TTW Nest · College, work, and adult lifeNovember 2026
Home / Can autistic people communicate just as effectively?
Let's Read Together

Can autistic people communicate just as effectively?

Two studies about autistic communication, and what happened when the researchers tried again

Geometric low-poly illustration accompanying the Can autistic people communicate just as effectively? chapter

Two studies. One clever method. A result that changed when the researchers tried again. If you've never read a journal article before, this is a good place to start. If you've read hundreds, you may still find yourself arguing about the conclusions.

We'll read a 2020 study by Catherine Crompton and colleagues, then a larger 2025 follow-up by Crompton and a partly overlapping research team. Both are open access. The questions below let you make a prediction before looking at the evidence. You can use them on your own, with a friend, or in a class.

First article · 2020Autistic peer-to-peer information transfer is highly effective

Crompton, Ropar, Evans-Williams, Flynn & Fletcher-Watson. Autism, 24(7), 1704–1712.

Read the full article ↗DOI: 10.1177/1362361320919286
Follow-up · 2025Information transfer within and between autistic and non-autistic people

Crompton and colleagues. Nature Human Behaviour, 9, 1488–1500.

Read the full article ↗DOI: 10.1038/s41562-025-02163-z
Bring the worksheet to the table.

The companion packet includes predictions, a data table, comparison questions, an optional story-passing activity, and notes for facilitators.

Get the packet ↓

1. Before you read: make a prediction

Imagine eight people sitting in a row. The first person hears a story and tells it to the second. The second tells it to the third, and so on. Nobody can go back and ask the original storyteller for another explanation.

Now imagine different sets of eight adults: one set all autistic, one all non-autistic, and one alternating autistic and non-autistic people. Where do you predict the most details will survive? Why?

Make your prediction, then open this

The researchers thought mixed groups might lose more information because communication could become harder when people have different experiences and expectations. That's related to Damian Milton's double empathy problem: misunderstandings can be reciprocal, rather than the fault of one person. A hypothesis is a prediction to test, not a fact the study is obliged to confirm.

Write it down before reading on. It's surprisingly easy to remember having predicted the result once you know what happened. Scientists aren't immune to that problem either.

2. What were the researchers asking?

Read the title and abstract of the 2020 paper. The core question was whether information sharing works differently in autistic, non-autistic, and mixed-neurotype chains.

The traditional deficit account might suggest that autistic people would struggle to communicate, including with other autistic people. The double empathy account suggests that difficulty could arise in the interaction between people with different ways of communicating. The researchers designed an experiment that could distinguish some predictions from those perspectives.

Pause and ask: Is this a study of communication in every possible setting? No. It is a study of how many details survive one particular storytelling task. The difference between those two statements is the difference between reading the title and reading the methods.

3. Read the methods: the telephone game becomes an experiment

The 2020 study recruited 72 adults, divided into nine chains of eight. There were three autistic chains, three non-autistic chains, and three mixed chains. Each group passed along a story. Researchers counted how many of 30 preidentified details appeared in each retelling. They also asked participants about how comfortable they felt communicating with their partner, called rapport.

This is called a diffusion chain. It lets researchers watch information change over repeated handoffs. It also creates a challenge: if one early person drops important details, everybody downstream has fewer details available to remember. The 2020 researchers noticed that issue and ran a proportional analysis to account for differing starting points.

Some important boundaries: participants were adults who communicated verbally and had IQ scores in the typical range. The sample was relatively small, and members of the chain knew one another's diagnostic status. The experiment wasn't designed to represent all autistic people or every classroom interaction.

Check your understanding: how many independent chains were there?

Nine chains, with eight people in each chain. That matters because the eight people in one chain are not eight fully independent experiments. Each person receives information shaped by what earlier people said. Counting seventy-two participants is not the whole story when evaluating how much evidence the three types of chains provide.

4. Look at the results before reading the discussion

Here's a simplified look at the reported average number of story details recalled, pooled across positions in each type of chain. These numbers are out of 30, and they are not final-person scores or a measure of everyday social ability.

