Bidirectional Relations between Parent Warmth and Criticism and the Symptoms and Behavior Problems of Children with Autism

Autism spectrum disorder (ASD) is a lifelong neurodevelopmental condition defined by impairments in social communication and restricted and repetitive interests and behaviors (American Psychiatric Association, 2013). In addition to these core ASD symptoms, nearly two-thirds of children with ASD exhibit co-occurring emotional and behavioral problems including inattention, disruptive behavior, and anxiety (Maskey, Warnell Parr, Le Coutuer, & McConachie, 2013). There is marked variability in the frequency and severity of the core ASD symptoms, as well as in co-occurring emotional and behavioral problems, both between children with ASD and within a child with ASD across time (e.g., Georgiades et al., 2013; Visser, 2017). The transactional theory of child development, originally proposed by Sameroff and Chandler (1975), posits that the family environment, including the emotional quality of the parent-child relationship, are linked in bidirectional ways with child functioning (e.g., Combs-Ronto, Olson, Lunkenheimer, & Sameroff, 2009; Sameroff & Chandler, 1975). Yet, to date, little research has examined the ways in which the emotional quality of the parent-child relationship both shapes and is shaped by the symptoms and emotional and behavior problems of children with ASD. Elucidating these pathways is critical for directing interventions aimed at fostering optimal child development and family well-being in families of children with ASD.

In support of the transactional theory of child development (Sameroff & Chandler, 1975), there is substantial evidence from longitudinal studies on the general population that child development and family environment alter each other in ongoing reciprocal ways. For example, the emotional quality of the parent-child relationship has been found to predict later change in multiple domains of child functioning, including academic, behavioral, and psychological outcomes (e.g., for review see Wilson & Durbin, 2012). In the opposite direction, although less studied, there is also evidence from longitudinal studies on the general population that child emotional and behavioral functioning predict later change in the emotional quality of the parent-child relationship. For example, McDaniel and Radesky (2018) found that child externalizing behavior problems predicted greater parent technology use during parent-child interactions, suggesting that parents, stressed by their child’s difficult behavior, may then withdraw from parent–child interactions with technology.

Across previous research studies, the emotional quality of the parent-child relationship has been measured in varied ways, including parent-reported surveys (e.g., Booth-LaForce & Oxford, 2008), parent interviews (e.g., Colman, Hardy, Albert, Raffaelli, & Crockett, 2006), observed parent-child interactions (e.g., Scaramella, Sohr-Preston, Mirabile, Robison, & Callahan, 2008), and Five Minute Speech Samples (FMSS; Magaña et al., 1986). In the FMSS, parents talk about their child and the parent-child relationship for five minutes. The emotional quality of the parent-child relationship is rated using coded dimensions (e.g., warmth and criticism). The FMSS was originally developed as a brief method to assess expressed emotion in the family relationship of parents and adult children with mental health conditions (Magaña et al., 1986), and uses coded dimensions such as warmth and criticism. These early studies found that the expressed emotions of the caregivers predicted relapse in psychiatric patients. Subsequent studies provided evidence that expressed emotion also predicted symptom levels and relapse rates across a range of other health conditions (e.g., Parkinson’s disease, schizophrenia [Wearden, Tarrier, Barrowclough, Zastowny, & Rahill, 2000; Leff & Vaughn, 1981]). The FMSS has sense been used to examine expressed emotion in parent-child relationships across a variety of populations. An advantage of the FMSS over other approaches is that it reduces response biases by capturing open-ended responses based on a general prompt, rather than asking value-laden questions that can trigger socially desirable responses. Moreover, the FMSS has been found to be associated with parent emotions and behaviors in observations of actual parent-child interactions (Weston, Hawes, & Pasalich, 2017) and has high reliability and validity in diverse populations (Magaña et al., 1986; Van Humbeeck, Van Audenhove, De Hert, Pieters, & Storms, 2002), including in parents of children with development disabilities (e.g., Hastings, Daley, Burns, & Beck, 2006) and specifically ASD (Greenberg, Seltzer, Hong, & Orsmond, 2006). In general population samples, the FMSS was also found to be robust to systematic differences based on child gender (Delvecchio et al., 2014; Kwon et al., 2006), child age (McCleary & Sanford, 2002), and family socio-economic status (Delvecchio et al., 2014).

To date, family research in the field of ASD has focused on describing the impact of the child ASD on parents, usually mothers. Less research has been aimed at investigating the reciprocal influence of parents on the development of the child with ASD. Yet, there is evidence that transactional connections between the parent-child relationship and child functioning are similar in families of children with vs. without developmental delays (Combs-Ronto et al., 2009; Neece, Green, & Baker, 2012; Rodas, Chavira, & Baker, 2017). The handful of longitudinal studies that have examined the emotional quality of the parent-child relationship using the FMSS and the functioning of children with ASD have been limited to the mother-child relationship, and samples of adolescents and adults with ASD. Smith and colleagues (2008) found that mother FMSS warmth (i.e., expressions of concern, empathy, and interest) and praise (i.e., expressions of praise) toward her son or daughter were negatively associated with the adolescent or adult with ASD’s severity of repetitive behavior 18 months later. In the opposite direction, the adolescent or adult son/daughter with ASD’s severity of repetitive behavior was negatively associated with mother FMSS praise toward the son or daughter 18 months later. In contrast, using this same sample, Baker, Smith, Greenberg, Seltzer, and Taylor (2011) found that mother FMSS rating of criticism (i.e., negative remarks and critical expressions) toward the son or daughter positively predicted an increase in the adolescent or adult son/daughter with ASD’s emotional and behavioral problems across 7 years, but not vice versa. Thus, it is possible that positive (i.e., warmth) aspects of mother-child relationship are bidirectionally related to child functioning but negative aspects (i.e., criticism) are unidirectional, with the parent-child relationship influencing the functioning of the son/daughter with ASD but not vice versa.

