The relationship between autism symptoms and arousal level in toddlers with autism spectrum disorder, as measured by electrodermal activity

Abstract

Electrodermal activity was examined as a measure of physiological arousal within a naturalistic play context in 2-year-old toddlers (N = 27) with and without autism spectrum disorder. Toddlers with autism spectrum disorder were found to have greater increases in skin conductance level than their typical peers in response to administered play activities. In the autism spectrum disorder group, a positive relationship was observed between restrictive and repetitive behaviors and skin conductance level increases in response to mechanical toys, whereas the opposite pattern was observed for passive toys. This preliminary study is the first to examine electrodermal activity levels in toddlers with autism spectrum disorder during play-based, naturalistic settings, and it highlights the potential for electrodermal activity as a measure of individual variability within autism spectrum disorder and early development.

Background

Autism spectrum disorder (ASD) is a developmental disorder characterized by deficits in social and communicative abilities as well as the presence of repetitive and restricted behaviors (RRBs; American Psychiatric Association, 2013). For toddlers on the autism spectrum, these communicative and social deficits are evident in difficulties with nonverbal communication skills such as reaching, pointing, and gestural use (Wetherby et al., 2004), and atypical object exploration (Ozonoff et al., 2008). While the ultimate cause of these atypical behaviors in ASD is unknown, it has been hypothesized that individuals on the spectrum could have disregulated or heightened sympathetic responses, especially when interacting with others, possibly making social activities aversive (Kylliäinen et al., 2012; Kylliäinen and Hietanen, 2006; Tinbergen, 1974).

The measurement of electrodermal activity (EDA) provides a window into the responses of the sympathetic branch of the autonomic nervous system (ANS; Dawson et al., 2007). EDA refers to changes in the electrical properties of the skin and includes changes in skin conductance level (SCL) or specific galvanic skin conductance responses (SCRs; Dawson et al., 2007). Changes in SCL and SCR are related to cognitive demand, anxiety, attention, sensory input, novelty, and affect (Dawson et al., 2007).

Previous studies disagree about whether individuals with ASD have heightened (Kylliäinen et al., 2012; Kylliäinen and Heitanen, 2006) or comparable (Hirstein et al., 2001; Levine et al., 2012; McCormick et al., 2014) autonomic reactions to social stimuli compared to their typical peers. For instance, children with ASD demonstrated responses of a greater magnitude than their typical peers when looking at pictures featuring direct gaze or eyes wide open (Kylliäinen et al., 2012; Kylliäinen and Heitanen, 2006). When watching a video meant to elicit repetitive and sensory behaviors, 3-year-olds with and without ASD showed no difference in EDA activity levels from their typical peers (McCormick et al., 2014). In a naturalistic social context, Hirstein et al. (2001) found that about two-thirds of their participants with ASD had heightened SCL during interaction with adults while the other third were hyporesponsive, suggesting that developmental heterogeneity in ASD may play a role in the inconsistent findings observed in EDA studies.

It is widely believed that the compounding influences and interactions of atypical social learning experiences heavily contribute to behavioral, clinical, and neurophysiological heterogeneity in ASD (Jones and Klin, 2009). Thus, examining features of ASD earlier in development would allow for investigations of the integrity of the ANS before confounding effects due to intervention and learned cognitive strategies take hold, potentially clarifying the nature of conflicting reports on EDA in ASD.

Therefore, this preliminary study examined EDA in toddlers with and without ASD during a semi-structured assessment of nonverbal and preverbal functioning, the Communication and Symbolic Behavior Scale (CSBS; Wetherby and Prizant, 2002). To that end, we investigated whether (1) toddlers with autism have a heightened autonomic response compared to their typical peers during this socially demanding assessment, (2) SCL changes differ based on the nature of the probe the child is engaged with during the assessment, and (3) a relationship exists between SCL changes and behavioral measures including RRBs or verbal functioning. We hypothesized that activities requiring more direct interaction with the clinician would lead to greater change in SCL as compared to those that were more passive; further, increasing atypicalities in behavioral phenotypes (e.g. greater autism symptoms, lower verbal functioning) would positively correlate with SCL change.

Methods

Participants

Participants included 15 toddlers with ASD (13 male; M = 21.7 months, SD = 3.29 months) and 12 typically developing (TD) boys (M = 22.75 months, SD = 3.73 months). Due to technical issues or hyporesponsivity, four participants (n_ASD = 2, n_TD = 2) were excluded from the analysis. In this study, we examined relative changes from baseline SCL; thus, hyporesponsive toddlers, who had baselines close to 0 microsiemens (μs), would have shown unstable values for relative change due to a small denominator effect. Though baseline levels of SCL can vary widely between individuals (Dawson et al., 2007), with some individuals being true hyporesponders (low EDA responses overall), in practice it is difficult to distinguish between true hyposensitivity and technical errors (e.g. skin-disconnected electrode) leading to reports of low EDA. Participants were recruited through a university clinic specializing in the diagnosis of ASD. All toddlers received a developmental assessment, the Mullen Scales of Early Learning (Mullen, 1995), and an assessment of nonverbal communication skills, the CSBS. Toddlers with ASD also received the Autism Diagnostic Observation Schedule (ADOS-G, Toddler Module; Luyster et al., 2009) from research-reliable clinicians (Table 1).

