← StudyX Devlog

Do Autistic People Communicate Well With Each Other? A 72-Person Telephone Game

Hiroki ObaraHiroki Obara · DeveloperBrain MondoBrainDojo

Brain Mondo · 11 of 23

Do Autistic People Communicate Well With Each Other? A 72-Person Telephone Game

When autistic people played a telephone game with each other, the story was passed on as accurately as when non-autistic people played with each other. The amount passed on dropped in groups that mixed autistic and non-autistic people. That is the result of a study by a team centred on the University of Edinburgh, published in the journal Autism in 2020.

In a Brain Mondo short video, I introduced this study in about 40 seconds. In this post I write about the details that did not fit in the video, the results of a large replication that came later, and what the study does and does not tell us.

Video: Autism facts: The telephone game (Japanese)

The video is in Japanese.

What the study set out to test

In clinical diagnostic criteria, autism is defined by difficulties in social communication. Applied directly, that definition predicts that autistic people find it hard to share information with others, and that two autistic people would find it harder still.

There is another view. It holds that communication breakdowns are not located only in the autistic person, but happen in both directions between autistic and non-autistic people. The paper introduces this as the “double empathy problem”. It notes several studies showing that non-autistic people also struggle to read autistic people’s mental states and facial expressions.

If this view is right, breakdowns should happen when the other person is a different kind of communicator, and might disappear between autistic people. Yet until then, no experiment had directly tested whether interactions between autistic people go well. Crompton and colleagues, who led the study, used a telephone game to find out.

What the experiment was

The participants were 72 adult volunteers. Thirty-six were autistic, and 33 of them had a clinical diagnosis. The other 3 were self-identified, and scored at a level consistent with a diagnosis on a questionnaire measuring autistic traits. The other 36 were non-autistic, with low autistic traits confirmed by questionnaire. All spoke English to a native level, IQ scores were within the typical range in every group, and the groups were matched on age, gender, years of education and IQ. The mean age was around 37.

The 72 people were split into nine chains of eight. Three chains were all autistic, three were all non-autistic, and three were mixed, with four autistic and four non-autistic people. In the mixed chains, autistic and non-autistic people alternated, and the first person was always non-autistic.

The story was about a bear on a surreal adventure, made up of 30 details. It was designed to be hard to predict and to contain nothing inherently social. On readability measures it was an easy text, at about the level for ages 9 to 10.

The procedure went like this. A researcher read the story to the first person and left the room. The second person came in, and the first person told them the story. The first person left, the third came in, and the second told the third. This went on to the eighth person, who finally recounted the story aloud alone. Participants waited in separate rooms for their turn so they could not hear the others. Everything was video recorded, and the team counted how many details each person passed on to the next.

Result: autistic chains did as well as non-autistic chains

Averaged by chain type, the number of details each person passed on, out of 30, was 12.40 in the non-autistic chains, 12.96 in the autistic chains and 8.92 in the mixed chains. There was no difference between the autistic and non-autistic chains, and both passed on more than the mixed chains.

As in any telephone game, the amount passed on fell further down the chain in every group. When the team analysed how fast it fell, the autistic chains fell at the same rate as the non-autistic chains, and only the mixed chains fell faster.

There is one detail worth noting. In the mixed chains, the first person remembered less to begin with than in the other chains. A weak start puts everyone after them at a disadvantage. So the team recalculated, taking the first person in each chain as 100% and asking what percentage of that each later person received. Even then, the autistic and non-autistic chains did not differ, and the mixed chains were lower than both. The weak start alone does not explain it.

Rapport was lower in mixed chains too

After the game, participants rated how well they had got on with the person who told them the story and the person they told it to. They gave scores out of 100 on five dimensions: ease, enjoyment, success, friendliness and awkwardness, and these were added together.

Rapport with the person who had told them the story was clearly lower in the mixed chains than in the non-autistic chains. It was also lower in the mixed chains than in the autistic chains. Between the autistic and non-autistic chains, no difference was found.

The authors conclude that autistic people have the skills to share information well with one another and to build good rapport, and that the problems were limited to interactions between autistic and non-autistic people. They describe autistic people as different without being deficient.

In 2020 the same team also reported an experiment in which people talked in pairs. Seventy-two participants were paired as two autistic people, two non-autistic people or one of each, and rated their rapport after each of three conversations. Rapport was highest in the non-autistic pairs, and autistic pairs had higher rapport than mixed pairs. When 80 observers watched recordings of the conversations, they rated mixed pairs lower than the other two, and rated autistic pairs higher than non-autistic pairs. This pattern was the same whether the observer was autistic or not.

Then, in 2025, a large replication of the telephone game study appeared in Nature Human Behaviour. It had 311 participants and was run as a Registered Report, a format in which the study plan is peer reviewed before the data are collected. The result differed somewhat from 2020. There was no difference in how much information was passed on between single-neurotype chains and mixed chains. Rapport was higher in the non-autistic chains, and telling participants about the diagnostic status of their chain improved rapport.

The authors say this result also challenges the assumption that autistic communication is deficient, while acknowledging that it contrasts with the 2020 findings. They suggest the greater variety of participants and differences in method may explain it.

So neither study produced results that support the assumption that autistic communication is deficient. The finding that mixing reduces how much gets passed on, however, was not repeated in the larger replication.

Limits and common misreadings

The authors of the 2020 paper list several limitations themselves.

  • The sample was not large. It was big enough to detect a moderate effect, but they wrote that replications were needed. The 2025 study above is that replication.
  • Participants knew what kind of chain they were in. Knowing whether the other person was autistic may have affected how they behaved.
  • Gender was unbalanced. Most participants were women, and the gender mix differed between chains.
  • The range of participants was limited. All of them communicated verbally, IQ scores were in the typical range, and many of the autistic participants were diagnosed as adults. Whether the same result holds for autistic people with intellectual disability is not known.
  • The mixed chains always started with a non-autistic person. Whether starting with an autistic person would change the result is left as an open question in the paper.

The common misreading is to take this as “autistic people have no difficulties”. What the study showed is that breakdowns are easier to explain as something that happens in the pairing between people than as a trait of the autistic person alone. The difficulties do not go away. What changes is where we see them happening.

Using it in daily life

Here are some ways to apply this study to daily life.

  • When a conversation does not land, do not blame only yourself. This is how the video ends. Breakdowns often happen between two people. There is a view that this applies equally to autistic and non-autistic people.
  • When something is not getting across, change the form. Rather than repeating the same explanation, you can change the form: write it down, break it into steps, or ask the other person to say the main point back to you.
  • Value time with people who communicate like you. The paper points to the good rapport between autistic people as support for more autistic peer support. If talking with someone on your wavelength feels easier, there is a real reason for that, and nothing to feel bad about.

Brain Mondo is a series of daily short videos about the brain, each based on a research paper. BrainDojo is an app for training calculation, memory, attention and more in a few minutes a day, and the once-a-day Brain Dojo Check gives you scores for processing speed, inhibitory control and short-term memory. You can use it to watch your own ups and downs rather than to compare yourself with anyone else.

References

Download on the App Store Get it on Google Play

BrainDojoBuild healthy brain habits