Characterizing difficulties with emotion regulation in toddlers with autism spectrum disorder

Abstract

Background:

Difficulties with emotion regulation (ER) underlie emotional/behavioral challenges and co-occurring psychiatric symptoms in autism spectrum disorder (ASD), yet very little is known about the early development of emotion dysregulation. The present study aimed to identify differences in positive and negative emotional reactivity and regulation strategies in toddlers with and without ASD.

Method:

Nine tasks from the Laboratory Temperament Assessment Battery (Lab-TAB) were completed with 37 toddlers with and without ASD (22–28 months). Video-recordings of these tasks were coded by research assistants using a behavioral coding scheme tapping facial, bodily, and vocal affect and the frequency of ER strategies. Mixed model analyses were performed to examine the mean affect and total regulation strategies across each task and t-tests were conducted to assess the types of ER strategies utilized.

Results:

Toddlers with ASD showed significantly less positive affect and greater frustration compared to non-ASD toddlers; reactivity was comparable between the groups for fear. Both groups used ER strategies in a similar pattern across tasks, with the exception that toddlers with ASD more often engaged in distraction to regulate. Effects of age and developmental level are discussed.

Conclusions:

Toddlers with ASD were robustly characterized by greater frustration and lower joy despite frequent and age appropriate attempts to regulate their emotions. This study provides preliminary evidence that observable indicators of emotion dysregulation are present by two years of age. Clinical implications and future research directions are discussed.

1. Introduction

Autism spectrum disorder (ASD) is a complex, neurodevelopmental disorder characterized by difficulties with social communication and interaction as well as the presence of restricted, repetitive behaviors (American Psychiatric Association, 2013). While emotional and behavioral problems are not core symptoms of ASD, they are often highly related to its clinical manifestation (Weiss et al., 2017). Recent estimates from the Centers for Disease Control and Prevention indicate that the majority of children with ASD have at least one co-occurring behavioral/psychiatric condition (Soke et al., 2018). Such symptoms are a critical reason that parents seek psychosocial or pharmacological services for their child (Croen et al., 2006; Madden et al., 2017), likely because emotional and behavioral challenges take a significant toll on family functioning (Herring et al., 2006; Nuske et al., 2018).

1.1. Emotion regulation in ASD

Poor emotion regulation (ER) is theorized to be a contributing and maintaining factor of psychiatric conditions in ASD (Mazefsky et al., 2013; White et al., 2014). ER involves the generation and automatic or effortful modulation of emotion to meet situational goals (Gross & Thompson, 2007; Mazefsky et al., 2013; Mazefsky, 2015; Thompson, 1994). A recent study of over 2000 school-aged children and adolescents found that children with ASD were four times more likely to exceed clinical cut-offs for emotion dysregulation characterized by rapidly increasing, strong, and poorly regulated negative emotion and two times more likely to present with clinically elevated dysphoria (poor upregulation of positive affect; Conner et al., 2020). Research has also demonstrated that older children, adolescents, and adults with ASD commonly utilize more maladaptive or ineffective strategies (e.g., avoidance) and fewer adaptive strategies (e.g., problem-solving) to regulate their emotions than individuals without ASD (Jahromi et al., 2012; Mazefsky et al., 2014; Samson et al., 2012), a pattern that is negatively correlated with psychiatric symptoms (Chandler et al., 2016; Mazefsky & White, 2014; Mazefsky et al., 2014, 2019; Rieffe et al., 2011; Samson et al., 2015; Sannar et al., 2017). ER has, therefore, been identified as an ideal candidate to target in psychosocial intervention (Weiss et al., 2017) to reduce the long-term economic and personal impact associated with psychiatric conditions in ASD (Croen et al., 2006; Joshi et al., 2010). It may also be preferable to intervene earlier, and potentially divert negative trajectories. However, the early manifestation of ER in toddlers and preschoolers with ASD is under-studied.

1.2. Early manifestations

The existing research suggests that emotion dysregulation in young children with ASD is characterized by high levels of negative affectivity and reduced positive expressions, although the latter is not consistently documented across studies (Garon et al., 2009, 2016; Gulsrud et al., 2010; Hirschler-Guttenberg et al., 2015; Jahromi et al., 2012; Macari et al., 2017, 2018; Nuske et al., 2017; Wetherby et al., 2004). Based on parent-report of the affective characteristics of temperament, toddlers with ASD demonstrate lower positive affect and higher negative affect than non-ASD toddlers (Garon et al., 2009, 2016; Macari et al., 2017). The few observational studies in toddlers with ASD have indicated greater distress relative to non-ASD toddlers (Gulsrud et al., 2010; Macari et al., 2018; Wetherby et al., 2004), with some opposing findings for positive affect. Studies focused on characterizing social communication have documented reduced joyful expressions (Dow et al., 2017, Wetherby et al., 2004), whereas a recent study found comparable joy among toddlers with and without ASD (Macari et al., 2018). Macari et al. (2018) were the first to systematically compare toddlers with ASD, developmental delay (DD), and typical development (TD) during tasks structured to elicit emotions. They also noted that toddlers with ASD demonstrated less fear than their non-ASD peers. Collectively, these studies document evidence of observable difficulties in emotional reactivity in children with ASD around two years of age yet additional replication of these findings in tasks meant to probe emotions are needed to clarify the role that regulation plays in emotional challenges in toddlers with and without ASD.

In typical development, toddlers 18–24 months are beginning to self-regulate their emotions using strategies such as seeking maternal assistance, problem-solving, and distraction to reduce distress (Calkins & Johnson, 1998; Calkins et al., 1998; Diener & Mangelsdorf, 1999). Based on parent-report, Macari et al. (2017) found that toddlers with ASD employed fewer regulation strategies than both the TD and DD groups, even after controlling for nonverbal developmental level. Other studies using similar tasks have found that young children with ASD (ages 2–6) attempt to regulate their emotions during stressful or frustrating tasks, but the strategies they employ are less effective than TD children (Jahromi et al., 2012; Nuske et al., 2017), consistent with developmental delays in ER (Hirschler-Guttenberg et al., 2015; Nuske et al., 2017).

