Specific Language Impairment in ASD: Exploring Language Phenotypes Beyond Standardized Testing

Background: Although current diagnostic criteria for ASD do not include language impairment, children with autism present with a wide-range of language abilities and deficits.  Language can vary independently of IQ in autism (Kjelgaard & Tager-Flusberg, 2001), and  some children with ASD evidently meet criteria for specific language impairment (SLI), as they present with impaired language despite intact nonverbal cognitive abilities (Tager-Flusberg, 2006).  However, it is currently unknown how common the SLI phenotype is, and which linguistic markers differentiate those children from other language phenotypes in this population.  Therefore, spontaneous language samples are an ideal method for investigating more detailed and comprehensive linguistic profiles (Dunn et al., 1996).

Objectives: To assess standardized-test subtyping of children with ASD by analyzing children’s language samples for SLI-relevant linguistic markers.

Methods:  98 participants (M= 68.63 months, SD = 12.06) from a large-scale study of autism phenotypes were classified into 4 groups based on standardized test scores (DAS or MSEL for Nonverbal, a composite of EOWPVT, PPVT, MSEL, DAS for Verbal). Groups included: Low (nonverbal/verbal <70), Low Normal (nonverbal/verbal between 70-85), Possible SLI (nonverbal >70, verbal more than 15 points below nonverbal), and Normal (nonverbal/verbal >85).  Recordings from the ADOS were transcribed, targeting tasks that afforded unprompted, spontaneous language production.  49 children were eliminated from further analysis because of video recording errors (n=16) and/or insufficient spontaneous language to transcribe (n=30); only 3 children in the Low group presented with sufficient language to transcribe so this group was excluded from further analyses.  Therefore, 49 children were included in the final sample (M= 67.8 months, SD = 11.5), across 3 of the groups: Low Normal (n=6), Possible SLI (n=8), and Normal (n=35).  Language samples were analyzed for frequency of language types and tokens, as well as MLU, grammatical errors, and omissions.

Results:  One-way ANOVAs revealed significant effects of group for MLU, frequency of grammatical errors, total word and verb types, and frequency of article errors (Fs >3.5, ps <.04).  However, post-hoc Scheffe tests revealed that the Possible SLI group differed from the Normal Group only in frequency of total grammatical errors (p=.003) and frequency of article errors/omissions (p=.03).  Interestingly, the Low Normal and Normal groups differed significantly in total word types (p=.02), total verb types (p=.04), and frequency of article errors.  Surprisingly, significant group effects were not observed for error frequencies with pronouns, plurals, and tense/agreement, which are the grammatical structures most often impaired in SLI.

Conclusions:  Standardized test results pointed to a possible subgroup of SLI in a large ASD sample; however, grammatical analyses of their speech samples did not fully support this subgrouping.  Overall, frequency of grammatical errors was higher in the Possible SLI group, but specific error types did not consistently reflect an SLI designation.  Further analyses will scrutinize the grammatical error patterns for each child, exploring different subgroup assignments based on these language samples. Our results suggest that including spontaneous language samples is critical for capturing the full extent of language impairments as well as designing targeted interventions for children with ASD.