Voice as biomarker: Using vocalizations to screen for autism

Researchers across the country and around the globe are using a technology to explore what children’s vocalizations tell us about autism spectrum disorder (ASD) and other intellectual disabilities (IDs).

What this means, says D. Kimbrough Oller, PhD, professor and Plough Chair of Excellence at the School of Communication Sciences and Disorders, University of Memphis, Tennessee, is that “autism has characteristics that can be recognized by completely objective means in the realm of vocalizations.” He continues, “These characteristics can be determined by automated methods to designate a child as being at very high risk” for the condition.

Oller remarks that the whole area of studying automated recordings of children’s vocalizations is not just a method or technology. “It’s a movement.” His study appearing in 2010 in the Proceedings of the National Academy of Sciences¹ is largely credited with getting the ball rolling. He explains that the “revolutionary technology” he used in his study “makes it possible for us to look inside the home or school and get a very quantitative picture of how kids talk and how people talk to them.”

His study included 1486 daylong recordings from 232 children, in all more than 3 million vocalizations. He and his colleagues found that the vocalization patterns, particularly the ability to form syllables, were a significant predictor of future language development and could distinguish children with ASD from those who were developing typically and, to a lesser degree, from those with language delay.

The technology that Oller refers to is LENA, which stands for Language ENvironment Analysis, both a group of products and a research foundation in Boulder, Colorado.

As for the products, the LENA Digital Language Processor is a small, lightweight digital recording device that fits into a pocket of a specially designed children’s vest and records up to 16 hours of continuous, high-quality audio. It records all the sounds produced by the child wearing the device, as well as externally produced sounds and vocalizations made to the child by anyone or anything within a 4- to 6-foot radius. Software then analyzes ideally 3 days of recordings from the child’s home language environment to identify patterns in the child’s vocalizations that pertain to such features as duration, pitch, and rhythm. These vocal features can then be used to discriminate a child with ASD from a typically developing one.

Research conducted by the LENA Foundation claims that the LENA Automatic Autism Screen discriminates between children with ASD, those who have language delays, and those who are typically developing with about 89% accuracy,² which, says the foundation, compares favorably to other nonautomatic, widely used screening measures, such the Method Checklist for Autism in Toddlers (M-CHAT).

“Of the kids who have autism, we can reliably distinguish them from others on the basis of vocalizations in kids as young as 24 months of age. We think that can be done in kids much younger, too,” says Jeff Richards, LENA’s chief statistician.

Steve Hannon, president of the LENA Foundation, says of this autism screening method, “We have indications that it can be extended to younger children. There’s great opportunity to start using the acoustic environment, the language environment, more widely.”

In fact, LENA is being used in all kinds of ways, not just to study autism. It’s becoming a cornerstone of community-based and citywide initiatives to improve literacy by focusing on increasing conversation with children before they enter school. Such programs are based on the premise that some children are exposed to up to 30 million fewer words than other children by the time they reach 4 years of age, and that those 30 million words can make the difference between a child thriving or merely surviving in the school environment and beyond. Such programs include the 30 Million Words initiative in Chicago, Illinois; the Providence Talks project in Rhode Island; and the newly launched Talk With Me Baby program in Atlanta, Georgia.

The LENA technology is permitting all kinds of novel research. Anne S. Warlaumont, PhD, assistant professor of Cognitive and Information Sciences at the University of California, Merced, is now studying the vocal interactions between children and their caregivers. Critical to language development, she, along with Oller and the researchers at LENA, are studying how these “social feedback loops” differ for children with ASD from those who are developing typically.

What they are finding is that, across the board, adults are more likely to respond immediately to children when children’s vocalizations are speech related (eg, words, babbling, singing, etc), rather than non–speech related (eg, crying, laughing, burping). In turn, the children are more likely to create more vocalizations.

“With autistic children, we are finding that the feedback loop is less potent,” explains Warlaumont. She says, “Children with ASD vocalize less to begin with, so they experience fewer exchanges and have fewer opportunities to learn from them.” In addition, the caregiver responses are less differentiated between the speech-related utterances and the non–speech-related vocalizations.

She says that the hope down the road is that all this information can be used to create better learning tools and optimize language development for all children.

REFERENCES

1. Oller DK, Niyogi P, Gray S, et al. Automate vocal analysis of naturalistic recordings from children with autism, language delay, and typical development. Proc Natl Acad Sci U S A. 2010;107(30):13354-13359. Available at: www.pnas.org/content/107/30/13354.long. Accessed March 28, 2014.

2. Richards JA, Xu D, Gilkerson J; LENA Foundation. Development and performance of the LENA automatic autism screen. LTR-10-1. August 2010. Available at: www.lenafoundation.org/Research/TechnicalReports.aspx. Accessed March 28, 2014.

Ms Hack is a medical writer and editor in New Jersey. She has 25 years’ experience in the field and is a long-time contributor to both Contemporary Pediatrics and sister publication Contemporary OB/GYN. She has nothing to disclose in regard to affiliations with or financial interests in any organizations that may have an interest in any part of this article.