Building on these findings, it is important to examine the directional associations between the emotional quality of the parent-child relationship and the functioning of children with ASD at other developmental stages. There is also a need to extend prior work by examining the directional associations between the emotional quality of the father-child relationship and the functioning of children with ASD. Both mothers and fathers of children with ASD report elevated levels of parenting stress relative to comparison groups (Hayes & Watson, 2013; Pisula & Porębowicz-Dörsmann, 2017; Valicenti-McDermott et al., 2015); thus, fathers are not immune to the effects of challenging child symptoms and emotional and behavioral problems. Moreover, in non-ASD populations, the quality of the father-child relationship has been shown to be linked to child functioning independent of the mother-child relationship (e.g., Ferriera et al., 2016). However, in cross-sectional research involving young children with ASD (aged 2–10 years), child functioning (ASD symptom severity) had a stronger association with self-reported perceptions of attachment in the mother-child than father-child relationship (Goodman & Glenwick, 2012). This finding may, in part, be driven by findings that mothers of children with ASD report more time per day in childcare than fathers of children with ASD (Hartley et al., 2014), and are less likely to be in paid employment than both fathers of children with ASD and mothers of children without ASD, with childcare being the most commonly cited reason for not being in the paid labor force (Callander & Lindsey, 2018; Hartley et al., 2014). As a result of spending more time with the child with ASD, it is possible that the association (in both directions) between the emotional quality of the mother-child relationship and child functioning may be stronger than the association between the emotional quality of the father-child relationship and child functioning.

Finally, previous research examining the emotional quality of the parent-child relationship using the FMSS and the functioning of children with ASD used parent-reported measures of child functioning (Abbeduto, 2004; Ciciolla, Gerstein, & Crnic, 2014; Greenberg, Seltzer, Krauss, Chou, & Hong, 2004; Orsmond, Seltzer, Greenberg, & Krauss, 2006). As a result, associations between parent-child relationship quality and child functioning may be conflated by shared method variance of having the same single reporter for both the FMSS and child functioning measures.

In the present study, we examined the bidirectional associations between FMSS warmth and criticism in the mother-child and father-child relationship and severity of ASD symptoms and emotional and behavioral problems in children with ASD across 12 months, using three time points of data collection. The current study built on previous research by examining these associations during middle and later childhood, assessing both the mother-child and father-child relationship, and including a teacher-reported measure of the child with ASD’s functioning. In total, 159 families of children with ASD (aged 6–13 years at Time 1) were included in the study. There were two study questions: 1) How is the emotional quality of the parent-child relationship, as assessed by the FMSS, associated with child severity of ASD symptoms and emotional and behavioral problems across 2 years? 2) Do the above associations differ for mothers versus fathers? Based on previous research (Baker et al., 2011; Smith et al., 2008), we hypothesized that a higher level of FMSS warmth and a lower level of FMSS criticism toward the son/daughter with ASD would be associated with a lower severity of child ASD symptoms and emotional and behavioral problems at the subsequent time point (12 months later). The quality of the mother-child relationship (i.e., mother FMSS) was predicted to have stronger associations (in both directions) with the child’s severity of ASD symptoms and emotional and behavioral problems than the father-child relationship (i.e., father FMSS), based on evidence that mothers spend more time in childcare than fathers in families of children with ASD (Callander & Lindsay, 2018; Hartley et al., 2014).

Method

Participants

Participants were part of an ongoing longitudinal study investigating family dynamics in the context of child ASD. Recruitment occurred through: 1) mailings to 10 schools and 3 childcare programs; 2) fliers posted at ASD clinics and community settings (e.g., libraries and grocery stores), and 3) by mailings sent to families of children with ASD in an ASD research registry (see citation removed for blind review). Interested families contacted the study team and were screened for inclusion criteria, including: 1) being aged ≥ 21 years; 2) being the parent of a child diagnosed with ASD who was aged 5–12 years; 3) being in a committed couple relationship (≥ 3 years) and both partners agreeing to be in the study; and 4) providing documentation of the child’s ASD diagnosis via records of a medical or educational evaluation that included the Autism Diagnostic Observation Schedule (Lord et al., 2000; Lord et al., 2012). In addition, parents completed the Social Responsiveness Scale (SRS-2; Constantino and Gruber, 2012) to verify the child’s current ASD symptoms. Five children with ASD received an SRS-2 Total t-score > 60; however, after reviewing medical or educational records, these families were included in the study as they were deemed to meet criteria for ASD.

Procedure

Figure 1 shows the Participant Flow Diagram. The FMSS sample was included in the study protocol at the second cycle of data collection. Current analyses focus on the 159 families of children with ASD who had FMSS data collected during the second cycle data collection within the larger study. For the purposes of the current analyses, we refer to the time points of FMSS data collection as Time 1, Time 2, and Time 3, which reflect the number of repeated measurement occasions for this specific measure in this study.

Figure 1.

