Short-Term Trajectories of Restricted and Repetitive Behaviors in Minimally-Verbal Children with Autism Spectrum Disorder

Abstract

Very little is known about the 30% of children with Autism Spectrum Disorder (ASD) who remain minimally verbal when they enter school. Restricted and repetitive behaviors (RRBs) are well-characterized in younger, preschool and toddler samples. However, the prevalence and impact of RRBs has not been characterized in older, minimally-verbal children. The goal of this study was to characterize this core diagnostic feature in minimally-verbal children with ASD ages 5 to 8 years over a 9-month period to better understand how these behaviors manifest in this crucially understudied population. RRBs were coded from caregiver-child interactions (CCX) at four timepoints. Upon entry into the study, children demonstrated an average of 17 RRBs during a 10-minute CCX. The most common category was Verbal. RRBs remained constant over six months, however a slight reduction was observed at the final timepoint. Compared to prior literature on younger samples, minimally-verbal children with ASD demonstrated higher rates of RRBs and higher rates of verbal RRBs. Further work is required to understand the function and impact of RRBs in minimally-verbal children.

Lay Abstract

Approximately one third of children with autism spectrum disorder (ASD) remain minimally-verbal at the time of school entry. In this study we sought to characterize the presence of restricted and repetitive behaviors (RRBs) in school-aged children (5 to 8) who were minimally verbal. Compared to prior studies, minimally-verbal children with ASD had higher frequencies of RRBs and demonstrated a different profile of behaviors, including more verbal RRBs.

Introduction

Nearly a third of children with autism spectrum disorder (ASD) have very limited functional spoken language skills by the time they enter school (Anderson et al., 2007; Bacon, Osuna, Courchesne, & Pierce, 2019; Rose, Trembath, Keen, & Paynter, 2016; Tager-Flusberg & Kasari, 2013). These children, characterized as ‘minimally verbal,’ are heterogeneous with respect to cognitive and social-communication abilities, language understanding, and language use (Tager-Flusberg & Kasari, 2013). The majority of research in this group has focused on characterizing their social-communicative profiles and developing language skills (DiStefano, Shih, Kaiser, Landa, & Kasari, 2016; Kasari et al., 2014; Paul, Campbell, Gilbert, & Tsiouri, 2013; Thurm, Manwaring, Swineford, & Farmer, 2015). Restricted and repetitive behaviors (RRBs), the other core diagnostic feature of ASD, have received much less attention and we know little about the typographies and trajectories of RRBs in this group. The goal of this study is to examine the profile of RRBs in a sample of school-aged minimally-verbal children with ASD over a short developmental window and how language abilities associate with these behaviors.

RRBs in Minimally-Verbal Children with ASD

RRBs, defined as the expression of repetitive body mannerisms, overriding preoccupations with objects/parts of objects, sensory behaviors and strict adherence to routines and rituals, are a core symptom of ASD (American Psychiatric Association, 2013). RRBs are an extremely heterogeneous group of behaviors, thus researchers have studied RRBs by topography, dividing behaviors by the manner in which they present (verbal, motor, or object-related; Barber, Wetherby, & Chambers, 2012; Harrop, McConachie, Emsley, Leadbitter, & Green, 2014; Watt, Wetherby, Barber, & Morgan, 2008). RRBs have also been classified as lower order behaviors (repetitive motor actions, repetitive object manipulations) or higher order (routines, insistence on sameness, circumscribed interests) behaviors (Turner, 1999). Studying RRBs in ASD is a somewhat controversial subject: parents report RRBs to be interfering, can cause social stigma, and, if left untreated, can escalate to self-injurious behaviors (Boyd, McDonough, Rupp, Khan, & Bodfish, 2011; Woodgate, Ateah, & Secco, 2008). However, autistic individuals also discuss the positive aspects to RRBs. For example, Kapp and colleagues (2019) discuss autistic adults’ experiences of stimming as a coping mechanism and form of self-regulation. Ascertaining the reasons behind RRBs is particularly challenging in minimally-verbal samples and therefore caution must be used with regards to intervening to reduce these behaviors, particularly if a clear understanding of their function cannot be determined.

