Stanford study finds children’s speech patterns may predict mental health risks

Research at Stanford University uncovers how analysing children’s language when discussing stress can help identify those at higher risk of depression and anxiety, potentially enabling earlier intervention.

Researchers at Stanford University say the way children describe painful or stressful experiences may help identify those most at risk of later depression or anxiety, offering a possible route to earlier intervention. The study suggests that a child’s speaking style may be more revealing than the events themselves, because it can show how the child organises and interprets what happened. Stanford has also been working on other early-warning signs of mental ill health in children, including sleep disturbance and brain-based markers of anxiety.

In the new work, more than 200 children aged 9 to 13 were asked to talk about stressful episodes in their lives. The researchers then used artificial intelligence tools to analyse both the language and the structure of the children’s responses. According to the study, certain speech patterns were linked to a higher likelihood of later internalising disorders, the term used for conditions such as anxiety and depression.

Among the strongest signals were a more categorical way of speaking, with words such as “always” and “never”, and a greater reliance on prepositions, which the researchers said appeared to reflect a more complex narrative structure. By contrast, children who used more third-person pronouns such as “he”, “she” and “they”, and who described other people and their surroundings rather than focusing only on themselves, tended to show a lower risk profile.

The content of the stories still mattered, though less strongly than the style of delivery. Accounts centred on conflict, fear and distress, including references to fighting or being attacked, were associated with greater risk. Stories about hobbies, games or ordinary activities pointed in the opposite direction. Chase Antonacci, the study’s lead author at Stanford, said the findings offer “a robust proof of concept” for developing tools that could spot risk markers before a formal diagnosis is made. The researchers say the challenge now is to turn that insight into practical, scalable screening methods for clinicians and schools.