To date, most observational studies have included children in the preschool age range (Hirschler-Guttenberg et al., 2015; Jahromi et al., 2012) despite preliminary evidence to document that emotional dysregulation may be part of the ASD phenotype as early as 12–24 months of age (Dow et al., 2017; Wetherby et al., 2004; Zwaigenbaum et al., 2016) or have only examined part of the ER construct (i.e., either emotional reactivity or regulation strategies; Gulsrud et al., 2010; Macari et al., 2018; Nuske et al., 2017). Studies in both ASD and in general population samples find that the initial emotional reactivity and the regulation of those emotions are best explained by a single factor (Mazefsky, Yu, et al., 2018; Mazefsky et al., 2020; Zelkowitz & Cole, 2016), which emphasizes the importance of measuring both aspects of emotion dysregulation in order to obtain a more complete understanding. The only observational study to characterize both emotional reactivity and regulation strategies found that 21–36-month-olds with ASD regularly displayed negative affect during free play with their mothers (Gulsrud et al., 2010). Although these toddlers attempted to regulate their distress, they infrequently used more advanced, adaptive verbal regulatory strategies that are commonly observed in TD toddlers (Gulsrud et al., 2010; Kopp, 1992; Stansbury & Zimmermann, 1999). Despite the merits of this study, the authors were unable to quantify the differences in ER from toddlers without ASD given that the study design did not include a control group.

1.3. Study aim

The primary aim of this study was to quantify emotional reactivity and use of regulation strategies in toddlers with and without ASD in both negative and positive situations. It was hypothesized that toddlers with ASD would exhibit a greater intensity of negative affect during negative situations (frustration, fear) and decreased positive affect during positive situations (joy) than non-ASD toddlers. It was also predicted that toddlers with ASD would employ fewer strategies to regulate their emotions and use less advanced ER strategies (e.g., avoidance, distraction, comfort-seeking) than toddlers without ASD.

The present study is unique in that it examines both emotional reactivity and regulation in toddlers with and without ASD. More research of this construct in early life is particularly important given that mental health difficulties can be emerge by four years of age (Chandler et al., 2016; Soke et al., 2018) and preliminary support that emotion dysregulation is tractable (Gulsrud et al., 2010; Scarpa & Reyes, 2011; Weiss et al., 2017) despite the otherwise persisting or worsening pattern of emotion dysregulation in children with ASD (Berkovits et al., 2016; Lecavalier et al., 2006). This research may help to advance our understanding on the early indicators of emotion dysregulation in ASD.

2. Methods

2.1. Participants

Thirty-seven toddlers (17 with ASD, 20 without ASD; see Table 1) between 22 and 28 months of age (M = 25.27 months) were recruited from the FIRST WORDS® Project, an ongoing, prospective longitudinal study focused on screening and early identification of ASD. Prior to this study, these children were screened for communication delays and ASD via the Early Screening for Autism and Communication Disorders (ESAC; Wetherby et al., 2021). The ESAC was either completed in primary care offices in northwest and southwest Florida or on an online screening portal (https://firstwordsproject.com) following provider referral or parent self-referral. Parents were then invited for further evaluation following the ESAC if: (1) his/her child received a positive screen for a communication delay or ASD based on parent-report; (2) he/she expressed concern about his/her child’s development; or (3) his/her child was randomly selected after receiving a negative screen. These children also completed a diagnostic evaluation between 18 and 26 months, which included the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) Toddler Module (Luyster et al., 2009), Mullen Scales of Early Learning (MSEL; Mullen, 1995), Vineland Adaptive Behavior Scales, Third Edition (VABS-3; Sparrow et al., 2016), Systematic Observation of Red Flags of ASD (SORF; Dow et al., 2020) based on a home observation, and parent-report questionnaires. Evaluations were conducted by masters or PhD-level clinicians who had extensive experience in early child development and ASD, including research reliability on the ADOS-2. Toddlers received a diagnostic determination based on the clinician’s clinical impression and assessment scores. Based on MSEL scores at the time of evaluation, 25 % of the non-ASD sample had a developmental delay (n = 5) whereas the remaining non-ASD sample were TD (n = 15).

Table 1.

Descriptive statistics of demographic and developmental characteristics.

	Non-ASD (n = 20)	ASD (n = 17)	Statistic
Age at Visit [months, M (SD)]	24.60 (1.79)	26.06 (1.71)	t = −2.52*
Sex [male, n (%)]	7 (35.0 %)	12 (70.6 %)	χ2 = 4.66*
Race [white, n (%)]	18 (90.0 %)	8 (47.1 %)	χ2 = 2.68
Ethnicity [non-Hispanic, n (%)]	19 (95.0 %)	8 (47.1 %)	χ2 = 3.13
Maternal education [4-year degree or higher, n (%)]	14 (70.0 %)	7 (41.2%)	χ2 = 0.89
Paternal education [4-year degree or higher, n (%)]	11 (57.9 %)	4 (22.2%)	χ2 = 3.39
ADOS-2 Toddler Module CSS [M (SD)]	2.40 (0.94)	7.59 (1.87)	t = 10.37*
MSEL ELC	105.90 (15.56)	71.29 (13.74)	t = 7.11*

Significant differences, p < .05, are noted by an asterisk (*). A small proportion of parents declined to report race (n = 5), ethnicity (n = 6), maternal education (n = 4), and paternal education (n = 7). ADOS-2 Toddler Module CSS = Autism Diagnostic Observation Schedule, Second Edition Toddler Module Calibrated Severity Score (range: 1–10). MSEL ELC = Mullen Scales of Early Learning Early Learning Composite (M = 100, SD = 15). t-values are reported for continuous data and χ² are reported for categorical data.

