after the stressor is removed, whereas prolonged negative affect or failure to recover is associated with the development of disturbance and psychopathology (Coifman & Bonanno, 2010). Thus, the ability to adjust to changing contextual demands is also considered developmentally appropriate and a marker of healthy adaptation. By observing child emotion expression with their parent from a negatively valenced task to a positively valenced task, we can evaluate the extent to which children are able to recover from distress and change their emotion expression to match the current contextual demands.

Although emotion researchers have argued the theoretical importance of studying emotion expression across multiple contexts for several decades (Cole et al., 1994, 2004), there remains a paucity of empirical work manipulating context in experimental design (Aldao, 2013). Particularly relevant for developmental work, researchers have recommended incorporating ecologically valid social contexts in emotion research (Aldao et al., 2016). In the present study, we draw on these recommendations by observing child emotion expression across negatively valenced and positively valenced parent-child discussions. We also integrate attachment theory and its conceptualization of the development of emotion within the parent-child relationship. Attachment theory allows us to generate specific hypotheses and draw clear conceptual connections between infant attachment quality and child emotion expression with their parent later in development.

Attachment and Emotion

Infancy

According to attachment theory, the regulation of infant emotion is a core function of early caregiving relationships. From this perspective, infants are biologically motivated to maintain proximity to caregivers for protection and to manage stress and other needs (Bowlby, 1969, 1982; Sroufe, 1996). It is thought that the experience of distress activates the attachment system, motivating infants to seek or maintain proximity with the parent and obtain needed assistance in regulation. However, infants vary in behavioral strategies employed to maintain proximity with attachment figures. These strategies fall into reliably coded categories (Ainsworth et al., 1978). When the system works well, infants effectively signal parents they need nurturance, parents provide sensitive and nurturing care, and infants are adequately regulated (i.e., return to homeostasis). Adapted to caregiving behaviors over the first year, infant strategies enhance or diminish the intensity of emotional expression in response to distress. Thus, the development of patterns of emotion expression with parents is also a central component of attachment theory. In this way, attachment theory dovetails with seminal theory and empirical work in the development of emotion regulation literature, which highlight the interplay between infant or child emotion expression and parent responses (Calkins, 1994; Cole et al., 1994, 2003; Eisenberg et al., 2010; Morris et al., 2017).

Infant attachment quality is typically assessed in the Strange Situation Procedure (Ainsworth et al., 1978). This assessment involves a series of separations from and reunions with the parent, with infant behaviors in the reunion episodes particularly relevant for understanding attachment quality. Separations are inherently distressing for infants, thus eliciting attachment behaviors upon reunion. When distressed, infants classified as securely attached to their parents make clear bids to their parents for reassurance or comfort and tend to be easily soothed by their parents. When observed in the Strange Situation, these behaviors are thought to reflect the infants’ expectations about the availability and responsiveness of their parents. That is, these infants may have learned that when they are distressed, they can depend upon their parents to nurture them and help regulate negative feelings. Over time, these children openly express and effectively regulate a range of emotions (Cassidy, 1994). These children may also be better able to tolerate negative emotions than children with insecure attachments to their parents (Sroufe et al., 2005). Because they have experienced effective regulation in early relationships, they may evaluate strong emotions as manageable, and over time they will learn to regulate emotions increasingly independently (Kopp, 1982; Sroufe, 1996).

Infants who are not classified as securely attached to their parents may be classified as insecure-avoidant, insecure-resistant, or disorganized. In the Strange Situation, infants who are classified as avoidant typically minimize the outward expression of emotion, whereas infants who are classified as having resistant attachments exhibit heightened emotion expression (Ainsworth et al., 1978). Infants who are classified as having disorganized attachments demonstrate a breakdown in or lack of relational regulatory strategy during the Strange Situation (e.g., displaying contradictory behaviors; Main & Solomon, 1990). Each of these attachment classifications is thought to confer risk for emotional and behavioral problems as well as poor relationship quality with parents and peers later in development (Cassidy, 1994; Sroufe et al., 2005; Zimmer-Gembeck et al., 2017). Examining associations between attachment and emotional competence in contexts beyond the Strange Situation and across development is an important step in understanding how infant attachment quality might be linked to these psychosocial outcomes.

Middle Childhood

Although attachment behaviors are most frequently assessed in infancy or early childhood, children continue to rely on parents as attachment figures as they become older (Kerns & Brumariu, 2016). In middle childhood, parents are still typically the primary attachment figures; however, the goal of the attachment system is thought to shift from maintaining proximity to parents to relying on confidence in caregiver availability (i.e., felt security) to engage in the social world. Ideally, parent-child relationships become increasingly collaborative as children rely more on self-regulation strategies than on external regulation from parents, while still seeking parental support as needed (Skinner & Zimmer-Gembeck, 2007). At this stage of development, children who have secure attachment histories are expected to seek out their parents when they are distressed and to send a clear signal to their parents that they are distressed (i.e., openly express negative affect; Bretherton, 1993). In secure dyads, parents address children’s distress, help them regulate or resolve distress, and facilitate recovery from distress and the return to positive affect (Kerns & Brumariu, 2016). Although secure attachment has been linked to emotion understanding and use of adaptive coping strategies more broadly (Cooke et al., 2016; Kerns & Brumariu, 2016; Parrigon et al., 2015), less is known about how attachment experiences may be associated with affective expression in specific emotional contexts with or without parents, particularly in middle childhood.

