The recognition, understanding, and regulation of emotions is important for healthy development and impairments in these processes are associated with the risk of internalizing and externalizing problems (e.g., Compas et al., Reference Compas, Jaser, Bettis, Watson, Gruhn, Dunbar, Williams and Thigpen2017; Eisenberg et al., Reference Eisenberg, Spinrad and Eggum2010; Houben et al., Reference Houben, Van Den Noortgate and Kuppens2015; Southam-Gerow & Kendall, Reference Southam-Gerow and Kendall2002). Although emotions are experienced within the individual, they often take place in and are shaped by interpersonal relationships. Parent–child relationships are a primary context in which children learn about the nature of emotions and acquire the skills needed to regulate their emotions, and emotional exchanges are a central feature of parent–child relationships (Tan et al., Reference Tan, Oppenheimer, Ladouceur, Butterfield and Silk2020). To a great extent, learning about emotions and the development of emotion regulation skills takes place in the context of direct, face-to-face interactions between parents and their children. In dyadic interactions, parents express their own emotions, react to the emotions of their children, and sometimes discuss the nature of emotions with their children (Eisenberg et al., Reference Eisenberg, Cumberland and Spinrad1998). These processes are present in the earliest phases of development and continue in important ways throughout adolescence (Bariola et al., Reference Bariola, Gullone and Hughes2011).

Children and adolescents experience and express positive and negative affect at levels similar to their parents, both across development and during discrete interactions (Halberstadt & Eaton, Reference Halberstadt and Eaton2002). For example, through observational coding of dyadic play interactions across 4 years, Sallquist et al. (Reference Sallquist, Eisenberg, Spinrad, Gaertner, Eggum and Zhou2010) found rank-order stability in the expression of positive affect in mothers and their young children. Describing mean (i.e., average) levels of parent and child affect provides insight into one aspect of the affective content of dyadic interactions (Lunkenheimer et al., Reference Lunkenheimer, Olson, Hollenstein, Sameroff and Winter2011; Southam-Gerow & Kendall, Reference Southam-Gerow and Kendall2002; Suveg et al., Reference Suveg, Southam–Gerow, Goodman and Kendall2007). But what about their structure? In a treatment study of aggressive youth, De Rubeis and Granic (Reference De Rubeis and Granic2012) found changes in mean affect during parent–child dyadic interactions from pretreatment to posttreatment to be unrelated to youth posttreatment externalizing symptoms. However, parent–child dyads with greater affective synchrony during interactions showed lower pretreatment levels of externalizing symptoms and demonstrated greater treatment gains compared to dysregulated dyads. In the current study, mean-level affect informs the content of parent–child interactions and affective synchrony provides insight into an important aspect of their dynamic structure. Examining both affective content and structure provides a more holistic picture of complex dyadic interactions and the best opportunity for elucidating the ways in which short-term affective exchanges relate to long-term youth emotional and behavioral functioning.

Understanding how systems change (and remain stable) is central to developmental psychopathology (Granic & Hollenstein, Reference Granic and Hollenstein2003), and extant research underscores the importance of examining the ways in which dyadic partners’ emotions are linked across time (Butler, Reference Butler2011; Kuppens et al., Reference Kuppens, Oravecz and Tuerlinckx2010). Myriad terms have been used to describe this process (e.g., emotional or affective contingency, attunement, coregulation, reciprocity, responsiveness; Bornstein, Reference Bornstein, Legerstee, Haley and Bornstein2013), and various statistical methods have been used to quantify moment-to-moment interpersonal coordination (e.g., linear regression, time series analysis, multilevel modeling, conditional probabilities; Butler, Reference Butler2011). In the present study, we use the term affective synchrony to reflect the covariation of parent and adolescent affect across time. The earliest evidence of affective synchrony is found in the interactions of parents and their infants and young children (Bornstein et al., Reference Bornstein, Manian and Henry2021; Feldman, Reference Feldman2007). In these developmental periods, higher levels of parent–child affective synchrony are related to better child self-control, social competence, and school performance, and less child aggression and social withdrawal (Cole et al., Reference Cole, Teti and Zahn-Waxler2003; Feldman et al., Reference Feldman, Greenbaum and Yirmiya1999; Harrist et al., Reference Harrist, Pettit, Dodge and Bates1994; Lunkenheimer et al., Reference Lunkenheimer, Hamby, Lobo, Cole and Olson2020). Despite the importance of extending research on parent–child affective synchrony across development, surprisingly little research on this topic has focused on adolescence (Lougheed, Reference Lougheed2020).

