A New Approach for Assessing Social Communication in ASD: Elsa

Background: Natural language samples can be a useful tool in the assessment of expressive language in ASD (Barokova & Tager-Flusberg, 2018; Tager-Flusberg et al., 2009). Expressive language is also central to social communication, which makes language sampling a great candidate for the evaluation of deficits in this domain – part of the core defining features of ASD. Yet, no study has examined the feasibility of using sample-derived measures in place of or as a complement to standardized assessments of communicative skills in ASD. Furthermore, sample-derived measures have been proposed to be good candidates for valid and reliable clinical outcome measures (Abbeduto, 2017; Barokova & Tager-Flusberg, 2018)

Objectives: We aimed to validate the utility of the Eliciting Language Samples for Analysis (ELSA) protocol for individuals with ASD as a tool for assessing social communication. We focused on the associations between ELSA-derived expressive language measures and scores on standardized assessments of communication and socialization – Vineland and SCQ.

Methods: We collected ELSA samples from 40 participants with ASD across a wide range of age, language ability, and symptom severity (Table-1). ELSAs consisted of activities with developmentally appropriate toys, eating a snack, watching a short movie, and a conversation about participants’ interests (Barokova et al., 2017). All samples were coded by naïve coders in real-time using ELAN (Sloetjes & Wittenburg, 2008). Frequency of speech utterances and number of conversational turns per minute were extracted from the coded files. Participants’ ASD diagnosis was confirmed with the administration of the ADOS. Vineland and SCQ questionnaires were collected from participants’ parents.

Results: Frequency of utterances and number of conversational turns per minute were significantly correlated with the Vineland Communication Standard Score (for frequency: r_s (38) = .592, p < .001, Figure-1; and for turns: r_s (38) = .595, p < .001) and with the Vineland Standard Socialization Score (r_s (38) = .514, p = .001 and r_s(38) = .503, p = .001). Frequency and conversational turns were strongly correlated with the SCQ Communication Subscore (r_s (33) = .665, p < .001, and r_s (33) = .666, p < .001) and moderately correlated with the SCQ Reciprocal Social Interaction Subscore, as well (r_s (33) = .422, p = .011, and r_s (33) = .404, p = .016).

Conclusions: The ELSA-derived expressive language measures were positively correlated with gold-standard measures of social communication, which provides preliminary support for the use of language sampling as an alternative method of assessment in this domain. These findings have laid the foundation for research focusing on the use of language sampling as a measure of autism symptoms that go beyond expressive language ability alone. Future work should examine the convergent and divergent validity of the frequency of utterances and conversational turn measures, as well as identify expressive language proxies for restricted interests (e.g., number of topics discussed) and repetitive behaviors (e.g., amount of verbal imitation).