Attachment and Emotion Expression in Context

A recent meta-analysis examined associations between attachment and child emotional competence across domains (Cooke et al., 2019). Within the range of emotion experience, the authors distinguished between global affect, or dispositional tendencies toward positive or negative affect, and elicited affect, or emotion expression in response to particular contextual stimuli. Meta-analytic findings indicated that secure attachment was significantly associated with less elicited negative affect than insecure attachment (Cooke et al., 2019). However, the authors note that the effect size and number of available studies were small. In addition, studies of elicited affect included a variety of measures, including questionnaires and observations of different types of tasks, some of which were parent-child interactions and some of which were not. It is likely that attachment quality would have differential effects on elicited affect depending on task characteristics such as whether the focus was a parent-child interaction and whether the eliciting task would be expected to activate the attachment system. The authors concluded that more research is needed in which affect is observed across multiple contexts to better understand these effects and to assess whether secure attachment may be associated with emotional flexibility (an ability to adapt emotion expression as contextual demands change). Of particular relevance to the present study, only a small subset of the studies identified measured elicited affect beyond infancy or early childhood, so it is not certain whether age at emotion observation may have been identified as a significant moderator if more studies in middle childhood had been available.

Emerging empirical evidence suggests that children with secure attachments show less avoidance of negative emotion expression with their parents than children with insecure attachments. One longitudinal study measured attachment quality in infancy using the Strange Situation, and when children were 12 years old asked them to complete several emotional challenges with their parents (Spangler & Zimmermann, 2014). Children who were classified as securely attached in infancy reported feeling angrier in an anger-induction challenge than children who were classified as insecurely attached in infancy. Children with histories of secure attachment were also observed to engage in better social emotion regulation, defined as expressing negative affect or seeking proximity to their parent during the anger induction, than children with histories of insecure attachment. To our knowledge, no other studies have examined the effects of infant attachment on elicited negative emotion in middle childhood. However, in a cross-sectional study in early childhood, 4-year-old children who were rated as more secure (measured with the Attachment Q Sort) exhibited less avoidance when discussing negative emotions with their parents than children who were rated as less secure (Waters et al., 2010).

In addition to open expression of negative affect, attachment security has been associated with affective recovery following a challenge among children in middle childhood. Specifically, higher attachment security (as measured by higher ratings of narrative coherence in the Child Attachment Interview) has been associated with greater reductions in negative affect (i.e., recovery) following a fear-potentiated startle paradigm among children in middle childhood than lower attachment security (Borelli et al., 2010). In addition, Abtahi and Kerns (2017) measured middle childhood attachment representations using a story stem assessment and asked children to report on their positive and negative affect throughout a social stressor. In this study, children’s stories were coded for the degree to which they reflected secure and insecure (ambivalent, avoidant, and disorganized) attachment patterns. Attachment security measured representationally was significantly associated with recovery of positive affect following the stressor, whereas higher insecure avoidance ratings were associated with less reactivity and recovery of negative affect (blunted responding), and higher insecure ambivalence ratings were associated with greater reactivity and recovery of negative affect (emotional lability; Abtahi & Kerns, 2017). Although these studies examined affective recovery outside of the parent-child context, results suggest that children with histories of secure attachment may exhibit emotion recovery following a stressor. Whether such recovery extends to the parent-child context is an empirical question.

Little is known about the effect of infant attachment quality on elicited positive affect. A recent meta-analysis failed to find an effect of secure attachment on elicited positive affect, and could not analyze the insecure subtypes because too few studies were available (Cooke et al., 2019). However, none of the studies identified included children older than two or three years of age. In addition, secure attachment was associated with more global positive affect than insecure attachment, and insecure-avoidant and disorganized attachment classifications were associated with reduced global positive affect (Cooke et al., 2019). Thus, if secure children generally experience more positive affect, including with their parents (Brumariu, 2015; Parrigon et al., 2015), it is plausible that they may exhibit greater positive affect with their parents in a positive discussion task in middle childhood.

The existing literature points to associations between attachment security measured behaviorally in infancy and through representations in middle childhood and children’s context-dependent emotion expression, but there are several important limitations. First, very few studies have been conducted with children in middle childhood, or with multiple contexts in the experimental design (for reviews and analyses across emotion-related constructs, see: Cooke et al., 2019; Parrigon et al., 2015). In addition, most studies reviewed have been cross-sectional rather than longitudinal. Further, most of the reviewed studies have relied on self-reports of affect (vulnerable to reporting bias) rather than observational coding. Direct observations of children’s emotion expression with their parents may provide important information about attachment processes and emotional development in middle childhood. Finally, these studies have not focused on emotional functioning of children at high risk for problems regulating emotions.

Maltreatment as a Risk Factor

Children who are maltreated are less likely to develop secure attachments and are more likely to develop insecure or disorganized attachments than children who are not maltreated (Baer & Martinez, 2006; Cyr et al., 2010). Theoretically, early experiences of frequent rejection could lead to avoidant attachment, experiences of inconsistent care may lead to resistant attachment, and abuse or frightening behaviors may lead to disorganized attachment (Ainsworth et al., 1978; Main & Solomon, 1990). Insecure or disorganized attachment and maltreatment are both risk factors for the development of psychopathology (Cicchetti & Olsen, 2012; Cicchetti & Toth, 2016; DeKlyen & Greenberg, 2008; Lyons-Ruth & Jacobvitz, 2016). One mechanism by which maltreatment may increase risk for psychopathology is via disruptions in the development of emotional competence in areas such as emotion regulation (Kim & Cicchetti, 2009) and emotion understanding (Rogosch et al., 1995). A recent meta-analysis indicated that maltreatment was associated with elevated self-reported and observed negative affect and decreased positive affect (Lavi et al., 2019). For example, teachers rate maltreated preschoolers higher on negative emotionality than non-maltreated preschoolers (Smith & Walden, 2001). Similarly, meta-analytic findings indicate that disorganized (versus organized) attachment is associated with more global negative affect and less global positive affect (Cooke et al., 2019). Further, in these analyses, insecure-avoidant attachment was also associated with less global positive affect, and insecure-resistant attachment was associated with more global and elicited negative affect than secure attachment (Cooke et al., 2019).