Children were eligible for the present study if they were between the ages of 22–28 months and were able to travel to one of the four evaluation sites (Tallahassee, Panama City, Tampa, Ft. Myers/Naples). In an effort to capture a similar level of heterogeneity in developmental level as the ASD group (Maenner et al., 2020), children with both DD and TD were included in the non-ASD comparison group. Children were excluded from the study if a diagnosis of ASD was deferred during the diagnostic evaluation (i.e., ASD could not be confirmed or ruled out by the clinician).

2.2. Procedures

2.2.1. Data collection

All families meeting the inclusion criteria were invited to participate following their child’s diagnostic evaluation.⁵ The visit lasted approximately 1–1.5 h and included a standardized protocol: select tasks from the well-studied Laboratory Temperament Assessment Battery (Lab-TAB; Goldsmith et al., 1999; Goldsmith & Rothbart, 1999; Planalp et al., 2017) in which the measurement of frustration, fear, and joy were of primary interest, periodic breaks, and a brief activity to allow the child to return to baseline following each task given physiological data were also collected (results will be reported elsewhere). The accompanying guardian (89.2 % female caregiver, 10.8 % male caregiver) was in the room for the duration of the visit.

Overall, parental acceptability of the procedures was high. Parents consented to all the tasks, with the exception of one parent who declined the child’s participation in the Scary Mask task only. No families ended the study before all Lab-TAB measures were presented. One child did not complete the Remote Control Spider task because of technical difficulties.

Informed consent for the parent’s and child’s participation was collected from the parent prior to the start of the study visit. Parents also consented to sharing archival data collected as part of their child’s participation in the FIRST WORDS Project. Procedures were approved by the Florida State University’s Institutional Review Board.

2.2.2. Behavioral coding

Trained undergraduate research assistants coded specified behaviors from video recordings of the Lab-TAB tasks using a detailed coding system tapping emotional reactivity and regulation. Coders were trained by the first author (TND) on coding scheme definitions and then calibrated with the first author and each other before proceeding to independent coding. Eight of the 37 videos were coded by two research assistances to assess inter-rater reliability The kappa values were as follows: 0.69 (0.41–0.83) for facial affect, 0.72 (0.51–0.94) for vocal affect, 0.69 (0.50–0.87) for bodily affect, and 0.53 (0.40–0.67) for regulation strategies, indicating moderate to substantial agreement (Gardner, 2000; Landis & Koch, 1977) for behavioral codes. The full coding manual is available upon request. All research assistants were blind to diagnostic status and to the hypotheses of the study.

2.3. Measures

2.3.1. Archival data

Children underwent a full diagnostic assessment prior to being recruited for the current study. Those data were made available to determine eligibility for the present study. The following measures were also included in data analyses: ADOS–2 Toddler Module calibrated severity scores (CSS; range: 1–10) to estimate ASD symptom severity (Esler et al., 2015), the MSEL Early Learning Composite (ELC; Mullen, 1995) to assess developmental level, and demographics collected via parent-report.

2.3.2. Overview of Lab-TAB tasks

A subset of tasks were selected from the Lab-TAB manuals (Goldsmith & Rothbart, 1999; Goldsmith et al., 1999) to reduce participant burden. The final protocol included nine tasks, which provided multiple instances to probe each emotion type given there were no a priori hypotheses about which tasks would most effectively capture ER. These tasks are intended to mimic responses in everyday activities (Goldsmith & Rothbart, 1999; Goldsmith et al., 1999). Minor modifications were made to tasks ensure developmental appropriateness for the current sample (e.g., parent remaining in room for duration of visit, but occupied with completing questionnaires and instructed only to respond to their child’s initiations). Task order was standardized in order to balance positive and negative tasks (Table 2; see Supplemental materials for a detailed explanation of each task.), with structured breaks after every third task. All negatively-valenced tasks incorporated a positive resolution phase. Adherence to the scripted protocol was prioritized when feasible (e.g., continuing to proceed with clean up even if child was distressed) in order to maintain standardization. Attempts to regulate the child’s distress in between tasks were permitted, such as distraction or parental comfort. The first author (TND) was the examiner for all tasks and could not be blinded to diagnostic status due to clinical training and specialization in the early diagnosis of ASD.

Table 2.

Task order and description.

Task	Emotion type	Description
Stranger Approach	Fear	An unfamiliar female adult (i.e., research assistant the child had not yet met) wearing a baseball hat and large sunglasses slowly approached and attempted to interact with the child.
Make That Car Go	Joy	The child and examiner raced cars down ramps for 2 min.
End of the Line	Frustration	The child was told to stop playing with a desirable toy and it was removed from his/her reach for 30 s.
Attractive Toy in a Transparent Box	Frustration	Two-minute task in which a desired toy is placed in a clear locked box. The child was provided the box with no means of unlocking it.
Popping Bubbles	Joy	The child was encouraged to use his/her hands and feet to pop bubbles.
Remote Controlled Spider	Fear	An unfamiliar animal moved towards and away from the child several times.
Scary Mask	Fear	The research assistant wore a scary mask and kneeled in front of the child for 30 s.
Modified Peek-a-Boo	Joy	The examiner played peek-a-boo with the child using the playhouse as the prop to hide.
Gentle Arm Restraint by Parent	Frustration	The parent physically restrained the child so that he/she could not reach an interesting toy for 30 s.

Description of all Laboratory Temperament Assessment Battery (Lab-TAB) tasks. Tasks are listed in order of how they were presented during each child’s lab visit.

2.3.3. Coding Protocol

The coding protocol characterized child emotional reactivity and use of regulation strategies during 10-second intervals for each of the nine tasks; only the time window for the intended emotion was coded and periods of task setup and recovery were not included. The coding protocol was modified from several different coding schemes, including those from the Lab-TAB manuals (Goldsmith & Rothbart, 1999; Goldsmith et al., 1999), the Autism & Developmental Disorders Inpatient Research Collaborative (ADDRIC) coding manual (Northrup et al., 2020), and Nuske et al.’ (2017) recent publication.