Table 1.

Sample characterization means (standard deviations).

Measure	ASD	TD
N	15	12
Age (months)	21.70 (3.29)	22.75 (3.73)
Verbal DQ	49.35 (30.75)	107.60 (20.70)
Nonverbal DQ	80.38 (14.85)	108.21 (13.15)
ADOS SA	14.87 (4.10)	n/a
ADOS RRB	4.27 (2.25)	n/a

ASD: autism spectrum disorder; TD: typically developing; ADOS: Autism Diagnostic Observation Schedule; RRB: repetitive and restricted behavior.

Procedures

A skin conductance sensor (Affectiva Q-Sensor) was placed on the lower calf of participants when they arrived at the lab, continuously recording throughout the visit and all assessments. Typically, EDA is measured on the palm of the hand or plantar surface of the foot (Dawson et al., 2007; McCormick et al., 2014), however, the lower calf has also been shown to accurately record data in young children and adults while allowing them full freedom of movement (Hedman et al., 2012; van Dooren and Janssen, 2012). The sensor measures EDA in microsiemens (μs) at up to 32 Hz (in this study EDA was collected at 8 Hz) as well as 3-axis acceleration and skin surface temperature. Data are stored on an internal flash drive for later analysis. The Q-Sensor has been shown to be comparable to traditional wired measures of EDA in cognitive, affective, and physical tasks (Fletcher et al., 2010) and has been successfully used in toddlers with placement on the lower calf (Hedman et al., 2012). In all cases, participants wore the sensor for at least 15 min prior to data collection in order to allow adequate warm-up time.

The CSBS is a 25- to 30-min semi-structured communication assessment, during which the clinician measures nonverbal communicative bids, including showing, pointing, and requesting as well as early linguistic skills (Wetherby and Prizant, 2002). The CSBS was administered by speech-language pathologists or a research assistant trained to administer it reliably. All participants were shown a subset of six toys, including bubbles, windup toys, a stretchable tube, a giraffe figurine, and picture books. In a typical session, the researcher presented a toy, allowed the toddler to play with it for several minutes while attempting to elicit social and or communicative bids, and then switched out the toy as the child lost interest. Some of these activities occurred while the child was seated at a table, and others involved movement around the room; this was largely dictated by the compliance and activity level of the child.

Data Analysis

Definition of toy and activity conditions

We examined changes in SCL as participants interacted with toys in the CSBS. Toys were categorized into five conditions: passive, books, mechanical, sensory, and temptation. Passive toys were still figurines that were not interactive, mechanical included cars and wind-up toys, sensory included stretchable, pliable toys, including a rubber ball with stringy arms, and temptation covered any activity that prompted the child to request for a desired item, for example, bubbles, balloon, and snack time. Following the assessment, coders matched timing of the CSBS with the output from the EDA sensor by watching video recordings of the assessment and marking when participants played with each toy. Because some toys were shown more than once, only the first presentation of each toy or activity was considered in the present analysis.

Data transformation

The dependent variable of interest was the relative change in SCL from baseline during each toy condition of the CSBS. The individual sessions associated with toy conditions evolved over time as the child interacted with the toys, and there was no constant baseline duration between activities. For this reason, an overall percentage of each trial was selected as a baseline of prestimulus activity due to the typical latency between stimulus presentation and subsequent SCR (Dawson et al., 2007) and on the basis of similar prior methodology (Miller et al., 1999). The baseline period included the SCL measurements in the first 20% of the presentation of each toy (M = 19.2 s, SD = 15.5 s). For each toy condition, the magnitude of relative change in EDA (Δ) was defined as the difference between the median SCL during baseline and the 95th percentile SCL during the remaining 80% of time during toy presentation (M = 76.9 s, SD = 62.1 s; see equation). The magnitude of change for each condition was then normed by dividing it by the median of the baseline for that condition, resulting in a measure of relative change. The log of this normed magnitude was taken to even out the distribution of the values (Kylliäinen et al., 2012) and to ensure normality in the underlying distribution. After transformation these relative changes within each toy type and diagnostic group passed both Kolmogorov Smirnov and Shapiro Wilks tests of normality, ps > 0.05, thus allowing for the use of parametric analysis.

$$\begin{array}{lll}\mathrm{\Delta }& =& \mathrm{Ln}(({95}^{\mathrm{th}}\mathrm{Percentile\; Value\; of}\\ & & \mathrm{Toy\; Condition}-\mathrm{Median\; of\; Baseline})/\mathrm{Median\; of\; Baseline})\end{array}$$