The nature of parent–child mean affect and affective synchrony may change over time and be particularly important over the course of adolescence. Adolescence is a critical transitional period. During childhood, a characteristic, predictable, and stable pattern of behavior develops between parent and child based on repeated interactions (Granic et al., Reference Granic, Dishion, Hollenstein, Adams and Berzonsky2003). In the transition from childhood to adolescence, close relationships with parents are maintained (Collins & Laursen, Reference Collins, Laursen, Lerner and Steinberg2004) but entrenched dyadic interactions are perturbed as the family system negotiates new issues (e.g., autonomy, less predictable behavior and emotion; Granic et al., Reference Granic, Dishion, Hollenstein, Adams and Berzonsky2003). Conflict discussions are key contexts in which the vertical relationships of childhood (i.e., parents hold knowledge and power and provide security and warmth) become more horizontal (i.e., interactions are equal, symmetrical, and reciprocal; Branje, Reference Branje2018), and so conflictual interactions may increase in frequency and intensity in adolescence (Granic et al., Reference Granic, Dishion, Hollenstein, Adams and Berzonsky2003). Accordingly, we might expect that mean (i.e., average) levels of parent and youth affect within conflict discussions become increasingly negative with adolescent age. However, the nature of the moment-to-moment coordination of parent and adolescent affect during conflict discussions is unclear. It is plausible to expect that parent–adolescent affective synchrony decreases across adolescence, as adolescents become more independent and autonomous. During conflict discussions, lower levels of parent–adolescent affective synchrony could also reflect efforts by parents to prevent escalation by maintaining neutral affect while adolescents have more extreme affective displays (van Bommel et al., Reference van Bommel, Van der Giessen, Van der Graaff, Meeus and Branje2019). Conversely, parent–adolescent affective synchrony may increase across adolescence, as adolescents mature and become more responsive to their parents’ emotions in constructive ways (Branje, Reference Branje2018). During conflict discussions, higher levels of parent–adolescent affective synchrony could reflect emotional acceptance or empathy by parents (e.g., Creavy et al., Reference Creavy, Gatzke-Kopp, Zhang, Fishbein and Kiser2020; Jospe et al., Reference Jospe, Genzer, Selle, Ong, Zaki and Perry2020).