Emotional reactivity was defined as expressed positive and negative affect. The coding system independently considered facial expressions, vocalizations, and body movements/positioning. Toddlers’ reactivity was characterized based on valence and intensity of affect, on a − 3 (highly negative) to + 3 (highly positive) scale. A code of 0 indicated no change in expression from the child’s baseline state (i.e., neutral or flat affect); a code of − 1 or + 1 indicated subtle, fleeting, or ambiguous negative/positive affect; a code of − 2 or + 2 indicated an obvious expression of negative/positive affect (e.g., a clear frown or smile); and a code of − 3 and + 3 indicated a high-intensity expression of negative/positive affect typically involving multiple indicators of emotion (e.g., big smile and raised eyebrows for intense joy). Given this coding scheme was applied to facial expressions, vocalizations, and body movements/positioning independently, this permitted for incongruence among affect types. For example, if a child was making a brief distressed grunting sound with a flat affect and no movement of his/her body, a code − 2 would be assigned for vocal affect and a code of 0 would be assigned for facial and body affect.

A comprehensive list of regulatory strategies was developed. Undergraduate coders indicated whether each strategy was present in the 10-second interval. Table 3 provides a summary of included regulation strategies.

Table 3.

Definitions and frequencies of regulation strategies.

Strategy	Definition	Non-ASD mean (SD)	ASD mean (SD)	t	d
Overfixation	Child becomes stuck and over-focused on an object during the task	0.15 (0.37)	4.71 (12.46)	−1.51	0.54
Carrying miscellaneous object	Child brings another object into the activity because unable to separate from it	2.25 (3.42)	7.71 (10.31)	−2.09	0.74
Approach toy/task	Reaching for toy/stimulus, walking toward it, leaning toward it	16.50 (12.11)	12.88 (9.25)	1.01	0.33
Distraction	Shifting attention away from stimuli to another toy/object	4.95 (3.63)	11.06 (6.72)	−3.51*	1.16
Avoidance	Attempting to escape or flee from the task/stimuli	6.65 (4.52)	8.00 (6.51)	−0.74	0.24
Perceptual Disengagement	Gaze aversion or looking/turning away to disengage from task/stimuli	0.95 (1.70)	2.00 (2.72)	−1.38	0.47
Freezing	Being paralyzed by fear or significant distress	0.85 (1.63)	1.12 (2.83)	−0.36	0.12
Self-soothing	Any behavior providing oral and/or tactile input	5.35 (11.59)	4.24 (5.84)	0.36	0.11
Self-stimulatory behavior	Hand/arm flapping, rocking, other repetitive or jerky body movements, banging/tapping on table, rocking on chair, spinning	0.15 (0.37)	1.71 (2.62)	−2.43^	0.87
Self-injurious behavior	Aggressive behavior directed towards their own body	0.00 (0.00)	0.06 (0.24)	−1.00	0.37
Aggression directed toward others	An intentional act to hurt or direct emotional reaction in an aggressive manner towards another person	0.05 (0.22)	0.24 (0.56)	−1.28	0.46
Aggression directed toward objects	An intentional act of directing an emotional reaction towards an object	0.15 (0.67)	1.06 (1.95)	−1.83	0.65
Social engagement with caregiver	Actively looking, talking, and/or using gestures directed to the caregiver	5.80 (5.21)	4.35 (4.70)	0.88	0.29
Social engagement with research staff	Actively looking, talking, and/or using gestures directed to the examiner or research assistant	23.95 (12.98)	13.24 (11.70)	2.62^	0.86
Physical comfort seeking with caregiver	Seeking comfort from the caregiver	10.45 (9.70)	8.82 (11.38)	0.47	0.16
Physical comfort seeking with research staff	Seeking comfort from the examiner or research assistant	0.55 (1.10)	0.29 (0.77)	0.81	0.27
Requesting to caregiver	Asking the caregiver for help through vocalization or gestures	2.60 (2.68)	0.82 (1.88)	2.36^	0.76
Requesting to research staff	Asking the examiner or research assistant for help through vocalization or gestures	3.30 (3.63)	2.35 (3.69)	0.79	0.26
Undirected or self-directed vocalizations	Singing, talking, counting, or otherwise vocalizing to self	7.80 (9.69)	7.59 (5.98)	0.08	0.03

Significant differences, p < .05, after correcting for multiple comparisons are noted by an asterisk (*). A carrot (^) indicates that p < .05, but the comparison was non-significant after the false discovery rate (FDR) correction. t-values and Cohen’s d are reported for each pairwise comparison.

A gestalt level of engagement for each task was also coded, in which the child had to show awareness of the task (e.g., briefly look, approach, or make some communicative bid to present adults or to relevant materials). If the child was disengaged in the task, reactivity and regulation codes were not applied because the emotional state was likely unrelated to the task itself.

2.4. Statistical analyses

All data were analyzed in SPSS (Version 24 for Mac). Emotional reactivity and regulation codes were averaged across each variable of interest – facial affect, vocalization affect, body affect – and a total sum of the 19 ER strategies was calculated. When children failed to engage in a task, those tasks were treated as missing data. For descriptive purposes, t-tests were conducted to examine group differences by emotion type (frustration, fear, joy).

To address the primary aim of the study, repeated measures mixed model analyses were conducted. Each model included the affect or ER strategies behavioral codes as the dependent variables; Lab-TAB task type, diagnosis (ASD or non-ASD), and an interaction for task-by-diagnosis as fixed effects; and task type as a repeated measure. Four total models were conducted with the following dependent variables: (1) facial affect, (2) vocal affect, (3) bodily affect, (4) ER strategies. Maximum likelihood was applied to estimate missing values. Significant main effects and interactions were probed using the EMMEANS function; a Sidak correction (Šidák, 1967) is commonly used in this type of model to control for multiple comparisons. The relationship with age and developmental level (i.e., MSEL ELC scores) was examined for each dependent variable. However, given the small sample size, we opted to do post-hoc correlational analyses to detect patterns as opposed to entering these variables as covariates. Significant correlations are reported.

