The interaction of fine motor, gesture, and structural language skills: The case of autism spectrum disorder

Abstract

Motor skill differences have been consistently reported in individuals with ASD. Associations between motor skill and social communication skills have been reported in both typical development (TD) and autism spectrum disorder (ASD). The current study extends these findings to characterize performance on a fine motor imitation task, probing skills as a predictor of social and communicative functioning, and co-speech gesture use. These research questions were addressed by a secondary analysis of data collected during a previous study characterizing a cohort of individuals who were diagnosed with ASD in early childhood but lost the autism diagnosis (LAD) by the time of adolescence. Fine motor imitation skills were compared between 14 individuals with LAD, 15 individuals with autism spectrum disorder (ASD), and 12 typically developing (TD) individuals. LAD and TD groups had more advanced fine motor imitation skills than the ASD group, and abilities were significantly associated with ASD symptoms and amount of gesture use (though there was a counterintuitive interaction between group and fine motor skill in the LAD and TD groups only, in which lower motor skills predicted more ASD symptoms; this relationship was of a small effect size and is likely driven by the compressed range of fine motor skills in these two groups). Findings suggest that fine motor skills normalize along with social communication skills and restricted and repetitive behaviors and interests in individuals who lose the ASD diagnosis, and that individuals with better fine motor abilities produce more co-speech gesture.

Introduction

Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by deficits in social and communication skills, as well as restricted and repetitive behaviors and interests. Children with ASD consistently demonstrate difficulties with both basic and complex motor skills (Jansiewicz et al., 2006; Mostofsky et al., 2006; Bhat et al., 2011; Macneil & Mostofsky, 2012; Whyatt & Craig, 2013; McAuliffe et al., 2017). Given that early motor skill development predicts important variability in social communicative functioning in typical development (Iverson et al., 2019; Walle & Campos, 2014) and ASD (Dziuk et al, 2007; Choi et al., 2018), the current study compares motor skills in youth with ASD, those who have lost the ASD diagnosis, and in typical development; additionally, we examine the relationships among motor skills, multiple aspects of social communication, and core ASD symptoms.

Early motor skill development is associated with language acquisition in typical development. Ejiri and Masataka (2001) found that in the period preceding canonical babbling, infant vocalizations and rhythmic arm movements co-occur with high frequency. Iverson and colleagues (2019) reported that caregivers are more likely to label an object when a child directly manipulates it, and that this facilitates word learning. Greater mobility and dexterity allow infants to interact with people and objects in the environment, providing opportunities for caregivers to label objects and events. For example, Karasik et al. (2014) found that infants who locomote can approach and show an object to a caregiver (e.g., make “moving bids”); in turn, caregivers are more likely to respond verbally and with action directives to such bids. Furthermore, verbal input is linked to object manipulation. Yu and Smith (2012) found that infants are more likely to learn the names of a novel object if it is labeled while it dominates the infant’s field of vision (most often, when the infant is holding the object). Given that motor skills facilitate word learning, the presence of motor skill differences could lead to reduced opportunities for labelling, and negatively impact word learning.

There is consistent evidence that differences in motor skills are associated with communication and social outcomes in ASD. Infants who later go on to develop ASD show slower growth in fine motor skills between six and 24 months compared to typically developing infants; furthermore, fine motor skills at six and 24 months predict language skills at age three years (Choi et al., 2018). Bedford and colleagues (2016) found that gross motor skills in children with ASD at age two predicted growth trajectories of expressive and receptive language skills between ages two and nine years. Another longitudinal study found that fine motor delays at age three were the strongest predictor of persistent language impairments at age 19 in two large (n=86 and n=181) cohorts of language-delayed children with ASD (Bal et al., 2019). Similarly, in two-year old children with ASD, motor skills were the best predictor of losing the diagnosis at age four (Sutera et al., 2007). Several recent meta-analyses have confirmed a highly likely association between motor and social skills in individuals with ASD (Ohara et al., 2019), as well as a strong relationship between motor and communication skill development in young children with ASD (West, 2019). These findings are consistent with the hypothesis that motor deficits are strongly related to language delays and social difficulties in ASD.

Imitation skills are also believed to play a critical role in language acquisition and social development. Infants imitate vocalizations, facial expressions, and gestures. By nine months of age, typically developing infants can imitate simple, goal-directed actions; by two years, they are able to imitate multi-step actions (Bauer & Hertsgaard, 1993; Herbert & Hayne, 2000) and tool use (Bushnell & Boudreau, 1998; Chen et al., 2000). Yang and colleagues (2010) found that infants can generalize actions learned through imitation from one set of cause-and-effect toys to another. These findings highlight the importance of imitation as a learning and social tool in infancy and early childhood, suggesting that a reduced capacity for imitation could increase the risk of delays in the acquisition of social, language, and play skills.

Children with ASD struggle with a variety of imitation tasks, including direct elicitation, interactive play, and observational learning (McDuffie et al., 2007). Imitation deficits are associated with poorer social and communicative skills in ASD (Stone, Ousley & Littleford, 1997; Stone & Yoder, 2001), and reduced imitation of adult actions predicts pragmatic language delays in ASD (Miniscalco et al., 2013). Based on findings that imitation skills and initiation of joint attention are directly related, Ingersoll (2008) proposed that reduced imitation of actions, facial expressions, and gestures leads to reduced interactions with caregivers and peers, and therefore reduced opportunities to understand social cause and effect events and to engage in social play. Because imitation provides a platform for critical forms of learning and social interaction, deficits or delays in the development of motor imitation may cascade into social and communicative delays in ASD.