When emotion expression is observed in parent-child interaction tasks, maltreated children may exhibit more negative affect and less positive affect than non-maltreated children. In a parent-child cleanup interaction, neglected preschoolers exhibited significantly more observed negative affect than non-maltreated preschoolers (Koenig et al., 2000). However, physically abused preschoolers exhibited similar negative affect to non-maltreated preschoolers. The authors posited that neglected children at this age may be more likely to exhibit exaggerated emotional displays to obtain parental attention, whereas physically abused children may not. In another study, maltreated children in foster care and non-maltreated children between one and four years old completed a variety of parent-child interaction tasks with their biological mothers (Robinson et al., 2009). Maltreated children exhibited significantly more observed anger and less positive affect than non-maltreated children, with no differences found between maltreatment types.

Maltreatment poses exceptional challenges for children. Although these children appear to express elevated negative emotion and low positive emotion when distressed (Lavi et al., 2019), they may not turn to parents for support. For example, maltreated preschoolers have been found to display elevated levels of negative affect (Koenig et al., 2000; Robinson et al., 2009), and yet parents of maltreated preschoolers report that their children are less likely to seek support at home compared with reports from parents of non-maltreated preschoolers (Smith & Walden, 2001). By middle childhood, maltreated children may have learned that displaying negative affect is not an effective strategy to receive parental support. When interviewed, maltreated children in middle childhood report that they expect less maternal support and more conflict in response to their negative emotion expression than non-maltreated children (Shipman et al., 2005; Shipman & Zeman, 2001). Maltreated children in middle childhood also report being more likely to inhibit their negative emotion expression with parents than non-maltreated children (Shipman et al., 2005).

However, attachment security may be a critical protective factor to support the development of emotional competence for maltreated children. Despite experiences of maltreatment, children may still develop secure attachment relationships with their parents whether by receiving an attachment-based intervention or by other individual or contextual supports. It is plausible that in middle childhood, maltreated children with histories of secure attachment may be more willing to openly express emotion with their parents when they are distressed to signal that they need support than maltreated children with histories of insecure attachment. In addition, maltreated children with secure attachment histories may exhibit greater emotion recovery or ability to adapt to changing contexts than maltreated children with insecure attachment histories, given emerging literature on attachment security and affective recovery (Abtahi & Kerns, 2017). The present study examines the association between secure attachment in infancy and context-dependent emotion expression in middle childhood during a parent-child interaction among families with a history of maltreatment risk.

Present Study

Children who are securely attached are expected to openly and flexibly express their emotions as task demands change, and to recover from distress through effective self-regulation or seeking parental support (Cassidy, 1994; Cole et al., 1994). The present study investigated whether secure attachment in infancy predicted children’s emotion expression during a two-part parent-child discussion in middle childhood. In the first part, dyads discussed a recent event that was upsetting to the child (distress discussion). In the second part, dyads planned the perfect day trip for the child (positive discussion). Children’s positive affect, negative affect, and appropriateness of emotion expression were coded for each task.

The present study aims were to investigate the association between infant attachment quality and 1) open and appropriate emotion expression within tasks and 2) flexible adaptation of emotion expression between tasks as demands changed. For the first aim, we hypothesized that children classified as securely attached in infancy would exhibit more task-congruent affect (i.e., negative affect in the distress discussion and more positive affect in the positive discussion) in appropriate ways and less task-incongruent affect than insecure children. For the second aim, we hypothesized that secure children would display a greater decrease in negative affect and increase in positive affect transitioning from the distressing discussion to the positive discussion, indicative of greater affective recovery or flexibility, than insecure children.

Method

Participants

Participants were children (46% female) enrolled in a larger randomized clinical trial of the Attachment and Biobehavioral Catch-up (ABC) intervention (Dozier & Bernard, 2019). Families were referred to the study from Child Protective Services (CPS) due to risk of maltreatment when children were infants. Families were then randomly assigned to receive ABC or a control intervention. Each intervention consisted of ten weekly home visiting sessions. Following the intervention, infant attachment quality was assessed. When children were in middle childhood, families were invited to participate in the middle childhood wave of assessments. Middle childhood data from the present study were collected when children were about nine years old (M = 9.47, SD = .35). Procedures were approved by the University of Delaware Institutional Review Board (Protocol 547621).

Of the children with available coding data (n = 98), twenty children were missing infant attachment data, resulting in the analytic sample size of 78 children. Of these children, eight children completed the assessment at age 9 with a caregiver other than the one who was part of the attachment assessment in infancy. Of these, three completed the assessment at age 9 with a biological parent, two with a biological grandparent, two with an adoptive parent, and one with a foster parent.

The majority of children in this sample were reported to be Black (69.2%) or more than one race (16.7%), and 14.1% were White. About one fifth of the children were reported to be Latino/Latina/Latinx (19.2%). In addition, the majority of parents (67.9%) reported receiving welfare or government assistance at the time of the middle childhood assessment. Parents reported an average household income of about $20,000 (M = $22,575, SD = $16,917). Finally, one third of the parents reported not completing high school (35.1%), and half of the parents reported that they had completed high school or received a GED (49.4%).

Procedure

Infant Attachment Quality

Infant attachment quality was assessed using the Strange Situation Procedure (Ainsworth et al., 1978) when children were about 20 months old (M = 20.16, SD = 6.24). This 24-minute procedure consists of two infant separations from and reunions with the primary caregiver, with infant behaviors during the reunion episodes particularly relevant to the coding of attachment quality.

