Autism spectrum disorders (ASD) are characterized by social and communication difficulties and the presence of repetitive behaviors and interests (DSM V, APA 2013
). In addition to these core impairments, up to 70% of the children and adolescents with ASD show comorbid psychiatric problems, like social anxiety and oppositional defiant disorder (e.g., Simonoff et al. 2008
). Recently, there is a growing interest in the role of emotion control as a possible underlying mechanism that may explain these internalizing and externalizing behavior problems (Mazefsky et al. 2013
The ability to control emotions is essential to navigate through daily hazards. It allows one to keep an optimal level of arousal in order to secure both social and personal goals (Chambers et al. 2009
). Its development is affected by social experiences, and modeled through social learning. Indeed, from childhood onwards we learn how to control our emotions in a socially and culturally accepted way (Gullone et al. 2010
; Morris et al. 2007
; Southam-Gerow and Kendall 2002
). However, children and adolescents with ASD have less access to the social learning environment and show deficits in the ability to control emotions (Mazefsky and White 2014
; Rieffe et al. 2012
; White et al. 2014
). The aim of the current longitudinal study is to examine the role of three indices related to emotion control—negative emotionality, emotion awareness and worry/rumination—on the development of internalizing and externalizing behavior problems in children and adolescents with ASD, as compared to a typically developing (TD) control group.
Emotion control is an umbrella term that is used to describe several aspects related to the ability to down regulate emotional over-arousal in emotion-evoking situations. Problems in emotion control can be related to impairments in both emotion generation and in the process that relates to the ability to deal with emotions (Sheppes et al. 2015
). Previous cross-sectional studies have shown that different indices of emotion control are related to internalizing and/or externalizing behavior problems. One index of problems in emotion control is the frequent experience of negative emotions, like fear, anxiety, and anger. This negative emotionality is a direct consequence of an inability to down regulate emotional over-arousal. Heightened levels of negative emotionality is associated with both internalizing (e.g., anxiety) and externalizing behavior problems (i.e., bullying and aggression) in TD youth and in children and adolescents with ASD (Pouw et al. 2013a
; Rieffe et al. 2012
; White et al. 2009
). Another index of emotion control is emotion awareness. Emotion awareness relates to the ability to know how you feel and to link this feeling to an emotion-evoking situation is critical for the experience and regulation of emotions (Barrett et al. 2001
; Rieffe et al. 2008
). Indeed, not being able to differentiate between emotions and focusing too much on bodily symptoms of an emotional experience is related to more depressive symptoms, anxiety symptoms and somatic complaints in TD children and children with ASD (Rieffe and De Rooij 2012
; Rieffe et al. 2008
). In a similar vein, several studies indicate a relation between alexithymia and emotional problems in children and adolescents with ASD (for a review, see Bird and Cook 2013
). Another index of emotion control that is related to the (in)ability to deal with emotional over-arousal is worry. Worry and rumination are highly related processes that are characterized by a chain of repetitive negative thinking, that increases the level of emotional over-arousal (Watkins 2008
). The role of worry/rumination in youth with ASD is relatively understudied. This is remarkable since individuals with ASD have a propensity to perseveration and may therefore be uniquely susceptible to worry/rumination (Mazefsky et al. 2012
; Patel et al. 2017
). Worry/rumination are typically associated with the development and maintenance of internalizing behavior problems in TD youth (Nolen-Hoeksema et al. 2008
). Likewise, we previously demonstrated a cross-sectional and longitudinal relation between worry/rumination and depressive symptoms in children and adolescents with ASD (Pouw et al. 2013b
; Rieffe et al. 2014
). Recently it has been shown that worry/rumination is also related to aggressive behavior in TD boys (McLaughlin et al. 2014
). It is however unknown whether worry/rumination also contributes to disruptive behavior problems in children and adolescents with ASD.
