Father- and Youth-Reported Family Affective Expression Differentially Predicts Youth Internalizing and Externalizing Symptoms

Abstract

Objective.

Temperamental approach is associated with adolescent internalizing and externalizing symptoms. Negative family affective expression, or problematic communication about emotions, is also associated with youth’s risk for symptoms. However, it is unclear whether negative family affective expression differentially predicts symptoms based on (a) youth’s temperamental approach and (b) informants’ perceptions of negative family affective expression. To address these issues, we explored whether mother-, father-, and youth-reported negative family affective expression moderated the relation between youth temperamental approach and symptoms.

Method.

Participants were 775 youth (71% male, 76% Caucasian) assessed at ages 10–12 (Time 1) and 12–14 (Time 2). Mothers, fathers, and youth reported on negative family affective expression and youth reported on temperamental approach at Time 1. Teachers reported on youth symptoms at Times 1 and 2.

Results.

Youth- and father-reported, but not mother-reported, negative family affective expression moderated the relation between youth approach and symptoms. When youth reported higher negative family affective expression, youth lower in approach exhibited higher internalizing symptoms than youth higher in approach. In contrast, when fathers reported lower negative family affective expression, youth lower in approach exhibited higher internalizing and externalizing symptoms than youth higher in approach.

Conclusions.

Assessments and interventions for youth symptoms should include not only temperamental features, but also multiple informants’ perspectives of family affective expression. Such efforts could promote greater family communication, address problematic family dynamics, and potentially attenuate risk for youth symptoms.

Internalizing and externalizing symptoms may develop in early adolescence and are often associated with numerous negative outcomes such as decreased social competence (Bornstein, Hahn, & Haynes, 2010), peer rejection (Keiley, Lofthouse, Bates, Dodge, & Pettit, 2003), and substance use (King, Iacono, & McGue, 2004). Internalizing problems can be defined as social withdrawal, anxiety, depression, and somatic complaints, whereas externalizing problems often manifest as hyperactivity, aggression, delinquent behaviors, and disinhibition. Consistent with ecological systems and developmental psychopathology approaches, the consideration of both children’s temperament (e.g., temperamental approach) and family features (e.g., negative family affective expression) is critical to understanding the development and maintenance of youth’s symptoms (Bronfenbrenner & Morris, 1988; Drabick & Steinberg, 2011). Such an approach can inform targeted interventions by elucidating child- and family-level characteristics that confer risk or resilience for youth psychological symptoms.

Temperament, defined as constitutionally based individual differences in reactivity and self-regulation, influences one’s emotional and behavioral responses to context and may confer risk or resilience for internalizing and externalizing symptoms (Thomas & Chess, 1984). One temperamental feature that has been robustly associated with these symptoms is the temperamental disposition to approach (or withdraw) from contexts and/or people (for a review, see Nigg, 2006). Youth higher in approach behaviors are generally more sensitive to appetitive stimuli, advance toward unfamiliar contexts, and display higher levels of positive mood (Nigg, 2006). However, they also may be lower in inhibitory control and more likely to disregard social rules and the rights of others, increasing risk for externalizing behaviors (Muris, Meesters, &Blijlevens, 2007; Muris & Ollendick, 2005; Nigg, 2006). In contrast, youth lower in approach behaviors are more likely to avoid aversive stimuli, be more sensitive to punishment, withdraw from novel situations, and exhibit higher levels of negative mood; thus, these youth may be at heightened risk for internalizing problems (Dollar & Stifter, 2012; Fox, Henderson, Marshall, Nichols, & Ghera, 2005).

Nevertheless, not all adolescents varying in approach behaviors develop symptoms. Consistent with biological sensitivity to context models, youth varying in temperamental approach may be differently affected by contexts such as the family environment, which may influence their susceptibility to developing symptoms (Boyce & Ellis, 2005; Rothbart, Posner, & Kieras, 2006). Some research has investigated to what extent parenting behaviors (e.g., overinvolvement, harsh parenting) increase risk for symptoms among adolescents varying in approach (e.g., Hudson, Dodd, Lyneham, & Bovopoulous, 2011; Williams et al., 2009); however, fewer studies have considered the broader family climate.

One candidate family process that may be associated with symptoms among youth varying in approach is family affective expression. Similar to the constructs of parent expressed emotion or family expressiveness, family affective expression refers to family members’ communication about emotions and affect, including the content, timing, and intensity of emotion expression (Boyum & Parke, 1995; Skinner, Steinhauer, & Sitarenios, 2000). Families with problematic affective expression may exhibit difficulty communicating emotions clearly and may not resolve conflicts in a timely manner. These family dynamics may create an unpredictable environment for children, contributing to youth’s overarousal or dysregulation (Cummings, Goeke-Morey, & Papp, 2002). Among families higher in negative affective expression, youth may exhibit difficulty expressing emotion, overreact to situations, and suppress negative emotions, increasing risk for symptoms (Adrian et al., 2009; Queen, Stewart, Ehrenreich-May, & Pincus, 2013; Suveg, Zeman, Flannery-Schroeder, & Cassano, 2005). However, families higher in positive affective expression may display higher levels of positive emotions, discuss their emotional experiences with youth, and encourage youth to reflect on their emotional states, mitigating risk for symptoms (Luebbe & Bell, 2014; Marakovitz, Wagmiller, Mian, Briggs-Gowan, & Carter, 2011; Saarni & Buckley, 2002).