Participant Flow Diagram

Current analyses are based on parent, child, and family sociodemographic information collected at Time 1. The FMSS (completed with both mothers and fathers) and teacher ratings of the child’s ASD symptoms and emotional and behavioral problems were completed at Time 1, Time 2, and Time 3. Each time point was spaced approximately 12 months apart (range: 12.0 to 15.2 months). At each time point, parents participated in a 2.5-hour lab or home visit in which they were interviewed about family socio-demographics and independently completed the FMSS, along with other measures. Parents were each paid $50 at each time point. Parents also provided contact information for a teacher or support staff person who knew the target child well, interacted with the child often, and could report on their child’s ASD symptoms and emotional and behavioral problems. Teachers or support staff were emailed and asked to complete questionnaires electronically. They were given a $5 gift card for participation. Informed consent was obtained from all individual participants prior to involvement in the study. All study activities were approved by the UW-Madison Institutional Review Board and are consistent with the 1964 Helsinki declaration and its later amendments.

Measures

Family Socio-demographics.

Family socio-demographics were reported on by parents at Time 1 and were included in models to control for any effect on study variables of interest. The child with ASD’s date of birth was used to calculate child age in years. Parents reported on the child’s sex, coded as female (1) or male (2). Children with ASD were considered to have intellectual disability (ID) if they had a medical diagnosis of intellectual disability and/or met criteria based on review of medical and/or educational records reporting IQ and adaptive behavior testing. Child ID status was coded as ID (1) or no ID (0). Parents reported on their household income, coded from 1–14, starting at ≤$9,999 (1) and increasing by $10,000 to $20,000 intervals to ≥$160,000 (14). Parents also reported on whether any of their other children had a neurodevelopmental or psychological condition. Based on data collected for each additional child in the family, a variable for having an additional child with a neurodevelopmental or psychological condition was created and coded as yes (1) or no (0).

Parent-Child Relationship.

The FMSS (Magaña et al., 1986) was administered to mothers and fathers at Time 1, Time 2, and Time 3 to assess warmth and criticism in the parent-child relationship. Mothers and fathers were individually asked to speak about their child with ASD for 5 minutes, addressing what kind of person their child is and how they got along together. The speech sample, was audio recorded, transcribed, and then coded using established criteria (Magaña et al., 1986; Vaughn & Leff, 1976) by a trained FMSS rater who was blind to study questions. An overall warmth rating based on (a) tone of voice; (b) spontaneity of expression of sympathy, concern, and empathy; and (c) expression of interest in the child with ASD, was globally considered across the entire speech sample on a 3-point scale: ‘no warmth’ (0), ‘borderline warmth’ (3), ‘high warmth’ (5) (according to Greenberg et al., 2006; Smith et al., 2008; Vaughn & Leff, 1976). An overall criticism rating was globally considered across the entire speech sample on a 3-point scale: ‘no criticism’ (0), ‘borderline criticism’ (1), or ‘high criticism’ (2) for each mother and father based on their responses in the FMSS (according to Magaña et al., 1986). Mothers and fathers were rated as ‘high criticism’ if they (a) made a negative opening remark, (b) negatively described their relationship with their child with ASD, or (c) made one or more critical comments about their child with ASD during the FMSS. Parents were rated as ‘borderline criticism’ if they made one or more statements of dissatisfaction about their relationship with their child with ASD but did not satisfy the requirements of high criticism and were rated as ‘no criticism’ if they did not make any critical comments throughout the FMSS. Past research has found a significant relation between these FMSS codes and observed criticism in actual parent-child interactions (McCarty, Lau, Valeri, & Weisz, 2004; Weston et al., 2017). All FMSSs were coded by the same trained FMSS rater. Inter-rater reliability of this trained rater with twelve other trained raters was high (mean inter-rater agreement was 93.3% [range 80–100%]).

Child ASD Symptoms and Emotional and Behavioral Problems.

The severity of the child’s ASD symptoms were rated by a teacher or support staff at Time 1, Time 2, and Time 3. Teachers or support staff completed the 65-item Social Responsiveness Scale – Second Edition (SRS-2; Constantino & Gruber, 2012) which rates the child’s ASD symptoms during the past 6 months from ‘Not True’ (1) to ‘Almost Always True’ (4). The SRS-2 Total t-score was used in analyses. The SRS-2 has been shown to have high internal consistency (Bruni, 2014). Cronbach’s alphas for teacher ratings on the SRS-2 in the current sample were all > .83 across time points. Although not the focus of current analyses, mothers and fathers also completed the SRS-2. Across time points, mother (r = .36–.50, p = .000) and father (r = .39–.56, p = .000) ratings of child ASD symptoms were moderately correlated with teacher ratings. In line with previous research that has found that parent ratings on the SRS-2 are often higher than those of teachers (e.g., Donnelly et al., 2018), mothers reported significantly higher SRS-2 scores than teachers at Time 1 (t = 6.08, p <.001), Time 2 (t = 6.41, p <.001), and Time 3 (t = 6.09, p <.001) in the current sample. Fathers also reported significantly higher scores on the SRS-2 than teachers at Time 1 (t = 4.95, p <.001), Time 2 (t = 4.90, p <.001), and Time 3 (t = 5.09, p <.001).

The frequency and severity of the child with ASD’s emotional and behavioral problems was also assessed at Time 1, Time 2, and Time 3 through the Teacher Rating Form (TRF) of child behaviors (Achenbach & Rescorla, 2001). The TRF uses a 3-point scale with responses ‘Not True’ (0), ‘Somewhat or Sometimes True’ (1) and ‘Very True or Often True’ (2). Ratings on the total problem items are summed to create the Total t-score, a measure of the total frequency and severity of emotional and behavioral problems. The TRF has been shown to have good internal consistency and construct validity in ASD samples (Duarte, Bordin, Oliveira, & Bird, 2003). In the current sample, the Cronbach’s alphas for TRF Total t-score were all > .93 across time points. Mothers and fathers also completed the parent versions of the Achenbach and Rescorla (2001) measure. Across time points, mother (r = .35–.44, p = .000–.001) and father (r = .36–.39, p = .001) ratings of child total behavior problems were moderately correlated with teacher ratings. There were no significant differences between parent (CBCL Total t-score) and teacher (TRF Total t-score) ratings of child behavior problems in the current sample for mother or father ratings.