Studies characterizing RRBs in ASD report changes over time, both with regards to frequency (Harrop et al., 2014; Honey, McConachie, Randle, Shearer, & Le Couteur, 2008; Wolff et al., 2014) and the form of expression (Esbensen, Seltzer, Lam, & Bodfish, 2009; Lam, Bodfish, & Piven, 2008). In young children, data from parent report and direct observation suggests that the most common form of RRBs are those involving repetitive actions with objects (Barber, Wetherby, & Chambers, 2012; Stronach & Wetherby, 2014; Watt, Wetherby, Barber, & Morgan, 2008). Behaviors categorized as lower order typically decrease over time, becoming less frequent with age and development, but not disappearing completely (Honey et al., 2008; Militerni, Bravaccio, Falco, Fico, & Palermo, 2002; Moore & Goodson, 2003; Richler, Huerta, Bishop, & Lord, 2010). More higher order RRBs have been reported in older and more cognitively advanced children (Moore & Goodson, 2003; Richler et al., 2010), such as intense interests and an insistence on sameness or routine. However, the majority of these data stem from toddler- and preschool-aged samples or older, verbal children. Therefore, it is unclear whether this is still the case when children reach school age and remain minimally verbal.

A number of studies have explored the associations between language and RRBs as a proxy for how language and RRBs co-occur with most studies reporting negative associations between language abilities and RRBs (Harrop et al., 2014; Lam et al., 2008; Ray-Subramanian & Ellis Weismer, 2012). Hus, Gotham and Lord (2014) reported higher raw RRB Autism Diagnostic Observation Schedule Module One scores (ADOS; Lord et al., 2012) in school age children (4 to 10 years) who were classified as “no words” on the ADOS. Verbal IQ accounted for the majority of variance in raw RRB totals, however the contribution of verbal ability to calibrated severity scores was less conclusive. Thurm et al. (2015) reported that RRBs remained high in preschoolers who remained minimally verbal over time, however RRBs were not a significant predictor of remaining minimally verbal one year after the initial assessment. Bal and colleagues (2016) also reported only slightly higher rates of RRBs in school-aged children with ASD with few to no words compared to children with a few words (ADOS-2 Module 1 classifications) and Wodka, Mathy, and Kalb (2013) found that RRBs were not associated with delayed speech acquisition.

Given the mixed findings around the association between language and RRBs and minimal literature classifying the profile of RRBs displayed by minimally-verbal, autistic children, the goal of this current study was to characterize RRBs displayed during a free play session. RRBs were coded at four timepoints in a nine-month time period in school-aged minimally-verbal autistic children. We anticipated higher rates of RRBs in this older, minimally-verbal sample, as well as a more varied profile of RRBs compared to toddlers and preschoolers. Based on our previous research, we expected some developmental change in RRBs with a reduction in total frequencies over time (Honey et al., 2008; Harrop et al., 2016; Harrop et al., 2014).

Method

Ethical approval for this research was obtained through Institutional Review Boards at the three study sites (UCLA, Vanderbilt, and Kennedy Krieger); caregivers gave written consent for their child to participation.

Participants

Sixty-one children with ASD and their caregivers were enrolled within a multisite, social-communication study (outlined in detail in Kasari et al., 2014; Shire et al., 2015; DiStefano et al., 2016; Clinical trial registration information-Developmental and Augmented Intervention for Facilitating Expressive Language (CCNIA); http://clinicaltrials.gov/; NCT01013545). Families were initially screened by phone and invited for screening assessments. Inclusion criteria included: (1) a clinical diagnosis of ASD confirmed by the administration of the Autism Diagnostic Observation Schedule Generic - Module 1 (ADOS-G; Lord et al., 2000); (2) age between 5 and 8 years; (3) absence of a known genetic disorder; (4) minimally verbal (defined as fewer than 20 spontaneous functional spoken words on a naturalistic language sample); (5) received at least two years of prior early intervention; and (6) nonverbal developmental age of at least 24 months derived from the Leiter International Performance Scale-Revised (Leiter-R; Roid & Miller, 1997).