Although negative family affective expression is associated with children’s externalizing and internalizing symptoms, this relation may depend on youth’s temperamental approach, although there is a dearth of research addressing this question. Families lower in negative affective expression may model appropriate expression and regulation of emotions, which may help youth (a) lower in approach develop strategies to modulate higher fearfulness and arousal in novel contexts, decreasing risk for internalizing symptoms; or (b) higher in approach manage impulses (e.g., desire for rewards), which may improve their inhibitory control and possibly attenuate risk for externalizing symptoms (Clark & Phares, 2004). Families that do not clearly communicate or that suppress emotions may provide inappropriate models for managing and expressing negative emotions. These family behaviors may increase social withdrawal tendencies among youth lower in approach and consequently risk for internalizing problems. Negative family affective expression may also interfere with youth’s understanding of others’ emotions; difficulties with processing emotional expressions may be particularly harmful among youth higher in approach who are more sensitive to rewards and more likely to violate social rules to obtain a goal (Nigg, 2006).

The associations among negative family affective expression, temperamental approach, and symptoms also may differ depending on the informant who reports on family affective expression, although there is a paucity of work in this area. Consistent with the Operations Triad Model, adolescents and parents may (a) agree on the extent of family processes (converging operations), (b) disagree about family functioning (diverging operations), or (c) report different levels of family functioning because of methodological differences in the assessment of family functioning (compensating operations) (De Los Reyes & Ohannessian, 2016). Most of the literature on family functioning has relied on single informant reports (e.g., parent or youth; Lucia & Breslau, 2006; Queen et al., 2013). Work that has considered multiple informants has indicated low to moderate correspondence between youth and parent reports of family functioning (e.g., conflict, cohesion, communication; Cavendish, Montague, Enders, & Dietz, 2014; De Ross, Marrinan, Schattner, Gullone, & 1999; Hartos & Power, 2000; Noller & Callan, 1986) and moderate associations between mother and father reports of the family environment (e.g., adaptability, cohesion, expressiveness; Noller & Callan, 1986; Ohannessian, Lerner, Lerner, von Eye, 1995; Queen et al., 2013).

Discrepancies in child and parent reports of family processes may stem from differences related to the developmental stages and roles of each family member, and different observations that family members may have (De Los Reyes & Ohannessian, 2016; Korelitz & Garber, 2016). For example, adolescence is characterized by increasing importance of peers and autonomy from family members (Drabick & Steinberg, 2011; Laursen & Collins, 2009). Consequently, adolescents may have conflicts with family members and, as a result, they may report lower levels of positive family functioning (e.g., communication, cohesiveness) and higher negative family affective expression compared to parents (Gonzales, Cuace, & Mason, 1996; Grotevant & Cooper, 1986; Ohannessian et al., 1995). In addition, although one parent may experience conflict with the adolescent, the partner may not be aware of the extent of this conflict or identify this conflict as less salient or frequent than the partner experiencing the conflict, which may result in different perceptions of negative family functioning. In contrast to adolescents, parents may rate family functioning in accordance with their parenting goals (e.g., maintaining family harmony and communication) and consequently report lower negative family affective expression (Dreman, 2003; Noller & Callan, 1986; Stuart & Jose, 2012).

Different perspectives of the family environment are important to consider because they may differentially predict adolescent internalizing and externalizing symptoms. Links between negative family affective expression (e.g., Noguchi & Ollendick, 2010) and other family processes (e.g., family cohesion; Dreman, 2003) with youth symptoms often are inconsistent across studies. For example, mother-reported family cohesion and adaptability (i.e., flexibility regarding family roles) are negatively associated with mother-reported internalizing and externalizing symptoms, but positively associated with youth-reported symptoms (Dreman, 2003). In addition, whereas female adolescent-reported family communication and family satisfaction are negatively associated with adolescent-reported anxiety symptoms, mother reports of these same processes are not significantly associated with these outcomes (Ohannessian & De Los Reyes, 2014).

In regard to associations between negative family affective expression and youth’s symptoms, only one study to our knowledge has investigated whether different informants’ reports of this family process are differentially predictive of youth symptoms. Specifically, Noguchi and Ollendick (2010) found that mother-reported family affective expression was negatively related, whereas father-reported family affective expression was positively related, to youth-reported anxiety symptoms. These findings indicate that the relations between negative family affective expression and youth symptoms may depend on the informant reporting on this family process. Child-specific features, such as temperamental approach, may contribute to these informant discrepancies and different family process-symptom associations; however, research addressing this possibility is wanting.

In the present study, we sought to address a number of gaps in the literature and had several aims. First, we explored a relatively understudied family feature, namely, family affective expression, which may be associated with youth symptoms. Second, we explored whether reports of negative family affective expression differed based on informant. Third, we investigated whether mother-, father-, and/or youth-reported family affective expression moderated the relation between youth approach and internalizing and externalizing problems during adolescence. This developmental period is characterized by myriad neurobiological, cognitive, and social changes during which the family environment may greatly impact adjustment (Yap, Allen, & Sheeber, 2007).

We made two hypotheses: (1) youth would report higher levels of negative family affective expression compared to fathers and mothers; and (2a) in the context of higher negative family affective expression, lower levels of approach would be associated with higher internalizing problems and higher levels of approach would be associated with higher externalizing symptoms; and (2b) in the context of lower negative family affective expression, levels of internalizing and externalizing symptoms would be similar regardless of the level of temperamental approach. We did not have informant-specific hypotheses related to negative family affective expression, youth approach, and symptoms given the dearth of research examining these constructs simultaneously.