Children classified as securely attached with their parents seek proximity to their parents and are easily soothed by their parents when they are distressed. Children in relationships not classified as secure may be categorized as avoidant, resistant, or disorganized. Children who are classified as insecure-avoidant typically do not seek proximity to their parents when distressed and may turn away from their parents. Children who are classified as insecure-resistant are unable to be soothed by their parents when distressed and exhibit a combination of proximity seeking and resistance behaviors. Children who are classified as disorganized demonstrate a breakdown in strategy for organizing their attachment behaviors (e.g., disoriented wandering, contradictory behaviors, freezing; Main & Solomon, 1990).

Strange Situation videos were coded by an expert coder, and one third of the videos were double coded by a second expert coder (k = .76). Disagreements were resolved by conference. In the present sample, 47.4% of children were classified as securely attached to their parents. Of the remaining infants, 32 were classified as disorganized, 8 were classified as avoidant, and 1 was classified as resistant. For the present study, a dichotomous secure attachment variable was created by scoring infants with a secure attachment classification as 1, and scoring infants with avoidant, resistant, or disorganized attachment classifications as 0.

Parent-child Interaction

When children were about nine years old (M = 9.44, SD = .35), they completed a video-recorded parent-child interaction task. The task consisted of two discussions – the first discussion is referred to as the distress discussion, and the second discussion is referred to as the positive discussion. In the distress discussion, children were asked to talk to their parents for eight minutes about an event or situation that had recently upset them. Prior to this discussion, children were privately interviewed by research assistants to select an event or situation that was primarily peer-related and was not a source of conflict with their parents. This discussion was designed to elicit negative affect and provide an opportunity for children to seek support and parents to provide it. In the positive discussion, parent-child dyads were asked to spend five minutes planning the perfect day for the child. This task was designed to allow for emotional recovery, and children were expected to exhibit more positive affect during this discussion than during the first discussion.

Children’s emotion expression and behavior were coded during each of the two discussions separately. Codes were adapted from the Family Study Attachment Coding Manual for children in middle childhood (Miller et al., 2019) and included ratings of children’s positive affect, negative affect, and appropriate expression of affect (both positive and negative affect were considered). Positive and negative affect ratings each ranged from one to five. A score of one indicated primarily neutral affect, and five indicated that the child expressed either positive or negative affect at high frequency and intensity. Ratings of appropriate expression of affect ranged from one to four, from “not adaptive” to “definitely adaptive.” A score of one indicated that the child was totally inappropriate in expressing emotions, either expressing excessive emotion or completely withdrawn. It was also considered an inappropriate expression of affect if children exhibited a notable mismatch between the tone of the conversation and their emotion expression (e.g., joking and laughing while discussing a recent death of a family member). A score of four indicated that the child openly expressed emotions, exhibited a range of emotions as appropriate, and freely disclosed thoughts or feelings to the parent.

All videos were double coded by trained coders and codes exhibited acceptable to good reliability (ICCs ranged from .68 to .85). Coders met weekly to discuss disagreements and resolved disagreements of greater than one point by conference. Only pre-conferenced codes were used in the calculation of reliability. For the present study, final codes consisted of the average of the two coders’ scores, or the conferenced code when the coders disagreed by greater than one point.

Analytic Plan

The present study aims were to investigate the association between infant attachment quality and 1) open and appropriate emotion expression within tasks and 2) flexible adaptation of emotion expression between tasks as demands changed. To address the first aim, t-tests were planned to examine associations between attachment classification and each emotion code in each discussion type. Secure attachment was expected to predict greater task-congruent affect (negative affect in the distress discussion, positive affect in the positive discussion), less task-incongruent affect (positive affect in the distress discussion, negative affect in the positive discussion), and more appropriate emotion expression.

To address the second aim, three repeated measures analyses of variance (ANOVAs) were planned (i.e., positive affect, negative affect, and appropriate expression of affect). An interaction of task type and attachment classification within subjects would indicate that the effect of attachment on emotion expression is dependent on task demands. Finally, regression analyses were planned to examine associations between infant attachment classification and change in emotion expression – regressing emotion expression in the positive discussion on emotion expression during the distress discussion. Secure attachment was expected to predict greater affective recovery or flexibility (decrease in negative affect and increase in positive affect from the distress discussion to the positive discussion), and more appropriate emotion expression across tasks than insecure attachment. Data were analyzed in SPSS version 26.

Results

Preliminary Analyses

Child gender, race (White/non-White), ethnicity (Latino/a/x or not Latino/a/x), age at the middle childhood assessment, and intervention condition were considered as covariates. In the present dataset, the ABC intervention was not significantly associated with infant secure attachment (χ² [1, N = 78] = 2.40, p = .12) or any of the behavioral codes, and so intervention condition was excluded from the models. In the full data set including all children, the intervention was associated with increased rates of secure attachment (Bernard et al., 2012). One reason for this discrepancy may be reduced power to detect intervention effects in this subsample. The present subsample represents 56% of the original group who completed the attachment assessment in infancy. The proportion of children who received each intervention was not significantly different between the original group who completed the attachment assessment and the present subsample (χ² [1, N = 139] = 0.03, p = .50). In this subsample, 35 children received the ABC intervention in infancy and 43 children received the control intervention.

Attrition analyses did not detect significant differences in demographic variables at enrollment (parent age, family income, parent education level, marital status, parent gender, parent race/ethnicity, child gender, and child race/ethnicity) between the present subsample and the original sample who completed the attachment assessment in infancy. In addition, there were no significant differences in demographic variables at enrollment between children who had attachment quality and follow-up data at age nine and children who were missing attachment quality or follow-up data at age nine.