To further evaluate differences between toddlers with and without ASD, t-tests were conducted on the reactivity and regulation variables. The False Discovery Rate (FDR) correction (Benjamini & Hochberg, 1995) was applied to control for multiple comparisons; this correction was used as it has greater power without increasing the number of Type I errors (Shaffer, 1995).

Notably, with 37 subjects and 9 repeated measurements (Lab-TAB tasks), the present study is sufficiently powered (i.e., ≥ 0.80) to detect small effects of f ≥ 0.16. However, posthoc t-tests were only powered to detect large effects of d ≥ 0.95. Subsequently, t-tests with medium, non-significant effects are reported in the figures.

3. Results

The demographic and developmental characteristics of the ASD and non-ASD groups are reported in Table 1.

3.1. Engagement

Five children did not engage in one task each; they engaged in all other tasks. One child with ASD did not actively participate in the Make That Car Go task (joy). Additionally, two children with ASD and two children without ASD did not shift their attention in any capacity during the Scary Mask task (fear).

3.2. Reactivity

Toddlers with ASD demonstrated fewer positive facial expressions (M = 1.00, SD = 0.87) during tasks probing joy than non-ASD toddlers (M = 1.65, SD = 0.49), t(24.44) = 2.70, p = 0.01, d = 0.94. The ASD subgroup also displayed fewer joyful body movements (M = 0.64, SD = 0.57) than the non-ASD subgroup (M = 0.97, SD = .42), t(35) = 2.04, p = .049, d = 0.67. At the aggregate level of emotion task type, there were no other significant differences in affect or ER strategies (Table 4).

Table 4.

Means and standard deviations of reactivity and regulation strategies aggregated by emotion type.

	Frustration				Fear				Joy
	Non-ASD	ASD	t	d	Non-ASD	ASD	t	d	Non-ASD	ASD	t	d
Facial Affect	−0.36 (0.99)	−0.92 (1.18)	1.58	0.52	−0.13 (0.92)	−0.28 (0.71)	0.55	0.18	1.65 (0.49)*	1.00 (0.87)*	2.70	0.94
Vocalizations Affect	−0.59 (0.51)^	−1.10 (1.05)^	1.81	0.63	−0.38 (0.54)	−0.33 (0.52)	−0.27	0.09	0.37 (0.42)	0.15 (0.50)	1.48	0.48
Body Affect	−0.78 (0.44)	−1.08 (0.74)	1.48	0.50	−0.42 (0.50)	−0.40 (0.63)	−0.09	0.04	0.97 (0.42)*	0.64 (0.57)*	2.04	0.67
Regulation Strategies	25.20 (7.41)	30.59 (10.79)	−1.74	0.59	31.50 (10.00)	28.82 (12.77)	0.72	0.24	35.75 (15.82)	32.82 (15.01)	0.57	0.19

Differences at p < 0.05 are noted by an asterisk (*); differences at p < 0.1 are noted by a carrot (^). t-values and Cohen’s d are reported for each pairwise comparison.

The mixed models analysis (depicted in Fig. 1) for facial expressions revealed there was a significant effect for task, F(8, 37.06) = 26.44, p < .001, and diagnosis, F(1, 35.96) = 4.48, p = .04. Although the task-by-diagnosis interaction was non-significant, F(8, 37.06) = 0.76, p = .64, a priori planned pairwise comparisons indicated that the groups demonstrated differing patterns of facial affect on two of nine tasks. Toddlers with ASD demonstrated greater negative affect (M = −0.66, SE = .18) than toddlers without ASD (M = −0.12, SE = .17) during Attractive Toy in a Transparent Box task (frustration), F(1, 37.00) = 4.64, p = .04. Additionally, there was a significant difference on the Modified Peek-a-Boo task (joy), in which children in the non-ASD group displayed more positive facial expressions (M = 1.87, SE = .27) than the ASD group (M = 0.98, SE = .30), F(1, 37.00) = 4.90, p = .03.

Fig. 1. Estimated means and standard errors from the mixed model analyses for facial affect, bodily affect, and vocal affect by task type.

The x-axis depicts the name of each task (Frustration: End of the Line, Attractive Toy in a Transparent Box, Gentle Arm Restraint by Parent; Joy: Make That Car Go, Popping Bubbles, Modified Peek-a-Boo; Fear: Stranger Approach, Remote Controlled Spider, Scary Mask) and the y-axis displays the affect rating (ranging from − 3 to + 3 consistent with the coding scheme). The bars are the estimated means by group and the error bars are the standard error from the mixed model analyses without any covariates. Tasks with an asterisk (*) on the x-axis indicate large-sized, significant group differences, p < .05, whereas tasks with a carrot (^) on the x-axis indicate medium-sized group differences that were non-significant.

The main effect of task, F(8, 35.18) = 12.47, p < .001, and the interaction for task-by-diagnosis, F(8, 35.18) = 4.23, p = .001, was significant when vocalization reactivity was the dependent variable (see Fig. 1). Diagnosis was not significant, F(1, 36.38) = 2.34, p = .13. During the End of the Line task (frustration), toddlers with ASD had more vocal distress (M = −1.28, SE = .21) than toddlers without ASD (M = −0.15, SE = .20), F(1, 37.00) = 14.96, p < .001.

With regard to bodily affect, there was a significant effect for task, F(8, 36.13) = 45.08, p < .001, and a significant task-by-diagnosis interaction, F(8, 36.13) = 2.43, p = .03 (see Fig. 1). The main effect of diagnosis was non-significant, F(1, 36.67) = 2.49, p = .12. Children with ASD displayed greater bodily frustration (M = −1.16, SE = .23) than non-ASD children (M = −.40, SE = .21) during the End of the Line task (frustration), F(1, 37.00) = 5.89, p = .02.

3.3. Regulation strategies

A total frequency of regulation strategies was computed and utilized as the dependent variable in a mixed model analysis (refer to Fig. 2). There was a significant effect of task, F(8, 36.62) = 23.52, p < .001. Diagnosis was not significant, F(1, 36.88) = 0.04, p = .85. The task-by-diagnosis interaction was also non-significant, F(8, 36.62) = 1.20, p = .33, as well as a priori planned pairwise comparisons.