Method

Participants

The sample was recruited from the project name removed, a longitudinal study that assessed children at differential risk for substance use disorder (SUD) based on the presence or absence of a lifetime diagnosis of SUD or other psychiatric disorders in the biological father. Biological fathers with and without histories of psychiatric diagnoses were recruited through substance dependence treatment programs and other venues using public service announcements, as well as newspaper and radio advertisements. Fathers who had a child aged 10–12 years (index youth) were eligible to participate. For the present study, index youth were grouped into one of two categories according to their biological father’s lifetime prevalence of mental health disorders: (a) paternal history of SUD or other psychiatric disorder (56%) or (b) no lifetime paternal history of psychiatric disorder (44%). Recruitment sources and procedures, as well as inclusion and exclusion criteria, are described in detail elsewhere (citation removed).

Assessments took place at two time points: when index children were aged 10–12 years (Time 1; N = 775; M = 10.95 ± 0.88 years; 71% male; 76% Caucasian, 21% African American, 3% “multiracial”) and 12–14 years (Time 2; n = 651; M = 12.99 ± 0.95 years old; 72% male; 76% Caucasian, 21% African American, 3% multiracial). The sample contains more males than females because recruitment of females began four years after project name removed was underway. Time 1 data were collected from 1991–2004. The median household income was $27, 311 (range = $4,999 – $123,128; M = $28,512, SD = $13,521).

At Time 1, mothers (n = 767) reported living with the biological parent of their child an average of 14.74 years (SD = 2.53; range = 0 years – 30 years), and co-parenting their children an average of 10.46 years (SD = 5.51; range = 0 years – 17 years). Fathers (N = 775) reported living with the biological parent of their child an average of 14.33 years (SD = 5.69; range = 0 years – 28 years) and co-parenting their children an average of 10.68 years (SD = 2.28; range = 0 years – 17 years). At Time 1, 91% of fathers reported that they were the head of the household and 9% of fathers reported that their spouse was head of the household; among mothers, 28% reported that they were head of the household, 71% reported that their spouse was head of the household, and 1% reported fathers as head of the household.

Procedure

The study was approved by a University Institutional Review Board. Prior to participation, participants were provided with a detailed study description. The primary caregiver provided informed consent and youth provided assent. Both the parent and youth were informed that their privacy was protected by a Certificate of Confidentiality issued to the project name removed from the National Institute on Drug Abuse. Caregivers provided consent for program staff to contact one teacher per child to obtain ratings. Teachers were mailed a packet with a cover letter, questionnaires, and a self-addressed, stamped envelope to return the questionnaires. Follow-up letters were sent if the teacher did not respond or complete all forms. Fathers, mothers, and children completed questionnaires related to family affective expression; children reported on their approach behaviors; and teachers reported on youth symptoms. Participants were financially compensated for participation.

Measures

Paternal Diagnostic Status.

Diagnoses were based on an extended version of the Structured Clinical Interview for DSM-III-R (Spitzer, Williams, & Gibbon, 1987), the most recent version available at the time of recruitment.

Family Affective Expression.

At Time 1, mothers, fathers, and children completed a 5-item subscale of family affective expression using the Family Assessment Measure (Skinner, Steinhauer, & Santa-Barbara, 1983; Skinner et al., 2000). Items are rated on a scale from 1 (strongly disagree) to 4 (strongly agree). Items are, “When someone in our family is upset, we don’t know if they are angry, sad, scared or what;” “We tell each other about things that bother us” (reverse coded); “I can let my family know what is bothering me” (reverse coded); “When someone is upset, we don’t find out until much later;” and “When our family is upset, we take too long to get over it.” Higher scores reflect more negative family affective expression. The scale demonstrated adequate internal consistency based on informant in the current sample (father-report: α=.70; mother-report: α=.68; child-report: α=.62). In other studies, the family affective expression subscale has demonstrated moderate internal consistency (adult-report α range = .61 – .74; child report α range = .61 – .71) (Skinner et al., 1983; Stewart, Kirisci, Long, & Giancola, 2015; Trosper & May, 2011). This subscale has also shown strong discriminant and construct validity among children and adults as it is negatively and positively associated with the FES family cohesion and family conflict subscales, respectively, and positively associated with the Minnesota Multiphasic Personality Inventory’s family discord and family problems subscales (Bloomquist & Harris, 1984; Skinner et al., 1983, 2000). The Family Assessment Measure has demonstrated strong internal consistency and moderate test-retest reliability; convergent validity with other measures of family functioning (e.g., Family Environment Scale, Family Assessment Device; Epstein, Baldwin, & Bishop, 1983; Moos & Moos, 1974); and strong discriminant validity among community and clinical samples (Skinner et al., 2000). Ninety-five percent of participants had both mothers and fathers complete the family affective expression subscale and 87% of families had father, mother, and youth reports of family affective expression.

Temperamental Approach Behaviors.

At Time 1, children reported on their approach behaviors using the 7-item approach/withdrawal subscale (α=.57; M inter-item correlation =.16) of the Dimensions of Temperament Survey-Revised (DOTS-R; Windle & Lerner, 1986). Items are rated on a scale from 1 (usually false) to 4 (usually true). Items are, “On meeting a new person, I tend to move towards him or her;” “I can make myself at home anywhere;” “I usually move towards new objects shown to me;” “My first reaction is to reject something new or unfamiliar to me” (reverse coded); “It usually takes me no time at all to get used to new people;” “I move towards new situations;” and “My first response to anything new is to move my head towards it.” Higher scores reflect higher approach behaviors. The approach/withdrawal subscale has shown (a) convergent validity with the extraversion subscale of Eysneck’s Personality Inventory (EPI); the activity and sociability subscales of the Emotionality, Activity, Sociability, and Impulsivity-II scale (EASI-II); and the strength of excitation (e.g., the ability to endure intense or long-lasting stimulation) and mobility of nervous system processes (e.g., ability to react quickly due to contextual changes) subscales of the Strelau Temperament Inventory (STI); (b) predictive validity with the extraversion and sociability subscales of the EPI and EASI-II, respectively; and (c) factorial validity with the EASI-II sociability subscale, the DOTS-R mood subscale, and the social ergonocity (e.g., the ability to engage in prolonged social interactions) subscale of the Strelau Temperament Inventory-Revised (Ruch, Angleitner, & Strelau, 1991; Windle, 1989;1992; Windle & Lerner, 1986;). This subscale has also shown moderate to high reliability in adolescent samples (α range = .73 – .85), and moderate test-retest stability (r range = .58 – .69) (Windle, 1992; Windle & Lerner, 1986). The DOTS-R has shown strong reliability and moderate levels of test-retest stability (Windle, 1992), convergent validity with other temperament measures (Goldsmith, Rieser-Danner, & Briggs, 1991), and concurrent validity with personality traits (Windle, 1989).