Our knowledge of the role of emotion control on the development of internalizing and externalizing behavior problems in children and adolescents with ASD relies mainly on cross-sectional data. Even though these studies provide essential information for our understanding of this relationship, longitudinal studies are key to advance our knowledge on whether these relations hold over time. Therefore, we conducted a longitudinal study to test the relation between negative emotionality, emotion awareness, and worry/rumination with internalizing and externalizing behavior problems in 9–15 year old boys with and without ASD. We focused on this age range, given that social and emotional problems often increase during adolescence (Kuusikko et al. 2008
; Paus et al. 2008
We investigated three clusters related to internalizing problems: depression, anxiety, and somatic complaints, and one general cluster of externalizing problems: disruptive behavior. Participants and their parents filled in questionnaires about different aspects of emotion regulation and overall well-being at three time points (9 months in between each wave). Specifically, we aimed to test in both groups (1) whether emotion control contribute to the prediction of internalizing and externalizing behavior problems 18 months later, (2) examine the developmental trajectory of internalizing and externalizing behavior problems over time and (3) test the co-occurrence of the developmental trajectory of emotion control with the developmental trajectory of internalizing and externalizing behavior problems.
Based on the literature, we expect that negative emotionality, poor emotion awareness, and worry/rumination are related to more internalizing problems in both boys with and without ASD (Aldao et al. 2010
; Barrett et al. 2001
; Rieffe et al. 2008
). For externalizing problems, we expect that negative emotionality and worry/rumination will have a positive predictive value for both groups (McLaughlin et al. 2014
). Given the importance of emotion regulation for social functioning, the ability to employ effective emotion control is relevant for both groups. Nevertheless, the social impairments related to ASD might leave children and adolescents with ASD more susceptible to develop problems with controlling their emotions for which adequate socialization seems to be critical. Thus, the lack of social learning opportunities may reduce their capacities to control emotions and add to their vulnerability for developing social and emotional problem behaviors. Hence, we expect that the predictive relation between emotion control with internalizing and externalizing behavior problems is stronger for boys with ASD compared to their TD peers.
The current study was part of a larger project investigating the social-emotional development of typically developing children and children with less access to the social environment (children with hearing loss and children with ASD) (e.g., Broekhof et al. 2017
; Netten et al. 2015
; Pouw et al. 2013a
; Rieffe et al. 2012
). For the purpose of the current study, we used the data of TD boys and boys with ASD from whom parent-reports and self-reports were available at least at one time point.
The high functioning ASD sample included 66 children at T1. Inclusion criteria were: (i) ASD diagnosis on T1 according to the DSM-IV (APA 1994
) based on the Autism Diagnostic Interview-Revised (Lord et al. 1994
) by a child psychiatrist, (ii) IQ score above 80 and (iii) no additional DSM-IV diagnoses. Participants were recruited from specialized diagnostic and treatment center for children with autism in the Netherlands. A group of 89 TD children was recruited from primary and secondary schools in the Netherlands (see Table 1
for sample characteristics). Inclusion criteria for the control group was: (i) IQ above 80 and (ii) no DSM-IV diagnosis. All procedures were approved by the Ethical Committee of Leiden University and all parents provided written informed consent.
Demographic characteristics of participants
No. of children
Mean—in age (SD) at T1
Socioeconomic status (SD)
IQ normscore picture arrangementd
IQ normscore block designd
Social responsiveness scale***
Two nonverbal subtests (i.e., block design and picture arrangement) of the Wechsler Intelligence Scale for Children-Third edition
(WISC III; Kort et al. 2002
) were used to calculate a general measure of intelligence. The obtained scores were converted into age-corrected norm scores. The grand population mean is set to 10. The IQ subtests were not administered in 2 ASD and 5 TD boys due to time constraints.