Results

A linear mixed model factorial analysis of Diagnosis (ASD vs TD) and Toy Condition (Passive, Books, Mechanical, Sensory, and Temptation) revealed a main effect of diagnosis (F(1, 18.1) = 8.9, p = 0.008, Cohen’s d = 1.27) and of toy condition (F(4, 15.6) = 3.1, p = 0.048; Figure 1). Toddlers with ASD showed significantly greater change in SCL than their typical peers. There was no difference in baseline SCL between the groups (F(1, 20.12) = 0.035, p = 0.853). All toddlers’ level of physiological arousal varied across conditions, with passive toys evoking the least change in SCL and temptation the most. No toy by diagnosis interaction was observed, which may be due to the relative small sample size for this study (F(4, 15.6) = .79, p = 0.55). Nevertheless, moderate to large effect sizes were observed between groups for most of the toy conditions, suggesting that an interaction might emerge with increased power (Table 2; Cohen, 1988).

Figure 1.

Bars indicate group means of change in EDA during presentation of five different toy conditions. Error bars indicate ±1SE. The ASD group exhibited larger changes in EDA than did the TD group, for all toy and activity conditions, indicating greater affective arousal.

Table 2.

Effect sizes for between group comparisons by toy condition.

Toy condition	Cohen’s d
Passive	0.19
Books	0.55
Mechanical	0.61
Sensory	1.75
Temptation	0.71

There was no significant toy × diagnosis interaction, so caution should be used in interpretation.

The CSBS is a semi-structured assessment, and the amount of time toddlers spent with each toy category was not consistent across participants. However, when the time of each toy play session was added into the model, we found that there was no statistically significant effect of the time spent playing with each toy on our EDA relative change measure (F(1, 74.3) = 1.4, p = 0.235).

We also predicted that ADOS scores would be related to change in SCL in the toddlers with ASD. There was no relationship between social affect or language scores and SCL change. However, correlations between EDA magnitude and ADOS RRB subscores revealed that for the passive toys, a higher magnitude of SCL change negatively correlated with ADOS RRB scores (r = −.714, p = 0.031). Conversely, during the mechanical toy sections of the CSBS, SCL change was positively correlated with ADOS RRB scores (r = .613, p = 0.045). There were no significant correlations between RRB scores and change during other toy conditions.

Discussion

Our hypothesis that there would be a significant difference in EDA between toddlers with ASD and their typical peers was supported: participants with ASD had a greater change in SCL in response to all activities in the CSBS, consistent with the theory that heightened autonomic arousal is related to the symptoms of ASD. Differences also emerged in SCL change across toy conditions for toddlers in both ASD and TD groups, where passive, non-interactive toys evoked the least amount of change in SCL and seemingly more exciting sensory and temptation activities evoked greater change.

In the ASD group, RRBs were related to arousal level when interacting with mechanical and passive toys. Change in SCL was positively correlated with RRBs during play with the mechanical toys, and negatively correlated with RRBs during play with the passive toys. One possible explanation for this differential response is that the children with ASD found more opportunities to engage in repetitive behaviors with the mechanical toys than the passive ones. Informal qualitative observation supports this idea. Participants with ASD often used mechanical toys (e.g. cars, planes, and spinning tops) in stereotyped or self-stimulating ways such as lying on the floor and moving the car back and forth quickly, or watching the spinning top while hand-flapping. In comparison, all toddlers (with and without ASD) spent the least amount of time with the passive toys (e.g. animal figurines) and toddlers with ASD often dropped the toys quickly or did not seem engaged with the examiner’s attempts at play. Thus, although our sample size was small, these findings suggest that EDA may relate to specialized interests and difficulties with imaginative play. Future research should determine which specific behaviors were related to a rise in SCL. This could be done by watching the CBCL videos and coding for RRBs during each interaction with the toys.

This preliminary study is the first to examine skin conductance in toddlers with ASD, and it offers support for the heightened sympathetic response theory (Kylliäinen et al., 2012; Kylliäinen and Heitanen, 2006; Tinbergen, 1974). Toddlers with autism seem to be experiencing a heightened level of arousal as measured by change in SCL, and this heightened arousal is also associated with an overall increase in RRBs as measured on the ADOS. At an early age, individuals with autism may be overstimulated by social play activities as demonstrated by an increase in SCL from baseline.

However, this study is limited to small sample of children, and it used a semi-structured context with many activities found in naturalistic settings to measure SCL. Future work with toddlers with ASD will need to employ larger sample sizes, and augment the analysis of EDA by looking at SCR as well as SCL, including peaks per minute, rise time, and duration (Dawson et al., 2007). Future studies should also examine the effects of motor movements, room temperature, and hydration status on SCL, and investigate generalization of our results to other environments. For example, matching individual observed behaviors time-locked with EDA responses could shed light on the emotional valence associated with an intense rise in skin conductance. In addition, while we used percentile-based statistics to reduce the impact of outlier data, movement artifacts were not accounted for in the data analysis and could have impacted results. Nevertheless, this preliminary work is an important step towards understanding the relationship between sympathetic arousal and ASD, especially early in development.