Internalizing and Externalizing Symptoms.

At Times 1 and 2, teachers reported on youth internalizing and externalizing symptoms using the Teacher Report Form of the Child Behavior Checklist (TRF; Achenbach, 1991). Teachers were asked to rate whether certain behaviors were characteristic of the student in the past six months on a scale from 0 (never) to 2 (almost always). A sample item from the internalizing scale is “unhappy, sad, or depressed,” and a sample item from the externalizing scale is “gets in many fights.” At Time 1, the scales demonstrated excellent internal consistency for internalizing symptoms (αs = .96 for boys and girls) and externalizing symptoms (αs = .97 for boys and girls). At Time 2, the composites demonstrated moderate to high internal consistency for internalizing (boys: α = .98, girls: α = .75) and externalizing (boys: α = .99, girls: α = .75) symptoms. The TRF is a common tool to assess youth emotional and behavioral symptoms and possesses strong discriminant and convergent validity, as well as internal consistency and test-retest reliability (Achenbach & Rescorla, 2001).

Statistical Analyses

Descriptive statistics and distributions of variables were examined and bivariate correlations were conducted to evaluate the relations among predictor and outcome variables using SPSS Version 23. A child sex variable was created with females coded as “0” and males coded as “1.” Individuals who were identified as Caucasian were coded as “0;” individuals identified as ethnic minority or multiracial were coded as “1.” Adolescents with fathers with no lifetime history of SUD or psychiatric disorder were coded as “0” and adolescents with fathers with a lifetime diagnosis of SUD or other psychiatric disorder were coded as “1.” We also ran regression analyses to ensure that the residuals of the predictor variables were normally distributed on the outcomes consistent with the homoscedasticity assumption of linear regression.

To address the first hypothesis, we conducted paired samples t-tests to investigate whether there were differences in reports of negative family affective expression based on informant. In addition, independent samples t-tests were conducted to examine whether father-reported negative family affective expression differed based on father diagnostic status. To address the second set of hypotheses, we conducted regression analyses using Mplus Version 7.11 (Muthén & Muthén, 1998–2014), which uses Full Information Maximum Likelihood (FIML) estimation to address missing data. FIML estimation assumes that missing data on a given variable are conditionally dependent on other observed variables in the data (i.e., missing at random). Unlike other methods (e.g., complete casewise analysis or listwise deletion) that may bias an analytic sample (Graham, 2009; Little & Rubin, 2002), FIML uses all available data to estimate model parameters without imputing values. This strategy maintains participants with missing data in the model estimation and produces smaller errors in parameter estimates and standard errors relative to other techniques for managing missing data (Enders, 2001; Graham, 2009; Newman, 2003).

Two regression analyses (one for internalizing and one for externalizing symptoms) were conducted to examine (a) the main effects of negative family affective expression and child approach behaviors and (b) whether negative family affective expression moderated the relations between child approach behaviors and youth symptoms. Negative family affective expression and child approach were z-scored (M=0, SD=1) before inclusion in the regression equations and interaction terms were created from the z-scored variables (Aiken & West, 1991). Although some studies use mean-centering, we chose to z-score predictor variables to put the SDs on a similar scale and thereby facilitate interpretation. For the regression equations, Step 1 included child age, sex, and ethnicity; father diagnostic status; father-, mother-, and child-reported negative family affective expression; child approach; and Time 1 symptoms that corresponded to the dependent variable. Step 2 included Step 1 variables and the three negative family affective expression × child approach interaction terms, one for each informant.

Post hoc probing was conducted using methods described by Aiken and West (1991) and Holmbeck (2002). A new conditional moderator variable (i.e., ± 1 SD from the z-scored values for negative family affective expression) was created to reflect higher and lower negative family affective expression. New interaction terms were created that included the conditional moderator. For significant interactions, the post hoc regressions involved simultaneous entry of the child approach variable, the conditional (± 1 SD) negative family affective expression variable, and the conditional negative family affective expression × child approach interaction in predicting internalizing or externalizing symptoms (Holmbeck, 2002). From the regression output, we derived unstandardized betas (slopes) and constants (intercepts). In graphing the interactions, we included the slopes and intercepts in regression equations that were 1 SD above and below the mean for negative family affective expression (Holmbeck, 2002).