In the current study, child gender and child ethnicity were significantly correlated with positive affect in the positive discussion, such that girls exhibited more positive affect than boys (r = −.23, p = .04) and Latino/a/x children exhibited more positive affect than children who were not identified as Latino/a/x (r = .23, p = .047). In addition, White children expressed significantly more negative affect in the positive discussion than non-White children (r = .26, p = .02). Notably, there were only seven children identified as White in the present sample and all seven were classified as insecurely attached in infancy. Thus, child gender and ethnicity were included in models of positive affect, and child race was included in models of negative affect. Correlations between study variables are presented in Table 1 and means and standard deviations of behavioral ratings by attachment classification are presented in Table 2.

Table 1.

Correlations of study variables and covariates.

	1	2	3	4	5	6	7	8	9	10	11	12
1. Intervention condition	--
2. Attachment security	.18	--
3. Gender	−.04	−.10	--
4. Race	−.10	−.30**	.11	--
5. Ethnicity	.02	−.20†	.06	.07	--
6. Age	.002	−.05	.22†	.19	−.06	--
7. Positive affect (distress)	.03	−.27*	−.15	.10	.10	.16	--
8. Positive affect (positive)	−.17	−.04	−.23*	−.11	.23*	−.09	.49***	--
9. Negative affect (distress)	.11	.08	−.04	−.09	−.21†	−.04	−.53***	−.50***	--
10. Negative affect (positive)	−.03	−.33**	.09	.26*	−.02	−.04	−.01	−.45***	.38**	--
11. Appropriateness (distress)	−.22†	.08	−.12	.21†	−.07	.13	.29**	.53***	−.55***	−.32**	--
12. Appropriateness (positive)	−.15	.35**	−.09	−.17	−.15	.08	.04	.35**	−.29*	−.64***	.45***	--

^†p < .1;

^*p < .05;

^**p < .01;

^***p < .001.

Note: For the Intervention condition variable, the Attachment and Biobehavioral Catch-up group was dummy coded as “1” and the control group was dummy coded as “0.” For the Attachment security variable, children classified as secure were dummy coded as “1” and children classified as insecure were dummy coded as “0.” For the Gender variable, male participants were dummy coded as “1” and female participants were dummy coded as “0.” For the Race variable, White participants were dummy coded as “1” and all other participants were coded as “0.” For the Ethnicity variable, Latino/a/x participants were dummy coded as “1” and non-Latino/a/x participants were dummy coded as “0.”

Table 2.

Means and standard deviations of study variables by attachment classification.

	Secure (n = 37)		Insecure (n = 41)		Avoidant (n = 8)		Resistant (n = 1)		Disorganized (n = 32)
	M	SD	M	SD	M	SD	M	SD	M	SD
Positive affect (distress)	1.97	0.90	2.55	1.15	1.94	0.41	3.75	--	2.66	1.12
Positive affect (positive)	3.05	0.74	3.12	1.11	2.83	1.27	4.00	--	3.16	1.08
Negative affect (distress)	2.75	1.02	2.59	1.11	2.91	1.22	2.00	--	2.52	1.10
Negative affect (positive)	1.64	0.60	2.19	0.93	2.09	0.96	1.25	--	2.24	0.93
Appropriateness (distress)	3.15	0.74	3.01	0.96	2.78	1.04	3.25	--	3.06	0.97
Appropriateness (positive)	3.42	0.47	2.94	0.78	3.25	0.48	1.75	--	2.9	0.82

Note: For this table and primary analyses, “Insecure” attachment included avoidant, resistant, and disorganized groups.

Primary Analyses

Does Secure Attachment Predict Emotion Expression Within Tasks?

First, t-tests were conducted to examine effects of secure attachment on positive affect, negative affect, and appropriate affect in each discussion type.

Analyses of emotion expression in the distress discussion revealed that children who were classified as securely attached in infancy exhibited significantly less positive affect (M = 1.97, SD = 0.90) than children who were classified as insecurely attached (M = 2.55, SD = 1.15; t(74.70) = −2.48, p = .02, 95% CI of the difference: −1.04, −0.11). However, no significant difference in expression of negative affect was detected between children who were classified as securely attached (M = 2.75, SD = 1.02) and children who were classified as insecurely attached (M = 2.59, SD = 1.11; t(76) = 0.68, p = .50, 95% CI of the difference: −0.32, 0.65). A significant difference in appropriateness was also not detected in this discussion between children classified as securely attached (M = 3.15, SD = 0.74) and children who were classified as insecurely attached (M = 3.01, SD = 0.96; t(74.27) = 0.74, p = .47, 95% CI of the difference: −0.24, 0.53).

In the positive discussion, children who were classified as securely attached displayed significantly less negative affect (M = 1.64, SD = 0.60) than children who were classified as insecurely attached (M = 2.20, SD = 0.93; t(69.07) = −3.16, p = .002, 95% CI of the difference: −0.90, −0.20). However, a significant difference was not detected in positive affect between children who were classified as securely attached (M = 3.05, SD = 0.74) and children who were classified as insecurely attached (M = 3.12, SD = 1.11; t(70.08) = −0.33, p = .75, 95% CI of the difference: −0.49, 0.35). In this task, children who were classified as securely attached in infancy were significantly more appropriate in their affective expression (M = 3.42, SD = 0.47) than children who were insecurely attached (M = 2.94, SD = 0.78; t(66.41) = 3.32, p = .001, 95% CI: 0.19, 0.77).

