Early Vocal Development in Infants At Risk of Autism: Prosodic Development, Social Interaction, and Outcome

Background:  Although deficits in receptive and expressive prosody are accepted as key symptoms of autism, the origin and the nature of their progression remain elusive. Typically developing infants transition from purely reflexive to more complex intonation patterns as they gain volitional control. An infant’s prosody is further shaped by engagement with the social world, particularly in response to the exaggerated prosody of a caregiver attempting to attract infant attention. Prosody thus becomes an important index of social attunement between infant and caregiver as infants respond to and learn from their caregiver’s voice. It is possible that specific developmental problems in social interaction and speech displayed by a child with autism at age two may be observable much earlier in atypical development of prosodic exchanges in infancy. 

Objectives:  The goal of this study is to explore the potential of prospective longitudinal measures of vocal behavior to inform clinical diagnosis by correlating prosodic development in the first two years of life with standard measures used for clinical assessment of autism. We test the hypothesis that abnormal early development of intonational interactions is predictive of later outcome.

Methods:  As part of an ongoing pilot project now funded through an NIH Autism Center of Excellence, we recruited 4 low-risk infants with no history of autism and 4 high-risk infants with older siblings diagnosed with autism. We collected high-quality day-long audio recordings of each infant in their home environment at monthly intervals from 2 months onward using a miniature digital audio recording device (LENA Foundation) worn in the child's clothing. We then collected a battery of clinical assessment  measures from each child at two years of age. At each time point, we extracted sequences of utterances containing infant-caregiver interactions and calculated three measures of prosodic development: the fundamental frequency contour, utterance duration, and relative timing between utterances. Using Functional Data Analysis to align the resulting densely sampled longitudinal profiles across individuals, we were able to quantify developmental changes in all of our measures. We were then able to compare differences in developmental trajectories with differences in clinical outcome measures.

Results:  Our final sample consisted of 4 typically developing children (TD), 2 children diagnosed with a language delay (LD), and 2 children diagnosed with broader autism phenotype (BAP). Group differences in fundamental frequency contour, utterance duration, and relative timing predicted categorization into TD and non-TD groups as determined by clinicians. Group-level categorizations based on prosodic measures were consistent with categorizations based on the ADOS summary scores. However, prosodic measures were not sufficient on their own to distinguish between LD and BAP subgroups. 

Conclusions:  Preliminary results suggest that typically developing and developmentally delayed infants may differ according to acoustic measures of atypical prosodic development. However, these vocal measures alone appear insufficient to discriminate between children with language delay and children with broader autism phenotype.