Receptive language and receptive-expressive discrepancy in minimally verbal autistic children and adolescents

Abstract

Among the approximately one-third of autistic individuals who experience considerable challenges in acquiring spoken language and are minimally verbal (MV), relatively little is known about the range of their receptive language abilities. This study included 1579 MV autistic children and adolescents between 5 and 18 years of age drawn from the National Database for Autism Research and the SFARI Base data repository. MV autistic children and adolescents demonstrated significantly lower receptive language compared to the norms on standardized language assessment and parent report measures. Moreover, their receptive language gap widened with age. Overall, our sample demonstrated significantly better receptive than expressive language. However, at the individual level, only about 25% of MV autistic children and adolescents demonstrated significantly better receptive language relative to their minimal expressive levels. Social skills explained a significant proportion of the variance in parent-reported receptive language skills, while motor skills were the most significant predictor of greater receptive-expressive discrepancy. Findings from this study revealed the heterogeneous language profiles in MV autistic children and adolescents, underscoring the importance of individualizing interventions to match their different communication strengths and needs and integrating multiple interconnected areas to optimize their overall development of language comprehension, socialization, and general motor skills.

Lay Summary

For the approximately one-third of autistic individuals who do not develop spoken language beyond a few words or phrases, little is known about how much they might understand spoken language. Our study found that minimally verbal autistic children and adolescents had much lower language comprehension compared to their non-autistic peers, and this gap widened as they got older. However, about a quarter of minimally verbal autistic children and adolescents understood significantly more language than they could produce. Those with better social skills had higher levels of understanding of spoken language, and those with better motor skills showed greater relative strength in understanding spoken language than producing it.

Beyond social interaction and communication challenges as well as restricted and repetitive behaviors (American Psychiatric Association, 2013), approximately 30% of autistic individuals experience considerable difficulties in acquiring spoken language and remain minimally verbal (MV) into adulthood (Tager-Flusberg & Kasari, 2013). While there is no consensus on the precise definition of MV, the term refers to the expressive language limitations of individuals who are over 5 years old (Bal et al., 2016; Kasari et al., 2013; Koegel et al., 2020; Muller et al., 2020). However, less is known about receptive language skills in MV autistic individuals, which have not been systematically studied in relation to their severely limited expressive language skills. The goal of this study is to investigate receptive language skills in a large sample of MV autistic children and adolescents, focusing on the range of their receptive abilities and factors that are related to their receptive-expressive language profiles. Understanding the discrepancies in MV autistic individuals’ abilities to comprehend and produce language provides insights into their heterogeneous language profiles, which are important for individualizing interventions to match their unique communication needs.

Receptive Language in MV Autistic Individuals

Language impairment is a common co-occurring problem in autism (Bishop, 2010; Kjelgaard & Tager-Flusberg, 2001). Delayed language is often the first developmental concern parents identify in their autistic children (Koegel et al., 2020; Talbott et al., 2015). Given that the development of speech by age 5 is a positive prognostic indicator of later social and adaptive functioning (Hampton & Kaiser, 2016; Sandbank et al., 2020; Tager-Flusberg & Kasari, 2013), expressive language has been an important focus of early intervention for autistic children. Nevertheless, despite access to early interventions, many children fail to acquire spoken language and remain non- or minimally verbal, using just a small number of words or fixed phrases in limited contexts. While there is growing literature on MV individuals’ expressive language and communicative repertoires (e.g., La Valle et al., 2020; Pecukonis et al., 2019), little is known about their receptive language profiles. Among the few relevant studies, Plesa Skwerer et al. (2016) found wide variability in receptive language skills among MV autistic children and adolescents across different methods of evaluating receptive language, including caregiver reports, standardized measures, and more experimental eye-tracking and touch screen tasks. Importantly, none of the participants scored within the normative range on the standardized direct assessment, the popular Peabody Picture Vocabulary Test (Dunn & Dunn, 2007). The MV participants often failed to achieve even the basal of the test, or their scores were at the floor. Similarly, Slušná et al. (2021) found that autistic individuals scored significantly below the population average in receptive vocabulary from middle childhood to late adulthood. However, cluster analyses showed some variation in receptive language in a younger MV autistic sample between 7 and 9 years old, such that some understood only a few words while others showed average or close to average comprehension at the lexical and sentence levels (Rapin et al., 2009). The variation in receptive language profiles at different ages highlights the critical need for a more comprehensive understanding of how receptive language changes with age in MV autistic individuals.

Receptive-Expressive Language Discrepancy in Autism

Discrepancies between receptive and expressive language in autism have been the focus of numerous studies. Compared with neurotypical children (Kover et al., 2013; Maljaars et al., 2012) or children with other developmental and intellectual disabilities (Barbaro & Dissanayake, 2012; Ellis Weismer et al., 2010; Seol et al., 2014), the receptive-expressive profile was found to be more discrepant in autistic children. However, findings across studies have not been consistent. Among the studies using age-equivalent scores to reflect the level of development, some reported an expressive advantage in autistic individuals during toddlerhood (e.g., Barbaro & Dissanayake, 2012; Ellis Weismer et al., 2010; Woynaroski et al., 2015), young childhood (Hudry et al., 2010; Volden et al., 2011), and school ages (Maljaars et al., 2012). In contrast, other studies showed a relative strength in parent-reported receptive language (Ellis Weismer et al., 2010; Luyster et al., 2008) or reported no discrepant receptive-expressive language profiles in autistic children (Jarrold et al., 1997). Studies using other scores, such as standard scores in Kjelgaard & Tager-Flusberg (2001) to reflect the degree of delay or advancement of language skills relative to chronological age expectations or vocabulary growth scores in Kover et al. (2013) to reflect the incremental differences in language levels also showed different receptive-expressive profiles in autistic children. These mixed findings may be related to various methodological factors and differences between the ages and ability levels of the participants. A meta-analysis conducted by Kwok et al. (2015) with 74 studies found no significant discrepancy in receptive and expressive language in autistic children and adolescents, even after controlling for their levels of cognitive abilities, types of measures, and methods of autism diagnosis. Despite the large number of studies that have investigated receptive-expressive discrepancies in autism, none, to our knowledge, has included an exclusively MV sample.

