Language Assessment in Minimally Verbal Children with ASD

Background: Language and communication impairments are common in children with ASD and are particularly salient in minimally verbal (MV) children. There is heterogeneity in language ability even among MV children, and there is still a lack of consensus in the way MV status is defined. As MV status is inherently shaped by the language assessments used with these children, understanding the methods and measures by which language is assessed with this population is one step toward constructing a clear definition of what it means for a child to be “minimally verbal.” Additionally, understanding concurrent predictors of language in this population may help inform both research and clinical practice.

Objectives: We present data exploring associations between receptive language (RL) and expressive language (EL) scores from different assessments and examine concurrent predictors of language skills in minimally verbal children.

Methods: Children (n=294) participated in multisite studies targeting social communication in MV preschool and school-aged children. The language measures used in the present study were collected as baseline measures for the original studies. The assessments analyzed were: Mullen Scales of Early Learning, Reynell Developmental Language Scales, Peabody Picture Vocabulary Test, Test of Early Language Development, Vineland Adaptive Behavior Scales and the MacArthur-Bates Communicative Development Inventories: Words and Gestures, and a naturalistic language sample. Concurrent predictors of language ability (using RL and EL composites) analyzed included child’s imitation ability, response and initiation of joint attention (JA), symbolic play, non-verbal cognitive ability, and chronological age.

Results: Using Spearman’s partial rank correlations controlling for age we found the correlations between EL raw scores to be moderate to strong and all significant at the 0.01, while RL raw scores were strongly correlated at the 0.01 level. We conducted discrepancy analyses using Friedman’s test and found that the age equivalents across EL and RL measures were significantly different. Regression models were used to predict concurrent expressive and receptive language. The results indicate that imitation (t=2.26, p=.025) and initiation of JA (t=3.94, p< .001) were significant predictors of concurrent RL (R²=.20). Additionally, imitation (t=2.51, p=.013), initiation of JA (t=2.38, p=.018), and nonverbal cognitive ability (t=3.35, p<.001) predicted EL (R²=.25).

Conclusions: Results indicate that there was strong convergent validity in raw scores for EL and RL measures, but the age equivalents produced were different. These findings demonstrate that there are different measures that can be used to effectively capture language abilities in MV children; however, researchers and practitioners should be cautious in using and interpreting the age equivalents given by these measures. Although results show JA, imitation, and nonverbal cognitive abilities are important skills for language, these only account for about 20-25% of variance. This suggests that language development is a complex process for MV children with ASD and needs further investigation.