Data Analysis Plan

Descriptive data and boxplots were used to examine the distribution of study variables. Paired samples t-tests were conducted to investigate mother-father differences in FMSS warmth and criticism within families. Structural equation model (SEM) in Analysis of Moment Structures (AMOS, version 24.0; Arbuckle, 2016) software was used to examine study questions. Prior to conducting the SEMs, we examined the association between family socio-demographics and our main study variables (FMSS and child functioning). Socio-demographic variables significantly associated with our main study variables were then controlled for in the SEM. Specifically, mother and father FMSS warmth and criticism and teacher SRS-2 and TRF scores were regressed on family socio-demographics using multiple linear regression. The unstandardized residual values were saved and then analyzed in the SEM analyses. To assess study questions, four cross-lagged panel SEMs were conducted in AMOS (Arbuckle, 2016) to test the bidirectional cross-lagged effects between FMSS warmth and criticism and teacher-reported child ASD symptoms and emotional and behavioral problems. AMOS analyzes the incomplete dataset via full-information maximum likelihood estimation (FIML), a robust method for handling missing data compared to imputation (Schlomer, Bauman, & Card, 2010). Separate models were run for FMSS warmth and criticism to reduce model complexity and avoid multicollinearity. Multicollinearity was assessed with variables of interest at each time point, including mother and father warmth and criticism, teacher SRS-2, and TRF for each time point. The Variance Inflation Factor (VIF) was used as an indicator of multicollinearity, which occurs when there is a strong linear correlation among the predictor variables. When these are highly correlated, it becomes difficult for the model to determine which predictor is affecting the response. The first two SEMs included mother and father FMSS (warmth and criticism in separate models) and teacher-reported SRS-2 Total t-score. The second two SEMs included mother and father FMSS (warmth and criticism in separate models) and teacher-reported TRF Total t-score. As cross-lagged panel models, all models examined the effects of the preceding time point (i.e., 12 months earlier) in both directions (e.g., FMSS predicting SRS-2 Total t-score 12 months later and vice versa).

Results

Sample Demographics

Sample demographics are displayed in Table 1. Of the 159 families (n = 318 parents) included in analyses, 35.2% of the children with ASD had intellectual disability, 86.2% were male, 76.7% were white, non-Hispanic, and their average age was 9.07 years (SD = 2.33) at the first time point included in current analyses. Mothers’ average age was 39.53 years (SD = 5.55) and 73.6% had a college degree. Fathers average age was 41.66 years (SD = 6.19), and 69.2% had a college degree. The average household income was $80–89K (SD = ~30K) and the average parent couple relationship length was 10.97 years (SD = 5.14). In three families, the child with ASD had been adopted (≥ 5 years prior). In nine families, one parent was a stepparent. More than one-third (39.8%) of families had an additional child with a neurodevelopmental disability or psychological disorder.

Table 1.

Parent, child, and family sociodemographic variable frequencies, proportions, means, and standard deviations

	Mother	Father	Child/Family
Variable	Mean (SD)	Mean (SD)	Mean (SD)
Parent Age	39.53 (5.55)	41.66 (6.19)
Child Age			9.07 (2.33)
Rel. Length			10.97 (5.14)
Household Income			8.94 (3.01)
Variable	Frequency (%)	Frequency (%)	Frequency (%)
Parent Race/Ethnicity (White, non-Hispanic)	143 (89.9)	141 (88.7)
Parent Education
Less than high school diploma or GED	3 (1.9)	10 (6.3)
High school diploma or GED	9 (5.7)	18 (11.3)
Some college	30 (18.9)	21 (13.2)
Associate’s Degree	16 (10.1)	17 (10.7)
Bachelor’s Degree	57 (35.8)	54 (34.0)
Some graduate school	10 (6.3)	8 (5.0)
Master’s degree	23 (14.5)	25 (15.7)
Doctoral/Law/Medical or equivalent degree	11 (6.9)	6 (3.8)
Child Gender (male)			138 (86.2)
Child Intellectual Disability status			56 (35.2)
Child Race/ethnicity (White, non-Hispanic)			122 (76.7)
Additional Child with neurodevelopmental disability or psychological disorder.			63 (39.8)

Note. Parents reported on parent and child age (calculated from date of birth and coded in years), couple relationship length (coded in years); household income (coded from 1–14, starting at ≤$9,999 [1] and increasing by $10,000 to $20,000 intervals to ≥$160,000 [14]), parent and child race/ethnicity (0 = White, non-Hispanic, 1 = other), parent education, child gender (1 = female, 0 = male), child with ASD’s intellectual disability status (1 = yes, 0 = no), and whether or not the family had an additional child with a neurodevelopmental disability or psychological disorder(1 = yes, 0 = no).

These 159 families did not significantly differ from the 24 families enrolled in the larger study but who did not participate at Time 1 in household income, child age, teacher-reportedSRS-2 Total t-score, or teacher-reported TRF Total t-score. Of these 159 families, those who did not participate at Time 2 or Time 3 differed in were more likely to have a target child who was male (F = 10.49, p = .002), fathers were less likely to be White, non-Hispanic (F = 15.20, p < .001)., and there was less likely to be an additional child with a neurodevelopmental disability or psychological disorder in the family (F = 5.08, p = .026).