While, by definition, all MV autistic individuals experience considerable difficulties in acquiring spoken language, it remains unclear what proportion of them may have significantly higher receptive language than their minimal expressive language. Little is known about how receptive and expressive language may differ in MV autistic individuals at an individual level, meaning the within-subject difference between their receptive and expressive language. Understanding receptive-expressive discrepancy at an individual level allows us to identify subgroups of MV autistic individuals with different developmental patterns of receptive and expressive language. While most previous studies used age-equivalent scores to investigate the receptive-expressive discrepancy in autistic children, these scores do not allow for direct comparisons between subscales at an individual level, and their psychometric limitations in accurately reflecting true levels of abilities have been well-documented (Maloney & Larrivee, 2007). Instead, standard scores are recommended for characterizing individual-level abilities (Mervis & Klein-Tasman, 2004). Using standard scores to characterize the receptive-expressive discrepancy can demonstrate the extent to which the development of receptive and expressive language is relatively advanced or delayed to each other comparing to the chronological age expectations, allowing us to identify which domain is a relative strength and weakness.

Factors Associated with Receptive Language Abilities in Autistic Individuals

Several child characteristics have been shown to be linked to receptive language in autistic children and adolescents with varying levels of language proficiency, including nonverbal cognitive abilities, social skills, autism severity, and motor skills. Nonverbal cognition is one of the most consistently reported predictors of language abilities in autistic children (Anderson et al., 2007; Bal et al., 2020; Wodka et al., 2013). In addition to predicting gains in spoken language over time in verbal (Anderson et al., 2007) and MV autistic children (Mazurek et al., 2012; Thurm et al., 2015; Wodka et al., 2013), nonverbal cognition was also positively correlated with concurrent receptive vocabulary knowledge in autistic children and youth who were MV (Muller et al., 2022), verbal (Haebig & Sterling, 2017), and who had intellectual disability (Goharpey et al., 2013). A longitudinal study also showed that nonverbal cognition at age 2 predicted receptive vocabulary at age 3, and nonverbal cognition at age 3 predicted more complex language comprehension at age 7 in verbal and MV autistic children (Charman et al., 2005). However, Slušná et al. (2021) found no significant association between nonverbal cognition and concurrent receptive vocabulary knowledge in nonverbal autistic individuals.

Social skills have also been associated with receptive language in autism, though none of the studies investigating this relationship focus on older MV children. Autistic children with stronger social skills showed better concurrent receptive vocabulary (Liss et al., 2001) and comprehension of spoken language and non-verbal communication behaviors, such as gestures and facial expressions (Yavuz et al., 2019). A longitudinal study also found that social skills at age 3 predicted word comprehension skills at age 7 in autistic children (Charman et al., 2005). However, after accounting for general cognitive ability, Kjellmer et al. (2012) found that social skills only explained some variance in word and phrase comprehension in younger 2-5-year-old autistic children.

Regarding the relation between autism symptoms and receptive language in autistic individuals, longitudinal studies found that more severe social communication and interaction difficulties (but not restricted and repetitive behaviors) at age 3 predicted lower receptive language levels at age 7 (Charman et al., 2005). Ellis Weismer & Kover (2015) also found that autism severity significantly predicted growth in language comprehension in autistic children across the preschool period, even after controlling for cognitive abilities, socialization, and joint attention. In contrast, a recent cross-sectional study showed that autism symptoms did not account for significant variance in concurrent receptive vocabulary in younger preverbal and older MV autistic children after controlling for nonverbal cognitive skills (Muller et al., 2022).

Motor delays and impairments, which are common in autism (Bhat, 2020; Bhat et al., 2022; Lloyd et al., 2013; Staples & Reid, 2010), are significantly correlated with language delays, including receptive language in the early years (e.g., Bedford et al., 2016; Bhat et al., 2012; Bruyneel et al., 2019; Franchini et al., 2018; Iverson, 2018). One study, which included MV autistic children, reported on the relationship between fine motor skills and expressive language (Butler & Tager-Flusberg, 2022). However, none has investigated the relationship with receptive language in MV autistic individuals, especially beyond infancy and early childhood.

Taken together, despite some mixed findings, receptive language in a broader autistic population has been shown to be related to nonverbal cognitive abilities, social skills, autism severity, and motor skills. However, it remains unknown whether these child characteristics are also significantly associated with receptive language in MV autistic children and adolescents or with their receptive-expressive language discrepancy.

Current Study

The first aim of this study was to investigate the range of receptive language abilities in MV autistic children and adolescents using both standardized assessments and parent report measures. In line with previous studies (Plesa Skwerer et al., 2016; Slušná et al., 2021), we hypothesized that MV autistic children and adolescents would demonstrate significantly lower receptive language levels compared to the population average. Our second aim was to investigate both group- and individual-level receptive-expressive language discrepancy in MV autistic children and adolescents. We hypothesized that some proportion of the participants would have discrepantly higher receptive language skills based on standard scores. Our third aim was to investigate factors associated with receptive language level and receptive-expressive discrepancy in MV autistic children and adolescents, including nonverbal cognitive abilities, social skills, autism severity, and motor skills. Based on studies including verbal autistic children (e.g., Bedford et al., 2016; Ellis Weismer & Kover, 2015; Haebig & Sterling, 2017; Yavuz et al., 2019), we hypothesized that higher receptive language scores compared to the group and receptive strength compared to one’s expressive language would be found in MV autistic children and adolescents with better nonverbal cognitive abilities, stronger social skills, lower levels of autism severity, and better motor skills.