Results

Participants who were missing any data (n = 474) did not differ (all ps > .05) from those with complete data (n = 301) in terms of age, t(773) = 0.11, Cohen’s d = .01; mother-reported negative family affective expression, t(697) = −.88, Cohen’s d = .07; father-reported negative family affective expression, t(707) = −1.33, Cohen’s d =.10; youth-reported negative family affective expression, t(698) = −0.24, Cohen’s d =.02; Time 1 internalizing problems, t(572) = −0.35, Cohen’s d = .03; Time 2 internalizing problems, t(416) = −0.24, Cohen’s d = .03; and Time 2 externalizing problems, t(416) = −1.67, Cohen’s d = .19. Participants who were and were not missing data differed in terms of Time 1 externalizing symptoms, t(572) = −2.37, p <.005, Cohen’s d = .20; child sex, χ²(1) = 8.91, p = .003, ϕ = .11; paternal diagnostic status, χ²(1) = 15.58, p < .005, ϕ = .14; and ethnicity, χ²(1) = 39.74, p < .005, ϕ = .23. Therefore, participants with missing data were more likely to be male and Caucasian, have a father who met diagnostic criteria for a psychiatric disorder, and exhibit elevated Time 1 externalizing symptoms.

Means, SDs, ns, and correlations are presented in Table 1. There was a low correlation between Times 1 and 2 internalizing symptoms (r = .20) and a moderate correlation between Times 1 and 2 externalizing symptoms (r = .63). The correlation from Time 1 to Time 2 externalizing symptoms was greater than the correlation for internalizing symptoms, Ζ = 4.61, p < .005¹. Temperamental approach was negatively correlated with Time 2 internalizing symptoms, but not correlated with Time 1 internalizing symptoms or externalizing symptoms at either time point. Child-reported negative family affective expression was positively correlated with mother-reported and father-reported negative family affective expression, though the magnitude was low. Mother- and father-reported negative family affective expression were moderately positively correlated.

Table 1.

Bivariate Correlations, Means, Standard Deviations, and n’s of Study Variables

Variable	1	2	3	4	5	6	7	8	9
1. Child Age	--
2. Child Approach Behaviors	.07	--
3. Father-Reported Negative Affective Expression	.01	.02	--
4. Mother-Reported Negative Affective Expression	−.03	−.03	.27***	--
5. Child-Reported Negative Affective Expression	.04	−.11**	.12**	.19***	--
6. Time 1 Internalizing Symptoms	.06	−.05	.13**	.14**	.02	--
7. Time 1 Externalizing Symptoms	−.13**	−.01	.12**	.13**	.13**	.65**	--
8. Time 2 Internalizing Symptoms	−.06	−.14**	.05	.16**	.18**	.20**	.26***	--
9. Time 2 Externalizing Symptoms	−.17***	−.05	.05	.09	.21***	.20**	.48***	.63***	--
M	10.95	19.47	10.52	9.80	10.34	5.10	15.72	5.26	13.06
SD	0.88	3.48	1.96	1.94	2.19	5.93	19.50	5.96	16.12
n	775	731	709	699	700	574	574	418	418

^*p <.05,

^**p <.01,

^***p <.001.

Hypothesis 1: Informant Reports of Negative Family Affective Expression

Consistent with the first hypothesis, youth reported higher levels of negative family affective expression compared to mothers, t(650) = −4.87, p < .005, Cohen’s d = .26; however, inconsistent with our first hypothesis, children’s and fathers’ reports of the family environment did not differ significantly, t(657) = 1.74, p = .082, Cohen’s d = .09. Unexpectedly, fathers reported higher levels of negative family affective expression than mothers, t(686) = 8.14, p < .005, Cohen’s d = .38. Independent samples t-tests also indicated that fathers without a SUD or psychiatric diagnosis reported higher levels of negative family affective expression (M = 10.88, SD = 2.02) compared to fathers with a SUD or psychiatric diagnosis (M = 10.27, SD = 1.88); t(593.71) = 4.08, p < .005, Cohen’s d =.31.

Hypothesis 2: Informant-Reported Negative Family Affective Expression × Child Approach

Results from the regression analyses with internalizing symptoms as the dependent variable (top of Table 2) indicate several significant main effects; lower levels of approach, higher levels of Time 1 internalizing symptoms, and higher child-reported negative family affective expression predicted Time 2 internalizing symptoms. The father- and child-reported negative family affective expression × child approach interactions predicted Time 2 internalizing symptoms; the mother-reported negative family affective expression × child approach interaction did not.

Table 2.

Summary of Hierarchical Regression Analyses Predicting Internalizing and Externalizing Symptoms from Negative Family Affective Expression and Child Approach

	Step 1					Step 2
Time 2 Internalizing Symptoms	B	SE B	β	R2	f2	B	SE B	β	R2	f2
Sex	−1.11	0.63	−0.09	0.10	0.11	−1.22	0.62	−0.09*	0.13	0.15
Age	−0.49	0.32	−0.07			−0.44	0.31	−0.07
Ethnicity	−0.32	0.74	−0.02			−0.40	0.72	−.0.03
Father Diagnostic Status	−0.03	0.60	−0.01			−0.21	0.59	−0.02
Child Approach	−0.61	0.29	−0.10*			−0.54	0.28	−0.09*
Father-Reported Negative Family Affective Expression	−0.10	0.31	−0.02			−0.00	0.30	0.00
Mother-Reported Negative Family Affective Expression	0.43	0.31	0.07			0.45	0.30	0.08
Child-Reported Negative Family Affective Expression	0.83	0.30	0.14**			0.86	0.30	0.15*
Time 1 Internalizing Symptoms	0.18	0.05	0.18**			0.17	0.05	0.17**
Father-Reported Negative Family Affective Expression × Approach						0.90	0.30	0.14**
Mother-Reported Negative Family Affective Expression × Approach						−0.27	0.31	−0.04
Child-Reported Negative Family Affective Expression × Approach						−0.70	0.28	−0.12*
Time 2 Externalizing Symptoms
Sex	0.87	1.58	0.02	0.30	0.43	0.85	1.56	0.02	0.33	0.50
Age	−2.01	0.79	−0.11*			−1.91	0.77	−0.10*
Ethnicity	0.45	1.87	0.01			0.15	1.84	0.00
Father Diagnostic Status	−3.17	1.48	−0.10*			−3.74	1.47	−0.11**
Child Approach	−0.09	0.70	−0.01			0.07	0.69	0.01
Father-Reported Negative Family Affective Expression	−0.29	0.75	−0.02			0.06	0.74	0.00
Mother-Reported Negative Family Affective Expression	−0.33	0.74	−0.02			−0.34	0.73	−0.02
Child-Reported Negative Family Affective Expression	2.59	0.74	0.16***			2.81	0.74	0.17***
Time 1 Externalizing Symptoms	0.37	0.04	0.45***			0.37	0.04	0.45***
Father-Reported Negative Family Affective Expression × Approach						1.96	.73	0.11**
Mother-Reported Negative Family Affective Expression × Approach						1.04	0.77	0.06
Child-Reported Negative Family Affective Expression × Approach						−2.27	0.68	−0.15*