Coding of Restricted and Repetitive Behaviors (RRBs)

Our primary outcome (RRBs) was observationally coded from videotaped caregiver-child interactions (CCX)- a videotaped play interaction between the child and their primary caregiver with a standard set of toys (including a ball, blocks, vehicles, dinosaurs, dishes with food, dolls and furniture, drum, puzzle, and shape sorter). Parents were instructed to play as they typically would. The CCX was recorded at all four study time points; Entry, Stage One (three months post entry), Exit (six months post entry) and Follow Up (9 months post entry/3 months post exit). Children who were randomized to receive the social communication intervention (JASP+EMT) with an augmented or alternative communicative (AAC; see Kasari et al., 2014 for further details) had access to a device during the CCX. This study used a speech generating device (SGD), such as an iPad or DynaVox, which displays symbols that produce voice output when selected. The SGD was programmed with relevant vocabulary for the CCX. Half (N = 31) of the children were randomized to the + AAC group and had access to the AAC during all CCXs. A further 6 children we re-randomized to the + AAC following Stage One assessments and had access to the AAC at Exit and Follow Up CCXs (see Kasari et al., 2014).

Each observation lasted between 10 and 12 minutes and the first 10 minutes of video was coded using the scheme developed by our team with some modifications based on the age of the children in this study (Harrop et al., 2016; Harrop, McConachie, Emsley, Leadbitter, & Green, 2014; Harrop, Gulsrud, & Kasari, 2015). Using Noldus Observer^® (Noldus, 1991), each RRB observed within the ten-minute play session was coded into one of nine categories: Motor, Visual, Repetitive Object Use (referred to as Object), Verbal, Sensory Aversion, Sensory Seeking, Repetitive AAC Use, Insistence on Sameness, Self-Injurious Behaviors (SIB) and Other (Table 1). For behaviors such as hand flapping or spinning, we coded bursts of activity rather than each individual action. An independent observer coded 25% of CCXs; inter-rater agreement was high: Chronbach’s Alpha for Total RRBs was 0.72 (individual categories range: 0.69-0.86).

Table 1:

Operational definitions and examples of RRBs coding during CCX (adapted from Harrop et al., 2015)

RRB Category	Operational Definition	Examples
Motor	Whole body movements (spinning, rocking, pacing, jumping, toe walking) that are repetitive and/or atypical OR complex and/or unusual mannerisms (hand flapping, finger/hand mannerisms). Each burst of behavior is coded.	Paces Spinning Hand flapping Hand posturing
Visual	Child visually inspects toys/objects for prolonged period and/or in an atypical manner (brings toys close to eyes, lays down and peers at toys)	Peers at toys from angle Brings toys close to eyes
Repetitive Object Use	Child plays/uses objects in a repetitive and/or non-functional way. This can include non-functional and/or repetitive play routines even if the play itself is relatively complex.	Arranges objects in a line Retains/hoards objects Pushes buttons over and over Uses restricted number of toys Bangs toys together
Verbal	Child uses repetitive and/or atypical vocalizations. Coding encompasses pitch, intonation, scripting, and echolalia.	Repetitive sounds Repetitive language Echolalia Scripted phrases
Sensory Aversion	Child has a noticeable reaction to sensory stimuli (touch, sound, visual).	Covers ears in response to sounds Physical reaction to sounds/actions
Sensory Seeking	Child displays an unusual interest in smell/touch/sounds.	Rubs toys to face Brings toys to ear to hear sounds Smells toys Rubs caregivers hair repeatedly Mouths toys
Insistence on Sameness	Child performs actions in a particular order or way or insists others behave in similar way.	Reacts negatively when caregiver plays with toy in different way Child protests if sequence of actions are disrupted Child insists toy remains in same location
Repetitive AAC Use	Child does not use AAC in a functional way, but uses repetitively and not for communication.	Presses buttons on AAC devise repeatedly and indiscriminately
Self-Injurious Behaviors	Child engages in self-injurious behaviors such as pinching self, hitting self with hand or object, banging head, skin picking.	Hits self repeatedly Pinches skin
Other	Child engages in behavior that is repetitive/restricted in nature but does not fall into the above categories

