According to research using machine learning, some of the speech variations
linked to autism are universal across languages while others are
language-specific. The study, which was written up in the journal PLOS One,
used different groups of Cantonese and English speakers.
Speech prosody problems frequently accompany autism spectrum disorder
(ASD). Speech prosody refers to linguistic features, such as rhythm and
intonation, that enable us to convey meaning and emote with our words. A
person's communication and social skills may be hampered by unusual speech
prosody, for instance by the possibility of miscommunication with others or
misinterpretation of oneself. Uncertainty surrounds the cause of these
speech variations that are frequently seen in autistic persons.
Author of the study Joseph C. Y. Lau and his team sought to clarify this
issue by examining prosodic traits linked to autism in two typologically
different languages.
As a multilingual speech scientist who was raised abroad, I've always been
fascinated by how exposure to many cultures and languages shapes people. A
cross-cultural and cross-linguistic approach might provide us with a wealth
of unique insights when studying autism, according to Lau, a research
assistant professor at Northwestern University and part of Molly Losh's
Neurodevelopmental Disabilities Laboratory.
The majority of research in this area has focused on English-speaking
populations, however prosody differs between languages. There is evidence to
suggest that the prosodic traits linked to autism are also
language-specific. Lau and his colleagues looked at the characteristics of
speech prosody that consistently correlate with autism across languages and
those that do not.
Cantonese speakers from Hong Kong and native English speakers from the
United States participated in the research. 55 of the subjects in the
English-speaking group were autistic, and 39 were neurotypical. There were
24 neurotypical and 28 autistic participants in the Cantonese group. Each
participant was required to tell the tale of a picture book without any
words. Their speech was captured, written down, and then separated into
individual utterances for additional analysis.
From the story excerpts, the researchers utilized a computer software to
extract the speech intonation and rhythm. While intonation relates to
differences in voice pitch, rhythm refers to variations in time and volume
of speech. The researchers next attempted to distinguish between people with
autism spectrum disorders and those with usual development by using machine
learning, a method that makes use of computer systems to evaluate and
interpret data.
The results showed that speech rhythm could accurately distinguish between
neurotypical and autistic participants in both the English and Cantonese
samples. However, in the English sample, speech intonation could only
distinguish between autistic and neurotypical subjects. Additionally, when
the researchers looked at a combined dataset of Cantonese and English
speakers, they discovered that the only factor that accurately distinguished
autistic people from neurotypical participants was speech rhythm.
These findings suggest that the machine learning system might identify
between speakers who are autistic and those who are neurotypical based on
characteristics of speech rhythm. This, according to the authors, is
consistent with other studies that claim autistic persons exhibit distinct
stress patterns, speaking rates, and volume levels. Furthermore, the results
imply that these variations hold true across two different languages.
"We can see there are features that are strikingly common in autistic
individuals from different parts of the world; meanwhile, there are also
some other features of autism that are manifested differently, as shaped by
their language and culture," Lau told PsyPost. "We use an AI-based analytic
method to study features of autism across languages in an objective and
holistic way.
In our study, we discovered that speech rhythm, or the regularity of speech
patterns, exhibited such similarities, although intonation, or the
fluctuation of pitch when we talk, revealed variances across linguistic
boundaries. Finding shared characteristics may open a door to investigating
the complex molecular underpinnings of autism, which profoundly impact
language and behavior in a consistent manner among autistic individuals
worldwide. On the other hand, aspects that differ between cultures or
languages may represent characteristics of autism that are more easily
modified by experience, which may reflect prospective areas that might
benefit from clinical intervention.
Notably, intonation did not indicate an autism diagnosis in the Cantonese
group and only did so in the English sample. Cantonese is a tone language,
meaning that words may have their meanings altered by pitch, according to
the study's authors, who speculate that this may be the case. The authors
speculated that the widespread use of linguistic pitch in tone languages may
have a compensating effect that lessens the effects of intonational
disparities in ASD. This may indicate that speech therapies that concentrate
on pitch and intonation can help autistic persons who speak non-tonal
languages, while more study is required in this area.
According to Lau, "We feel the cross-linguistic similarity of rhythmic
patterns of autistic speech included in our work points to an interesting
follow-up field of enquiry: if rhythm is a potentially universal component
of ASD that is less changeable by experience than language." "Testing this
theory will need for many additional languages, including languages from
other language families all across the world, and a considerably bigger
sample size. We look forward to growing our research initiatives and
building global partnerships that would eventually enable such an
examination.
The researcher continued, "Although the direct benefits of this study to
the autism community may appear modest at this time, we do hope that, beyond
its theoretical implications, this machine learning study can inspire future
scientific and technological advancements that can provide more direct
benefits to the autism community, such as in the area of AI-assisted
healthcare.
The study, “Cross-linguistic patterns of speech prosodic differences in autism: A
machine learning study”, was authored by Joseph C. Y. Lau, Shivani Patel, Xin Kang, Kritika
Nayar, Gary E. Martin, Jason Choy, Patrick C. M. Wong, and Molly Losh.