Several disorders increase in prevalence during adolescence, including depression and some forms of disruptive behavior disorders (Kessler et al., Reference Kessler, Angermeyer, Anthony, De Graaf, Demyttenaere, Gasquet, de Girolam, Gluzman, Gureje, Haro, Kawakami, Karam, Levinson, Medina Mora, Oakley Browne, Posada-Villa, Stein, Adley Tsang, Aguilar-Gaxiola and Üstün2007). Deficits in individual ability to regulate emotions is a key mechanism in the development of psychopathology (Compas et al., Reference Compas, Jaser, Bettis, Watson, Gruhn, Dunbar, Williams and Thigpen2017; McLaughlin & Lambert, Reference McLaughlin and Lambert2017; Yap et al., Reference Yap, Allen and Sheeber2007), and emotion regulation skills are shaped in part through repeated exchanges with parents (Tan et al., Reference Tan, Oppenheimer, Ladouceur, Butterfield and Silk2020). Accordingly, both mean-level (i.e., average) affect and affective synchrony during parent–adolescent interactions may be related to adolescent symptoms of internalizing and externalizing disorders (Granic, Reference Granic2005; Morris et al., Reference Morris, Cui, Criss, Simmons, Cole and Hollenstein2018). Broadly, parental expressions of negative affect and behaviors (e.g., hostility, criticism) during parent–child interactions are linked to youth psychopathology (Peris & Miklowitz, Reference Peris and Miklowitz2015). In childhood, fewer behavioral problems have been observed in youth whose dyadic interactions with their parents are characterized by more positive affect and less negative affect (Lunkenheimer et al., Reference Lunkenheimer, Olson, Hollenstein, Sameroff and Winter2011) and greater affective synchrony (Harrist et al., Reference Harrist, Pettit, Dodge and Bates1994; Kerr et al., Reference Kerr, Rasmussen, Smiley, Buttita and Borelli2021; Lougheed et al., Reference Lougheed, Hollenstein, Lichtwarck-Aschoff and Granic2015). Still, the ways in which mean-level affect and affective synchrony relate to youth internalizing and externalizing symptoms are not entirely clear, especially during adolescence. For example, Schwartz et al. (Reference Schwartz, Byrne, Simmons, Whittle, Dudgeon, Yap, Sheeber and Allen2014) found that maternal aggressive responses to both adolescent aggression (synchronous) and adolescent positive affect (asynchronous) are related to the onset of adolescent depression. While parent–adolescent interactions characterized by high and escalating levels of negative emotions may be associated with higher levels of adolescent internalizing and externalizing problems (Granic & Patterson, Reference Granic and Patterson2006; Yap et al., Reference Yap, Schwartz, Byrne, Simmons and Allen2010), so might patterns of parental nonresponsiveness to adolescents’ emotions (Yap et al., Reference Yap, Allen and Ladouceur2008). Associations between affective synchrony and internalizing and externalizing symptoms during adolescence have not been a major focus of previous research (Lougheed, Reference Lougheed2020).

The measurement of affective synchrony ranges from macro- to micro-level methods, including parent and adolescent self-reports of emotions over moments, weeks, months, and years (e.g., Kim et al., Reference Kim, Conger, Lorenz and Elder2001) and observations and ratings of emotion by trained coders during parent–adolescent interactions (e.g., van Bommel et al., Reference van Bommel, Van der Giessen, Van der Graaff, Meeus and Branje2019). In terms of time scale, measurement of parent and adolescent affect on a moment-to-moment basis is arguably the most powerful and sensitive method to capture affective synchrony. With regard to reporting, both self-report and observational coding have methodological benefits and drawbacks. Observational coding by independent raters captures expressed emotion rather than experienced emotion and a large body of research has shown that these constructs are distinct (Ekman, Reference Ekman and Cole1972, Reference Ekman1993; Greenaway & Kalokerinos, Reference Greenaway and Kalokerinos2019); both are important but inform separate research questions. Whereas self-reports capture subjective emotional experience, asking individuals to freely recall their affect after events occur may contribute to measurement error, including biases in retrospective recall (Feldman-Barrett, Reference Feldman-Barrett1997; Kahneman, Reference Kahneman, Kahneman and Tversky2000; Rosenberg & Ekman, Reference Rosenberg and Ekman1994).

A potentially valuable alternative method is video-mediated recall (VMR), which involves obtaining reports from members of dyads about their emotions while they view recordings of their face-to-face interactions (Welsh & Dickson, Reference Welsh and Dickson2005). VMR can be conceptualized as an amalgam of self-report and observational coding giving rise to a set of advantages achieved by neither method alone. VMR allows for the use of specific, controlled tasks that provide similar contexts for emotions within and across parent–adolescent dyads (inherent to observational coding), informs internal, covert emotional processes that are not accessible to external observers (inherent to self-reports), and minimizes susceptibility to errors in recall as compared to retrospective reports over longer periods of time. VMR has been used successfully in studies of affective synchrony in married couples (Rohrbaugh et al., Reference Rohrbaugh, Shoham, Butler, Hasler and Berman2009) and randomly paired undergraduate students (Butler et al., Reference Butler, Gross and Barnard2014). VMR has also been used to understand individual differences in emotion dynamics in adolescents (Yang et al., Reference Yang, Ram, Lougheed, Molenaar and Hollenstein2019). To our knowledge, however, this approach has only been used in one study with parent–adolescent dyads (see Main et al., Reference Main, Kho, Miramontes, Wiebe, Çakan and Raymond2021).