^*p < .05,

^**p <.01,

^***p < .001.

Note. Step 2 included Step 1 variables and the three family affective expression × child approach interaction terms, one for each informant.

In regard to the father-reported negative family affective expression × child approach interaction, post hoc probing indicated the simple slope associated with higher negative family affective expression was not significant; that is, teachers reported similar internalizing symptoms among children regardless of their approach levels when fathers reported higher levels of negative family affective expression, inconsistent with hypothesis 2a (Figure 1A). In the context of lower father-reported negative family affective expression, teachers reported elevated internalizing symptoms among youth lower in approach than youth higher in approach, inconsistent with hypothesis 2b.

Figure 1.

A. Father-reported negative family affective expression (FFAE) moderates the relation between child approach behaviors and internalizing symptoms. B. Child-reported negative family affective expression (CFAE) moderates the relation between child approach behaviors and internalizing symptoms. C. Father-reported negative family affective expression (FFAE) moderates the relation between child approach behaviors and externalizing symptoms. D. Relation between child-reported negative family affective expression (CFAE), child approach behaviors, and externalizing symptoms.

Post hoc probing of the child-reported negative family affective expression × child approach interaction indicated that, unlike father-reported negative family affective expression, the slope was significant in the context of higher child-reported negative family affective expression (Figure 1B). Consistent with hypothesis 2a, among youth who reported higher levels of negative family affective expression, teachers reported elevated internalizing symptoms among youth lower in approach compared to youth higher in approach. In the context of lower child-reported negative family affective expression, the slope was not significant. Thus, teachers reported relatively low and similar levels of internalizing symptoms regardless of children’s approach levels in the context of lower child-reported negative family affective expression, consistent with hypothesis 2b.

With regard to externalizing symptoms (bottom of Table 2), there were several significant main effects, indicating that younger child age, having a father without a lifetime history of psychiatric disorders, higher child-reported negative family affective expression, and higher Time 1 externalizing symptoms were associated with Time 2 externalizing symptoms². Child temperamental approach was not predictive of Time 2 externalizing symptoms. As with internalizing symptoms, only the father- and child-reported negative family affective expression × child approach interactions predicted Time 2 externalizing symptoms³.

Post hoc probing of the father-reported negative family affective expression × child approach interaction indicated that the simple slope was not significant in the context of higher father-reported negative family affective expression (Figure 1C). Inconsistent with hypothesis 2a, when exposed to higher father-reported negative family affective expression, teachers reported similar levels of externalizing symptoms regardless of children’s approach levels. In the context of lower father-reported negative family affective expression, teachers reported elevated externalizing symptoms among youth lower in approach than youth higher in approach, inconsistent with hypothesis 2b.

Post hoc probing of the child-reported negative family affective expression × child approach interaction in predicting externalizing problems indicated that the simple slopes for higher and lower negative family affective expression were not significant (Figure 1D). Inconsistent with hypothesis 2a and 2b, teachers reported similar levels of externalizing symptoms among youth varying in approach, regardless of the levels of negative family affective expression that they reportedly experienced.

Discussion

Although temperamental approach and negative family affective expression independently predict adolescent internalizing and externalizing symptoms (e.g., Betts, Gullone, & Allen, 2009; Dollar & Stifter, 2012; Halberstadt et al., 1999; Muris et al., 2007; Queen et al., 2013), the current study is the first to examine whether negative family affective expression differently predicts symptoms based on (a) youth’s temperamental approach and (b) the informant used to index family affective expression. In the context of higher youth-reported negative family affective expression, teachers reported higher internalizing symptoms among youth lower in approach than youth higher in approach. We expected higher negative family affective expression to be associated with higher teacher-reported externalizing symptoms among youth higher in approach, and higher internalizing symptoms among youth lower in approach (see hypothesis 2a); however, this was not the case. When fathers reported higher, and youth reported lower, negative family affective expression, teachers reported similar symptom levels regardless of youth’s levels of approach. However, when fathers reported lower negative family affective expression, teachers reported greater externalizing and internalizing symptoms among youth lower in approach, inconsistent with hypothesis 2b.