Methods

Participants

A total of 1579 MV autistic children and adolescents (321 Female; M_age=8.75 years, SD=3.6, range=5-18) were included in this study. The inclusion criteria were 1) between 5 and 18 years old; 2) meeting the autism spectrum cut-off on the Autism Diagnostic Observation Schedule (ADOS, Lord et al., 2000; Lord et al., 2012) Module 1, which confirmed that the participants were autistic and met the widely accepted definition of MV (Bal et al., 2016); 3) with at least one age-appropriate standardized measure of both receptive and expressive language including the Vineland Adaptive Behavior Scales, Second Edition (VABS-II, Sparrow et al., 2005) or the Preschool Language Scale, Fifth Edition (PLS-5; Zimmerman et al., 2011). See Table 1 for more demographic details.

Sources of Data

Data were obtained from the National Database for Autism Research (NDAR) and the SFARI Base from the Simons Foundation Autism Research Initiative. This study was approved by the University Institutional Review Board (IRB) at Boston University. NDAR is a research data repository sponsored by the US National Institutes of Health (NIH) to promote collaboration through data sharing and querying (https://nda.nih.gov). Each participant from NDAR has a global unique patient identifier (GUID) that can be linked to individual-level data from different datasets. The final sample of this study included 986 participants from 59 studies on NDAR, in which parents provided informed consent, approved by each study’s university IRB, to allow their data to be shared with future studies through NDAR.

SFARI Base (http://sfari.org/resources/sfari-base) is another research data repository funded by the Simons Foundation Autism Research Initiative, which contains large-scale behavioral and biological data of autistic individuals. Three cohorts from the SFARI Base were included in this study, including Simon Searchlight, Version 11.0 (released in March 2019), Simons Simplex Collection Version 15.3 (SSC; released in August 2013), and Autism Inpatient Collection (released in December 2017). Each participant in SFARI Base has an SFARI ID that can be used to locate individual-level data. At each participating site, parents provided informed consent approved by each university IRB. The final sample included 593 participants from SFARI Base.

Measures

Language Skills

The VABS-II (Sparrow et al., 2005) is a standardized measure of adaptive functioning. Participants’ receptive and expressive language were measured by the Receptive and Expressive subscales on the VABS-II Communication domain, which demonstrates good internal consistency reliability ranging from .68 to .92, test-retest reliability from .76 to .90 for the age group of 5 to 18, and excellent validity (Sparrow et al., 2005). The Receptive subscale includes three content categories: understanding, listening and attending, and following instructions. Most test items measure responding to voice, language, and verbal instructions. The Expressive subscale includes pre-speech expression, beginning to talk, interactive speech, speech skills, and expressing complex ideas. A significant receptive-expressive discrepancy was determined by meeting or exceeding the VABS-II pairwise-comparison critical value (Sparrow et al., 2005). Both the Parent Survey Interview Form and the Parent/Caregiver Rating Form were included in this study, given that their content are similar enough to warrant a combination for analyses (Sparrow et al., 2005).

The PLS-5 (Zimmerman et al., 2011) is a standardized direct assessment to measure developmental language skills in children from birth to 7 years 11 months. It demonstrates good to excellent test-retest reliability with average coefficients from .86 to .95, excellent internal consistency with an average coefficient of .91, and good validity (Zimmerman et al., 2011). Participants’ receptive and expressive language were measured by the Auditory Comprehension and Expressive Communication subscales. A significant receptive-expressive discrepancy was determined by meeting or exceeding the PLS-5 pairwise-comparison critical value.

Autism Severity

The ADOS (Lord et al., 2000; Lord et al., 2012) is a semi-structured observational assessment for the diagnosis of autism spectrum disorder. This study included both the ADOS-Generic (Lord et al., 2000) and the ADOS-2 (Lord et al., 2012) Module 1. The Social Affect and Restricted and Repetitive Behavior domain raw scores and total raw score were converted to the ADOS calibrated severity scores (CSS) on a Likert scale of 1-10 (1-3 indicates non-spectrum, 4-5 indicates autism spectrum disorder, and 6-10 indicates autism), which were only normed in autistic individuals aged 2-14 years in Module 1 (Gotham et al., 2009; Hus et al., 2014). Higher CSS scores indicate greater autism severity.

Nonverbal Cognition

The Differential Ability Scales, Second Edition (DAS-II; Elliott, 2007) is a standardized assessment for assessing the cognitive abilities of children and adolescents from ages 2:6 to 17:11 and has demonstrated good reliability (e.g., internal consistency reliability for nonverbal abilities is .89 for the Early Years upper level and .92 for the School Age battery) and good validity (Elliot, 2007). Both the Early Years Battery Upper Level (for children aged 3:6 to 6:11) and the School Age Battery (for children aged 7:0 to 17:11) were included in this study. The DAS-II Early Years ratio nonverbal IQ (NVIQ) score was calculated by averaging the age equivalent scores of the Matrices and Picture Similarities subscales and dividing by chronological age, multiplied by 100. Similarly, the DAS-II School Age ratio NVIQ score was calculated by averaging the age equivalent scores of the Matrices and Sequential and Quantitative Reasoning subscales and dividing by chronological age, multiplied by 100. Higher ratio NVIQ scores indicate higher levels of nonverbal cognitive abilities.