Fig. 2. Estimated means and standard errors from the mixed model analysis for total regulation strategies by task type.

The x-axis depicts the name of each task (Frustration: End of the Line, Attractive Toy in a Transparent Box, Gentle Arm Restraint by Parent; Joy: Make That Car Go, Popping Bubbles, Modified Peek-a-Boo; Fear: Stranger Approach, Remote Controlled Spider, Scary Mask) and the y-axis displays the total frequency of regulation strategies. The bars are the estimated means by groups and the error bars are the standard error from the mixed model analysis without any covariates. There were no significant group differences by task, i.e., all p > .05, but the task with a carrot (^) denoted had a medium-sized group difference that was non-significant.

The top three ER strategies for toddlers with and without ASD were: (1) approach toys/task, M = 14.84, SD = 10.90; (2) social engagement with research staff, M = 19.03, SD = 13.38; and (3) physical comfort seeking with caregiver, M = 9.70, SD = 10.39. In comparison, the least frequently utilized regulatory strategies were: (1) self-injurious behavior, M = 0.03, SD = 0.16; (2) physical comfort seeking with research staff, M = 0.43, SD = 0.96; and (3) aggression directed toward others, M = 0.14, SD = 0.42. There was a differential pattern on how often the groups employed distraction (ASD: M = 11.06, SD = 6.72, non-ASD: M = 4.95, SD = 3.63), t (35) = −3.51, p = .001.

During the tasks probing fear, physical comfort seeking with caregiver, avoidance, and approach toy/task emerged as the top three strategies, with distraction occurring at significantly different frequencies for the ASD and non-ASD groups (ASD: M = 4.76, SD = 2.80; non-ASD: M = 1.50, SD = 2.37), t(35) = −3.84, p < .001. Top strategies during the tasks probing frustration and joy included approaching toy/task and being social with the research staff. In addition, children most commonly engaged in distraction during the tasks probing frustration and carried around a miscellaneous object during the tasks probing joy. There were no significant differences for the tasks probing frustration and joy after correcting for multiple comparisons.

Among the tasks where differential reactivity emerged, toddlers with and without ASD were similar in the amount of regulation strategies used during the End of the Line task (frustration). Relative to the other frustration tasks, children in the sample had the highest rate per minute of employed regulation strategies during End of the Line (frustration; M = 5.59, SD = 2.83 for 30 s), with distraction being the most frequently used (ASD: M = 0.60, SD = 0.94; non-ASD: M = 1.00, SD = 1.17), p > .05. In comparison, toddlers with and without ASD used an average of 17.92 (SD = 8.13) regulation strategies during the 2-minute Attractive Toy in a Transparent Box task (frustration), p > 0.05. Children were most likely to approach the box to regulate their emotions, which was comparable between the ASD (M = 2.24, SD = 2.46) and non-ASD groups (M = 2.70, SD = 2.79), p > .05. Finally, toddlers with and without ASD did not differ on the total regulation strategies utilized during Modified Peek-a-Boo (joy). ER during the Modified Peek-a-Boo task was generally low, averaging 7–8 strategies throughout the entire episode. Toddlers most commonly engaged in social interaction with the researcher, although more so for toddlers without ASD (M = 4.20, SD = 1.77) than toddlers with ASD (M = 2.47, SD = 2.15), t(35) = 2.69, p = .01.

3.4. Relationships with age and developmental level among all children

Age was significantly correlated with the following: (1) facial affect during End of the Line (frustration), r(35) = −0.40, p = .01, Modified Peek-a-Boo (joy), r(35) = −0.44, p = .01, and Popping Bubbles (joy), r(35) = −0.44, p = .01; (2) vocal affect during End of the Line (frustration), r(35) = −0.34, p = .04, Remote-Controlled Spider (fear), r(35) = −0.45, p = .01, and Modified Peek-a-Boo (joy), r(35) = −0.45, p = .01; and (3) body affect during Modified Peek-a-Boo (joy), r(35) = −0.39, p = .02. This pattern of findings indicate that younger children were less likely to show negative affect during the End of the Line task (frustration) and more likely to show positive affect during the Modified Peek-a-Boo task (joy) on at least two of the three measured affect constructs. Additionally, younger children were more likely demonstrate joyful facial expressions during the Popping Bubbles task (joy) and more negative vocalizations during the Remote-Controlled Spider task (fear).

Developmental level was related to the following: (1) vocal affect during End of the Line (frustration), r(35) = −.34, p = .04; and (2) amount of regulation strategies during End of the Line (frustration), r(35) = −.40, p = .01, and Popping Bubbles (joy), r(35) = .35, p = .03. This pattern of findings indicate that children with lower developmental scores had fewer negative vocalizations and fewer attempts to regulate their emotions during the End of the Line task (frustration). Additionally, more developmentally delayed children less commonly attempted to engage in regulation strategies during the Popping Bubbles task (joy).

4. Discussion

Impairments in ER may underlie emotional and behavioral challenges as well as co-occurring psychiatric symptoms in ASD (Beck et al., 2020; Mazefsky et al., 2013), yet very little is known about the early development of emotion dysregulation in ASD. The presence of atypical affect and regulation by 12–24 months in ASD (Bryson et al., 2007; Clifford et al., 2013; Dow et al., 2017; Garon et al., 2016; Zwaigenbaum et al., 2005) suggests that ER difficulties are part of the very early ASD phenotype (Zwaigenbaum et al., 2015) and highlights the importance of studying early ER development. To our knowledge, this was the first study to conduct a systematic observation of reactivity and regulation strategies in toddlers with and without ASD.