Preliminary Analyses and Zero-Order Correlations

As shown in Table 2, paired-sample t-tests indicated that, within families, mothers expressed a significantly higher level of FMSS criticism than fathers at Time 1 (t = 2.49, p = .014) and Time 2 (t = 2.57, p = .011). There was not a significant mother-father difference in FMSS warmth at any of the time points. Table 3 presents Pearson correlations examining the associations between family socio-demographics (i.e., child age, sex, ID status, household income, and presence of additional child with neurodevelopmental or psychological condition), mother and father FMSS warmth and criticism, and teacher SRS-2 and TRF scores. Child age was significantly negatively associated with teacher SRS-2 Total t-score (r = −.24, p = .016) at Time 2. Child sex was significantly associated with teacher SRS-2 Total t-score at Time 1 (r= − .31, p = .005) and Time 2 (r = −.27, p = .010), with males having lower SRS-2 scores. Child ID status was significantly associated with teacher SRS-2 Total t-score at Time 1 (r = .45, p < .001), Time 2 (r = .50, p < .001), and Time 3 (r = .46, p < .001), and teacher TRF Total t-score at Time 3 (r = .34, p = .002). Household income was significantly associated with teacher SRS-2 Total t-score at Time 1 (r = −.27, p = .007) and Time 3 (r = −.29, p = .007). Having an additional child with a neurodevelopmental disability or psychological disorder was not significantly associated with teacher ratings of child functioning or emotional and behavioral problems at any time point. Mother and father FMSS warmth and criticism were not significantly associated with socio-demographic variables. Multiple linear regressions were used to regress mother and father FMSS warmth and criticism and teacher SRS-2 and TRF scores separately onto child age, child sex, child ID status, household income, and presence of multiple affected children. The unstandardized residual values from these analyses were then analyzed using SEM.

Table 2.

Means, standard deviations, and paired samples t-tests of primary study variables

	Time 1	Time 2	Time 3
Variable	Mean (SD)	Mean (SD)	Mean (SD)
Mother Warmth	3.83 (1.2)	3.69 (1.4)	3.73 (1.3)
Father Warmth	4.01 (1.0)	3.79 (1.3)	3.90 (1.4)
Mother Criticism	0.28 (0.6)	0.34 (0.6)	0.27 (0.6)
Father Criticism	0.16 (0.4)	0.19 (0.5)	0.20 (0.5)
Teacher SRS-2	67.66 (10.6)	67.85 (10.5)	67.30 (11.4)
Teacher TRF	62.03 (6.8)	62.22 (7.1)	62.13 (7.6)

Note. SRS-2 = Social Responsiveness Scale, 2^nd Edition; TRF = Teacher Rating Form. Mothers were rated as having a significantly higher level of FMSS criticism than fathers at Time 1 (t = 2.49, p = .014) and Time 2 (t = 2.57, p = .011).

Table 3.

Correlations between variables of interest (N = 159 couples)

Variable	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18
Time 1
1. Mother Warmth	--
2. Father Warmth	.23**	--
3. Mother Criticism	−.51***	−.17*	--
4. Father Criticism	−.22**	−.38***	.31***	--
5. Teacher SRS-2	−.07	.01	.04	−.10	--
6. Teacher TRF	−.23*	.05	.19	.05	.64***	--
Time 2
7. Mother Warmth	.47***	.30***	−.34***	−.17	−.10	−.08	--
8. Father Warmth	.22*	.38***	−.26**	−.36***	.01	−.12	.30***	--
9. Mother Criticism	−.45***	−.29**	.51***	.26**	.05	.09	−.60***	−.25**	--
10. Father Criticism	−.21*	−.31***	.28**	.44***	−.07	.07	−.23**	−.47***	.34***	--
11. Teacher SRS-2	−.24*	−.16	.12	−.04	.67***	.32**	−.05	−.00	.06	−.02	--
12. Teacher TRF	−.30***	−.12	.07	.02	.43***	.43***	−.19	−.10	.12	.01	.67***	--
Time 3
13. Mother Warmth	.41 ***	.22*	−.48***	−.18	−.03	−.07	.55***	.26**	−.50***	.32***	−.18	−.16	--
14. Father Warmth	.29**	.31***	−.25**	−.24**	−.11	−.08	.30**	.39***	−.30**	.39***	−.08	−.26*	−.41***	--
15. Mother Criticism	−.39***	−.16	.38***	.13	.16	.19	−.46***	−.23*	.33***	.21*	.23*	.28**	−.44***	−.25**	--
16. Father Criticism	−.21*	.42***	.38***	.32***	−.02	.03	−.45***	−.35***	.45***	.49***	−.04	.12	−.45***	−.51***	.39***	--
17. Teacher SRS-2	−.14	−.15	.13	.09	.57***	.35**	−.13	.38	.09	.05	.66***	.38***	−.10	−.17	.03	−.00	--
18. Teacher TRF	−.23*	−.02	.38***	.20	.41***	54***	−.20	.35	.16	.09	.43***	.42***	−.29**	−.28**	.28**	.30**	.61***	--
Child Age	.06	−.06	.01	.00	.03	.09	−.13	.66	.07	.11	−.24*	−.13	−.02	−.11	−.02	.08	−.11	−.15
Child Sex	.07	.04	−.06	.08	−.31**	−.04	.02	.72	−.04	.09	−.27**	−.14	.20	.02	−.05	.04	−.18	.06
Child ID status	−.01	−.07	.04	−.11	.45***	.17	.02	.09	−.10	−.13	.50***	.16	.06	.11	.02	−.13	.46***	.34**
Household Income	−.12	.01	.05	.03	−.27**	−.09	−.15	.09	.03	−.11	−.14	−.13	.05	.06	−.05	.02	−.30**	−.15
Additional child with ASD or other condition	−.11	−.09	.04	.01	.16	.13	−.08	.10	.04	−.16	.06	.02	.04	−.06	−.05	.12	.16	.13