In support of the first hypothesis, youth reported higher levels of negative family affective expression than mothers, though unexpectedly, levels reported by youth and fathers did not differ. These results are aligned with research indicating that adolescents report higher negative family functioning (e.g., family conflict, control) than mothers, the effect size for which is low (De Ross et al., 1999), but conflicts with other work indicating moderate effect sizes for the difference in mothers’ and children’s reports of parenting (e.g., degree of behavioral and psychological control; Korelitz & Garber, 2016). Our results conflict with research indicating that adolescents report higher levels of negative parenting/family processes compared to fathers, the effect size for which is low (Korelitz & Garber, 2016). The present findings are also consistent with research indicating low (e.g., Cavendish et al., 2014; Ohanessian et al., 1995) to moderate (e.g., Korelitz & Garber, 2016; Stuart & Jose, 2012) positive associations between youth- and parent-reported family processes. Adolescents’ desire for independence may conflict with parental rules or contribute to problematic family interactions, and youth may thus perceive their family environment as conflictual (Grotevant & Cooper, 1986; Ohannessian et al., 1995). In contrast to adolescents, mothers may have a vested interest in maintaining family harmony and accordingly report lower levels of negative family affective expression (Noller & Callan, 1986; Stuart & Jose, 2012). Relative to mothers, fathers may take on more of a disciplinarian role, which may evoke conflictual exchanges particularly among adolescent youth (Halberstadt et al., 1999). Thus, the relationships between fathers and adolescents may be particularly contentious, leading both adolescents and fathers to report elevated negative family affective expression.

In addition, finding moderate positive associations between mother and father reports of family processes is consistent with some work (Queen et al., 2013), but conflicts with other research indicating high positive correlations between mother- and father-reported family features (e.g., De Ross et al., 1999; Ohannessian et al., 1995). Moderate associations among maternal and paternal reports of family affective expression likely may involve caregivers’ adoption of gender-role stereotypes, differences in what constitutes appropriate emotional expression, or differences in experiences or observations within the family unit (De Los Reyes & Ohannessian, 2016). For example, one parent may have frequent contentious exchanges with their children during which their partner is not present or the partner may identify these interactions as less significant, which may result in differences in parents’ perceptions of family functioning. Moreover, mothers may be more encouraging of children’s expression of positive and negative emotions, and may thus report lower levels of problematic family affective expression. In contrast, fathers may react more negatively to youth’s negative emotional expressions (Garside & Klimes-Dougan, 2002; Klimes-Dougan et al., 2007); this behavior may create greater conflict in the family and fathers may thus perceive higher negative family affective expression (Noguchi & Ollendick, 2010).

Consistent with hypothesis 2a, in the context of higher youth-reported negative family affective expression, teachers reported heightened internalizing symptoms among youth lower in approach compared to youth higher in approach. This finding mirrors previous results that indicate in the context of higher negative family affective expression, youth higher in negative affectivity and inhibition exhibited elevated internalizing symptoms (Hilt et al., 2012; Marakovitz et al., 2011). The current results also indicate that in the context of lower father-reported negative family affective expression, teachers reported higher internalizing and externalizing symptoms among youth lower in approach than youth higher in approach, inconsistent with hypothesis 2b. We expected that in the context of higher negative family affective expression, teachers would report greater levels of externalizing behaviors among youth higher in approach. Our results are consistent with work examining similar constructs that have linked (a) child shyness to peer reports of children’s anger (Eisenberg, Pidada, & Liew 2001), and (b) conflicted shyness (fearfulness to interact socially) with parent-reported externalizing symptoms among boys (Kopala-Sibley & Klein, 2016). Youth lower in approach, who are more temperamentally fearful or inhibited, may react aggressively when faced with real or perceived provocation (Vitaro, Barker, Boivin, Brendgen, & Tremblay, 2006). It is also possible that the increased social classroom demands (e.g., expressing oneself in class) may result in anger, frustration, and acting out among youth lower in approach who may have difficulty asserting themselves in the classroom (Eisenberg et al., 2005); hence, teachers may report elevated externalizing symptoms among these youth relative to youth higher in approach.

Findings of significant moderation among father- but not mother-reported family affective expression may be because mothers and fathers often have discrepant perspectives regarding appropriate emotional expression. For example, some research indicates that relative to mothers, fathers tend to discourage, react more punitively to, and minimize youth’s negative emotional displays (e.g., sadness) (Cassano, Perry-Parish, Zeman, 2007; Garside & Klimes-Dougan, 2002). Biases resulting from psychological difficulties also may influence fathers’ perceptions of family functioning, particularly given the recruitment approach for the present sample. Indeed, fathers with a psychiatric diagnosis reported lower negative family affective expression than fathers without a SUD or psychiatric diagnosis, suggesting that fathers with a lifetime history of psychological difficulties may under-report negative family processes because they are outside of the home (e.g., for treatment, legal difficulties) and consequently do not observe family functioning, experience biases secondary to their diagnoses, or report fewer negative family processes because their histories have influenced their threshold for identifying behavior as problematic (Lander, Howsware, & Byrne, 2013). Nevertheless, although fathers with a lifetime history of psychiatric disorders reported lower levels of negative family affective expression relative to fathers without a history of psychological problems, the means of negative family affective expression for both groups of fathers were relatively similar and it would be useful to determine whether this difference can be replicated in other samples.

Our results also indicate that teacher-reported externalizing symptoms had greater continuity over time compared to teacher-reported internalizing symptoms. It is notable that Time 1 externalizing symptoms had the greatest prediction to Time 2 externalizing symptoms (β = .45, p < .005) compared to the other variables in Steps 1 and 2 of the regression equations. Given the variance accounted for by Time 1 externalizing symptoms, the ability of other variables to predict Time 2 symptoms was attenuated, which may be why temperamental approach did not predict Time 2 externalizing symptoms, counter to prediction. Greater continuity of externalizing symptoms also may be a result of the developmental period considered. For example, during adolescence, the neurological mechanisms underpinning cognitive control are still developing, which may contribute to risk-taking, impulsive, and/or oppositional behaviors that are likely to persist into early adulthood (Steinberg, 2008). Moreover, teachers may be more aware of aggressive behaviors that interfere with the classroom environment and report higher externalizing symptoms relative to internalizing symptoms, which are generally less observable and may be more episodic.