One aspect of communication that may be particularly sensitive to underlying motor and imitation difficulties is co-speech gesture. Co-speech gestures are hand or body movements accompanying fluent speech, that arise spontaneously and serve a communicative purpose (McNeill et al., 1992). Categories of co-speech gesture include iconic gestures, which are physical depictions of concrete objects or events (e.g., rounded hands showing the size and shape of a coconut;, Lord et al., 2012), categorized as descriptive gestures on the Autism Diagnostic Observation Schedule (ADOS; Lord et al., 2012, the gold-standard diagnostic assessment tool for ASD; metaphoric gestures, which refer to abstract concepts or relationships (e.g., both hands oriented with the palms facing upwards, moving up and down in alternation, referring to the concept of justice as a scale); deictic gestures, used to orient others to a referent (e.g., pointing), categorized as instrumental gestures on the ADOS; and beat gestures, which are rhythmic hand movements that typically convey emotion or emphasis (coded as emotional/emphatic gestures on the ADOS). Co-speech gestures enhance listeners’ comprehension of spoken language. Gestures often convey “complementary” information that is not present in speech (Hostetter, 2011), and are important in language acquisition (Cook & Goldin-Meadow, 2006; Iverson & Goldin-Meadow, 2005; Mumford & Kita, 2014). It is well documented that infants and toddlers later diagnosed with ASD show delays in the development of preverbal declarative gestures such as pointing (Mundy et al., 1990; Stone, Ousley, Yoder, et al., 1997; Filipek et al., 1999).

Recent studies of co-speech gesture in verbally fluent adults with ASD have demonstrated mixed results with regard to quantity of gesture, with some studies finding reduced gesture production in children, adolescents and adults with ASD (de Marchena & Eigsti, 2014; Silverman et al., 2017; So et al., 2014) and others finding no difference (de Marchena & Eigsti, 2010; de Marchena et al., 2018). However, qualitative differences in the gestures in ASD are reliably reported, including decreased temporal synchrony with accompanying speech (de Marchena & Eigsti, 2010; Morett et al., 2016), slowing in gesture processing (Silverman et al., 2010), reduced clarity of gesture meaning (Silverman et al., 2017), and a higher proportion of unilateral to bilateral (i.e. one-handed to two-handed) gestures (de Marchena et al., 2018). Deaf, signing children with ASD make frequent palm reversal errors; that is, they make manual signs that look correct from their own perspective, but are reversed from the perspective of the person they are communicating with (Shield & Meier, 2012). This tendency may be indicative of deficits in the ability of individuals with ASD to map others’ motor movements onto their own bodies. Deficits in motor skills and/or motor imitation abilities likely contribute to atypical co-speech gesture in verbally fluent individuals with ASD (McAuliffe et al., 2017); however, the relationship between spontaneous co-speech gesture and motor imitation skills has not been studied in this group.

Loss of Autism Diagnosis (LAD).

Although ASD is typically thought of as a life-long condition, the literature suggests that between 3 and 25% of children diagnosed with an ASD in early childhood no longer meet criteria for the disorder by adolescence (Helt et al., 2008). Our team has followed a group of children with a documented ASD diagnosis before age five, who had no symptoms of ASD in adolescence (Fein et al., 2013); this outcome, previously called “optimal outcome” is described here as a loss of autism diagnosis (LAD). Individuals in the LAD group had typically developing friends and were mainstreamed in regular education classroom without services related to social skills; IQ, adaptive, and language abilities within the average range; and no current symptoms of ASD as measured by the ADOS. In a series of studies, the LAD group was compared with age- and IQ-matched children with a current diagnosis of ASD and with typical development (TD). Overall, results indicate that LAD individuals had typical or even enhanced scores on structured and observational assessments of social skills (Orinstein et al., 2015; Suh et al., 2016), communication (Tyson et al., 2014; Suh et al., 2014; Canfield et al., 2016), and restricted and repetitive behaviors (Troyb et al., 2014).

Current study.

The current study examines relationships among fine motor imitation skills, language skills, gestures, and core ASD symptomatology, in individuals with ASD, LAD, and typical development. In addition to informing about ASD, this study serves to illuminate our understanding of how these factors interact in the course of typical language acquisition and language processing. Currently, there are no published accounts of fine motor imitation abilities in individuals with LAD; it is unknown whether this skill normalizes along with social affect and restricted and repetitive behaviors in this group. Furthermore, this is the first study to directly examine the relationship between fine motor imitation skill and co-speech gesture in individuals with LAD or ASD. Data were collected as part of the 2013 study described above (Fein et al. 2013). Based on many prior studies (Jansiewicz et al., 2006; Dziuk et al., 2007; Bhat et al., 2011; Macneil & Mostofsky, 2012; Whyatt & Craig, 2013; McAuliffe et al., 2017), and given that motor skills are a strong predictor of outcomes in children with typical development (Iverson, 2010; Peyre et al., 2019 and ASD (Sutera et al., 2007; Helt et al., 2008; Bedford et al., 2016), we hypothesized that fine motor imitation skills would be intact in the LAD group, and comparable to the TD group, and that the ASD group would show fine motor imitation differences. We also hypothesized that fine motor imitation skills would independently relate to language, gesture production as rated on the ADOS, and autism symptomatology. Finally, exploratory analyses examined relationships among vocabulary, grammar/syntax, fine motor imitation, and autism symptomatology.