Social Skills

The VABS-II (Sparrow et al., 2005) Socialization domain standard score was used to reflect the participants’ social skills. The VABS-II Socialization domain includes three subdomains of social skills, including how the individual interacts with others (i.e., Interpersonal Relationships subscale), plays and uses leisure time (i.e., Play and Leisure Time subscale), and demonstrates responsibility and sensitivity to others (i.e., Coping Skills subscale). The VABS-II Socialization domain demonstrates excellent internal consistency reliability ranging from .88 to .95, good test-retest reliability from .74 to .88 for the age group of 5 to 18, and excellent validity (Sparrow et al., 2005). Both the Parent Survey Interview Form and the Parent/Caregiver Rating Form were included. A higher score indicates a higher level of social skills.

Motor Skills

The Developmental Coordination Disorder Questionnaire (DCDQ; Wilson et al., 2009) is a 15-item parent questionnaire using a 5-point Likert scale to screen for coordination disorders and measure gross and fine motor skills in children between 5 and 15 years old. The DCDQ includes a total score (ranging from 15 to 75) reflecting the general motor skills and scores from three subscales measuring control during movement, fine motor, and general coordination. The measure is reported to have high internal consistency and sensitivity (Green et al., 2009; Van Damme et al., 2022). Given that the DCDQ does not have standard scores, we used the total raw score to represent the developmental level of motor skills in this study, where higher scores indicate more advanced motor skills.

Data Preparation and Analytic Plan

The distributions of the standardized measures scores were first tested for normality through the Kolmogorov-Smirnov test and visual inspection on univariate histograms and the normal Q-Q plots. Based on the normality tests, the scores of receptive language, nonverbal cognition, social skills, autism severity, and motor skills all violated normality (see Table 2). As parametric analysis of transformed data has more statistical power than nonparametric analysis (Rasmussen & Dunlap, 1991), and the means are better representations of central tendency for the distribution of our data (i.e., receptive language, social skills, motor skills, and NVIQ) than the medians, we transformed the skewed data for parametric analyses. Specifically, we applied a square root transformation to the scores of the VABS-II receptive and expressive language, NVIQ, and social skills due to moderately, positively skewed data (Tabachnick & Fidell, 2007) and transformed chronological age and the DCDQ motor total score using a base-ten log transformation due to strongly, positively skewed data (Tabachnick & Fidell, 2007). These data transformations effectively reduced skewness in all these variables and yielded normal distribution, except that the PLS-5 auditory comprehension and expressive communication scores remained skewed after transformation. Transformed data were used for correlation and regression analyses. The parametric correlation test (Pearson’s r) was used to examine bivariate correlations among the continuous variables, including the VABS-II receptive and expressive language, social skills, NVIQ, motor skills, and age, and the nonparametric correlation test (Spearman’s rank-order correlation) was used for the ADOS CSS, given its ordinal level of measurement, and the PLS-5 scores, given its significant skewness.

Descriptive statistics were calculated for the participants’ demographic characteristics. Preliminary analyses were conducted to examine the associations between receptive language and demographic variables to identify any significant covariates that would be controlled for in subsequent analyses. For the first aim of the study, we used the VABS-II receptive v-scale score and the PLS-5 auditory comprehension standard score to describe the receptive language profiles in MV autistic children and adolescents. For the second aim of the study, we used the established pairwise-comparison critical values on the VABS-II (Sparrow et al., 2005) to determine individual-level significant receptive-expressive language discrepancy. If the difference between the receptive and expressive subscales standard scores met or exceeded the pairwise-comparison values, the receptive-expressive discrepancy was considered significant for that participant. We used standard scores to identify the extent to which the development of receptive and expressive language is relatively advanced or delayed to each other when comparing to chronological age expectations. For the third aim, correlation analyses were computed to examine the bivariate associations between receptive language and the four empirically selected child characteristics, including nonverbal cognition, social skills, autism severity, and motor skills. A multiple regression model was then constructed to identify the concurrent predictors of receptive language in MV autistic children and adolescents by examining the relative predictive values of all variables that were significantly correlated with receptive language. All assumptions of the regression models were examined before carrying out the analyses, and our data met all the assumptions. Last, a multiple regression model was constructed to identify the concurrent predictors of the magnitude of the receptive-expressive discrepancy (a continuous variable) with all of these child characteristics. All assumptions of the regression model were examined and met.

Results

Receptive Language in MV Autistic Children and Adolescents

On the VABS-II, on average, the participants had a receptive v-scale score of 7.53 (SD = 2.62) and an expressive v-scale score of 5.37 (SD = 2.43). Compared to the VABS-II norms (i.e., the mean receptive and expressive v-scale scores of 15 (SD = 3); Sparrow et al., 2005), one-sample t-tests showed that MV autistic children and adolescents in this study had much lower receptive and expressive language skills, t (1316) = −103.39, p < .001, and t (1313) = −143.42, p < .001, indicating around 2.5 to 3 SDs below the population averages. With respect to the PLS-5, on average, the participants demonstrated an auditory comprehension standard score of 50.97 (SD = 3.78) and an expressive communication standard score of 50.42 (SD = 1.62). Similarly, compared to the PLS-5 norms (i.e., auditory and expressive communication standard scores of 100 (SD = 15); Zimmerman et al., 2011), participants in this study had significantly lower receptive and expressive language skills than their age-matched neurotypical peers, t (152) = −160.64, p < .001, and t (152) = −378.30, p < .001, indicating around 3 SDs below the population averages. However, while only 6.9% of the participants scored at the floor on the VABS-II receptive language subscale (i.e., with a standard score of 1 or 2 depending on the age), 89% were at the floor on the PLS-5 auditory comprehension subscale (i.e., with a standard score of 50). Given that the PLS-5 failed to capture the variation in receptive language skills in MV autistic children and adolescents, it was excluded from analyses related to other aims.

