Research is now tracking early infant vocal communication using all-day recordings evaluated with human coding and in limited cases with automated analysis at very large scales. Results reveal unexpected features of development, illustrating foundations of language. Much of this work conducted at the University of Memphis (UM) is founded in an infrastructural theory of language development that has produced notable successes by supplying a widely-used stage model of vocal development and by providing the basis for determining clinically useful markers, especially associated with the onset and consolidation of canonical babbling (CB), a key foundation for spoken language. The UM empirical work within the infrastructural framework, based primarily on human-coding of longitudinal all-day recordings, has also begun to produce a new account of vocal foundations and vocal interaction both in the 1st months after full-term birth and in infants born prematurely by >2 months and still in the hospital. This work points toward possible indicators of risk for disorders in the first months of life. Further development of the new account holds promise for illuminating the development of vocal language and supplying predictors for conditions such as autism spectrum disorders (ASD). The proposed research will expand the UM effort to much larger scale and address prediction of language development and of ASD through an ASD sibling study. The work will combine forces between the UM and Emory University's Marcus Autism Center (MAC), which both independently and through its Autism Center of Excellence (ACE) Program Project from NIH, is in the process of collecting a massive set of longitudinal all-day recordings of infants 1) showing typical development (TD), 2) at risk for ASD because they have an older sibling with ASD, and 3) at risk for language delay (LD) due to low socioeconomic status (SES). The MAC/ACE research includes analysis of the recordings with automated acoustic methods, but does not include funding for human coding of the massive recording set. Extensive human coding is in fact crucial for the ACE project, because without it, there is insufficient gold-standard data in terms of which to advance the accuracy and relevance of automated methods for developmental prediction and for detection of developmental communication disorders. Furthermore, many aspects of vocal patterning and interaction are as yet simply not accessible by automated means-human coding is the only workable option. The proposed research will combine efforts of UM and MAC, with 1) MAC supplying longitudinal all-day recordings across the 1st year of life to UM from the 3 groups, 2) UM performing coding and human-assisted acoustic analysis on representatively sampled segments from the recordings, and 3) both sites working together to predict infant outcomes at 24 months based on the human coding/analysis and based on the automated analysis that will result from supervised machine learning made possible by the human coding/analysis. The work targets improving prediction of language development and early identification of ASD.
This work will provide scientific foundations on vocal development and language in three groups of children, including ASD siblings, infants of low SES, and controls (low risk). The work will lay new foundations for automated large-scale monitoring of infant vocalizations.
|Yoo, Hyunjoo; Bowman, Dale A; Oller, D Kimbrough (2018) The Origin of Protoconversation: An Examination of Caregiver Responses to Cry and Speech-Like Vocalizations. Front Psychol 9:1510|