Note. SRS-2 = Social Responsiveness Scale, 2^nd Edition; TRF = Teacher Rating Form. Parents reported on child age (coded in years), child sex (1 = female, 2 = male), child with ASD’s ID status (1 = yes, 0 = no), household income (coded from 1–14, starting at ≤$9,999 [1] and increasing by $10,000 to $20,000 intervals to ≥$160,000 [14]), and having an additional child with a neurodevelopmental or psychological condition (coded a yes [1] or no [0]).

^*p-value = p ≤ .05,

^**p ≤ .01,

^***p ≤ .001.

Cross-lagged Panel Models.

Figures 2, 3, 4, and 5 display the unstandardized path coefficients for the four cross-lagged panel SEMs addressing the study questions. Means and standard deviations for main measures at each time point are included in Table 3. All analyses were assessed for significance at the p < .05 level. The Variance Inflation Factor (VIF) ranged from 1.02–1.89 at Time 1, 1.01–1.94 at Time 2, and 1.04–2.08 at Time 3. Thus, multicollinearity was deemed not a concern based on the standard VIF cut-off of 3 (Hair, Black, Babin, & Anderson, 2010). Fit statistics for each of the SEMs are presented in Table 4. In order to determine model fit, global measures of fit, including the χ² statistic and the root mean squared error of approximation (RMSEA) were examined. RMSEA values less than .08 were considered an adequate fit. Incremental fit indices, including the Tucker–Lewis index (TLI) and the comparative fit index (CFI) were also examined. Values above .90 on the TLI and CFI are often considered acceptable fit values (Hu and Bentler, 1999). Fit was improved for all models with the addition of paths between mother and father criticism and warmth variables both at the same time point and across time points. In addition, lag-2 effects (from Time 1 to Time 3) were added for teacher SRS-2 and CBCL variables. All final model RMSEA values fell below .08, with models 2 and 3 demonstrating closer fit with RMSEA values less than .05. All models had fit values higher than .90 on the CFI. Models 2 and 3 had fit values higher than .90 on the TLI, whereas models 1 and 4 had less than adequate fit with TLI values of .81 and .67, respectively, which may be due in part to the weaker correlations seen among variables included in these models.

Figure 2.

Cross-lagged model for mother and father warmth and teacher-rated child ASD symptoms

Note. SRS-2 = Social Responsiveness Scale, 2^nd Edition. Unstandardized estimates and standard errors represent associations with teacher SRS-2 ratings (i.e., child ASD symptoms). Significant effects are marked in solid lines. *p ≤ .05, **p ≤ .01, ***p ≤ .001.

Figure 3.

Cross-lagged model for mother and father criticism and teacher-rated child ASD symptoms

Note. SRS-2 = Social Responsiveness Scale, 2^nd Edition. Unstandardized estimates and standard errors represent associations with teacher SRS-2 ratings (i.e., child ASD symptoms). Significant effects are marked in solid lines. *p ≤ .05, **p ≤ .01, ***p ≤ .001.

Figure 4.

Cross-lagged model for mother and father warmth and teacher-rated child behavior problems

Note. TRF = Teacher Rating Form. Unstandardized estimates and standard errors represent associations with teacher TRF ratings (i.e., teacher rating form of child behavior problems). Significant effects are marked in solid lines. *p ≤ .05, **p ≤ .01, ***p ≤ .001.

Figure 5.

Cross-lagged model for mother and father criticism and teacher-rated child behavior problems

Note. TRF = Teacher Rating Form. Unstandardized estimates and standard errors represent associations with teacher TRF ratings (i.e., teacher rating form of child behavior problems). Significant effects are marked in solid lines. *p ≤ .05, **p ≤ .01, ***p ≤ .001.

Table 4.

Cross-lagged model fit statistics

Model	χ2 (df)	P	RMSEA	TLI	CFI
1	14.61 (8)	0.07	0.07	0.81	0.97
2	10.44 (8)	0.24	0.04	0.93	0.99
3	10.84 (8)	0.21	0.04	0.91	0.98
4	18.45 (8)	0.02	0.08	0.67	0.94

Note. Model 1 = Cross-lagged model for mother and father warmth and teacher-rated child ASD symptoms; Model 2 = Cross-lagged model for mother and father criticism and teacher-rated child ASD symptoms; Model 3 = Cross-lagged model for mother and father warmth and teacher-rated child behavior problems; Model 4 = Cross-lagged model for mother and father criticism and teacher-rated child behavior problems. χ² = Chi-squared test of absolute model fit; P = χ² p-value; RMSEA = Root Mean Square Error of Approximation; TLI = Tucker-Lewis Index; CFI = Confirmatory Fit Index. All fit statistics reported were for the default model.