The current study used VMR on a moment-to-moment basis to capture the dynamics of affect in interactions between parents and their children during a laboratory conflict-topic discussion task. Our sample of youth ranged from 10 to 15 years old, and accordingly we tested our hypotheses in early to mid-adolescence. We addressed the following three aims and four hypotheses. Aim 1. We examined the association between parent and adolescent affect over the course of the conflict-topic discussion task. As predictors, we used moment-to-moment changes in each partner’s affectFootnote ¹ as well as each partner’s mean level of affect. In the context of multilevel modeling, these associations represent the relation of variables within person but across time (level 1) and the relation of variables after pooling across time (level 2), respectively. In hypothesis one, we expected mean parent and adolescent affect to be positively related to partner affect. In hypothesis two, we expected positive associations between partners’ concurrent, moment-to-moment affect; in other words, we expected to observe affective synchrony. Aim 2. We explored whether affective processes in parent–child interactions differ as a function of adolescent age. For hypothesis three, we expected that parent and adolescent mean affect would be inversely related to adolescent age. Considering limited research on affective synchrony in adolescence, analyses examining the association of age and moment-to-moment affect were exploratory. Aim 3. We were interested in the ways in which associations between parent and adolescent affect differed based on levels of adolescent internalizing and externalizing symptoms. For hypothesis four, we expected that mean affect would be more negative for both parents and adolescents when levels of adolescent internalizing and externalizing symptoms were higher. Again, due to scant research on affective synchrony in adolescence, analyses examining the interaction between adolescent symptoms and moment-to-moment affect were exploratory.

Method

Participants and procedure

Parents and adolescents in a southern metropolitan area were recruited to participate in a study about stress and emotions in families from October 2018 to February 2020 (recruitment was suspended in March 2020 due to the COVID-19 pandemic). In order to enroll a sample that varied regarding adolescents’ emotional and behavioral problems, recruitment occurred primarily through listserv postings and outpatient clinic referrals. Parents who expressed interest in the study were screened on the phone prior to enrollment. Parent–adolescent dyads were excluded from participation if the adolescent was outside the age range of 10–15 years old, the parent was not the legal guardian of the adolescent, or the dyad did not live together at least 50% of the time for the past 6 months. Dyads were also excluded if mothers reported a diagnosis of schizophrenia in themselves or their adolescent or a diagnosis of autism spectrum disorder in the adolescent, as these disorders may impede an individual’s ability to complete some of the study tasks. In families with multiple adolescents meeting inclusion criteria, the older adolescent was invited to participate in the study.

Fifty-six dyads were enrolled and participated in the study. One dyad did not complete the conflict-topic discussion task due to an abbreviated lab visit, and their data were excluded listwise from the current analyses. Thus, 55 dyads contributed partial or full data to the current analyses. VMR data were missing from two parents and three adolescents due to research assistant error during the study visit (e.g., errors in installing the software or saving the data); no families refused to complete the VMR task. Continuous VMR affect ratings from the 55 dyads were divided into sixty 10-s intervals (see Data Analytic Strategy for details), resulting in 3,300 observations for analysis.

Adolescents were between 10 and 15 years old (M = 12.27; SD = 1.67), and 55% were female. Sixty-nine percent were White, 24% Black, 6% Asian, and 2% more than one race. Parents were primarily female (93%)Footnote ² , married or living with a partner (69%), and ranged in education (37% completed a graduate degree, 9% some graduate education, 30% college graduate, 22% some college, 2% high school graduate). See Table 1 for additional information about sociodemographic characteristics of the sample.

Table 1. Sociodemographic characteristics of adolescents and parents

Parent–adolescent dyads completed a 4- to 5-hr laboratory visit, which included a series of questionnaires, interviews, and laboratory tasks. At the beginning of the visit, parents and adolescents independently completed an adapted version of the Issues Checklist (Robin & Foster, Reference Robin and Foster1989). The Issues Checklist asks respondents to indicate which of 44 topics they discussed with their partner in the last 4 weeks (e.g., coming home on time, helping out around the house), and how they felt during those discussions (1 = calm, 5 = angry). Research assistants selected one topic that was rated highly by both the parent and adolescent. If the parent and adolescent did not rate the same topic highly, research assistants selected the topic that was rated highest by the adolescent. Dyads participated in a videotaped discussion task, during which they were instructed to discuss the selected topic for 10 min. The dyad was asked to (a) describe the issue, (b) explain how they feel about it, (c) discuss why it has become a source of conflict, and (d) attempt to resolve the issue. After the discussion, parents and adolescents viewed the videotape in separate rooms and provided moment-to-moment ratings of their own in-task affect (Girard, Reference Girard2014).