Preliminary analyses showed that chronological age was negatively correlated with VABS-II receptive language, r (1317) = −.423, p < .001, and PLS-5 auditory comprehension, r_s (153) = −.270, p < .001, indicating a widening gap between receptive language skills in MV autistic children and adolescents and their non-autistic peers as they aged. Similarly, age was negatively correlated with VABS-II expressive language, r (1314) = −.660, p < .001, and PLS-5 expressive communication, r_s (153) = −.296, p < .001. Given the steeper negative slopes of expressive language compared to receptive language with age on both the VABS-II and the PLS-5, expressive language fell much further behind typical development than receptive language in older MV autistic children and adolescents. Female and male MV autistic children and adolescents did not differ significantly in their receptive language scores on the VABS-II, t (1315) = −.725, p = .469, or the PLS-5, t (151) = −.797, p = .426. Therefore, sex was not considered a covariate in further analyses.

Receptive-Expressive Language Discrepancy

At the group level, the VABS-II receptive language scores (M = 7.53, SD = 2.62) were significantly higher than expressive language scores (M = 5.37, SD = 2.43), t (1310) = 38.84, p < .001. At the individual level, 317 of 1311 MV autistic children and adolescents demonstrated significantly higher receptive than expressive language scores based on the VABS-II pairwise-comparison critical values, representing 24.18% of those with both VABS-II receptive and expressive scores. The receptive-expressive discrepancy ranged from −5 to 9 (M = 2.16, SD = 1.99). The subgroup with a discrepant receptive-expressive language profile had significantly better receptive language (M = 8.74, SD = 2.42) than those whose receptive and expressive language were not discrepant (N= 993, M = 7.14, SD = 2.57), t (1309) = 9.722, p < .001. The discrepant group also had significantly better motor skills than the non-discrepant group, t (361) = 2.51, p = .013. Only one participant showed discrepantly higher expressive than receptive language on the VABS-II (0.08% of the sample).

Concurrent Predictors of Receptive Language in MV Autistic Children and Adolescents

The VABS-II receptive language was significantly correlated with nonverbal cognitive abilities measured by the DAS-II, r (324) = .478, p < .001, social skills on the VABS-II, r (1291) = .676, p < .001, ADOS-2 overall CSS, r_s (971) = −.08, p = .011, and motor skills on the DCDQ, r (363) = .287, p < .001. Given that the DCDQ total raw score did not take age into account, the association between receptive language and motor skills was verified by a partial correlation controlling for age. The association remained significant, r (360) = .297, p < .001. These findings indicate that MV autistic children and adolescents with better nonverbal cognitive abilities, social skills, motor skills, and lower levels of autism severity had significantly better receptive language reported by parents. To account for multiple comparisons, the Benjamini-Hochberg procedure (Benjamini & Hochberg, 1995) with a 5% false discovery rate was used (see Table 3). All of these associations remained significant after controlling for multiple comparisons. Given that all four child characteristics (i.e., NVIQ, social skills, autism severity, and motor skills) were significantly correlated with receptive language on the VABS-II, a multiple regression model was constructed to examine their relative predictive values. All assumptions of the regression models were examined and met. Specifically, although some of the child characteristics were correlated with one another (see Table 3), variance inflation factors were all below 4, ranging from 1 to 1.9, which indicated absent to weak multicollinearity that did not require further investigation. Overall, a significant regression model emerged, R² = .44, F (5, 226) = 35.43, p < .001. Social skills was the only salient factor predicting receptive language, β = .507, p < .001, above and beyond all other child characteristics and chronological age (see Table 4).

Concurrent Predictors of Receptive-Expressive Language Discrepancy

A multiple regression model was constructed to identify the predictive values of NVIQ, social skills, autism severity, and motor skills on the receptive-expressive discrepancy in MV autistic children and adolescents controlling for chronological age. A significant model emerged, R² = .171, F (5, 203) = 8.358, p < .001. Better motor skills on the DCDQ significantly predicted the receptive-expressive discrepancy, β = .148, p = .041, above and beyond NVIQ, social skills, and autism severity. The covariate chronological age also demonstrated a significant main effect, β = .369, p < .001. When adding the interaction between age and the DCDQ total score to the model, the R² change was not significant, p = .636, indicating that motor skills and age predicted receptive-expressive discrepancy independently. A greater receptive-expressive discrepancy was associated with higher motor skills and older age (see Table 5).

Discussion

This study investigated the range of receptive language abilities, the receptive-expressive language discrepancy, and their concurrent predictors in a large-scale MV autistic sample. Overall, we found that MV autistic children and adolescents not only demonstrated much lower receptive language compared to their non-autistic peers based on both direct assessments and parent report measures, but their receptive language gap widened with age. Across the sample, receptive language scores were significantly higher than expressive language. At the individual-level discrepancy, about one-quarter of MV autistic children and adolescents demonstrated significantly better receptive language relative to their minimal expressive levels. Social skills explained a significant proportion of the variance in receptive language scores reported by parents, while motor skills predicted a more discrepant language profile with significantly better receptive than expressive language, above and beyond all other factors.

