It’s More Than Sounds – How ASD Affects Word Choice and Speech Rate


Even without words, young children communicate with parents, family members, and peers using coos, babbles, cries and fusses, and emerging research suggests that children with Autism Spectrum Disorder (ASD)** may have differences in these vocalizations. As children grow older and begin to develop more complex language, differences in spoken language and social conversation begin to become evident to parents and clinicians. Language differences can difficulties in speech production, lack of reciprocity in a conversation, difficulty understanding humor and sarcasm, and conversational quirks like atypical inflections, unusual word choice, or repetitive speech.

With these language differences and difficulties in mind, a team of scientists from the Quantitative Linguistics Lab (QL2) at the Center for Autism Research (CAR) at the Children’s Hospital of Philadelphia (CHOP), led by Julia Parish-Morris, PhD, are examining speech samples from hundreds of children with the goal of determining how verbal interactions change over developmental time, how these interactions differ in individuals with autism, and what kinds of interventions can help individuals reach their maximum potential. “We aim to collect many different types of language samples, in a variety of contexts, so that we can detect the features that are intrinsic to social communication in autism,” says Dr. Parish-Morris.

By focusing on how autistic individuals use language, as well as the acoustic properties that make up their vocal signature, Dr. Parish-Morris and her team’s research has found variations in both the rate and content of speech in toddlers and school-aged children. Conducting their research in a more natural environment (unstructured “get-to-know-you” conversations with a non-expert), the QL2 studies have shown that as children grow older, they tend to speak faster, produce more words, and engage in discussion about social topics, whether or not they have an ASD diagnosis. In addition, children with ASD often differ from typical controls in their choice of words. For example, research from the QL2 team found that autistic children tend to use fewer personal pronouns when speaking than children without autism (study lead: Amber Song, UPenn student). This finding became more evident as ASD symptoms increased, with children with more severe symptoms using fewer personal pronouns than children with less severe symptoms. “Our research suggests that personal pronouns might be a useful way to monitor how much value and attention a child places on social topics, which may help us track response to social communication interventions.

Some may not think how frequently “I” or “me” is used in conversation would make a noticeable difference to a non-professional ear, however, according to study results from the QL2 team, it took individuals without clinical training just five minutes to identify social communication challenges in children with ASD during a “get-to-know-you” natural conversation. This isn’t only research to show non-experts can “hear” ASD. Using cutting edge machine learning analysis, a QL2 research team led by postdoctoral fellow Michael Hauser, PhD, found that specific language features common among children and teens with ASD can be used to predict ASD diagnosis, even when the conversation is conducted by a non-expert. In children, these features included pronoun use, pause duration, and “friend”-category words. For teens, the contextual use of language and frequent negative word choice were found to be markers. “Developmental sensitivity in machine learning is critical for the future of digital phenotyping; this new study suggests that machine learning approaches can benefit from domain knowledge about children vs. adolescents, while providing new insights into critical characteristics that separate ASD conversations from typical conversations in each time period,” explains Dr. Parish-Morris.

The real world implications of these language studies reflect what families already know about the impact of ASD – autistic children often struggle making and maintaining friendships. This can lead to bullying and social isolation, difficulty with romantic relationships and even trouble in the workplace. As with most ASD interventions, early intervention to support social language is key. In a recent QL2 study led by doctoral candidate Emily Ferguson, preschoolers placed in an inclusive early intervention classroom heard more peer talk and an equivalent amount of teacher speech compared to toddlers in mixed-disability or autism-only settings. “With practice building skills and supportive, appropriately inclusive settings, children with ASD can increase their comfort and confidence communicating in social situations,” says Dr. Parish-Morris.

** Please note, CAR uses both person-first and identity- first language. Recent research has demonstrated that “autistic” is the preferred term in the autism community, with many others preferring “on the autism spectrum” or “person with autism” (Kenny et al., 2016, "Which terms should be used to describe autism? Perspectives from the UK autism community" Autism).