Can Statistical Learning Predict Children’s Language Skills?

For many of us, our first word was either “dada” or “mama.” A concept known as “statistical learning” likely played a role in helping us reach that milestone. 

Statistical learning, or SL for short, is a foundational theory of language development used to explain the process by which a person gains the understanding to detect, extract and predict word patterns by taking in sound sequences and other forms of information. 

The theory was first explained decades ago, when researchers found that infants were implicitly able to pick up on words and sound patterns by passively listening to people speak, explains Zhenghan Qi, assistant professor for the Department of Communication Sciences and Disorders and the Department of Psychology. 

“Imagine you’re a 10-month-old baby just sitting there playing while you continually hear sound sequences,” Qi says. “After a few minutes, some words just pop out from that sequence because they always occur together, like ‘ba·by,’ ‘ba·by,’ ‘ba·by,’ ‘dia·per,’ ‘dia·per,’ ‘dia·per.’ These syllables, they always occur together, so your own brain is more attuned to that combination.” 

This fall, Qi will be investigating the role statistical learning plays in informing language and literacy development in school-aged children with and without autism as part of a longitudinal study supported by the National Institute on Deafness and Communication Disorders. The university is leading the study in partnership with three other universities.

“Our study has three major goals, Qi says. “In six-year-old autistic children and their non-autistic peers, by using miniature artificial languages: One, we hope to test whether children’s statistical learning abilities can predict their long-term improvement of language and literacy skills in school. Two, we hope to decipher how children automatically pick up patterns from speech and prints in their environment by monitoring the oxygenated blood flow in their brains during learning. Lastly, we hope to prove that children’s learning in the lab is reflecting the language patterns they have learned over the years from their native language.” 

Qi notes that previous studies have shown autistic learners to be a heterogenous group, each with their own set of strengths and challenges. 

“Even with social communication challenges, some of them reach adequate or even superior language skills,” Qi says. “But others seem to struggle more.”

About 30% of those with autism have severe cases, Qi explains. And that population tends to be the one that has profound language impairments. 

What role could statistical learning play in understanding autistic children’s language development long term?

Leave a Reply

Your email address will not be published. Required fields are marked *

Skip to content