When examining the range of receptive language abilities in MV autistic children and adolescents, there were evident floor effects on the PLS, meaning that most of the participants were unable to complete an adequate number of test items to achieve a standard score above the minimum on this direct measure. It is not rare to see floor effects, which pertain to standard scores, on direct measures in MV autistic individuals (Bal et al., 2016; Kasari et al., 2013). However, when it is necessary to use standard scores (e.g., to compare individual-level receptive and expressive language scores), the VABS is more appropriate for MV individuals compared to the PLS, given that the VABS provides normed-referenced v-scale scores as low as almost 5 standard deviations below the mean (Sparrow et al., 2005), allowing for more valid estimates of MV individuals’ significantly delayed language abilities that cannot be adequately assessed through direct measures. Moreover, the VABS assesses communication abilities beyond structural language. MV individuals may be more responsive to the social communication bids from their parents in everyday situations than those from an unfamiliar examiner in the laboratory or clinic setting. Therefore, parents may report higher receptive abilities in their MV autistic child than what is revealed in decontextualized language measures with unfamiliar examiners. Given that language measures are often used to inform intervention goals and progress, future research should consider including the VABS to capture language skills in MV autistic individuals.

In this study, we found a negative association between receptive language standard scores and age, indicating a widening gap in receptive language abilities in older MV autistic children and adolescents relative to their non-autistic peers. While MV autistic individuals’ receptive language abilities may improve as they age, their language growth does not keep pace with typical development. Similar findings were reported in verbal autistic children aged 7-18 years old who failed to make gains in communication and socialization skills at a rate commensurate with their chronological age (Klin et al., 2007). It may be that, compared to neurotypical peers who receive more complex and a greater amount of language input from school instruction and more reading and writing activities in later childhood and adolescence, MV autistic individuals’ communication difficulties may limit them from participating or engaging in such learning activities, resulting in their receptive language development plateauing or falling behind the pace of typical development. Thus, when communication demands increase in everyday life, poor receptive language becomes more evident in MV autistic individuals.

Regarding the receptive-expressive language discrepancy, we found that MV autistic children and adolescents demonstrated significantly better receptive than expressive language as a group, but only approximately a quarter of them demonstrated receptive strength compared to their minimal expressive levels, indicating some but not most MV autistic children and adolescents understand significantly more than they can produce. This finding contrasted with Woynaroski et al. (2016), who found that expressive language was higher than receptive language in preverbal autistic children from 2 to 4 years old. The contrasting findings may be due to different age groups who are at different stages of language development (i.e., preverbal vs. MV) between Woynaroski et al. (2016) and our study. Previous research with verbal autistic children showed a developmental change in receptive-expressive profile, such that expressive language was higher than receptive language at 30 months of age, but the discrepancy steadily decreased from then to 44 months of age (Davidson & Ellis Weismer, 2017). In contrast to these findings with preverbal and verbal autistic children during toddlerhood and early childhood, our study highlights a different receptive-expressive language profile in MV autistic children and adolescents beyond age 5, who demonstrated better receptive than expressive language. Given the cross-sectional nature of our study, we cannot investigate the developmental changes in the receptive-expressive profile in our sample. However, we found a significant main effect of age on receptive-expressive discrepancy, indicating that more discrepant receptive-expressive profiles were observed in older MV autistic individuals between 5 and 18 years of age. It may be that while both receptive and expressive language gaps widen with age in MV autistic children and adolescents compared to non-autistic peers, the expressive gap is larger than the receptive gap, contributing to a more discrepant receptive-expressive profile in older MV individuals.

In our investigation of factors associated with receptive language skills, we found that social skills emerged as the most significant factor for MV autistic children and adolescents, above and beyond other child characteristics. This finding was in line with previous studies with verbal autistic children that those with stronger social skills demonstrated better language comprehension concurrently and longitudinally (Charman et al., 2005; Liss et al., 2001; Yavuz et al., 2019). In the VABS, a broader range of adaptive social skills are captured, including those important for developing and maintaining interpersonal relationships (e.g., expressing and recognizing emotions; friendship and dating) and playing with others (e.g., sharing and cooperating; going places with friends; recognizing social cues). These adaptive social competencies are crucial for maintaining high-quality reciprocal social interactions, expanding exposures to different conversational contexts, and facilitating engagement in social activities, which in turn influence the quantity and quality of language inputs MV individuals receive and ultimately affect their receptive language development. Importantly, previous longitudinal research showed that the association between social competence and language abilities (both receptive and expressive) is reciprocal between autistic children aged 2 to 4 years (Bennett et al., 2014). It may also be that, for MV autistic children and adolescents, poor social skills affect their receptive language abilities by reducing social interactions and language exposure, while their poorer receptive language also prevents them from engaging in a range of social contexts and experiences, as captured on the VABS Socialization scale, creating reciprocal interactions between reduced social interactions and poorer receptive language development over time.