Methods

Procedure.

This study was carried out in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Declaration of Helsinki as revised in 2000. For participants under 18, parent consent and child assent were obtained prior to testing. For participants 18 and over, informed consent was obtained.

Participants.

A subset of participants, ages 8.5–21.2, from the larger Fein et al. (2013) sample completed the Imitating Hand Positions subtest of the NEPSY-II; the assessment was not available for all participants because of time limitations. Only participants who completed the NEPSY-II task were included in the present analyses. The present study includes 15 individuals with ASD (1 female), 14 individuals with LAD (no females), and 12 typically developing individuals (no females). All participants had verbal, nonverbal, and full-scale IQ scores within 1.5 SD of the average of 100 (e.g., 78 or higher). Participant details are shown in Table 1. Inclusion in the LAD group also required inclusion in a mainstream classroom without ASD-related accommodations, no or minimal symptoms of ASD as measured by the ADOS, and having typically developing friends. Some individuals in the LAD group were judged to show mild social impairment that was secondary to anxiety, depression, inhibition, or inattention/impulsivity, which were not judged to have an autistic quality (Fein et al., 2013); recruitment details and other specifics are described in the 2013 paper.

Table 1.

Descriptive statistics

	Autism spectrum disorder (ASD), n = 15		Loss of ASD diagnosis (LAD), n = 14		Typically developing (TD), n = 12		F/ H	p	t / Mann-Whitney U
	M (SD)	Range	M (SD)	Range	M (SD)	Range
Age (yrs)	13.5 (3.1)	9.0 – 20.0	13.6 (3.8)	8.5 – 21.2	14.1 (3.0)	10.7 – 21.2	0.11	0.9
CELF-FS*	10.6 (2.5)	5.0 – 14.0	12.2 (1.6)	9.0 – 14.0	12.9 (1.0)	11.0 – 14.0	7.11	.03	TD > ASD
PPVT	108 (15)	81 – 127	112 (12)	89 – 130	118 (9)	102 – 134	2.32	0.11
NVIQ	110 (11)	92 – 127	112 (11)	92 – 129	110 (8)	90 – 124	0.08	0.92
ADOS CSS**	6.5 (2.7)	1.0 – 10.0	1.6 (1.0)	1.0 – 4.0	1.1 (0.3)	1.0 – 2.0	41.33	<.001	ASD > LAD, TD

Note: CELF-FS = Clinical Evaluation of Language Fundamentals, Formulated Sentences; PPVT = Peabody Picture Vocabulary Test; NVIQ = Wechsler Abbreviated Scale of Intelligence Nonverbal IQ; ADOS = Autism Diagnostic Observation Schedule Calibrated Severity Score. The p value tested differences across three groups; t or Mann-Whitney U tested pairwise group differences.

Materials and Design.

Participants completed an array of measures as part of the larger LAD study, a cross sectional, descriptive study. Data from the following tasks were included in the present study, including previously unanalyzed motor tasks. As a measure of grammatical abilities, participants completed the Formulated Sentences subtest of the Clinical Evaluation of Language Fundamentals-IV (CELF-IV; Semel et al., 2003). This task requires examinees to construct grammatically and semantically correct sentences based on presented words and images. Total possible scores range from a minimum of zero points to a maximum score of 56. Criterion validity for the CELF-IV has been established with the CELF-III (Semel et al., 2003); interrater reliability of subtests ranges from .88 to .99 (Semel et al., 2003).

The Peabody Picture Vocabulary Test-III (PPVT-III; Dunn & Dunn, 1997) provided a measure of receptive (comprehension) vocabulary. The PPVT-III consists of 228 items, each of which requires the participant to identify a target word from among four images. The PPVT-III has good concurrent validity with the verbal IQ composite of the Weschler Intelligence Scale for Children-III (WISC-III; Hodapp & Gerken, 1999), and has good inter-rater and test-retest reliability (Dunn & Dunn, 1997).

The Wechsler Abbreviated Scale of Intelligence Second Edition (WASI-II; Wechsler, 2011) consists of four subtests that generate both verbal and nonverbal intelligence indices, in addition to a full-scale intelligence quotient. Inclusion for the current study required a minimum score of 78 or higher; in fact, all participants had scores at least in the low-average range (80–89), ranging up to the gifted range (130 or above), across verbal, nonverbal, and full domains. The WASI-II demonstrates concurrent validity with various measures including the Wechsler Individual Achievement Test (WIAT) and the Wide Range Intelligence Test (WRIT) (Wechsler, 2011. According to the WASI-II manual (2011), reliability coefficients for subtests range from .81 to.97 for children, and from .84 to .98 for adults.