Of the 61 children recruited, 57 had videotaped CCX samples at Entry. Fifty-five completed Stage 1 assessments (3 months post entry) and 51 remained at Exit (6 months post entry). Of those, 44 returned for Follow Up assessments (9 months post entry).

Standardized Assessments and Language Measures

Participants completed a range of standardized assessments upon entry into the study to confirm eligibility and to describe the participants’ important characteristics (Table 2). In addition to the ADOS-G used for screening (Lord et al., 2000), children completed a range of language measures; The Peabody Picture Vocabulary Test (PPVT-4; Dunn & Dunn, 2007); The Test for Early Language Development (TELD-3; Hresko, Reid, & Hammill, 1999); Leiter International Performance Scale (Leiter-3; Roid & Miller, 1997); and a Naturalistic Language Sample (NLS); a 20-minute standardized, adult-child interaction with a specific set of toys that is designed to evaluate a child’s spontaneous expressive language ability (Tager-Flusberg et al., 2009). The primary NLS variable for this study was total number of spontaneous communicative utterances (TSCU), which included comments, requests, and protests. Scripted and nonsocial utterances were excluded. The NLS procedures, fidelity and variables are outlined in detail in Kasari et al (2014). The ADOS-G was completed at Entry only. The TELD-3 and Leiter-3 were collected at Entry, Exit and Follow Up. The NLS and PPVT-4 were collected at all timepoints.

Table 2.

Participant Characteristics

		Entry N=57	Stage 1 N=55	Exit N=51	Follow-up N=44
Demographics
Gender	Number of Males	49	46	44	38
Chronological Age	In Years	6.79 (1.35)	7.10 (1.30)	7.23 (1.30)	7.33 (1.23)
Standardized Assessments
TELD-3	Receptive Age Equivalency (Years)	2.03 (0.62)		2.36 (0.95)	2.48 (0.91)
	Expressive Age Equivalency (Years)	1.72 (0.39)		2.02 (0.66)	2.02 (0.65)
Leiter-R	Brief IQ	67.74 (19.05)			66.23 (17.22)
PPVT-4	Age Equivalency	2.63 (0.68)	2.91 (0.92)	3.30 (1.78)	3.23 (0.99)
ADOS	Social Affect	14.31 (3.62)
	RRB	5.65 (1.77)
Language Sample
	Total Spontaneous Utterances	31.53 (28.84)	47.78 (34.65)	49.04 (35.68)	54.00 (41.35)
	Number of Spontaneous Word Roots	17.23 (15.71)	27.27 (25.98)	29.27 (26.57)	31.91 (27.44)
	Spontaneous MLU	1.30 (0.49)	1.48 (0.58)	1.47 (0.59)	1.50 (0.58)

Note. Values refer to Mean (SD)

Data Analysis

To model the short-term trajectories of Total RRBs, generalized linear mixed models (GLMM; Mcculloch & Neuhaus, 2014) were applied including the main effect of time. We included subject level random intercepts and controlled for chronological age, non-verbal mental age, language (TSCU), and site. Separate models were fit for each outcome (total and individual categories of RRBs) where time was modeled as a categorical variable (Entry, Phase One, Exit and Follow Up).

The same control variables and subject level random effect were included for the individual RRB category outcomes and each of these outcomes were modeled separately. Modifications to models fit for the individual RRB categories were needed to account for the highly positively skewed count outcomes. Negative binomial models with random effects were used for each of the individual RRB category outcomes, these models are well suited to handle zero inflated data that is overdispersed.