Moreover, motor skills emerged to be closely linked to the receptive-expressive discrepancy in MV autistic children and adolescents, above and beyond nonverbal cognitive abilities, social skills, or autism severity. This finding reveals a unique link between motor skills and individual-level receptive-expressive discrepancy in MV autistic children and adolescents, which is not driven by the developmental level or autism symptoms. As suggested in previous studies with neurotypical toddlers, achieving new motor milestones, such as learning to crawl and walk, changes the way infants interact and communicate with others and the environment, which in turn alters how others respond to them (Iverson, 2018; Karasik et al., 2014). Similarly, motor skills provide a necessary foundation for MV autistic children and adolescents, even when they get older, to participate and engage in different activities, in which social skills begin to impact the quality of these participations and interactions. Those with better motor skills have a greater chance of exploring different environments, interacting with more people, objects, and other environmental stimuli, and possibly participating in a wider range of activities, such that they are exposed to more complex and diverse language, which eventually promotes their language development. For example, the DCDQ measures if a child “runs as fast and in a similar way to other children of the same gender and age”, has the motor ability to move on playground equipment or build a structure with blocks, and their “printing, writing, or drawing in class is fast enough to keep up with the rest of the children in the class.” For school-aged MV autistic children and adolescents, if they can run or move around the playground as fast and smoothly as their peers and write or draw as fast as others in the class, it is more likely for them to have higher-quality reciprocal social interactions and more engaged learning experiences than those without an adequate level of motor skills, leading to more diverse language exposure that contributes to their language development. By definition, all MV autistic individuals experience considerable difficulties acquiring spoken language, potentially attributed to speech-specific motor limitations (Chenausky et al., 2019) or oromotor impairments (Maffei et al., 2023), which are not captured by the DCDQ used in this study. While oromotor-specific impairments may have hindered MV autistic individuals’ abilities to speak, their language exposures facilitated by more advanced general motor skills may result in more pronounced receptive language development, leading to a more discrepant receptive-expressive discrepancy. This may be one of the reasons why we observe a receptive strength in MV autistic children and adolescents with better motor skills. In addition, the longitudinal relationship between early motor skills and later receptive language development has been reported in autistic children and children with other language difficulties. For example, Bedford et al. (2015) found that autistic children with better early gross motor skills at 2 years old had faster receptive language development from 2 to 9 years. Viholainen et al. (2006) found that children with familial risk for dyslexia who had slower early motor development during toddlerhood had a significantly smaller receptive vocabulary at the ages of 3.5 and 5 years old and slower reading of words at the end of the first grade. These longitudinal studies showed that early delays in motor development have a long-lasting impact on receptive language development. Given that autistic individuals with motor difficulties during childhood tend to have increasingly poorer motor performance from adolescence into adulthood (Travers et al., 2017), it is possible that MV autistic individuals in our study may also have atypical motor development from the early years. While our finding regarding the association between motor skills and receptive-expressive discrepancy are grounded in their concurrent relationships, it may reflect a long-lasting impact of early motor development on the current motor skills and receptive language. More research is needed to disentangle the developmental mechanism underlying the longitudinal receptive language-motor relationship in autism.

Although the precise mechanisms underlying the relation between motor skills and receptive strength remain unknown, numerous neurophysiological findings suggest that the motor cortex plays a significant role in language processing and perception, which may help explain the relation (Fischer & Zwaan, 2008; Pulvermüller & Fadiga, 2010). During active or passive speech perception, some areas of the motor cortex, such as the inferior frontal premotor cortex, are activated (Vigneau et al., 2006; Wilson et al., 2004). A broader area of the motor cortex was also found to be involved in word and sentence semantic comprehension (Fischer & Zwaan, 2008; Pulvermüller & Fadiga, 2010). In addition, previous studies found that the cerebellum, the brain region primarily responsible for motor control, was activated during sentence comprehension (Stowe et al., 2004) and semantic decision-making (McDermott et al., 2003). Based on these findings, the motor cortex and the cerebellum are significantly involved during different levels of language processing and perception, which may indicate the neural bases of the close link between motor skills and receptive language in autistic individuals.

Limitations and Future Directions

One limitation of the current study is the reliance on the parent report as the sole measure of language abilities in MV autistic children and adolescents due to the floor effects on the direct measure in our sample. Although parent report measures provide valuable insights into the language abilities demonstrated by MV autistic children and adolescents during everyday situations, it is important to acknowledge that parents’ subjective perceptions may sometimes under- or over-estimate their child’s language abilities. This pattern of estimation could also extend to other subscales (e.g., VABS Socialization), potentially contributing to a strong relation between parent-reported receptive language and social skills. Moreover, our analyses were constrained to the variables available in the extant data. Future research should consider investigating how other child characteristics beyond the ones included in the current study may be related to receptive language and receptive-expressive discrepancy in MV autistic individuals. Last, given the cross-sectional nature of this project, our findings highlight that future research is needed to examine the longitudinal association between social skills, motor skills, and receptive language development and its underlying mechanisms in MV autistic individuals.

Conclusions

Using large-scale datasets, this study was the first known attempt, to the best of our knowledge, to investigate the receptive-expressive language discrepancy in an exclusively MV autistic sample. Although all MV autistic individuals experience considerable challenges in acquiring spoken language and their language comprehension skills fall behind typical development, approximately a quarter of them demonstrate relative strength in receptive language compared to their minimal expressive level. Findings in this study advance our understanding of the heterogeneous language profiles in MV autistic children and adolescents, underscoring the importance of individualizing interventions to match their different communication strengths and needs. Given that those with stronger social skills had higher levels of receptive language and those with better motor skills had more discrepant receptive-expressive profiles with a receptive strength, we suggest that language interventions for MV autistic children and adolescents may benefit from incorporating social and motor skills training, as this holistic approach holds the potential to optimize the overall language outcomes by integrating several interconnected areas of development.

Appendix

Table 1.