Participants completed the Autism Diagnostic Observation Schedule Second Edition (ADOS-2; Lord et al., 2012) to confirm diagnosis and to provide measures of spontaneous language and gesture production; depending on the child’s developmental level, Module 3 or Module 4 was administered. The ADOS consists of a series of tasks and semi-structured activities that are designed to elicit reciprocal social behavior and imaginative activities. The ADOS is scored by the examiner based on observed behavior. There are 29 scorable items in Module 3 and 32 items in Module 4. These items fall into the broad categories of communication, reciprocal social interaction, imagination, stereotyped behaviors and restricted interests, and other abnormal behaviors; each item receives a score ranging from 0 (most typical) to 3 (most atypical). If the total sum is eight or higher, criteria are met for ASD. ADOS classifications have good concurrent validity with clinician diagnosis (Lord et al., 2012). Interrater reliability for both modules is 88 % (Lord et al., 2012). Cronbach’s alphas are highest for the Social domain (.86-.91), slightly lower for the Communication domain (.74-.84), and lowest for Stereotyped Behaviors and Restricted Interests (.47-.56; Lord et al., 2012). In order to minimize the effects of individual characteristics such as age and IQ when comparing scores among youth who completed different modules of the ADOS, we calculated a standardized calibrated severity score for each individual (CSS; see Gotham et al., 2006 for details).

Module 3 has only one measure of co-speech gesture, item A10 (descriptive, conventional, instrumental, or informational gesture). For children in the current study who completed Module 4, items A9 (descriptive, conventional, instrumental, or informational gesture) and A10 (emphatic or emotional gestures) were averaged for a measure of co-speech gesture. Thus, gesture assessments comprised a single score with a range of 0 (most typical) to 3 (most atypical).

Finally, the Imitating Hand Positions (IH) subtest of the Developmental Neuropsychological Assessment II (NEPSY-II; Korkman et al., 2007) provided a measure of fine motor programming, imitation, and visuospatial abilities. Participants imitated increasingly complex hand and finger positions that were demonstrated by the examiner, with a maximum possible score of 12 points for each hand. This subtest is normed for ages 6 – 13 years. As many participants were 14 years or older, the present analyses used raw rather than standard scores. Despite having poor test-retest reliability for children ages seven and older (Brooks et al., 2010), likely due to floor effects, this subtest is a face-valid assessment of fine motor imitation skills with excellent construct validity (Schmitt & Wodrich, 2004). Analyses focused on the dominant because this is where individual variability is likely to be the greatest.

Data analyses.

Two participants with ASD and one LAD participant were missing CELF-FS scores due to examiner error or time constraints during testing; there were no other missing data points. These data points were omitted from relevant analyses. A one-way Kruskal-Wallis analysis of variance was used to test for group differences in fine motor imitation skills, because NEPSY-II scores had a skewed distribution; Mann-Whitney U tests were used for follow-up analyses. Hedges’ g (following Cohen’s criteria) was calculated as a measure of effect size (Huedo-Medina & Johnson, 2013; Hedges, 1981; Cohen, 1988).

Exploratory correlational analyses tested relationships among fine motor imitation skills, vocabulary skills, grammatical abilities, and gesture scores. Because PPVT (vocabulary) scores were normally distributed, we ran a multiple regression analysis with fine motor imitation skills and nonverbal IQ as predictors, with PPVT as the dependent variable. ADOS CSS scores were not normally distributed (scores for the ASD group were much higher than those for the LAD and TD groups); similarly, structural language abilities (CELF-FS scores) were skewed positively as individuals in the ASD group scored lower on average than individuals in the LAD and TD groups. Therefore, a generalized linear model (GLM) with gamma distribution and log link function was used to test the relationship among fine motor imitation, ADOS CSS and CELF-FS scores as more robust models when normality and homoscedasticity assumptions due to small sample sizes or other factors cannot be met (McCullagh & Nelder, 1989). To test the hypothesis that fine motor imitation skills were specifically associated with ASD symptomatology over and above the role of general cognitive ability, we ran a GLM with fine motor imitation skills and NVIQ as predictors and ADOS CSS as the dependent variable and compared the fit of this model to a model including group (ASD vs LAD and TD together) as a factor. The TD and LAD groups were combined because both groups currently do not display symptoms of ASD. We also assessed for effects of an interaction between group and fine motor imitation skills in predicting ADOS CSS using another GLM.

Finally, we tested the fit of a generalized linear model with ADOS CSS as the dependent variable and NEPSY-II, NVIQ, CELF-FS, and PPVT as predictors in order to assess the relative contributions of vocabulary, grammatical ability, and fine motor imitation skills to ASD symptomatology. Generalized linear models are robust even with small sample sizes, and the number of predictors did not exceed the sample size. Therefore, this model was appropriate. To test the hypothesis that fine motor imitation skills were specifically associated with language abilities, we ran another generalized linear model with CELF-FS scores as a dependent variable and fine motor imitation skills and NVIQ as predictors. The models were run with both log and identity link functions, and fit statistics were compared to determine which link function produced a model with better fit. In order to assess the role of fine motor imitation skills in predicting gesture scores on the ADOS, a multinomial cumulative probit link model was used because item-level gesture scores are on an ordinal scale.

We used MultNonParam, a kwpower function in R, to calculate power for Kruskall-Wallis tests. Our analyses had 90% power to detect differences in IH performance among groups, and more than 95% power to detect differences in gesture scores among groups. Power for generalized linear models was calculated using PROC POWER in SAS. Our analyses had 75% power to detect the effect of CELF-FS as a predictor of ADOS scores (small effect size), and 85% power to detect the effect of NEPSY scores on ADOS scores, also a small effect. When analyses were run with grammatical abilities as the outcome variable, our analyses had 40% power to detect the small effect size of NEPSY-II as a predictor, and 70% power to detect the effect size of group, which was also small. Finally, we had 80% power to detect the effect of NEPSY-II on gesture as an outcome variable.