The mood list (Rieffe et al. 2004
) is a self-report questionnaire that was used to assess children’s negative mood over the past 4 weeks. We used three subscales of the mood list: anger, fear and sadness. Each subscale consisted of 4 items on a 3-point rating scale (e.g., I never/sometimes/often feel angry). We used a total score for negative mood by taking the sum score of the negative items. Higher scores indicated dysregulated emotion experience. Previous studies has shown good reliability and validity of this measure, this questionnaire was not previously administered in ASD populations. Internal consistency in the current study was good (0.90 ≥ α ≥ 0.79).
Children rated their awareness and understanding of their own emotions on two subscales of the Emotion Awareness Questionnaire (EAQ; Rieffe et al. 2008
): differentiating emotions and bodily awareness of emotions. The subscale differentiating emotions contained 7 items and measured whether children were able to differentiate between their own emotions (e.g., “I am often confused or puzzled about what I am feeling [reversed-scored]”). Ratings were made on a 3-point scale ranging from 1 = (almost) not true to 3 = always true. A high score indicates good ability to differentiate between emotions. The subscale Bodily Awareness of Emotions measures whether children are aware of bodily changes related to emotions and consists of 5 items (e.g., “I don´t feel anything in my body when I am scared or nervous”). A high score indicated low bodily awareness, which was associated with more emotion awareness based on factor analysis (Rieffe et al. 2008
). A total score of the 2 subscales were used as index for emotion awareness. The EAQ has shown to have good reliability and validity (Lahaye et al. 2011
) and has been previously administered in children and adolescents with ASD (Rieffe et al. 2011
). In the current study, internal consistency of this measure was acceptable (0.81 > α ≥ 0.69).
The worry/rumination questionnaire for children (Jellesma et al. 2005
; Miers et al. 2007
) is a self-report measure, which assess the tendency of children to dwell on a problem instead of dealing with it in terms of solving or coping adaptively with the emotional impact of the situation. The questionnaire comprises 10 items and children are asked to rate the degree to which each item (e.g., When I have a problem, I think about it all the time) is true about them on a 3-point scale (1 = not true, 2 = sometimes true, 3 = often true). A high score indicates a high level of worry/rumination. This questionnaire has good reliability and validity and was previously administered in children and adolescents with ASD (Rieffe et al. 2011
). In the current study, internal consistency was good (0.81 ≥ α ≥ 0.89).
Disruptive Behavior Problems
The Child Symptom Inventory (CSI; Gadow and Sprafkin 1994
; Dutch version by; Theunissen et al. 2012
) is a behavior rating-scale to assess childhood disorders based on DSM-IV criteria. The parent-checklist was used to assess problems related to attention deficit hyperactivity disorder (ADHD), oppositional deviant disorder (ODD) and conduct disorder (CD). Seventeen items assessed the symptoms of ADHD (e.g., “Is quickly distracted”), eight items assessed symptoms of ODD (e.g., “Does things to deliberately annoy others”) and 15 items assessed symptoms of CD (e.g., “Has deliberately started fires”). Parents were asked to rate each symptom on a 4-point scale (1 = never and 4 = very often). A higher score indicated more disruptive behavior. Previous studies indicate that the CSI has satisfactory reliability and validity in community and ASD samples (Gadow and Sprafkin 2009
). In the current study, internal consistency was high (0.93 > α ≥ 0.90).
For internalizing symptoms, three indices were taken: anxiety, depression, and somatic complaints.
The CSI was also used to assess problems related to generalized anxiety. Parents rated children’s generalized anxiety symptoms in the last six months on 7 items. Ratings were made on a 4-point scale ranging from 1 = never to 4 = very often. We used a total score of the 7 items. A higher score indicated more anxious feelings. Internal consistency was sufficient (0.82 > α ≥ 0.74).
