The Same Skill Shows Up Differently in Conversation

What the research found

Theory of mind is the everyday ability to work out what another person is thinking or feeling. It is often described as something autistic people lack. This study, published in the journal Autism in 2022, tested that idea in a more demanding setting than usual: real conversation between two children.

The researchers recorded 50 pairs of children and adolescents talking freely for a few minutes. In total 100 young people took part, aged 8 to 16 (average age 13.3). Thirty-two were autistic and 68 were typically developing. The pairs were mixed: 25 were two typically developing children, 18 paired an autistic child with a typically developing one, and 7 were two autistic children. All the participants had an IQ of at least 80 and were verbally fluent, so the findings speak to that group and not to autistic children with intellectual disability or limited speech.

The team built a new rating scale to score theory of mind as it actually appears in talk. It had two halves. The Positive scale counted moments when a child referred explicitly to their partner’s thoughts or feelings. The Negative scale counted violations of ordinary conversational norms: the kind of slips that suggest a child has lost track of what the other person knows, wants, or expects. The scoring was reliable between independent raters.

The two halves told different stories. On the Positive scale, the autistic and typically developing children were indistinguishable (t = 0.02, p = 0.98). They spoke about minds, their own and each other’s, at essentially the same rate. On the Negative scale, the autistic children scored significantly higher, meaning more norm violations (t = 3.25, p = 0.003). So the difference was not that autistic children failed to engage with other minds. It was concentrated in one specific place: the smooth, moment-to-moment management of a conversation.

The researchers also gave each child two established laboratory measures of theory of mind. One, a visual-affective task, asks children to read emotions from faces and voices. The other, a spontaneous task using moving shapes, measures how readily a child attributes intentions unprompted. Both predicted fewer conversational norm violations. Better face-and-voice reading was linked to fewer violations (B = -0.24, p = 0.01), as was more spontaneous attribution of intention (B = -0.30, p = 0.02). Notably, the autistic and typically developing groups differed on the face-and-voice task (autistic children scored lower, p < 0.001) but did not differ at all on the spontaneous task (p = 0.83). The picture is uneven, not a blanket deficit.

What it means for you and your child

The useful takeaway is a shift in framing. A child who struggles in conversation is not necessarily a child who cannot understand other minds. In this sample, autistic children talked about thoughts and feelings just as much as their peers. Where they differed was in the rapid, largely unconscious upkeep of a conversation: tracking what the other person already knows, sensing when to hand over the turn, noticing when a comment has landed awkwardly. Those are describable, teachable behaviours rather than evidence of a missing capacity.

That distinction matters for how support is aimed. If the underlying ability were simply absent, there would be little to work with. Because the study suggests the ability is present but applied unevenly under the pressure of live conversation, the sensible focus is the conversation itself: making the implicit rules explicit, rehearsing turn-taking, and giving a child room to check what a partner is thinking rather than assume it. This is closer to coaching a skill than remediating a deficit, and it fits the kind of patient, one-to-one attention that suits many autistic learners.

It is worth being honest about what this single study can and cannot show. The two lab measures predicted fewer norm violations, but this is a correlation. It does not prove that training face-reading or intention-reading would cause conversations to improve. The design is a snapshot, not a trial, and cause and effect cannot be separated from it. Treat the finding as a well-supported clue about where difficulty concentrates, not as a prescription.

The sample also sets real limits. Only 32 of the 100 children were autistic, which the authors note likely reduced the variation they could measure. The autistic group was 84% White, the sample was mostly boys, and every child was verbally fluent with an IQ of 80 or above. None of this weakens the core observation, but it does mean the findings should not be stretched to autistic girls, to minimally verbal children, or to those with intellectual disability, whose profiles may look quite different. The coding also rested on audio and transcripts, so facial expression and eye contact went unmeasured.

For a parent, the practical message is measured but genuinely encouraging. When your child speaks about what others think and feel, that is theory of mind in action, and it may be steadier than a difficult conversation makes it appear. The place to concentrate attention is the machinery of conversation itself, and that is precisely the sort of thing that can be practised, one exchange at a time.