Results

A Kruskall-Wallis analysis of variance revealed a significant group difference in NEPSY imitation scores for the dominant hand, H(2) = 8.35, p = .015. Mann-Whitney U tests revealed that scores in the ASD group were significantly lower than scores of the LAD group, U(N_ASD = 15, N_LAD = 14) = 49.50, p = 0.01, and of the TD group, U(N_ASD = 15, N_TD = 12) = 44.50, p = 0.02; scores are shown in Table 2. The LAD/TD group scores did not differ, U(N_LAD = 14, N_TD = 12) = 65.50, p = 0.31. A generalized linear model (Table 3) demonstrated that dominant hand scores remained significantly different between ASD youth and LAD/TD youth even after controlling for age and nonverbal IQ, p < .001. Group differences were not significant for the nondominant hand, H(2) = 3.58, p = .17. Levene’s test showed that variances among the three groups were equal for both dominant hand scores, F(2,38) = 3.00, p = .06 and nondominant hand scores, F(2,37) = 1.58, p = .22. One youth in the LAD group and two in the ASD group scored below the 2^nd percentile for population norms for age 13 years.

Table 2.

Group differences in motor imitation and gesture scores

	ASD, n = 15	LAD, n = 14	TD, n = 12	Kruskall-Wallis H	Mann-Whitney U	Hedges G
NEPSY Dom hand	9.1 (2.3); 3 – 12	10.9 (1.2); 8– 12	10.8 (0.8); 10– 12	8.35, p = .02	TD, LAD > ASD	0.9 for ASD vs LAD; 0.9 for TD vs ASD
NEPSY Nondom hand	9.5 (2.8); 2 – 12	10.6 (2.4); 3– 12	10.6 (0.9); 9–12	3.58, p = .17	-	-
ADOS Gestures	0.7 (0.7); 0 – 2	.04 (.10); 0 – 0.5	.08 (0.3); 0– 1	14.11, p < .001	ASD > LAD, TD	1.3 for ASD vs LAD; 1.1 for ASD vs TD

Note: ASD = Autism spectrum disorder; LAD = Loss of ASD diagnosis; TD = typical development. Scores are presented as M (SD), Range. NEPSY = Neuropsychological Assessment, Imitating Hand Positions subtest; ADOS Gesture = Composite of descriptive and emphatic gesture item-level scores on the ADOS; higher scores indicate more impaired performance; maximum score = 3

Table 3.

Gamma log model group differences NEPSY scores

Fit statistics
Pearson χ2		AIC	AIC-C	BIC	Omnibus Test
1.08		179	180.69	187.55	0.026
Model summary
Predictors	B	Exp(B)	S.E.	Wald	df	Sig.	95% CI	Effect size
NVIQ	0.004	1	0.0031	1.65	1	0.2	−.002 –.010	Small
Age	3.67E-05	1	2.71E-05	1.83	1	0.18	−1.641E.-5 – 8.980E-5	Small
Group**	−0.17	0.85	0.061	7.44	1	0.006	−.29 – −.05	Small

Note: AIC = Akaike’s Information Criterion; AIC-C = Finite Sample Corrected AIC; BIC = Bayesian Information Criterion; Larger χ² indicates better model fit; lower AIC, AIC-C and BIC indicates better model fit; Wald = Wald Chi-Square Test; NEPSY= Neuropsychological Assessment, Imitating Hand Positions; NVIQ = nonverbal IQ score; Group = ASD vs LAD/TD; Exp(B) = exponentiation of the B coefficient (odds ratio; measure of effect size);

^*p< .05

^**p< .01

^***p< .001.

Simple correlational analyses, shown in Table 4, revealed that fine motor imitation, vocabulary, and grammatical skills were each significantly correlated with ADOS CSS, such that better abilities were associated with lesser ASD severity. The effect size of this correlation was moderate for fine motor imitation skills and vocabulary skills, and large for grammatical skills. Fine motor imitation skills, vocabulary skills, and grammatical skills were also significantly correlated with gesture use during the ADOS, such that better abilities were associated with more use of co-speech gesture. Again, the effect size was moderate for fine motor and vocabulary skills, and large for grammatical skills. Vocabulary (PPVT) and structural language (CELF score) skills were also significantly correlated. Fine motor imitation skill was correlated with vocabulary and grammatical skills at a trend level.

Table 4.

Correlations among study variables across groups

	NVIQ	PPVT	CELF-FS	NEPSY	Gesture	ADOS
PPVT	0.26	1
CELF-FS	−0.05	.30‡	1
NEPSY	0.17	0.30‡	0.31‡	1
Gesture	−0.22	−0.35*	−0.56***	−0.42**	1
ADOS CSS	−0.20	−0.41**	−0.51***	−0.40**	0.71***	1

Note: NVIQ = nonverbal IQ score; PPVT = Peabody Picture Vocabulary Test; CELF-FS = Clinical Evaluation of Language Fundamentals, Formulated Sentences subtest; NEPSY = Neuropsychological Assessment, Imitating Hand Positions subtest; ADOS CSS = Autism Diagnostic Observation Schedule Calibrated Severity Score; Gesture = ADOS descriptive and emphatic gesture composite

^‡p< .10

^*p< .05

^**p< .01

^***p< .001

Multiple regression analysis.