Despite all children participating in a social-communication intervention, we did not examine treatment effects on RRBs. Given the focus of the intervention on social communication, and that all children received the same behavioral intervention, we did not anticipate changes on RRBs. In studies (including our own) that have examined change in RRBs in younger samples, minimal change in RRBs has been observed through participation in social-communication interventions (Dawson et al., 2010; Green et al., 2010; Harrop, Gulsrud, Shih, Hovsepyan, & Kasari, 2016b). Further, in the only (to our knowledge) randomized controlled trial targeting RRBs as the primary outcome through parent trainings, researcher rated change in RRBs was minimal (Grahame et al., 2015).

Results

Profile of RRBs in Minimally-Verbal Children with ASD

At each timepoint, all children demonstrated at least one codable RRB (Table 3). At Entry, children displayed an average of 16.86 RRBs (SD = 9.14) during the 10-minute sample of the CCX, however this was highly variable with a range between 1 and 45 RRBs within the CCX (Figure 1). The most common RRB at all timepoints was Verbal with between 84 and 94% of children demonstrating at least one instance of a Verbal RRB at each timepoint (Figure 2; Table 3). At Entry, over half of verbal RRBs were categorized as atypical vocalizations (3.50), compared to a roughly even spilt of echolalia (1.50) and scripting (1.42). Repetitive Object Use was the second most common RRB at Entry, with over 70% of children demonstrating at least once instance at each timepoint (Table 3). Due to their low occurrence, Sensory Seeking and Sensory Aversion were combined together into a broad “Sensory” category. “Other” RRBs were not coded at any timepoint and thus dropped from the following analyses. Due to low occurance, SIB, insistence on sameness, and repetitive AAC use were grouped together into a broad “Other” category in the following analyses and in Figure 2.

Table 3.

RRBs by Type and Timepoint

	Entry N=57			Stage 1 N=55			Exit N=51			Follow-up N=44
	Mean (SD)	% with behavior	Range	Mean (SD)	% with behavior	Range	Mean (SD)	% with behavior	Range	Mean (SD)	% with behavior	Range
Total	16.86 (9.14)	100%	1-45	17.89 (9.52)	100%	3-52	17.49 (8.49)	100%	3-37	11.57 (7.49)	100%	2-31
Motor	3.11 (4.66)	73.3%	0-18	3.76 (6.05)	69.1%	0-31	3.41 (4.03)	70.6%	0-18	1.95 (3.06)	52.3%	0-13
Visual	2.29 (4.09)	48.2%	0-23	2.96 (3.93)	70.9%	0-16	2.18 (3.40)	64.7%	0-15	1.43 (3.17)	40.9%	0-15
Repetitive Object Use	2.64 (3.15)	76.8%	0-17	3.18 (4.69)	72.7%	0-29	3.02 (3.54)	78.4%	0-19	2.20 (3.63)	70.5%	0-22
Verbal	6.48 (5.01)	94.6%	0-25	6.05 (4.90)	94.5%	0-24	6.63 (6.37)	94.1%	0-36	4.57 (4.24)	84.1%	0-15
Sensory Seeking	1.84 (3.99)	50%	0-25	1.35 (2.03)	43.6%	0-9	1.73 (2.79)	52.9%	0-13	1.14 (2.41)	40.9%	0-14
Sensory Aversion	0.21 (0.71)	12.5%	0-4	0.16 (0.63)	9.1%	0-4	0.12 (0.38)	9.8%	0-2	0.02 (0.15)	2.2%	0-1
Repetitive AAC Use	0.29 (0.73)	16.1%	0-3	0.13 (0.43)	9.1%	0-2	0.33 (1.26)	11.8%	0-8	0.11 (0.49)	6.8%	0-3
Self Injurious Behavior	0.02 (0.13)	1.8%	0-1	0.02 (0.13)	1.9%	0-1	0.02 (0.14)	2%	0-1	0.02 (0.15)	2.3%	0-1
Insistence on Sameness	0.17 (0.47)	14.3%	0-2	0.27 (0.91)	12.8%	0-5	0.06 (0.24)	5.9%	0-1	0.11 (0.75)	2.3%	0-5

Mean(SD)

Figure 1.