Prior to the laboratory visit, adolescents completed the Youth Self Report (YSR; Achenbach & Rescorla, Reference Achenbach and Rescorla2001) using REDCap survey software (Harris et al., Reference Harris, Taylor, Thielke, Payne, Gonzalez and Conde2009). Parents received $100 for their participation in the study, and adolescents received $50. All procedures were approved by the Vanderbilt University institutional review board.

Data analytic strategy

Planned analyses

We used SPSS Version 26 for all analyses. In preliminary analyses, we computed zero-order correlations to understand basic associations among the study variables. We used multilevel modeling to test all hypotheses. Full information maximum likelihood estimation was implemented in multilevel modeling analyses to account for missing data. First, intraclass correlation coefficients (ICC) were estimated in separate null models predicting parent affect and adolescent affect to establish need for multilevel modeling (Pornprasertmanit et al., Reference Pornprasertmanit, Lee and Preacher2014). See Supplementary Information for null model equations. Second, we conducted two sets of univariate analyses, one predicting adolescent affect as the dependent variable ( $A{A_{td}}$ ; henceforth referred to as the “Adolescent Affect Model”) and the other predicting parent affect as the dependent variable $(P{A_{td}}$ ; henceforth referred to as the “Parent Affect Model”). The Adolescent Affect Model is displayed below. See Supplementary Information for a parallel equation that reflects the Parent Affect Model.Footnote ^3.

$${\rm{Level}}{\mkern 1mu} {\rm{1}}{\mkern 1mu} {\rm{equation}}:{\mkern 1mu} A{A_{td}} = \;{\beta _{0d}} + \;{\beta _1}(P{A_{td}} - {\rm{ }}{\overline {PA} _{.d}}) + {e_{td}}$$

$${\rm{Level\, 2\, equation\!\!:}}\,{\beta _{0d}} = \;{{\it{\gamma}} _{00}} + {\gamma _{01}}{\overline {PA} _{.d}} + \;{u_{0d}}$$

$${\beta _{1d}}{\rm{\;}} = {\rm{\;}}{\gamma _{10}}{\rm{\;}} + \;{u_{1d}}$$

\begin{align}\rm{Mixed \ model\!:}\ {\it{A{A_{td}}}} &= \;{\gamma _{00}}{\rm{\;}} + {\rm{\;}}{\gamma _{01}}{\overline {{\it{PA}}} _{.d}}{\rm{\;}} + {\rm{\;}}{\gamma _{10}}({\it{P{A_{td}}}} - {\overline {{\it{PA}}} _{.d}}){\rm{\;}}\\ &\quad + \;{{\it{u}}_{0{\it{d}}}}{\rm{\;}} + \;{{\it{u}}_{1{\it{d}}}}({\it{P{A_{td}}}} - {\overline {{\it{PA}}} _{.d}})\end{align}

$${e_{td}}\sim N(0,\, \sigma _e^2)$$

$$\left[ {\matrix{ {{u_{0d}}} \cr {{u_{1d}}} \cr } } \right]\sim N\left( {\left[ {\matrix{ 0 \cr 0 \cr } } \right],\left[ {\matrix{ {{\tau _{00}}} & {} \cr {{\tau _{10}}} & {{\tau _{11}}} \cr } } \right]} \right)$$