Of course, correlations across groups might only reveal group differences rather than specific relationships. To address this possibility, we ran a multiple regression analysis with vocabulary (PPVT) scores as the dependent variable and dominant hand fine motor skills (NEPSY) and nonverbal IQ as independent variables. The result just missed significance, R²=.14, F(2,38)=2.97, p= .06; independent of nonverbal IQ, NEPSY scores were not a significant predictor of vocabulary, β =.27, p = 0.09.

Generalized linear models.

Fit statistics and model summaries for generalized linear models assessing the relationships among fine motor imitation skills, NVIQ, diagnostic group, language skills, and autism symptomatology are summarized in Tables 5–6. When NEPSY, NVIQ, CELF-FS and PPVT scores were included as predictors, only CELF-FS (grammatical skills) remained a significant predictor of ASD symptomatology, with a small effect size, Exp(B) = 0.83. We ran a separate model where group was included as a predictor, with LAD/TD participants forming a single group (given the homogeneity of their scores) that contrasted with the ASD group. Results suggested that NEPSY (Exp(B) = 1.41, small effect), Group (Exp(B) = 236.52, large effect), and a NEPSY*Group interaction (Exp(B) = 0.71, small effect) were significant predictors of ASD symptomatology. Follow-up analyses revealed an unexpected positive relationship between NEPSY-II scores and ADOS scores for the “typical” group (LAD and TD participants; Exp(B) = 1.41, small effect), such that better fine motor imitation skills were associated with more ASD symptomatology; see Table 5. The model for the ASD group was not significant.

Table 5.

Gamma log models with ADOS as outcome: Fit statistics

Pearson χ2	AIC	AIC-C	BIC	Omnibus
Model 1 (NVIQ, NEPSY CELF-FS*, PPVT)
22.20	158.6	161.31	168.43	.001
Model 2 (NEPSY**, Group***, NEPSY*Group interaction*)
8.05	124.3	126.0	132.90	<.001
Model 3 (NEPSY**, TD and LAD only)
5.55	45.67	46.76	49.44	0.011
Model 4 (NEPSY, ASD only)
2.50	80.07	82.25	82.20	0.63

Note: AIC = Akaike’s Information Criterion; AIC-C = Finite Sample Corrected AIC; BIC = Bayesian Information Criterion; Larger χ² indicates better model fit; lower AIC, AIC-C and BIC indicates better model fit

^*p< .05

^**p< .01

^***p< .001.

Table 6.

Gamma log models with ADOS as outcome: Model summaries

Predictors	B	Exp(B)	S.E.	Wald	df	p	95% CI	Effect size
Model 1
NEPSY	−0.09	0.91	0.07	1.67	1	0.20	−.24 to .05	Small
CELF-FS*	−0.16	0.85	0.06	5.92	1	0.01	−.28 to −. 03	Small
PPVT	−0.01	0.99	0.01	1.9	1	0.17	−.03 to .01	Small
Model 2
Group***	4.13	62.15	1.03	16.05	1	<.001	2.11 to 6.15	Large
NEPSY**	0.21	1.24	0.08	6.77	1	0.009	.05 to .37	Small
Group*NEPSY*	−0.24	0.71	0.10	6.00	1	0.015	−.44 to −.05	Small
Model 3 (TD and LAD only)
NEPSY**	0.21	1.24	.08	7.77	1	0.005	.06 to .36	Small
Model 4 (ASD only)
NEPSY	−0.03	0.97	0.08	0.03	1	0.629	−.15 to .09	Small

Note: Wald = Wald Chi-Square Test; Exp(B) = exponentiation of the B coefficient (odds ratio; measure of effect size); NEPSY= Neuropsychological Assessment, Imitating Hand Positions; CELF-FS = Clinical Evaluation of Language Fundamentals, Formulated Sentences; PPVT = Peabody Picture Vocabulary Test; ADOS = Autism Diagnostic Observation Schedule Calibrated Severity Score

^*p< .05

^**p< .01

^***p< .001.

Generalized linear models with CELF-FS scores as the dependent variable are summarized in Tables 7–8. A model with NEPSY and Group (again, TD/LAD vs. ASD) as predictors was significant and had better fit than the model with NEPSY and NVIQ as predictors. Effect sizes for all models were small.

Table 7.

Gamma log models with CELF-FS (grammatical abilities) as outcome: Fit statistics

Pearson χ2	AIC	AIC-C	BIC	Omnibus
Model 1 (NVIQ, NEPSY*)
1.05	172.88	174.09	179.43	0.15
Model 2 (NEPSY, Group*)
0.966	169.213	170.425	175.763	0.025

Note: AIC = Akaike’s Information Criterion; AIC-C = Finite Sample Corrected AIC; BIC = Bayesian Information Criterion; Larger χ² indicates better model fit; lower AIC, AIC-C and BIC indicates better model fit

^*p< .05

^**p< .01

^***p< .001.

Table 8.

