Marisha Speights

Marisha Speights noticed that speech and language disorders were under-diagnosed in communities of color. 

Not only did she do her doctoral research on the disparity, she made it her life’s work. 

Speights established a lab at Northwestern University that does pioneering research dedicated to advancing the early detection of communication disorders in young children. 

A speech scientist and speech-language pathologist, Speights has continually seen a difference in the diagnosis of speech and language issues. 

“In my practice and in my research, I would keep seeing a troubling pattern. There was a difference in results and resources for children,” the 20-year speech and language pathologist said. “I would see different things, even now.  

“I looked at the data again and again, and I would see that white children from higher economic areas would be identified (with a speech and language issue)  by a screener. Then in Head Start programs, where more kids are at risk and they are from lower socioeconomic environments, there are less kids identified with a problem. 

“This has been historically seen. We are at a critical point. How can we develop a more accurate measure that can be used across more demographic groups?” 

Speights established the Pediatric Speech Technologies and Acoustics Research lab at Northwestern

Marisha Speights (foreground) works with students at the Pediatric Speech Technologies and Acoustics Research lab at Northwestern University. Speights’ PedzSTAR lab is working on the disparity in youth being identified as having speech and language issues. (Provided photo)

in 2021. The Evanston lab focuses on developing computer-based technologies that accurately identify speech and language difficulties, particularly in historically under-served communities. Through research and partnerships, PedzSTAR aims to transform clinical practices and ensure equitable access to early intervention services. 

Access to services

The longtime speech pathologist said that increasing accuracy in language testing could get more children needed access to services. 

“We know the current measures had bias in how they were developed,” Speights said. “In screenings, the measures used to identify a problem don’t work for everyone.” 

She noted that because early speech and language studies were done predominantly with white children, variables are skewed. 

Speights said testing models don’t account for variation in speech and language in different races because of the ‘whiteness’ in testing designs and scoring. 

“They set up studies in the past and they had a measure, but who were they measuring? Whites from higher economic areas,” Speights said. “There were important details missing. I doubt we understand the full impact of that measure.” 

PedzSTAR is partnering with the Childcare Network of Evanston and local nonprofit Communication Health, Advocacy & Therapy to create the Communication Justice Program to bring speech-language services into under-served communities. 

“We think about how to bring access to early intervention services to community organizations, how to navigate insurance,” Speights said. “We take information from the schools on what some of the barriers are to reaching families. We are not just working on development of services. We are trying to make sure we are a community partner, that people see as a members of the community. 

“People can come into a preschool, get early intervention information and leave. Our core value is to make sure the community has the benefit of what we do to support services. When researchers go into area preschools, we want to make sure the research is grounded with many voices.” 

Speights said when evaluators develop a new measure for speech and language disorders, they often don’t talk to stakeholders to ensure that the measure has a “multi-voice perspective.” 

Closing the gap

She believes that the disparity gap is closing with the help of technology 

As principal investigator of the PedzSTAR lab, she and her team develop computational tools to classify speech disorders, refine automatic speech recognition models, and leverage machine learning for equitable early screening. 

“We need a universal measure for all children,” Speights said. 

She noted that artificial intelligence machines can be taught to detect when certain youth could be having speech difficulties. 

“We can use technology to look at the speech signal,” the Northwestern professor said. “As youth are developing, there is a natural progression, their first words, etc. From 1 year of age to 5 years of age, there is a normal path of development. Most kids do it rather naturally. If we can identify early difficulties in speech and language, we can address them in their early education years. 

“That is why we need to have accurate early detection.” 

She said work through AI thus far has been more than 90 percent accurate in identifying which children would have a speech disorder. 

“If we can automate into technology and have sound research behind it, it opens up the world,” Speights said. “If technology can help with early identification of potential speech problems, we can get those same children early intervention services.” 

The speech pathologist admitted that increasing the ability to accurately find youth with speech issues does not address the problem “downstream of providing the care.” 

“Maybe with the technology, youth get some interim care while they are in the waiting phase,” Speights said. “Once they are on the waiting list for Early Intervention services, it could be a long wait. Another side of the problem is addressing our thinking and how we pivot to provide services.” 

She said technology could also be used to support youth once speech and language issues have been identified. Speights envisions apps that could be used by caregivers and teachers to assist youth with language issues. 

“However, we still need to address the bottleneck in the pipeline,” the Northwestern professor said. “There is a lot to address. I hope we can automate things and it will increase the number of available pathologists.” 

Central to Speights’ research is the Speech Exemplar and Evaluation Database, a child speech study aimed at filling a data gap. Supported by the Bill and Melinda Gates Foundation and global collaborations, Speights aims to advance AI-powered, culturally responsive tools for early speech disorder detection and equitable access to intervention. 

“We want to make sure the data collected is representative of different cultural groups,” Speights said. “We want to make sure we have a diverse collection of speech samples. We are building concern of fairness for all.” 

kbeese@chronicleillinois.com