In the Adolescent Affect Model, parent affect ratings for all 10-s intervals were person-mean centered. In other words, each parent’s mean affect score across the 10-min interaction was subtracted from their affect ratings at each 10-s interval $(P{A_{td}} - {\overline {PA} _{.d}}$ ). The parent person mean-centered affect score $(P{A_{td}} - {\overline {PA} _{.d}}) $ was included as a level-1 predictor, and the parent mean affect score $({\overline {PA} _{.d}}) $ was included as a level-2 predictor. Accordingly, ${\gamma _{01}}$ provides a test of hypothesis 1, the level-2 associations of mean affect. Further, ${\gamma _{10}}$ provides a test of hypothesis 2, the level-1 associations of affective synchrony. Third, we built upon our two sets of univariate analyses to address Aims 2 and 3. We included adolescent age, adolescent internalizing symptoms, and adolescent externalizing symptoms as level-2 predictors of level-1 slopes and intercepts. All slopes and intercepts were random effects. The Adolescent Affect Model is displayed below. See Supplementary Information for an equation that reflects the Parent Affect Model.

$${\rm{Level\, 1\, equation\!\!:}}\,A{A_{td}} = {\rm{\;}}{\beta _{0d}} + {\rm{\;}}{\beta _1}(P{A_{td}} - {\overline {PA} _{.d}}) + {\rm{\;}}{e_{td}}$$

\begin{align*}{\rm{Level \,2 \,equation\!\!:}}\,{\beta _{0d}} & = \;{\gamma _{00}} + {\gamma _{01}}{\overline {PA} _{.d}} + {\gamma _{02}}ag{e_d} + {\gamma _{03}}{\rm{inter}}{{\rm{n}}_d} \\ & \quad +\; {\gamma _{04}}{\rm{exter}}{{\rm{n}}_d} + {\gamma _{05}}{\rm{ag}}{{\rm{e}}_d}{\overline {PA} _{.d}} \\ &\quad + \;{\gamma _{06}}{\rm{inter}}{{\rm{n}}_d}{\overline {PA} _{.d}} + {\gamma _{07}}{\rm{exter}}{{\rm{n}}_d}{\overline {PA} _{.d}} \\ & \quad + \;{u_{0d}}\end{align*}

$${\beta _{1d}}{\rm{\;}} = {\rm{\;}}{\gamma _{10}} + {\gamma _{11}}{\rm{ag}}{{\rm{e}}_d} + \;{\gamma _{12}}{\rm{inter}}{{\rm{n}}_d} + {\gamma _{13}}{\rm{exter}}{{\rm{n}}_d} + \;{u_{1d}}$$

\begin{align*}{\rm{Mixed \,model\!\!:}}\,A{A_{td}} &= {\gamma _{00}} + {\gamma _{01}}{\overline {PA} _{.d}} + {\gamma _{02}}{\rm{age}}{_d} + {\gamma _{03}}{\rm{intern}}{_d} \\ &\quad +\; {\gamma _{04}}{\rm{extern}}{_d} + {\gamma _{05}}{\rm{age}}{_d}{\overline {PA} _{.d}} \\ &\quad +\; {\gamma _{06}}{\rm{intern}}{_d}{\overline {PA} _{.d}} + {\gamma _{07}}{\rm{extern}}{_d}{\overline {PA} _{.d}} \\ &\quad +\; {\gamma _{10}}(P{A_{td}} - {\overline {PA} _{.d}}) \\ &\quad+\; {\gamma _{11}}{\rm{age}}{_d}(P{A_{td}} - {\overline {PA} _{.d}}) \\ &\quad +\; {\gamma _{12}}{\rm{intern}}{_d}(P{A_{td}} - {\overline {PA} _{.d}}) \\&\quad +\; {\gamma _{13}}{\rm{extern}}{_d}(P{A_{td}} - {\overline {PA} _{.d}}) + {u_{0d}} \\ &\quad +\; {u_{1d}}(P{A_{td}} - {\overline {PA} _{.d}})\end{align*}

$${e_{td}}\sim N(0,\, \sigma _e^2)$$

$$\left[ {\matrix{ {{u_{0d}}} \cr {{u_{1d}}} \cr } } \right]\sim N\left( {\left[ {\matrix{ 0 \cr 0 \cr } } \right],\, \left[ {\matrix{ \hskip-15pt{{\tau _{00}}}{} \cr {{\tau _{10}}}\ {{\tau _{11}}} \cr } } \right]} \right)$$