Overall, these analyses demonstrated that secure attachment was associated with less task-incongruent affect across tasks and more appropriate emotion expression than insecure attachment in the positive discussion. However, secure attachment did not predict more task-congruent affect than insecure attachment in either discussion type.

Does Secure Attachment Predict Emotion Expression Between Tasks?

Positive Affect.

In the repeated measures ANOVA model of positive affect, there was a significant within-subjects effect of task type (F(1,74) = 20.37, p < .001, η_p² = .22), reflecting that on average, children expressed more positive affect in the positive discussion (M = 3.07, SD = 0.91) than in the distress discussion (M = 2.15, SD = 1.03; model estimated mean difference: 0.82, 95% CI: 0.60, 1.05). Regarding between-subjects effects, gender was a significant predictor of positive affect expression (F(1,74) = 4.94, p = .03, η_p² = .06), and estimated marginal means indicated that girls expressed more positive affect (M = 3.07, SE = 0.19) than boys (M = 2.43, SE = 0.18). A main effect of attachment classification on overall positive affect was not detected (F(1,74) = 2.52, p = .12, η_p² = .03; model estimated mean difference: 0.31, 95% CI: −0.08, 0.69), but there was a significant interaction between task type and attachment classification (F(1,74) = 5.94, p = .02, η_p² = .07). Significant interactions were not observed for task and child ethnicity or task and child gender. See Figure 1 for a depiction of the estimated marginal means of positive affect by task and attachment classification (Insecure Task 1: M = 2.56, SE = 0.16; Insecure Task 2: M = 3.10, SE = 0.14; Secure Task 1: M = 1.97, SE = 0.17; Secure Task 2: M = 3.07, SE = 0.15). Taken together, results indicated that the effect of attachment quality on positive affect depended on task type, such that secure attachment was associated with less positive affect in the distress discussion than insecure attachment, but not in the positive discussion.

Figure 1.

Model estimated marginal means are presented with standard errors.

To further examine the association between infant attachment classification and change in positive affect expression from the distress discussion to the positive discussion, positive affect in the positive discussion was regressed on positive affect in the distress discussion and secure attachment, controlling for child ethnicity and child gender. In this model, secure attachment was not a significant predictor of positive affect in the positive discussion (β = .12, p = .26, 95% CI for B: −0.17, 0.61), and neither was child gender (β = −.17, p = .10, 95% CI for B: −0.68, 0.06). Positive affect in the distress discussion was positively associated with positive affect in the positive discussion (β = .47, p < .001, 95% CI for B: 0.24, 0.60) and Latino/a/x ethnicity was positively associated with positive affect in the positive discussion (β = .21, p = .03, 95% CI for B: 0.04, 0.98).

Negative Affect.

In the repeated measures ANOVA model of negative affect, there was again a significant within-subjects effect of task type (F(1, 75) = 47.72, p < .001, η_p² = .39), indicating that on average, children expressed more negative affect in the distress discussion (M = 2.69, SD = 1.07) than in the positive discussion (M = 1.93, SD = 0.84; model estimated mean difference: 0.75, 95% CI: 0.53, 0.99). No between-subjects effects were significant, and the interactions between task type and race were not significant. However, a significant interaction effect was observed for task type and attachment classification (F(1,75) = 6.05, p = .02, η_p² = .08). See Figure 2 for a depiction of the estimated marginal means of negative affect by task and attachment classification (Insecure Task 1: M = 2.61, SE = 0.17; Insecure Task 2: M = 2.15, SE = 0.13; Secure Task 1: M = 2.73, SE = 0.18; Secure Task 2: M = 1.68, SE = 0.13). Results indicated that the effect of secure attachment on negative affect depended on task type, such that secure attachment was associated with less negative affect in the positive discussion than insecure attachment, but not in the distress discussion.

Figure 2.

Model estimated marginal means are presented with standard errors.

To further examine the association between infant attachment classification and change in negative affect expression from the distress discussion to the positive discussion, negative affect in the positive discussion was regressed on negative affect in the distress discussion and secure attachment, controlling for child race. In this model, secure attachment was a significant predictor of negative affect in the positive discussion (β = −.31, p = .003, 95% CI for B: −0.84, 0.17), such that children who were classified as securely attached in infancy exhibited greater decreases in negative affect from the distress discussion to the positive discussion than children who were classified as insecurely attached in infancy. Negative affect in the distress discussion was also associated with negative affect in the positive discussion (β = .42, p < .001, 95% CI for B: 0.18, 0.48) and White race was associated with negative affect in the positive discussion (β = .21, p = .046, 95% CI for B: 0.01, 1.18).

Appropriateness of Affect.

In the repeated measures ANOVA model of appropriate affective expression, a significant within-subjects effect of task type was not detected (F(1, 76) = 1.21, p = .28, η_p² = .04). That is, the appropriateness of children’s affective expression did not differ significantly between the distress discussion (M = 3.09, SD = 0.82) and the positive discussion (M = 3.12, SD = 0.67; model estimated mean difference: −0.10, 95% CI: −0.28, 0.08). However, a significant main effect of attachment classification emerged (F(1, 76) = 4.46, p = .04, η_p² = .06) such that children who were securely attached in infancy exhibited more appropriate emotional expression in middle childhood than children who were insecurely attached (model estimated mean difference: 0.31, 95% CI: 0.02, 0.61). In addition, there was a marginally significant interaction between attachment classification and task type (F(1, 76) = 3.34, p = .07, η_p² = .04). See Figure 3 for a depiction of the estimated marginal means of appropriateness of emotion expression by task and attachment classification (Insecure Task 1: M = 3.01, SE = 0.14; Insecure Task 2: M = 2.94, SE = 0.10; Secure Task 1: M = 3.15, SE = 0.14; Secure Task 2: M = 3.42, SE = 0.11).