Gamma log models with CELF-FS (grammatical abilities) as outcome: Model summaries

Predictors	B	Exp(B)	S.E.	Wald	df	p	95% CI	Effect size
Model 1
NVIQ	−0.002	1.00	0.003	0.37	1	0.52	−.008 to .004	Small
NEPSY*	0.034	1.04	0.017	4.02	1	0.045	.001 to .067	Small
Model 2
Group*	−0.142	0.87	0.068	4.32	1	0.04	−.275 to .008	Small
NEPSY	0.013	1.01	0.019	0.47	1	0.49	−.024 to .049	Small

Note: Wald = Wald Chi-Square Test; Exp(B) = exponentiation of the B coefficient (odds ratio; measure of effect size); NEPSY= Neuropsychological Assessment, imitating hand positions subtest; CELF-FS = Clinical Evaluation of Language Fundamentals, formulated sentences subtest

^*p< .05.

A generalized linear model with NEPSY as predictor and Gesture composite score as the outcome variable was significant with a moderate effect size (see Table 9). The relationships among all domains are summarized in Figure 1.

Table 9.

Gamma log model with ADOS gestures as outcome

Pearson χ2	AIC	AIC-C	BIC	Omnibus
5.75	47.70	48.34	52.84	0.002
Predictor	B		Exp(B)	SE	Wald	df	p	95% CI	Effect size
NEPSY*	−0.094		0.91	.03	9.83	1	0.002	−0.15 to −0.04	Small

Note: AIC = Akaike’s Information Criterion; AIC-C = Finite Sample Corrected AIC; BIC = Bayesian Information Criterion; Larger χ² indicates better model fit; lower AIC, AIC-C and BIC indicates better model fit; Wald = Wald Chi-Square Test; Exp(B) = exponentiation of the B coefficient (odds ratio; measure of effect size); NEPSY = Neuropsychological Assessment II, Imitating Hand Positions subtest

^*p< .05.

Figure 1. NEPSY imitating hand positions dominant hand score distributions.

Note: ASD = Autism spectrum disorder; LAD = Loss of ASD diagnosis; TD = typical development.

Summary of results.

Consistent with predictions, analyses indicated that adolescents with current ASD had significantly weaker fine motor imitation skills compared to their peers with no current ASD symptoms or with typical development. Further, correlational analyses indicated that motor abilities were associated with broader ASD symptoms, as well as ADOS gesture scores specifically; and that gesture scores were associated with grammatical abilities. Regression analyses suggested that variance in both language abilities and ASD symptoms was non-unique (shared across variables) in this small but well-characterized sample. Motor imitation skills remained a significant predictor of current ASD symptomatology after controlling for NVIQ. There was a significant interaction between diagnostic group and fine motor imitation skills, such that for the combined TD and LAD group, higher scores on the NEPSY-II fine motor imitation task were associated with more symptoms of ASD, whereas this was not the case for the ASD group. Across regression analyses and generalized linear model analyses, fine motor scores just failed to reach significance as predictors of distinct aspects of language ability after controlling for other variables.

Discussion

Given the strong relationship between motor skills and social communication skills, the current study probed fine motor imitation skill, in the dominant hand, as a covariate of language and autism severity, in youth with ASD, LAD, or typical development. Findings revealed that youth who had lost the autism diagnosis had motor imitation skills similar to their typically developing peers. In contrast, youth with a current ASD diagnosis had weaker fine motor imitation skills than both LAD and TD groups. This extends previous reports of fine motor and action imitation difficulties in ASD (Jansiewicz et al., 2006; Dziuk et al., 2007; Bhat et al., 2011; Whyatt & Craig, 2013), indicating that motor skills appear to normalize along with social skills, communication abilities, and restricted and repetitive behaviors and interests. When individuals lose the ASD diagnosis, their fine motor skills appear to be indistinguishable from those of their typically developing peers.

This outcome reflects at least two distinct possibilities. First, children with better motor imitation skills may show a heightened response to intervention, or an otherwise more dramatic behavioral change in ASD symptoms, making them more likely to lose the diagnosis altogether; this change could occur specifically because motor skills play a causal role in the amelioration of ASD symptoms. That is, the presence of better motor skills early in development might lead children to benefit more (compared to children with more limited motor skills) from intervention targeting language or social domains. Such a relationship between motor skills and improvement in other domains has been established in a study of children with Down syndrome, for example (Yamauchi et al., 2019), as well as in a study of infants at high and low familial risk for ASD (LeBarton & Landa, 2019). Second, the change in ASD symptoms could lead children to engage more in physical games and play, to participate in structured sports and related activities, and thus to display significant improvement in the motor imitation domain. That is, improved social and language skills might facilitate participation in activities that in turn lead to fine and gross motor strength and dexterity. Of course, a cross-sectional study cannot distinguish between these alternatives; a longitudinal study, currently underway, may shed further light on the relationship between these domains.

Our exploratory analyses revealed that fine motor imitation skills were correlated with both grammatical skills and vocabulary skills at the trend level. When controlling for NVIQ, fine motor skills did not significantly predict vocabulary scores, while the relationship was at a trend level for grammatical abilities. This suggests that, although fine motor skills in early childhood predict subsequent language development in individuals with ASD and in typical development (Choi et al., 2018; LeBarton & Landa, 2019), the relationship between these skills may become weaker over the course of development, particularly for verbally fluent adolescents, for whom there is a limited range of variability; these skills may come to be influenced by a myriad of experiential factors.

