2-Year-Olds with Autism Spectrum Disorder Direct a Smaller Proportion of Vocalizations Toward Others

Background: Socially-directed vocalizations such as cries, laughter, and babbling are early infant communicative signals that, when atypical or diminished, could indicate risk for autism spectrum disorder (ASD). Importantly, vocalizations can be socially directed before words are acquired, and may carry information about social development across a variety of populations, from typically developing to language-delayed. Despite the clinical relevance of socially directed vocalization in young children with ASD, only a small number of studies have quantitated rates of socially directed vocalizations in infants at high familial risk (HR) for ASD. Additionally, relationships between quality of early vocalizations and social directedness of early vocalizations have not been systematically studied. This relationship is important to understand in HR populations given evidence that toddlers with ASD produce atypical rates of speech-like versus non-speech vocalizations, relative to non-ASD peers.

Objectives: To test whether the proportion of socially directed vocalizations differs in HR toddlers with ASD vs. without ASD, and measure whether rates of social directedness differ between groups separately by vocalization type (speech-like and non-speech).

Methods: Participants were 10 LR and 20 HR (10 HR-ASD; 10 HR- [non-ASD]) 2-year olds from the Infant Brain Imaging Study (IBIS). Groups were matched on Mullen Visual Reception score, maternal education, and sex ratio. Audio-visual recordings of the Communication and Symbolic Behavior Scales (CSBS) administration were available for all participants. Using ELAN, two trained raters segmented child vocalizations as (1) speech-like, (2) non-speech, and (3) vegetative (excluded from analysis). Discrepancies in segmentation and annotation were resolved through consensus. Child vocalizations were further classified as socially directed or non-directed on the basis of orienting, use of gesture, making a request, or responding to a question, by two raters with discrepancies resolved by a third rater. Inter-rater reliability was acceptable for ratings of speech-like, non-speech, and vegetative vocalizations (Cohen’s κ=0.74) and directed versus non-directed vocalizations (κ=0.65). Proportions of directed vocalizations out of the total number of vocalizations were treated as dependent variables in linear regressions with diagnosis as a dummy-coded factor, with ASD as the reference group.

Results: There were no significant differences between ASD, LR and HR- for recording duration, vocalization rate, or proportion of speech-like to non-speech vocalization. The ASD group produced a significantly smaller proportion of socially directed vocalizations (M=0.40, SD=0.14) than the LR group (M=0.53, SD=0.12, p=0.04, Cohen’s d=1.0), and a marginally smaller proportion than the HR- group (M=0.52, SD=0.15, p=0.05, d=0.83). The ASD group also directed a significantly lower proportion of non-speech vocalizations (M=0.28, SD=0.09) than LR (M=0.49, SD =0.20, p=0.007, d=1.35), but not HR- (M=0.36, SD=1.7, p=0.25). The ASD group trended toward a smaller proportion of directed speech-like vocalizations (M=0.42, SD=0.15) compared to LR (M=0.54, SD=0.10, p=0.07, d=0.94) and HR- (M=0.54, SD=0.15, p=0.07, d=0.8).

Conclusions: This study provides preliminary evidence that when 2-year-olds with ASD vocalize, they direct fewer of their vocalizations toward others than typical peers. With additional video annotation (ongoing), we aim to increase our sample size and test interactions between diagnostic group, vocalization type, and manner of social directedness.