Results

Descriptive statistics and bivariate correlation analyses

Table 2 displays means and standard deviations for study variables (adolescent affect, parent affect, adolescent internalizing symptoms, adolescent externalizing symptoms). For both adolescents and parents, affect ratings ranged from very negative (−100) to very positive (100), and the mean affect ratings were close to neutral (adolescent affect: M = 4.23, SD = 44.13; parent affect: M = 5.65, SD = 41.78). Distributions of adolescent and parent affect ratings were extremely similar. Approximately 10.1% of adolescent ratings and 10.1% of parent ratings were very negative (−100 to −50); 33.6% of adolescent ratings and 29.3% of parent ratings were moderately negative (−50 to −1); 9.6% of adolescent ratings and 9.2% of parent ratings were neutral (0); 31.3% of adolescent ratings and 36.3% of parent ratings were moderately positive (1 to 50); and 15.4% of adolescent ratings and 15.1% of parent ratings were very positive (51 to 100). Of the 43 participants for whom we were able to calculate symptom T scores (i.e., adolescents > 10 years old), adolescent symptoms were approximately one-half standard deviation above the normative mean for internalizing symptoms (M = 56.79, SD = 9.80) and externalizing symptoms (M = 53.05, SD = 9.93).Footnote ⁴

Table 2. Means, standard deviations, ranges, and bivariate correlations for study variables

Note. Raw scores are presented for adolescent internalizing symptoms and adolescent externalizing symptoms.

*** p<.001.

Zero-order correlations are also presented in Table 2. As hypothesized, adolescent and parent affect were positively correlated with a medium effect. For adolescents, higher levels of internalizing and externalizing symptoms were associated with more negative in-task affect with small effects. Higher levels of adolescent internalizing and externalizing symptoms were associated with more negative in-task parent affect with small effects. As these correlations were in the expected directions, they provide evidence of construct validity of parent and adolescent affect as measured by VMR.

Multilevel modeling

The ICC derived from the null univariate models predicting parent and adolescent affect indicated that 38% of the observed variation in parent affect and 48% of the observed variation in adolescent affect was due to differences among dyads (Table 3). These values suggest that multilevel modeling is an appropriate analytic method for these data.

Table 3. Null models predicting adolescent affect and parent affect

Note. PE = parameter estimate; SE = standard error; ICC = intraclass correlation coefficient.

Tests of hypotheses 1 and 2 (Aim 1) are presented in Table 4. In support of hypothesis 1 (level-2 analyses), multilevel models demonstrated that more positive mean parent affect ratings were associated with more positive adolescent affect ratings ( ${\gamma _{01}}$ = .56, p = .001), and more positive mean adolescent affect ratings were associated with more positive parent affect ratings ( ${\gamma _{01}}$ = .35, p = .001). In support of hypothesis 2 (level-1 analyses), multilevel models showed that more positive parent affect ratings were associated with more positive concurrent adolescent affect ratings ( ${\gamma _{10}}$ = .24, p < .001), and more positive adolescent affect ratings were associated with more positive concurrent parent affect ratings ( ${\gamma _{10}}$ = .28, p < .001). That is, results provide evidence of parent–child affective synchrony in adolescence.

Table 4. Multilevel models predicting adolescent affect and parent affect

Note. PE = parameter estimate; SE = standard error; Adolescent affect _t = adolescent affect at the current interval; Parent affect _t = parent affect at the current interval.

^a Person mean centered.

With regard to Aim 2, adolescent age moderated parent–adolescent synchrony, but only in the Parent Affect Model ( ${\gamma _{11}}$ = .07, p = .001). As depicted in Figure 1, parent–adolescent synchrony was stronger (i.e., the slope was steeper) when adolescents were older (region of significance: adolescent age > −2.29). Results are displayed in Table 5. Exploratory analyses did not provide evidence of an Adolescent Age × Mean Affect interaction on parent or adolescent affect.