Change in total RRBs over time

Figure 2.

Change in RRBs over time by type

Trajectories of RRBs over time

Total RRB’s remained stable from Entry (M = 16.86, SD= 9.1468) to Phase One (M = 17.89, SD=9.52) and from Entry to Exit (M = 17.49, SD=8.49); during which the children were all in the active treatment phase of the study (p value’s > .05; Table 4). However, there was a significant decrease in total RRB’s from Entry to Follow Up (B= −4.91, SE_B = 1.47, p = 0.001; Table 4), with the mean number of total RRBs reducing from 16.86 (SD = 9.14) to 11.57 (SD = 7.49).

Table 4.

Parameter Estimates and Confidence Intervals for Longitudinal Models

	Total			Verbal			Visual			Object			Motor			Sensory			Other
	Est	CI	p	Est	CI	p	Est	CI	p	Est	CI	p	Est	CI	p	Est	CI	p	Est	CI	p
Intercept	16.56	12.04 21.07	<0.001	1.38	1.00 1.76	<0.001	−0.04	−0.89 0.82	0.931	0.62	0.01 1.22	0.048	1.25	0.49 2.01	0.001	0.84	0.08 1.60	0.030	−1.80	−3.07 −0.52	0.006
Site	0.14	−1.97 2.24	0.900	0.22	0.05 0.39	0.010	0.22	−0.18 0.63	0.282	0.07	−0.20 0.35	0.607	−0.39	−0.76 −0.02	0.039	−0.31	−0.66 0.05	0.088	0.42	−0.10 0.94	0.110
Phase 1	1.01	−1.73 3.75	0.471	−0.08	−0.38 0.21	0.569	0.24	−0.25 0.74	0.339	0.11	−0.28 0.49	0.590	0.18	−0.20 0.56	0.351	−0.10	−0.68 0.47	0.729	−0.01	−0.96 0.94	0.982
Exit	0.78	−2.03 3.59	0.587	0.00	−0.29 0.30	0.980	−0.15	−0.68 0.37	0.565	0.14	−0.27 0.55	0.496	0.19	−0.20 0.58	0.349	−0.06	−0.64 0.51	0.833	−0.08	−1.07 0.90	0.869
Follow-Up	−4.91	−7.84 −1.98	0.001	−0.35	−0.67 −0.03	0.031	−0.72	−1.30 −0.13	0.016	−0.30	−0.74 0.13	0.172	−0.37	−0.80 0.06	0.094	−0.45	−1.09 0.18	0.160	−0.64	−1.76 0.48	0.260
CAge	1.65	−0.25 3.55	0.088	0.28	0.13 0.43	<0.001	−0.04	−0.39 0.30	0.805	0.34	0.08 0.60	0.010	−0.09	−0.43 0.24	0.581	−0.07	−0.39 0.25	0.669	0.11	−0.32 0.54	0.616
Leiter Brief IQ	−0.72	−2.57 1.13	0.446	0.18	0.04 0.33	0.014	−0.13	−0.47 0.20	0.436	0.02	−0.23 0.27	0.864	−0.21	−0.54 0.11	0.200	−0.36	−0.67 −0.04	0.026	0.19	−0.31 0.69	0.454
NLS TSCU	−0.02	−1.38 1.35	0.979	0.01	−0.12 0.14	0.891	0.19	−0.06 0.44	0.133	0.05	−0.14 0.24	0.615	−0.17	−0.39 0.05	0.124	−0.17	−0.44 0.11	0.239	−0.14	−0.55 0.27	0.498

Note. Reference group for time variable is Entry.

