A neuroimaging study in Hong Kong found that a brain region involved in language and social processing responded less strongly to emotionally meaningful sentences in autistic children than in non-autistic children. In non-autistic children, the left superior temporal sulcus responded more strongly to sentences describing people in emotional situations than to sentences about objects, but this tendency was reduced in autistic children. The paper was published in NeuroImage: Clinical.
Autistic children often differ from their non-autistic peers in social interaction, emotional processing, and communication. However, the neural differences underlying these patterns are not yet fully understood. While researchers have extensively investigated how autistic individuals process language and recognize emotions in facial expressions and tone of voice, much less attention has been paid to how they process the social and emotional meaning of words and sentences themselves.
Everyday language is filled with expressions describing relationships, social interactions, and feelings. Understanding words such as “friend,” “hug,” or “grief” involves more than recognizing their grammatical role and literal meaning; it also involves processing the social and emotional experiences they represent. Research suggests that autistic individuals often perform similarly to non-autistic individuals when explicitly asked to judge whether a word is positive or negative, but differ more consistently in how automatically they process emotional meaning.
Neuroscientists have identified parts of the brain’s superior temporal cortex, particularly the left superior temporal gyrus and the neighboring superior temporal sulcus, as important for processing both language and social information. A previous study of non-autistic adults found that these regions responded more strongly to sentences describing people in emotional situations than to sentences describing inanimate objects, a pattern called a social-emotional bias.
Study author Hanlin Wu of the Chinese University of Hong Kong and colleagues hypothesized that this social-emotional bias would already be present in childhood but would be weaker in autistic children than in their non-autistic peers. To test this, they used functional magnetic resonance imaging (fMRI) to examine how the brains of autistic and non-autistic children responded to spoken Cantonese sentences describing either people in emotional situations or objects.
The researchers recruited 78 right-handed children, all native Cantonese speakers attending mainstream schools in Hong Kong. Twenty children were excluded because they moved their heads too much during scanning, which can distort brain imaging data. The final sample included 58 children between the ages of about 8 and 12 (average age 10), of whom 37 were autistic and 21 were not, and 48 were boys.
The children completed assessments of autism symptoms (the Autism Diagnostic Observation Schedule, Second Edition), language comprehension (the Hong Kong Cantonese Oral Language Assessment Scale), and non-verbal intelligence (the Test of Nonverbal Intelligence, Fourth Edition). Parents also rated their children’s proficiency in Cantonese and English.
During scanning, the children simply listened to sentences without responding. Social-emotional sentences described people in emotional situations, half positive and half negative (e.g., “My older sister laughed at the joke” or “Chi-Keung feels he is no good”), while object sentences described non-human objects, events, or concepts (e.g., “Banana peels are yellow” or “The sun rises in the east”). The children also heard sentences played backward, which served as a non-language sound for comparison.
All sentences were eight syllables long and were matched on word frequency and grammatical complexity. They were generated with text-to-speech software using a single synthetic voice, so the two types of sentences differed in their meaning but not in tone of voice or other vocal cues.
Across both groups, the left superior temporal sulcus responded more strongly to social-emotional sentences than to object sentences, indicating that the social-emotional bias seen in adults is already present in children. When the groups were analyzed separately, non-autistic children showed a robust bias, while in autistic children it was weaker and did not reach statistical significance.
In one of four regions the researchers had selected in advance based on the earlier adult study, the middle portion of the left superior temporal sulcus, autistic children showed a weaker social-emotional bias than non-autistic children, a moderate-to-large difference. The groups did not differ in brain responses to backward speech or to object sentences, suggesting that general hearing and language processing were similar.
The difference appeared to be driven mainly by negative sentences, for which the gap between groups was largest. However, the statistical test comparing positive and negative sentences was not significant, so the authors treat this pattern as suggestive.
The group difference was not explained by children’s language ability. It could not be statistically separated from autism symptom level, because diagnosis and symptom scores overlapped so heavily, and within the autistic group, children with more severe symptoms did not show weaker brain responses.
“This reduced activity in the left STS [superior temporal sulcus] in response to social-emotional semantics may contribute to autistic children’s social communication challenges,” study authors concluded.
The study contributes to the scientific understanding of how autistic children’s brains process social and emotional meaning in language. However, the number of study participants was modest by current neuroimaging standards, and the non-autistic group was particularly small. Because of this, the study could reliably detect only relatively large differences between groups, and when the researchers compared the groups across the entire brain rather than in the preselected regions, no difference survived statistical correction.
Head movement led to the exclusion of 31% of autistic children compared with 13% of non-autistic children, so the findings may not extend to autistic children who have difficulty lying still in a scanner. Most participants were boys, and most autistic participants had at least one co-occurring condition, most commonly ADHD. Because the social-emotional sentences differed from the object sentences in both social content and emotional tone, the study also cannot separate these two factors.
The paper, “Autistic children show reduced activity in left superior temporal cortex in response to social-emotional meaning in language,” was authored by Hanlin Wu, Matthew Ganquan Shi, Phoebe Wai-Yu Chan, Peilun Song, Siu-Fung Lau, Gary Yu-Hin Lam, Xin Kang, Patrick C.M. Wong, and Xiujuan Geng.