We found that better fine motor imitation skills were significantly correlated with lesser autism symptom severity (ADOS CSS), with a medium effect size. Motor imitation skills remained a significant predictor of ASD symptom severity when controlling for nonverbal IQ with a small effect size. This main effect was informed by a significant interaction between diagnostic group and fine motor imitation skills in predicting ASD symptom severity. In the ASD group, the relationship was in the expected direction but did not reach significance, likely due to low power; this is broadly consistent with other findings in a similarly-aged sample (e.g., Dziuk et al., 2007. For the TD and LAD groups, higher scores on the fine motor imitation task were, somewhat surprisingly, associated with more symptoms of ASD. Interpretation of this relationship must recognize the low number of total symptoms in the TD and LAD groups (total mean scores of 0.8 and 2.1, respectively), as well as the small effect sizes of these findings. Given the small sample sizes described here, this finding calls for replication in a larger group.

We found that fine motor imitation skills predicted item-level gesture scores on the ADOS, such that those with better fine motor imitation skills produced more co-speech gestures. These results extend previous studies of praxis (i.e., difficulties with skilled motor movement) in individuals with ASD (Dziuk et al., 2007; Dowell et al., 2009), where the extent of motor difficulties (also known as dyspraxia) predicts ASD symptomatology (Stone & Yoder, 2001). MacNeil & Mostofsky (2012) found that even controlling for more basic motor skills, children with ASD showed differences in praxis relative to TD children. Mostofsky and colleagues demonstrated that praxis differences in children with ASD are not limited to imitation, and are also present in gesture-to-command and tool use tasks (2006). Given time limitations, it was not possible to test the relationship between more complex motor skills and ASD symptomatology in this study. Future work will address these questions, while also examining associations of motor skills with structural language abilities, vocabulary scores, and co-speech gesture.

Limitations.

The small sample size of this study limits conclusions, given that we likely had insufficient power to detect the hypothesized relationships between motor and language skills, and to test associations within diagnostic groups. It should be noted as well that the current analysis was hypothesis-driven but exploratory, based on the existence of relevant data. A further limitation of the current study is the nature of the NEPSY imitating hand positions task. By definition, action imitation tasks require both motor skills and the ability to attend to and cognitively represent others’ gestures. Rogers and Pennington (1991) proposed that deficits in executive functioning may underlie impairments in imitation for individuals with ASD. In support of this hypothesis, patients with frontal lesions show deficits in motor imitation tasks (Kimberg & Farah, 1993). Because the NEPSY IH task involves both a motor and cognitive component, the results of the current study may reflect group differences in cognitive domains such as executive functioning or attention, rather than motor skills per se. The current preliminary results provide motivation for testing these associations in a larger sample with a more extensive battery of motor functioning (as described in a pre-registration: https://osf.io/uewmp/?view_only=a6dc8fee70e94115b6154114effb42b9). A more fundamental limitation is that a cross-sectional design cannot test causal relationships. The present study provides a motivation for the necessary longitudinal research.

Implications

The results of the present study (summarized in Figure 3) are consistent with the hypothesis that youth with stronger praxis skills are better at gesture and other nonverbal aspects of communication such as mirroring, contributing to more “typical” scores on the ADOS (although the opposite relationship was present when analyses were limited to the TD and LAD groups, with a small effect size). Results were also consistent with a non-specific model in which individuals with stronger skills in Domain A are more likely to have stronger skills in Domain B (e.g., the “Matthew effect”); common underlying biological or neural differences in individuals with ASD may contribute to both dyspraxia and core ASD symptomatology. The present study suggests an important role for fine motor skills in the core symptoms of ASD, and further emphasizes the role of motor skills in language and gesture production more generally. Results also make the case for the inclusion of motor skill goals in early intervention for ASD; furthermore, understanding how motor skills and social communication are linked in ASD may inform our understanding of how these domains interact in development.

Figure 3. Summary of relationships among fine motor imitation skills, language skills, and symptoms of ASD.

Note: Exp(B) = odds ratio (measure of effect size), r = correlation coefficient; ADOS = Autism Diagnostic Observation Schedule; ‡ p< .10, *p< .05, **p< .01, ***p< .001.

While it is relatively unsurprisingly that motor skills and gesture production might be linked, the fact that motor skills and spoken language abilities are related raises further interesting possibilities for future research. For example: this linkage might be an indirect one, whereby motor skills influence gesture production, which in turn impacts language acquisition. Alternatively, infants and young children with better motor imitation skills may have more synchrony in their social interactions; social synchrony, or the temporal matching of verbal and nonverbal social behaviors with an interactive partner, has been posited to be essential for typical social development (Harrist & Waugh 2002; Feldman, 2007). In turn, these more synchronous social interactions may be more conducive to gesture and language development. Longitudinal approaches will be needed in order to better understand this developmental cascade.

Figure 2. NEPSY imitating hand positions nondominant hand score distributions.

Note: ASD = Autism spectrum disorder; LAD = Loss of ASD diagnosis; TD = typical development

Highlights.

• Children who lost their autism diagnosis (LAD), and those with a history of typical development (TD), had better fine motor imitation skills in the dominant hand than children with a current ASD diagnosis.

• Dominant hand fine motor imitation skills were predictors of ASD symptom severity; language skills were an even stronger predictor.

• Dominant hand fine motor imitation skills were predictors of the production and quality of conversational (co-speech) gestures.