Figure 3.

Model estimated marginal means are presented with standard errors.

Discussion

Results of the present study indicate that secure attachment in infancy predicts child context-dependent emotion expression during a discussion with their parent eight years later. Specifically, children were expected to exhibit task-congruent affect in age- and situationally-appropriate ways, and minimize task-incongruent affect, in a distress discussion and a subsequent positive discussion (Aim 1). In the present study, secure attachment was associated with less task-incongruent affect than insecure attachment within each discussion type. That is, in a distressing parent-child discussion task, children classified as securely attached in infancy exhibited significantly less positive affect than children classified as insecurely attached. In the positive parent-child discussion task following the distressing discussion, children classified as securely attached exhibited significantly less negative affect and more appropriate emotion expression than children classified as insecurely attached. Contrary to study hypotheses, attachment effects were not observed for task-congruent affect (i.e., negative affect in the distress discussion, positive affect in the positive discussion). In addition, children were expected to display decreases in negative affect and increases in positive affect from the distress discussion to the positive discussion, which would indicate emotion flexibility or recovery – adapting to changing contextual demands (Aim 2). Children classified as securely attached exhibited a greater decrease in negative affect from the distressing discussion to the positive discussion than children classified as insecurely attached, but there was not a significant effect of attachment quality on change in positive affect.

These findings have important implications for understanding the previously observed links between infant attachment and later psychosocial outcomes for children. For example, one mechanism by which infant attachment may contribute to healthy emotional development is through experiences of successful dyadic regulation in caregiving relationships that build self-regulatory capacities. Children in middle childhood are expected to respond to a given situation with the appropriate emotion expression and at the appropriate intensity for the context (Parrigon et al., 2015). When upset, it is appropriate for children to express distress, clearly seek support from parents, and ideally recover from distress. The present study provides a window into the way these interactions may be supporting children’s developing emotion regulation abilities in middle childhood. First, children with histories of secure attachment exhibited less task-incongruent affect and were rated as significantly more appropriate in their emotion expression than children with histories of insecure attachment. Notably, the “appropriateness” code in the present study taps task-congruent affective expression as well as age- and situationally-appropriate behaviors. In addition, secure attachment was associated with greater decreases in negative affect from the distress discussion to the positive discussion. These findings support the idea that secure attachment promotes developmental competence in emotion expression – on average, children with histories of secure attachment seemed to be able to appraise situations with changing demands (perhaps reflecting emotion understanding; Cooke et al., 2016) and modulate the valence and behavioral expression of their affect. The present study also demonstrates how secure attachment in infancy may serve as a protective factor for maltreated children, who are at elevated risk for emotion dysregulation. Perhaps children with histories of secure attachment were able to reduce their negative affect from the distress discussion to the positive discussion in part because the parent-child relationship was functioning to help children process distress and return to baseline.

Children in middle childhood are expected to seek out their parents for support when they are distressed, but experiences of maltreatment and/or disrupted attachment may interfere with this process. In the current study, maltreated children with histories of secure attachment and maltreated children with histories of insecure attachment exhibited similar levels of negative affect with their parents during a distressing discussion. This finding was in contrast to our predictions that secure attachment would be associated with greater negative affect in the distress discussion than insecure attachment. However, insecure attachment was associated with greater expression of positive affect during a distressing discussion than secure attachment. This finding may indicate that children with histories of insecure attachment were unwilling to engage in the distressing discussion as long or as fully as children with histories of secure attachment. For example, insecure children may have changed topics to avoid continuing a distressing discussion. Thus, results are mixed with respect to the idea that secure attachment leads to more open expression of negative emotion with parents. Measuring discrete emotions may be better suited to testing this hypothesis. For example, by measuring discrete emotions like anger and sadness, we may find that insecure children are expressing anger and frustration toward their parents, and secure children are expressing sadness about an event to seek support. However, by only measuring negative affect, both of those scenarios could result in similar negative affect scores. The present study is limited in that a non-maltreated comparison sample was not available. It may be that associations between attachment security and emotion expression with parents depends on maltreatment history, such that secure attachment in infancy may not be sufficient to promote open expression of negative affect with parents among maltreated children. Examining attachment processes in both high-risk and low-risk samples is important to more comprehensively test attachment theory in this regard.

In addition, theory and empirical evidence indicate that secure attachment may be associated with greater emotional flexibility or recovery from stressors than insecure attachment. The current study is largely consistent with this hypothesis, as secure attachment was associated with a greater decrease in negative affect from the distressing discussion to the positive discussion than insecure attachment. Results indicate that secure children may be better able to down-regulate their distress following a challenge with their parents than insecure children, or that parent-child processes in secure dyads are more effective at helping children down-regulate distress than processes in insecure dyads. However, secure attachment was not associated with change in positive affect from the distress discussion to the positive discussion. In contrast, Abtahi and Kerns (2017) found that attachment security was associated with recovery of positive affect following a social stressor among children in middle childhood. There are several differences between the current study and that of Abtahi and Kerns (2017) which may help explain this discrepancy. First, the current study involved two parent-child discussions rather than a social stressor and then a removal of that stressful stimulus. It may be that children who complete emotional tasks with their parents experience less distress than children who complete stressful tasks without their parents. In this case, attachment effects on affective recovery may only emerge when the elicited negative affect is strong enough - the presence of a parent may prevent children from experiencing strong negative affect, at least for the task in the present study. Future research would be strengthened by examining child affect with and without the parent to examine the impact of parental presence. Next, the current study measured attachment quality in infancy rather than attachment representations in middle childhood. Attachment representations in middle childhood are more proximal in time to children’s emotional experiences measured in the lab, and so effects may be more likely to emerge. Third, the current study observed affect during the task rather than relying on child self-report. More research is needed to determine the conditions in which attachment security may predict emotion flexibility or recovery.