Figure 1 Adolescent Age Moderates the Association between Current Adolescent Affect and Current Parent Affect

Note. Minimum adolescent age (10 years old): simple slope = 0.11 (0.06), z = 1.99, p = 0.05. Average adolescent age (12 years old): simple slope = 0.27 (0.04), z = 7.67, p <.001. Maximum adolescent age (15 years old): simple slope = 0.47 (0.07), z = 6.67, p <.001. Region of significance: adolescent age > −2.29.

Table 5. Multilevel models predicting adolescent affect and parent affect, including interactions

Note. PE = parameter estimate; SE = standard error; Parent affect _t = parent affect at the current interval; Adolescent affect _t = adolescent affect at the current interval.

^a Person mean centered.

In Aim 3, we examined how associations between parent and adolescent affect, both in terms of mean affect and affective synchrony, may differ based on levels of adolescent internalizing and externalizing symptoms (Table 5). In partial support of hypothesis 3 (level-2 analyses), externalizing symptoms interacted with mean partner affect to predict affect in both the Adolescent Affect ( ${\gamma _{07}}$ = −.07, p = .02) and Parent Affect ( ${\gamma _{07}}$ = −.04, p = .05) models. As depicted in Figure 2, parent mean affect predicted adolescent affect when externalizing symptoms were below the mean in the Adolescent Affect Model (region of significance: adolescent externalizing symptoms < −0.22). Similarly, Figure 3 shows that adolescent mean affect predicted parent affect when externalizing symptoms were below the mean in the Parent Affect Model (region of significance: adolescent externalizing symptoms < −0.32).

Figure 2 Adolescent Externalizing Symptoms Moderates the Association Between Parent Mean Affect and Current Adolescent Affect

Note. Low externalizing symptoms (1 SD below average): simple slope = 0.77 (0.21), z = 3.68, p < 0.001. Average internalizing symptoms: simple slope = 0.29 (0.16), z = 1.84, p = .07. High internalizing symptoms (1 SD above average): simple slope = −0.19 (0.30), z = −0.62, p = 0.54. Region of significance: externalizing symptoms < −.22.

Figure 3 Adolescent Externalizing Symptoms Moderates the Association Between Adolescent Mean Affect and Current Parent Affect

Note. Low externalizing symptoms (1 SD below average): simple slope = 0.50 (0.14), z = 3.54, p < .001. Average internalizing symptoms: simple slope = 0.21 (0.12), z = 1.81, p = .07. High internalizing symptoms (1 SD above average): simple slope = −0.08 (0.22), z = −0.35, p = 0.73. Region of significance: externalizing symptoms < −.32.

Figure 4 Adolescent Internalizing Symptoms Moderates the Association Between Current Parent Affect and Current Adolescent Affect

Note. Low internalizing symptoms (1 SD below average): simple slope = 0.13 (0.05), z = 2.40, p = 0.02. Average internalizing symptoms: simple slope = 0.23 (0.04), z = 6.50, p < .001. High internalizing symptoms (1 SD below average): simple slope = 0.33 (0.05), z = 6.67, p < .001. Region of significance: internalizing symptoms > −10.33.

Analyses examining the interaction of adolescent internalizing and externalizing symptoms and moment-to-moment affect (level-1 analyses) were exploratory. We found that internalizing symptoms, but not externalizing symptoms, moderated parent–adolescent synchrony. These results were consistent in both the Adolescent Affect Model ( ${\gamma _{12}}$ = .01, p = .01) and Parent Affect Model ( ${\gamma _{12}}$ = .01, p = .01). As depicted in Figures 4 and 5, parent–adolescent synchrony was higher (i.e., the slope was steeper) when the adolescent had more internalizing problems (region of significance: Adolescent Affect Model, internalizing symptoms > −10.33; Parent Affect Model, internalizing symptoms > −12.84).

Figure 5 Internalizing Symptoms Moderates the Association Between Current Adolescent Affect and Current Parent Affect

Note. Low internalizing symptoms (1 SD below average): simple slope = 0.17 (0.05), z = 3.24, p = .001. Average internalizing symptoms: simple slope = 0.27 (0.04), z = 7.67, p < .001. High internalizing symptoms (1 SD above average): simple slope = 0.38 (0.05), z = 7.59, p < .001. Region of significance: internalizing symptoms > −12.84.