Problems related to depression were measured with an adapted Dutch version of the Children’s Depression Inventory (CDI) (Kovacs 1985
; Dutch version by; Timbremont et al. 2002
). This self-report questionnaire includes 27 items that are related to specific depression symptoms (e.g., “I am sad
. Ratings were on a three-point scale ranging from never/hardly true (1) to very true (3). The item pertaining to suicidal ideation was removed from the measure. In the analyses we used the total score of the 26 items. Higher scores on the CDI indicates higher depressive mood. CDI has good reliability and validity and has previously been administered in ASD populations (Lerner et al. 2012
). Internal consistency in the current study was sufficient (0.86 > α ≥ 0.66).
was measured by the Somatic Complaint List (SCL) (Jellesma et al. 2007
). Children rated the frequency with which they experience certain somatic complaints such as a headache in the past four weeks on a 5-point scale (1 = never to 5 = very often). The scoring was reversed for the two positively formulated items. The SCL consists of 21 items; a high total score indicated more somatic complaints. Previous studies have shown that the SCL has good reliability and validity (Jellesma et al. 2007
) and has previously been administered in ASD populations (Rieffe et al. 2011
). Internal consistency in the current study was sufficient (0.83 > α ≥ 0.72).
Children were visited three times with approximately a 9-month time interval at home (MT1 to 2 = 8.77 months; SDT1 to 2 = 1.27; MT2 to 3 = 9.23, SDT2 to 3 = 1.17), at school or at their institution. As part of a larger study, children were asked to fill in several questionnaires on a laptop and to perform some experimental tasks. The test sessions took approximately 1 h each. It was emphasized that their responses would be anonymous. Parents were asked to complete questionnaires online or with paper and pencil. All participants were invited for the second and third wave. Nine participants (ASD: n6; TD: n = 5; attrition rate: 7.1%) in the second wave and 26 participants in the third wave (ASD: n = 9; TD: n = 17; attrition rate: 16.8%) indicated that they could not or did not want to participate anymore.
Statistical analyses were performed using the statistical software package for social sciences version 21.0 (SPSS Inc., Chicago). Sample characteristics were analyzed by independent t test. To test whether indices of emotion control at T1 predict problem behavior at T3, we used a hierarchical regression analyses for each problem behavior separately. In these analyses we entered as predictors diagnostic group (dummy coded: 1 = ASD, 0 = TD) in the first step, negative emotionality, emotion awareness and worry/rumination in the second step and the interaction between diagnostic group and the indices of emotion control in the third step. Outcome variables were indices of internalizing and externalizing behavior. All predictors were centered to the mean before entered in the regression analyses. Even though our predictors were moderately correlated with one another (see Supplementary Table 3), multicollinearity diagnostics indicated adequate tolerance levels. Note, controlling for IQ did not alter the results.
To analyze the developmental trajectory of behavior problems in boys with ASD compared to TD boys, we used a multilevel model approach (Singer and Willett 2003
). One advantage of a multilevel approach is that it allows for hierarchy within data, such as observed in longitudinal data. In a longitudinal data set, time points are nested within participants and multilevel modeling can account for this data dependency. Another advantage of multilevel modeling is that it can handle missing data. In a multilevel model, cases with complete data at every time point are weighted more heavily. Importantly, as long as one time point of measurement is available, the case is included in the estimation of effects. In these analyses, time was treated as within-individual variable (t
) (level 1) and group was included as between-individual variable (level 2). All mixed-models followed a formal model-fitting procedure. That is, we started with an unconditional means model that only included a fixed and random intercept, to allow for individual differences in starting points and account for the repeated nature of the data. The unconditional means model was compared to additional models that tested the grand mean trajectory of age [centered around 9 years (age of the youngest child)]. Thereafter we included diagnostic group and the interaction between age and diagnostic group, to examine whether the developmental trajectory of individuals with ASD differed from TD individuals. Preferred models had significantly lower Akaike Information Criterion values (AIC; Akaike 1974
) and Bayesian Information Criterion (BIC; Schwarz 1978
Finally, we tested whether changes in indices of emotion control could explain change in problem behavior (i.e., score of measurement 1–3 minus the score of measurement 1 for the indices of emotion control and problem behavior). We used a step-wise procedure including (1) change score of each predictor and (2) including the interactions with diagnostic group. In the results the best fitted models are described.