Demographic Characteristics of Child Participants and Parents (N = 1579)

Demographic Characteristics	n	%
Child Gender
Male	1258	79.7
Female	321	20.3
Child Ethnicity (n = 769 reported)
White	482	62.7
Black or African American	74	9.6
Native American or Alaska Native	3	0.4
Native Hawaiian or Pacific Islander	2	0.3
Asian	57	7.4
More than one race	72	9.3
Not specified or other	79	10.3
Father Highest Education Level (n = 764 reported)
Less than high school	57	7.5
High school graduate	160	20.9
Some college	112	14.7
Associate or bachelor’s degree	213	27.9
Graduate degree or professional training	222	29.1
Mother Highest Education Level (n = 759 reported)
Less than high school	41	5.4
High school graduate	140	18.4
Some college	137	18.1
Associate or bachelor’s degree	259	34.1
Graduate degree or professional training	182	24.0
Annual Household Income (N = 713 reported)
Less than $20,000	70	9.8
$20,000 - $50,000	158	22.2
$50,000 - $80,000	143	20.1
$80,000 - $100,000	95	13.3
Over $100,000	247	34.6

Table 2.

Means and Standard Deviations of Child Characteristics Variables

Child Characteristics	n	M	SD	Min	Max	Komogorov- Smirnov Test
VABS-II Receptive v-scale score	1317	7.53	2.62	1	18	.15***
VABS-II Socialization score	1331	55.42	10.73	6	100	.06***
DAS-II NVIQ	341	58.96	21.04	29	114	.07***
ADOS overall CSS	1167	7.52	1.44	4	10	.17***
DCDQ total score	363	32.55	11.68	15	75	.11***

Note. VABS-II = the Vineland Adaptive Behavior Scale, Second Edition; DAS-II = the Differential Ability Scale, Second Edition; NVIQ = nonverbal ratio IQ score; ADOS = the Autism Diagnostic Observation Schedule; CSS = calibrated severity score; DCDQ = the Developmental Coordination Disorder Questionnaire. Scores of the VABS-II receptive language, VABS-II socialization, DAS-II NVIQ, and DCDQ motor skills were transformed for parametric analyses due to data skewness.

^*p < .05.

^**p < .01.

^***p < .001

Table 3.

Correlations between Receptive Language, Nonverbal Cognition, Social Skills, Autism Severity, and Motor Skills in MV Autistic Children and Adolescents (N = 1579)

	1	2	3	4	5
1.VABS-II Receptive Language	--
2. DAS-II Nonverbal Cognition	.478 ***	--
3. VABS-II Socialization	.676 ***	.558 ***	--
4. ADOS Overall CSS	−.08 *	.038	−.07*	--
5. DCDQ Motor Skills	.287 ***	.367 ***	.312 ***	−.04	--

Note. VABS-II = the Vineland Adaptive Behavior Scale, Second Edition; DAS-II = the Differential Ability Scale, Second Edition; ADOS = the Autism Diagnostic Observation Schedule, Second Edition; CSS = calibrated severity score; DCDQ = the Developmental Coordination Disorder Questionnaire. Bolded p-values remain significant after performing the Benjamini-Hochberg procedure (Benjamini & Hochberg, 1995) with a 5% false discovery rate to account for the impact of multiple comparisons.

^*p < .05.

^**p < .01.

^***p < .001

Table 4.

Multiple Regression Results for Nonverbal Cognition, Social Skills, Autism Severity, and Motor Skills in Predicting Receptive Language in MV Autistic Individuals Controlling for Chronological Age (N = 231)

Variable	B	95% CI for B		SE B	β	p-value
		LL	UL
Constant	−2.842	−9.398	3.713	3.327
Age	−1.672	−3.745	.401	1.052	−.115	.113
Nonverbal Cognition	.126	−.092	.344	.111	.08	.258
Social Skills	1.617***	1.195	2.039	.214	.507***	<.001
Autism Severity	−.006	−.159	.147	.078	−.004	.940
Motor Skills	1.127	−.522	2.776	.837	.076	.179

Note. CI = confidence interval; LL = lower limit; UL = upper limit; Nonverbal cognition represented by the Differential Ability Scale, Second Edition NVIQ score; Social skills represented by the Vineland Adaptive Behavior Scale, Second Edition, Socialization domain standard score; Autism severity represented by the Autism Diagnostic Observation Schedule calibrated severity score; Motor skills represented by the Developmental Coordination Disorder Questionnaire total score.

^*p < .05.

^**p < .01.

^***p < .001

Table 5.

Multiple Regression Results for Nonverbal Cognition, Social Skills, Autism Severity, and Motor Skills in Predicting Receptive-Expressive Discrepancy in MV Autistic Individuals Controlling for Chronological Age (N = 231)

Variable	B	95% CI for B		SE B	β	p-value	R2	Δ R2
		LL	UL
Step 1							.171	.171***
Constant	−10.115	−17.577	−2.653	3.785		.008
Age	4.745***	2.300	7.189	1.24	.369***	<.001
Nonverbal Cognition	−.025	−.287	.237	.133	−.018	.851
Social Skills	−.018	−.497	.46	.243	−.007	.939
Autism Severity	.976	−.158	2.11	.575	.109	.091
Motor Skills	1.939*	.083	3.795	.941	.148*	.041
Step 2							.172	.001
Age*Motor Skills	−2.641	−13.612	8.33	5.564	−.486	.636

Note. CI = confidence interval; LL = lower limit; UL = upper limit; Nonverbal cognition represented by the Differential Ability Scale, Second Edition NVIQ score; Social skills represented by the Vineland Adaptive Behavior Scale, Second Edition, Socialization domain standard score; Autism severity represented by the Autism Diagnostic Observation Schedule calibrated severity score; Motor skills represented by the Developmental Coordination Disorder Questionnaire total score.

^*p < .05.

^**p < .01.

^***p < .001.

Data Availability Statement

The data that support the findings of this study are available from the NIMH Data Archive and SFARI Base. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the NIMH Data Archive and SFARI Base with permission.