In addition, it is notable that father diagnostic status predicted externalizing symptoms more so than children’s temperamental approach (Δβ = .11, p < .05). More specifically, fathers without a SUD or psychiatric history were more likely to have children with higher externalizing symptoms. This finding was unexpected given that paternal SUD is associated with adolescent externalizing and internalizing symptoms (Blackson et al., 1999; Ohannessian et al., 2004). It is possible that among families with a father with a history of psychiatric problems, adolescents may feel a greater obligation towards the family or take on more adult roles, which in turn may attenuate expected normative developmental changes (e.g., risk taking, externalizing behaviors) associated with adolescence. Moreover, children may be reluctant to behave defiantly in the school setting for fear of upsetting their father and/or contributing to their father’s psychological difficulties. Alternatively, fathers who received psychological treatment may be more likely to engage in positive parenting behaviors that may attenuate externalizing problems, though the current data did not allow us to test this possibility. Regardless of the reason for this association, these results highlight that father’s mental health status may be a unique predictor of externalizing problems and that assessments and interventions should consider the psychological functioning of fathers in addition to other family members.

The current study has several strengths. First, although studies typically use single informants to assess family functioning, we incorporated youth, mother, and father reports of negative family affective expression to capture differences in family perceptions that may be differentially predictive of youth’s symptoms (De Los Reyes et al., 2016) and to minimize biases inherent in mono-rater methods. Second, we investigated whether reports of negative family affective expression differed depending on the informant report considered. Third, we examined whether a relatively understudied family feature, family affective expression, was associated with both internalizing and externalizing symptoms. Fourth, we explored whether family affective expression moderated the relation between youth temperamental approach and internalizing and externalizing symptoms. Identifying youth lower in approach who may be at risk for symptom development in the context of more problematic family affective expression can guide intervention work aimed at family processes associated with youth’s symptoms. Fifth, we used the same measure to assess perceptions of family functioning to ensure that differences in assessment measures did not contribute to informant discrepancies of family processes (De Los Reyes, Thomas, Goodman, & Kundey, 2013).

There are also limitations to consider. One limitation is that the sample consisted of mostly Caucasian males and over half of the sample included youth with fathers with a lifetime diagnosis of SUD or a psychiatric disorder. Thus, the generalizability of our findings may be limited and future work should consider replicating these results in other samples (e.g., community, clinical). Additionally, individuals with missing data were significantly different from those with present data, although FIML was used to address missing data based on observed data. Further, the reliabilities of the family affective expression and approach subscales were relatively low; future research should replicate our results in other samples or use other scales that may have better psychometric properties (e.g., family expressiveness subscale of the Family Environment Scale; Moos & Moos, 1974; the sociability subscale of the EASI-II scale; Buss & Plomin, 1975). Given mixed findings regarding the outcomes among youth varying in temperamental approach in the context of negative family environments (e.g., Kopala-Sibley & Klein, 2016; Marakovitz et al., 2011), future research should incorporate multiple indices of temperamental approach such as questionnaires (e.g., Behavioral Inhibition and Approach Scales; Carver & White, 1994; Approach Subscale of the DOTS-R, Windle & Lerner, 1986) or observations/lab tasks to examine whether the method used to assess temperamental approach influences study findings. In addition, we considered data derived from questionnaires only; thus, the pattern of findings may have been influenced by shared method variance or informant social desirability. Indeed, some work indicates that mothers and fathers have a greater social desirability bias when reporting about negative, as opposed to positive, parenting (Bornstein et al., 2015). Future work thus should consider using multiple informants (e.g., parents, siblings) to assess negative family affective expression, youth temperamental approach, and adolescent symptoms. In addition, bidirectional relations between youth symptoms and/or approach behaviors and family affective expression likely occur. Symptoms and temperamental approach may predict levels of family affective expression in a reciprocal, transactional process and future research should consider exploring these pathways using a cross-lagged design. In addition, we did not examine whether temperamental approach and family negative affective expression interacted to predict within-person change in symptoms. Future work should consider how informant perceptions of the family context and youth temperament interact to predict between- and within-person changes in symptoms across developmental periods using a multi-level modeling framework.

The current study adds to the literature by indicating that negative family affective expression is associated with internalizing and externalizing problems among youth varying in approach during adolescence and that family functioning differentially predicted symptoms depending on the informant considered. The low to moderate correspondence among informants of negative family affective expression suggests that assessments and interventions should address discrepancies among family members regarding the affective quality of the family environment. Clinicians that target the expression and regulation of negative emotions within the family context may help family members to express emotions appropriately and problem-solve in affective situations across different settings. Such an approach may promote greater understanding and communication among family members, which in turn may influence youth and family adjustment (Guo & Slesnick, 2013; O’Gorman, 2012).

Our findings indicate that when youth reported higher negative family affective expression, teachers reported higher internalizing symptoms among youth lower in approach than youth higher in approach; when fathers reported lower negative family affective expression, teachers reported elevated internalizing and externalizing symptoms among youth lower in approach than youth higher in approach. These findings suggest that clinicians should not only include different informants of family processes during the assessment process and multiple family members in the intervention, but also assess youth’s child-specific features (e.g., temperamental approach). This approach may help to identify whether youth varying in temperamental features are differentially susceptible to conflictual family interactions and symptom development. Future work should also investigate factors that contribute to divergent perceptions of the family environment, such as informant diagnostic history, that may differentially predict youth’s externalizing and internalizing symptoms. In addition, future studies should include multiple informants’ reports of other family processes (e.g., cohesion, conflict) to identify whether informants differentially predict internalizing and externalizing symptoms among youth varying in other temperamental features (e.g., effortful control, negative affectivity).