The overarching aim of this investigation was to characterize emotional reactivity and regulation strategies for toddlers with and without ASD during their participation in nine Lab-TAB tasks. Results partially supported the hypothesis that toddlers with ASD would exhibit decreased positive affect during tasks probing joy and increased negative affect during tasks probing fear and frustration. Young children with ASD showed significantly less positive affect during tasks probing joy compared to young children without ASD. Additionally, children with ASD demonstrated significantly greater negative affect during two of the three tasks probing frustration, but comparable negative affect during tasks probing fear relative to the non-ASD group. The results did not support the hypothesis that toddlers with ASD would employ fewer and less sophisticated ER strategies, with the exception that children with ASD more often engaged in distraction. In general, the ASD and non-ASD groups utilized a comparable amount and similar types of regulation strategies across tasks.

Regarding emotion type, toddlers without ASD were more likely to display positive facial expressions and body movements during tasks intended to probe joy than children with ASD. However, Modified Peek-a-Boo was the only joyful task to differentiate the subgroups, such that young children without ASD displayed significantly more smiles and more often attempted to socially engage the examiner to up-regulate affect than young children with ASD. Given the non-ASD group was significantly younger than the ASD group, it is important to consider this finding in context of the relationship of age; results of the present study indicated that younger children were more likely to smile and make joyful sounds, consistent with developmental expectations of peek-a-boo play (Miller & Commons, 2007). Although reduced shared enjoyment is considered a symptom of ASD (American Psychiatric Association, 2013; Zwaigenbaum et al., 2016), documented differences in positive affect are variable in the ASD literature (Garon et al., 2009; Macari et al., 2017, 2018). The present study provides some evidence that toddlers with ASD show less positive affect than their non-ASD peers, but additional investigation is needed regarding how children with ASD deviate from a typical developmental trajectory and how context influences the display of joyful affect.

With regard to negative affect, group differences were observed during End of the Line and Attractive Toy in a Transparent Box. These tasks are designed to probe frustration and are rather realistic to real-world experiences – having a toy taken away from the child (End of the Line) and having difficulty accessing a desired item (Attractive Toy in a Transparent Box). Young children with ASD were significantly tenser and more vocally distressed than their non-ASD peers during End of the Line, suggesting that this task was effective at elucidating frustration in a laboratory setting. Additionally, toddlers with ASD demonstrated more negative facial expressions than toddlers without ASD during Attractive Toy in a Transparent Box. However, the End of the Line task was sensitive to age and developmental level. In an independent study of youth with ASD who were admitted to an inpatient unit for severe emotion dysregulation (Northrup et al., 2020), younger children (6–13 yrs vs. 14–21 yrs) demonstrated greater negative affect during End of the Line, which documents the opposite relationship found in the present study with children who were younger and showed less negative affect. Northrup and colleagues (2020) also documented that verbal cognitive functioning impacted how reactivity manifested in youth with ASD. Collectively, these results highlight the importance of longitudinal studies to better understand the development of negative affect throughout early childhood and into adolescence as well as the impact of age, language, and cognitive functioning.

One notable finding was the difference across type of negative task. Whereas some group differences emerged in the frustration tasks, the results did not support differences in affect during tasks probing fear. This could be attributed to low reactivity occurring in both groups during the fear tasks (i.e., all children showed minimal reaction to these tasks). There is some evidence that laboratory procedures evoke less fear than parent report of child fear in everyday situations (Talge et al., 2008). However, this result may also emphasize the importance of context and type of elicitation probe. For example, as we did, Macari et al.’ (2018) findings documented high negative affect during frustrating tasks in toddlers with ASD. However, in a social communication task, Wetherby et al. (2004) documented more comparable levels of distress in toddlers with and without ASD. Overall, this may suggest that conclusions regarding emotional reactivity may vary based on the situation, and especially based on the paradigm for laboratory studies.

An unexpected finding was that the groups used a comparable amount of regulation strategies throughout all tasks. However, this finding paired with greater negative affect present in toddlers with ASD suggests that employed regulation strategies were at least partially ineffective, consistent with previous studies in older children with ASD (Jahromi et al., 2012; Mazefsky et al., 2014; Samson et al., 2012). Although there were varying levels of regulation strategies utilized between the tasks, robust group differences were not documented within task. For example, young children with and without ASD collectively utilized a small amount of regulation strategies when expressing joy during Modified Peek-a-Boo. Additionally, both groups frequently employed ER strategies during Attractive Toy in a Transparent Box. This task was adapted from the original to remove the use of keys (unless the child requested them as part of a regulation strategy) to be best suited for 22–28-month-olds (Goldsmith et al., 1999; Mullen, 1995); approaching the box to explore may have been toddlers’ attempt to problem solve how to access the desired toy.

Across all tasks, the most common strategies included approaching the task or a presented toy with curiosity, engaging the examiner in a social interaction, and seeking comfort from a parent. Distraction was also frequently utilized during the tasks probing frustration. These strategies are consistent with what Gulsrud and colleagues (2010) documented in toddlers with ASD. The present study did not find differences between the ASD and non-ASD groups; however, this finding should be interpreted in context of the small sample size. The findings are also generally in line with previously documented emotion regulation studies of typically developing children, such that 18–24-month-olds frequently problem solve and engage in distraction (Calkins & Johnson, 1998; Calkins et al., 1998; Diener & Mangelsdorf, 1999); however, TD toddlers also frequently seek assistance, which was not a common strategy used by toddlers in the present study. Data from this study indicated that toddlers with ASD were more likely to engage in distraction than their non-ASD peers, particularly during the tasks probing fear; this finding is the only one in support of the hypothesis that children with ASD would use less sophisticated strategies than children without ASD. Distraction was operationalized as the child shifting his/her attention from the stimuli to another toy/object. Although distraction can serve as an ER strategy, it is also possible that there was a higher prevalence in toddlers with ASD consistent with the ASD phenotype. Previous studies have documented that young children with (or with increased likelihood) for ASD show an attentional bias towards objects of interest (Jones & Klin, 2013; Sasson & Touchstone, 2014).