Significant stability effects were found for each model with FMSS warmth and criticism at most time points predicting warmth and criticism at the subsequent time point for both mothers and fathers. Further, teacher SRS-2 Total t-score and TRF Total t-score significantly predicted teacher SRS-2 Total t-score and TRF Total t-score between Time 1 and 2. In the first SEM, mother (b = −4.22, SE = 2.06, β = −0.18, p = .04) and father (b = −5.65, SE = 2.46, β = −0.20, p = .02) FMSS warmth at Time 1 predicted teacher SRS-2 Total t-score at Time 2. In the second SEM, teacher SRS-2 Total t-score at Time 2 predicted mother FMSS criticism at Time 3 (b = 0.01, SE = 0.00, β = 0.19, p = .04). In the third SEM, mother FMSS warmth at Time 1 predicted teacher TRF Total t-score at Time 2 (b = −4.27, SE = 1.88, β = −0.27, p = .02). In turn, Time 2 teacher TRF Total t-score predicted father FMSS warmth at Time 3 (b = −0.01, SE = 0.01, β = −0.23, p = .02). In the fourth SEM, teacher TRF Total t-score at Time 2 predicted mother FMSS criticism at Time 3 (b = 0.01, SE = 0.00, β = 0.22, p = .03). Finally, mother FMSS warmth at Time 3 predicted teacher TRF Total t-score at Time 3 in Model 3 (b = −5.81, SE = 2.21, β = −.28, p = .01) and father FMSS criticism at Time 3 predicted teacher TRF Total t-score at Time 3 in Model 4 (b = 16.33, SE = 6.23, β = 0.29, p = .01).

Discussion

Family research in the field of ASD has given relatively little attention to understanding the potential bidirectional associations between the emotional quality of the parent-child relationship and child functioning. In line with the transactional theory of child development (Sameroff & Chandler, 1975), and research involving children without neurodevelopmental conditions (Combs-Ronto et al., 2009; Neece et al., 2012), the emotional quality of the parent-child relationship and child functioning are likely to be transactionally related across time in families of children with ASD. In other words, the emotional quality of the parent-child relationship is likely to be both influenced by child functioning and also plays a key role in shaping child functioning across time in families of children with ASD. The current study used cross-lagged panel models in SEM to examine the directional associations between FMSS warmth and criticism in the mother-child and father-child relationship and the severity of ASD symptoms and emotional and behavioral problems in children with ASD (initially aged 6–13 years) across 12 month periods, using three time points of data collection.

Overall, we found evidence of bidirectional associations between the emotional quality of both the mother-child and father-child relationships and the functioning of the child with ASD. In line with previous FMSS studies with families of adolescents and adults with ASD (Greenberg et al., 2006; Smith et al., 2008), Time 1 mother warmth toward the child with ASD significantly negatively predicted child severity of ASD symptoms and co-occurring emotional and behavioral problems at Time 2. In the same fashion, Time 1 father warmth toward the child with ASD significantly negatively predicted child severity of ASD symptoms at Time 2. However, in contrast to previous findings for adolescents and adults with ASD (Baker et al., 2011), neither mother nor father criticism toward the child with ASD predicted later child functioning. Thus, it is possible that increased parental warmth is a stronger determinant of declines in the symptoms and behavior problems individuals with ASD during middle- and later childhood than is increased parental criticism.

In the opposite direction, we also found that child functioning predicted later emotional quality of the parent-child relationship. The child with ASD’s frequency and severity of emotional and behavioral problems at Time 2 predicted (in negative and positive directions) father warmth and mother criticism toward the child at Time 3. Similarly, the child with ASD’s severity of ASD symptoms at Time 2 positively predicted mother criticism toward the child at Time 3. Thus, the current findings indicated that a high level of child ASD symptoms and emotional and behavioral problems, which have been shown to be associated with heightened level of parenting stress (Ciciolla et al., 2014; Hastings et al., 2006; Valicenti-McDermott et al., 2015), may lead to increased criticism in mothers but reduced warmth in fathers in families of children with ASD.

In the current study, the effect of the emotional quality of the parent-child relationship on child functioning was apparent in the early time points (Time 1 to Time 2), whereas the effect of child functioning on emotional quality of the parent-child relationship occurred at the later time points (Time 2 to Time 3). This pattern may be capturing a change in the directional flow of effects between the parent-child relationship and child functioning across development in children with ASD. Children with ASD started in middle to later childhood (aged 6–13 years) at Time 1 and ended in late childhood and adolescence (aged 8–15 years) by Time 3. It is possible that difficult child behaviors may be more likely to affect the emotional-quality of the parent-child relationship, as their son/daughter with ASD approaches and goes through adolescence. Although research has shown that many individuals’ ASD symptoms and behavior problems decline with age (Shattuck et al., 2007), caregiver burden remains high for parents of children with ASD during the transition to adolescence (Cadman et al., 2012), likely due to the changing expectations for behaviors and further divergence in functioning from their typically developing peers (Rosenthal et al., 2013). Further, in a qualitative study, parents voiced difficulties related to shifting how they parent and manage difficult behaviors as their son/daughter with ASD became an adolescent (van Esch et al., 2018). Larger studies are needed in the future to formally test child age and understand potential developmental shifts in the pathways between the emotional quality of the parent-child relationship and child functioning.