Note. Sensory refers to Sensory Seeking and Sensory Aversion which were combined for this analysis

Other refers to Repetitive AAC Use, Self-Injurious Behaviors, and Insistence on Sameness which were also combined due to their low frequency.

There were no significant changes from entry to phase one and entry to exit for any of the individual categories (all p’s >.05; Tables 3 and 4). However, from entry to follow up there was a significant reduction in Verbal (B= −0.35, SE_B = 0.16, p = 0.03) and Visual RRB’s (B= −0.72, SE_B = 0.30, p = 0.02). Verbal RRBs reduced from an average of 6.48 (SD=5.01) at entry to 4.57 (SD=4.24) at follow up, whereas Visual RRBs reduced from an average of 2.29 (SD= 4.09) to 1.43 (SD= 3.17). All other categories remained stable between entry and exit (all p’s > .05; Tables 3 and 4).

Discussion

The primary goal of this study was to characterize the frequency and distribution of RRBs in children with ASD who remain minimally verbal at school entry and to assess the stability of these behaviors within a 9-month developmental window. Overall, our sample demonstrated an average ~17 RRBs upon entry into the study. This frequency remained constant throughout the active treatment stage of the study, only decreasing to below baseline levels at follow up. The most common RRB was verbal, followed by motor RRBs and RRBs with objects.

Despite an average of nearly 17 RRBs at Entry, total number of RRBs ranged greatly from 1 to 45 within a short naturalistic observation. This frequency is higher than that using a similar coding scheme and caregiver child context with younger children (a range of 0-30 was reported by Harrop et al., 2015), suggesting that, school-aged minimally-verbal children demonstrate more RRBs than younger children. Higher frequencies of RRBs may have implications for treatment progress for children, possibly impacting the therapist or parent’s ability to engage with intervention protocols. Conversely, RRBs, particularly those with objects, could also be leveraged as mechanisms to engage children within intervention. For example, in a previous study we reported parent’s responses to their children’s RRBs with parents who engaged in a social-communication focused intervention more likely to redirect or build upon their child’s object RRBs while ignoring motor RRBs (Harrop et al., 2016b). However, minimal changes were observed on child RRBs. Therefore, a greater number of RRBs, particularly with objects, could provide opportunities for engagement even if no reduction in RRB frequency is observed. Further research is required to understand why this group of children display more RRBs compared to younger, verbal or pre-verbal children and the root of these behaviors. It is possible that these RRBs were expressed due to anxiety, frustration, or excitement and serve a communicative function.

This hypothesis is potentially supported by the fact the most frequent RRBs in the current sample were verbal behaviors (atypical vocalizations, scripting and echolalia). This finding is in contrast to toddler and preschool samples, where the most frequent RRB involved object use (Harrop et al., 2014, 2015; Barber et al., 2012; Watt et al., 2008). In our sample, between 33 and 39% of RRBs were coded as Verbal, compared to between 6 and 18% in studies with younger children. In contrast, between 16 and 19% of RRBs were coded as Object, compared to 37 and 69% in previous studies with younger children. The most frequent type of verbal RRB was atypical vocalizations. These accounted for over half of the verbal RRBs displayed. Further research is required to understand the function of these vocalizations and whether they could be leveraged as potential segues into more functional communicative attempts. Further, interventions need to assess whether verbal RRBs actually represent communicative attempts that can be expanded upon within intervention, and not behaviors that need to be extinguished.