It was surprising that attachment did not significantly predict children’s task-congruent emotion expression. Several theoretical reasons for this have been outlined already – perhaps insecure children were expressing negative affect (e.g., anger) in the distress discussion but were not appropriately seeking support, or perhaps the effect of adversity on children’s comfort expressing negative affect with their parents was too strong to be buffered by secure attachment. Alternatively, perhaps present study findings are best explained by attachment effects on emotion understanding (Cooke et al., 2016) – children with histories of insecure attachments may have a deficit in understanding the emotional context of a situation, thus exhibiting greater task-incongruent affect than children with histories of secure attachments. Another explanation relates to collapsing the subtypes of insecure attachment (insecure-avoidant, insecure-resistant, and disorganized). Attachment theory and prior literature have linked avoidant attachment to blunted emotion expression and resistant attachment to exaggerated emotion expression, so collapsing these groups may obscure effects (Cooke et al., 2019). Although these classifications have been observed consistently across samples, the proportion of children who fall into each category is often small, making it difficult for researchers to examine differences among the insecure and disorganized classifications. Because the present study included a disproportionately large number of children who were classified as disorganized in infancy relative to insecure-avoidant and insecure-resistant subtypes, it is possible that results primarily reflect differences between secure and disorganized attachment. However, follow-up analyses with secure, avoidant, and disorganized groups did not reveal significant differences in positive affect, negative affect, or appropriateness of emotion expression. Means and standard deviations of emotion codes across subtypes are presented in Table 2, which may help readers generate hypotheses about differential effects of the insecure subtypes and disorganized attachment on emotional development. In addition, little is known about the effect of disorganized attachment on emotion expression. If disorganized infants demonstrate a breakdown in strategy when distressed, could they exhibit both blunted and exaggerated emotion expression within one context? More nuanced coding of emotion constructs could improve understanding of the effect of disorganized attachment in particular on elicited emotion with and without parents. Although a continuous rating of disorganization was not available in the present dataset, including continuous disorganization in future research may help elucidate potential effects. Future research should also aim to include a larger sample than used here, perhaps varying in risk for maltreatment, to produce adequately sized subgroups for more fine-grained analyses.

In addition, observed parent behavior during the discussions was not included in the present study. Future research may include parent behavior to better understand the dyadic processes occurring in emotional parent-child interactions. Another limitation is the lack of an emotionally neutral discussion task before or after the more emotional discussions; because both discussion types were emotionally evocative, we were unable to test infant attachment as a predictor of “baseline” emotion expression in a neutral parent-child interaction context or as a predictor of recovery to baseline. Finally, the present study did not include measurements of children’s subjective experiences of emotion or physiological reactivity. Given literature on associations between attachment security and emotional coherence, it is plausible that secure attachment may be associated with greater coherence of multimethod indicators of emotional experience than insecure attachment. For example, avoidant or dismissing attachment has been associated with reduced outward expression of emotion relative to physiological reactivity (Borelli et al., 2014; Roisman, 2007; Zelenko et al., 2005). In addition, it would be interesting to learn whether children’s subjective reports of emotion experience converge with or diverge from observed emotion expression. Given literature on maltreatment and attachment effects on emotional competence in general and emotion understanding in particular, one might hypothesize that secure attachment would be associated with greater coherence of self-reported emotion and observed emotion, indicative of greater emotion understanding, than insecure attachment. Further, qualitative interviews following an activity like the parent-child discussion task in the present study would be interesting to explore whether insecure children are able to articulate if they felt comfortable expressing difficult feelings with their parents or not, and whether observed emotional expression is consistent with child report of the emotions expressed.

The present study has several important strengths. First, this study used a gold standard measure of attachment in infancy, the Strange Situation, as a longitudinal predictor of children’s emotional functioning in middle childhood. Examining these longitudinal associations is critical to understanding how relations between attachment and emotion may unfold over time. Another strength of the study is the use of observational coding of emotion rather than self-report. In addition, emotion expression was measured during specific emotional discussion types within a parent-child interaction, increasing interpretability and relevance to attachment theory. The measurement of affect and appropriateness of affective expression in two ecologically valid and attachment-relevant contexts allowed us to extend previous work which has examined attachment effects on global affect or single contexts. For researchers and clinicians, it is of great interest to understand the effects of attachment on complex emotional experiences that are known to depend on many contextual factors. Further, this study measured attachment and emotion in the context of maltreatment, which is an established risk factor for the development of emotion dysregulation and psychopathology (Cicchetti & Olsen, 2012; Kim & Cicchetti, 2009).

Conclusion

The present study empirically demonstrated links between infant attachment quality and context-dependent emotion expression in middle childhood. Specifically, nine-year-old children who were classified as securely attached in infancy exhibited less positive affect in a distressing parent-child discussion, as well as less negative affect and more appropriate emotion expression in a positive parent-child discussion, than children classified as insecurely attached. In addition, secure children exhibited greater decreases in negative affect from the distressing discussion to the positive discussion than insecure children. Findings support the importance of promoting secure attachment among dyads at risk for maltreatment to promote healthy emotion expression and regulatory abilities. Additional avenues for intervention may include emotion coaching for maltreated children or children with histories of insecure attachment to increase skills in appropriateness of emotion expression and emotion regulation in the context of parent-child interactions, or with maltreating parents to increase their abilities to respond with nurturance when children express negative affect.