We found both similarities and differences in the associations between the emotional quality of the parent-child relationship and the functioning of children with ASD for mothers versus fathers. Bidirectional relations between the FMSS and child functioning were found for both the mother-child and father-child relationship. However, in line with our hypothesis, the current findings suggested that the emotional quality of the mother-child relationship had more widespread associations (four significant pathways of effect) than the father-child relationship (two significant pathways of effect). We believe that this difference is due to mothers’ being the primary child caregiver more often than fathers in families of children with ASD (Callander & Lindsay, 2018; Hartley et al., 2014). Given their prominent role in daily child caregiving, the emotional quality of the mother-child relationship may more strongly shape and be shaped by the functioning of children with ASD than fathers, on average. Mechanisms driving our finding that child symptoms and behaviors may lead to increased criticism in mothers but instead reduced warmth in fathers in families of children with ASD are not clear and should be the focus of future research. On average, mothers were rated as having a higher level of criticism toward the child in FMSS than fathers, within families. It is also possible that restricted range in FMSS criticism in fathers hindered our ability to detect effects. In other populations, research suggests that mothers are more likely than fathers to be involved in their child’s education (Hill, 2015). The more robust associations between the emotional quality of the mother-child relationship and child functioning may be a result of mothers being more aware of their child’s symptoms and emotional and behavioral problems exhibited in school because of their increased involvement and highlights the valuable contribution that the current study offers to the field as behaviors that occur within the school context may be more likely to elicit emotional responses from parents.

There were several strengths to the current study. We used a direct measure of the emotional quality of the parent-child relationship via the FMSS, which has been shown to mirror actual family interactions (e.g., McCarty et al., 2004; Weston et al., 2017). We also assessed both positive (i.e., warmth) and negative (i.e., criticism) indicators of the emotional quality of the parent-child relationship. In addition, we used an outside (teacher report) measure of child ASD symptoms and emotional and behavioral problems; thus, avoiding the conflation of effects by having parents as single reporters for both the FMSS and child functioning measures. SEM models simultaneously examined the mother-child and father-child relationship and controlled for earlier level of child functioning and FMSS warmth and criticism when examining effects across time points by controlling for effects of the preceding time points. The study also had limitations. The sample was fairly homogeneous in race/ethnicity and socio-economic status; it is not clear if findings generalize to families of children with ASD from minority racial/ethnic groups and/or of lower socio-economic status. Further, the cross-lagged panel models only included three time points and spanned middle childhood. We are not able to make conclusions about the longer-term directional relations between the emotional quality of the parent-child relationship and child functioning or how these associations look at other developmental stages or shift over time. Finally, while teacher-report offered the advantage of a non-parent rating of child functioning, it may also have reduced our ability to detect associations of interest. Child behaviors often vary between the home and school setting (Tung & Lee, 2018) and the links between parent-child relationship and child functioning may be most evident for child symptoms and behaviors in the home environment.

Future studies are needed to understand nuances in the pattern of findings and to determine if this pattern varies based on specific family and child characteristics. For example, the current study did not make distinctions about the type of ASD symptoms (e.g., social communication vs. restricted and repetitive behaviors) or emotional or behavioral problems (e.g., depressed affect vs. aggression) related to the emotional quality of the parent-child relationship. Further, additional longitudinal research should examine other aspects of the parent-child relationship as the child with ASD ages, such as praise and emotional over-involvement measured by the FMSS, as these factors may become more predictive of child outcomes as the child enters adulthood. These are important areas for further work. There is also a need to examine the impact of the presence of an additional child with ASD or other neurodevelopmental or psychological condition (which was true for 39.8% of our sample) in influencing pathways between the parent-child relationship and child functioning. Although, in the current study, having multiply-affected children did not directly correlate with parent warmth and criticism, mothers of multiple children with ASD were more likely to show low criticism in their FMSS about the child of focus at the final time point of the study (F = 5.82, p = .018), and fathers of multiple children with ASD were more likely to show low warmth about the child of focus in their FMSS at the final time point of the study (F = 6.14, p = .015). It is likely that other aspects of the parent-child relationship may vary depending on whether the family has an additional child with special health care needs. Child factors including age, ID status, and gender were also associated with teacher ratings of child ASD symptoms and emotional and behavioral problems in the current study. Future research should examine the extent to which these child factors influence transactional associations between the parent-child relationship and child development across time.

In summary, our findings indicate that the emotional quality of the mother-child and father-child relationships are bidirectionally related to the symptoms and emotional and behavioral problems of children with ASD across middle childhood to adolescence. Findings have important implications for promoting optimal developmental outcomes for children with ASD and supporting the well-being of the broader family system. Interventions should educate parents on the ways in which the emotional quality of the parent-child relationship is intertwined with the child with ASD’s functioning. These intervention efforts should engage both mothers and fathers, as both the mother-child and father-child relationship are transactionally related to the functioning of children with ASD. Specific interventions should support parents in developing a high level of warmth toward the son/daughter with ASD and to maintain this warmth and avoid increased criticism in the face of difficult child ASD symptoms and emotional and behavioral problems. Such interventions may need to draw on both direct strategies (e.g., parenting training aimed at replacing maladaptive attributions about child behavior or trainings on adaptive ways to manage child behavior) as well as indirect strategies (e.g., financial support, stress management, and respite care) that reduce mediating factors such as parenting stress that get in the way of fostering a parent-child relationship marked by high warmth and low criticism. In non-ASD populations, there is growing evidence that such direct and indirect interventions can contribute to improvements in the parent-child relationship (e.g., mindfulness parenting) and subsequently lead to improved child outcomes (Coatsworth, Timpe, Nix, Duncan, & Greenberg, 2018; Neece, 2014). Another promising intervention, driven by expressed emotion constructs and supported by research done with families of adults with schizophrenia, is multiple family group therapies (Jewell, Downing, & McFarlane, 2009). Evidence-based practices that have stemmed from FMSS research in other populations (e.g., families of adults with schizophrenia) such as multiple family group theories (e.g., Jewell et al., 2009) may also be useful for families of children with ASD. In these therapies, family members learn strategies to reduce stress and be supportive of one another as a means of negating expressed emotions in the home.