Overall, the present study indicates a general pattern of toddlers with ASD demonstrating greater levels of distress despite utilizing age expected ER strategies and fewer indicators of joy than toddlers without ASD. Findings corroborate previous studies of parent-report data that young children with ASD display lower positive affect and higher negative affect than their non-ASD peers (Garon et al., 2009; Garon et al., 2016; Macari et al., 2017). Results differed from the only other study of toddlers with and without ASD in which peak emotional expressiveness was assessed during Lab-TAB tasks (Macari et al., 2018). Varying findings are potentially due to differences in behavioral coding approaches. However, the conclusions across both studies - that toddlers with ASD have atypical emotional reactions - was consistent. The present study also supports a previously documented pattern that children with ASD consistently employ ER strategies albeit ineffectively (Jahromi et al., 2012; Mazefsky et al., 2014; Samson et al., 2012). These findings demonstrate the importance of considering both reactivity and regulation when assessing ER as a construct, commensurate with findings from several previous studies (Mazefsky, Yu, et al., 2018; Mazefsky et al., 2020; Zelkowitz & Cole, 2016).

4.1. Strengths

There were numerous strengths of this study. First, the present study included toddlers who were between 22 and 28 months of age with in-depth confirmation or rule-out of ASD. One challenge to studying early emotion dysregulation in ASD is that the average age of ASD diagnosis in the United States is around 4.5 years (Maenner et al., 2020) despite the capacity to make diagnoses that are valid and stable in children as young as 14 months (Guthrie et al., 2013; Pierce et al., 2019). Participants from this study were ascertained from a study focused on ASD screening during the first two years of life, which allowed children to participate in the emotion-focused protocol shortly after early diagnosis around two years of age.

Second, prior investigations in preschool-aged children (Hirschler-Guttenberg et al., 2015; Jahromi et al., 2008; Nuske et al., 2017) have utilized typically developing children as the comparison group. We included TD and DD children in the non-ASD comparison group in an effort to capture a similar level of heterogeneity in developmental level as the ASD group (Maenner et al., 2020). Macari and colleagues (2018) did not find significant ER differences between their DD and TD groups, providing support for the utility of a non-ASD comparison group. Further, we also examined and reported the relationship with dependent variables and MSEL scores, although future work in this area is necessitated.

Third, observational methods allow for objective assessment of ER (Cibralic et al., 2019; Weiss et al., 2014). Previous studies of emotion dysregulation in young children with ASD have primarily relied on parent-report and have been predominantly limited to focus on reactivity, whereas the present study was able to characterize both reactivity and regulation to more comprehensively study the ER construct.

4.2. Limitations

There are several limitations to be considered. The sample size was rather small, resulting in some analyses being underpowered. To account for this, only differences that were at least medium in size are reported in tables/figures. This study also incorporated many comparisons and analyses given its novelty and exploratory nature; Sidak and False Discovery Rate corrections were applied to reduce Type I error. However, it is possible by doing so that key, underpowered findings did not emerge. Further, the ascertained sample was homogeneous with regard to race and there were significant differences for sex and age. Collectively, these factors impede the generalizability of the present results. Replication with larger, more diverse samples is needed, particularly given previous literature documenting sex and sociocultural differences in emotion dysregulation-related constructs (e.g., Macari et al., 2018; Sameroff et al., 1982).

Another limitation is that tasks in the laboratory may not be realistic or probe the desired emotion (Gardner, 2000). However, previous studies have documented the relationship of Lab-TAB tasks with parent-reported temperament and psychiatric/behavioral symptoms in ASD samples (Macari et al., 2018; Northrup et al., 2020). Moreover, the primary clinician (first author) was not blind to diagnostic status due to her expertise in the early ASD phenotype; however, all undergraduate research assistants were blind to diagnostic status at time of visit and coding. Finally, results indicated that age and developmental level played a significant role in the present findings. This was a young sample and the groups included children with varied developmental challenges from ASD to non-ASD DD to TD and the nature of these developmental challenges may change across time. The sample size, cross sectional design, and included measures did not allow for further investigation of the impact of age and developmental level, particularly in exploring the role that communication plays in ER (Kasari et al., 2006; Prizant et al., 2003). Replication studies with larger samples as well as longitudinal studies with a focus on developmental changes would address these limitations.

4.3. Implications

The present study used a developmentally appropriate framework for characterizing emotion dysregulation in toddlers with ASD. Several tasks differentiated the ASD and non-ASD groups – End of the Line, Attractive Toy in a Transparent Box, and Modified Peek-a-Boo. In particular, reactivity during the End of the Line task in which children had a toy removed was robust, as children with and without ASD differentially responded in two of three domains of affect. In addition, four regulatory strategies were most commonly utilized in this sample of children: approach toy/task, social engagement with examiner, physical comfort seeking with caregiver, and distraction. Further research should be conducted to identify the best tasks and most important ER strategies for this age range (e.g., applying machine learning).

Additional research is needed to investigate if toddlers with ASD have observable differences in baseline reactivity or if there are nuances in the way in which ER strategies are employed. Future investigations could also include: (1) looking at changes within the task and/or recovery time after the task; (2) examining regulation strategies utilized at the peak intensity of affect; (3) quantifying other markers of emotion dysregulation, such as electrodermal activity, heart rate variability, and parent-reported behavioral/emotional challenges in daily life. Moreover, extending this study to examine the role of parental responses during negative and positive situations may be beneficial, as previous studies have demonstrated that parental responsivity influences ER in young children (Gulsrud et al., 2010; Hirschler-Guttenberg et al., 2015; Smith et al., 2006). Although the present study provided some preliminary findings regarding observable differences in ER among toddlers with and without ASD, further characterization of ER in the first few years of life is crucial. Such data could provide a clearer understanding on the relationship between dysregulation and early behavioral challenges and provide key baseline measurements for future longitudinal studies examining ER as a theorized mechanism of co-occurring psychiatric symptoms in ASD (Beck et al., 2020; Chandler et al., 2016; Damiano et al., 2014; Mazefsky et al., 2013; Mazefsky, Day, et al., 2018; Northrup et al., 2020; Weiss et al., 2017).