There were a number of behaviors that were not frequently observed within the CCX, namely sensory aversion, insistence on sameness, repetitive AAC use and SIB. These behaviors (with the exception of repetitive AAC use) are reported as common in ASD (Baranek, Boyd, Poe, David, & Watson, 2007; Boyd et al., 2010; Chezan, Gable, McWhorter, & White, 2017; Soke et al., 2018). Sensory aversions are reported in between 45 to 95% of individuals with ASD (Baranek et al., 2007; Ben-Sasson et al., 2007; Tomchek & Dunn, 2007). Estimates of SIB in ASD vary from 27% to 50% (Baghdadli, Pascal, Grisi, & Aussilloux, 2003; Bodfish, Symons, Parker, & Lewis, 2000; Soke et al., 2017; Soke et al., 2016; Soke et al., 2018). Insistence on sameness is also frequently reported in ASD, though typically through parent report rather than direct observation (Cuccaro et al., 2003; Lam et al., 2008; Richler, Bishop, Kleinke, & Lord, 2007; Schertz, Odom, Baggett, & Sideris, 2016; Stone, Hoffman, Lewis, & Ousley, 1994). SIB in ASD is associated with limited communication and lower IQ scores (Soke et al., 2017). However, these behaviors may not be readily observed during a CCX and may only occur when demands are placed on a child or when situations or stimuli “trigger” these behaviors (such as loud noises). The low occurrence of these behaviors during the CCX was surprising, especially given their associations with communication abilities and IQ, and future work should combine caregiver report with observational measures to understand the frequency and impact of these classes of RRBs in minimally-verbal children with ASD.

As with toddler and preschool samples, we observed an effect of time, with a general reduction of RRBs post treatment. It is possible that change in RRBs was not apparent during the active treatment phase of the study (entry through exit) due to the ongoing acquisition of social communication and it is only following the conclusion of treatment and increased social-communication, a reduction was observed. In children with ASD in general, very few controlled studies have targeted RRBs as a primary outcome (Grahame et al., 2015). While a handful of RCTs have included measures of RRBs as secondary outcomes (Dawson et al., 2010; Green et al., 2010), change in RRBs following intervention is less well understood than change in the domain of social communication. Further, as a field we do not understand what causes RRBs, and this is particularly difficult to ascertain when children are minimally verbal. Therefore, caution must be used when interpreting reductions in RRBs as these behaviors may serve important functions for individuals with ASD (Kapp et al., 2019).

Limitations

A few limitations are worthy of discussion. While our coding scheme measures the frequency of RRBs, it does not measure the duration, intensity, or potential impact of RRBs. Therefore, children may display many behaviors during a short time period, but these behaviors do not interfere with the ongoing interaction or last very long.

This study provides one of the most detailed descriptions of the presentation of RRBs in minimally-verbal autistic children; however, further work is required to determine how suitable this coding scheme is for older, minimally-verbal youth. All prior work has been with children under six and many of whom have phrase speech (Harrop et al., 2014, 2016b, 2015; ), therefore caution must be exercised in interpreting our findings. Future work should combine observational and parent report measures or use methods such as motion sensors or more detailed language coding to fully characterize RRBs in this population.

Due to the nature of the overall treatment study (Kasari et al., 2014), not all children had access to the AAC during the CCX sessions. Half of the children had access to the AAC at all timepoints, whereas the children in the social communication only arm did not have access to the AAC at Entry or Stage One. Six children from the social communication only group were re-randomized following Stage One assessments to receive the AAC device and thus had access to the device during Exit and Follow Up CCXs. Therefore, children only had access to the AAC if it was intentionally incorporated into their intervention. This difference between the groups was intentional, so that children and parents who were not familiar with the AAC were not distracted during the CCX. The intentional use of the SGD for communication within the context of the behavioral intervention potentially allowed for fewer opportunities for repetitive use.

Conclusions

Minimally-verbal autistic children are poorly understood, particularly with regards to behaviors outside of social-communication (Tager-Flusberg & Kasari, 2013). Results of this study suggest elevated rates of RRBs relative to younger toddler and preschool samples and a different concentration of RRBs that are more verbal- than object-focused (Barber et al., 2012; Harrop et al., 2014; Watts et al., 2008). Despite maintained rates of RRBs through treatment, children significantly reduced in their expression of RRBs three months post intervention suggesting developmental change in this core set of behaviors. Our results add to a small, yet growing, literature characterizing children who remain minimally verbal at school entry.