The analysis found that the autistic and non-autistic chain averages were not statistically distinguishable by the reported test. The mixed chains retained fewer details on average. A separate analysis examined how details disappeared as the story moved through the chain and again found a steeper loss for the mixed condition. The researchers also reported lower rapport in some mixed-condition comparisons.

Before you read their conclusion: what can we say so far?

In this task, these autistic participants successfully passed information along autistic-only chains, with results comparable to non-autistic-only chains. Mixed chains performed worse on the measured detail-transfer outcome. That is meaningful evidence against the assumption that autistic adults must communicate poorly with each other. It is not evidence that all autistic people are equally effective communicators in all settings, or that any mixed group will necessarily communicate worse.

One tiny statistical sentence that matters

When the paper says two groups didn't differ significantly, that does not prove they were identical. It tells us the analysis didn't find convincing evidence of a difference using its model and threshold. That's worth remembering in the next paper.

5. Read the discussion without letting it drive the car

Now read the 2020 paper's discussion. The researchers argue that their findings challenge the idea of a general autistic social-communication deficit and support examining communication as a two-way process.

That's a reasonable interpretation to consider. But notice what is directly measured: detail retention in a story-passing task and self-rated rapport. The study didn't measure friendship quality over years, successful group projects, or college persistence. Those broader possibilities might inspire future research, but they're not outcomes from this experiment.

The authors themselves point out limitations: nine chains, diagnostic status known to participants, uneven gender composition, and an adult sample that excludes many autistic people's experiences. Different stories or social environments might produce different patterns.

Here's the difference between an observation and a leap.

Observation: In the 2020 experiment, mixed chains lost more story details. Leap: Mixed-neurotype study groups are inherently ineffective. The researchers didn't test actual classroom groups, and one small experiment cannot settle that claim.

6. Five years later: the larger study

Now open the 2025 follow-up. This time researchers recruited across three sites. 311 adults attended the research sessions: 154 autistic and 157 non-autistic participants. They again studied story transfer in autistic, non-autistic, and mixed chains, and added conditions including whether participants were informed about other people's diagnostic status and whether the story was fictional or factual.

It was a Registered Report. The researchers arranged for a study plan and analysis approach to be reviewed and accepted before seeing the study results. That reduces some opportunities for deciding what counts as an important finding after the outcome is known. It doesn't guarantee the hypotheses will be correct or the study perfect.

Their prediction: mixed-neurotype chains would lose information faster and have lower rapport.

Predict again: do you expect the earlier result to repeat?

There's no penalty for predicting the 2020 result will hold up, or that it won't. The important thing is being able to say why. A larger sample can help estimate effects more reliably, but changes in participants, tasks, and procedures can also change what a study observes.

What they found

The larger study found no difference in information transfer between single-neurotype and mixed-neurotype chains. On that outcome, the striking mixed-chain disadvantage from 2020 did not repeat.

There were still social differences worth discussing. Non-autistic chains reported higher rapport, and informing participants about diagnostic status improved rapport. But disclosure in this controlled experiment is not a recommendation that students disclose diagnoses publicly. It tells us something about these participants and these interactions, not what any individual ought to share with a professor or employer.

2020 · 72 participants

Mixed chains lost more details

Autistic-only and non-autistic-only chains performed comparably on the measured task. Mixed chains had lower retention.

2025 · 311 participants attended

No mixed-chain transfer penalty found

Transfer results did not differ between mixed and single-neurotype chains. Rapport and information about diagnostic status still mattered in other ways.

7. So was the 2020 study wrong?

This is where good science gets interesting. The first paper documented what happened in that sample, using that design and analysis. The second paper used a larger, related but not identical design and did not find the same transfer difference. That means we should be less confident that mixed-neurotype information transfer is generally worse than we might have been in 2020.

But it does not erase the finding that autistic people can share information effectively with other autistic people. That broad challenge to simplistic assumptions about communication survives the follow-up, even though one specific explanation of the mixed-group pattern becomes less secure.

Nor does either study conclusively prove or disprove the full double empathy account. That is a broader framework about mutual understanding, not a single prediction that rises or falls on a particular telephone-game result.

What could explain the difference? The 2025 authors discuss the more heterogeneous participant sample and methodological differences. You can also ask whether the earlier result was a chance finding, whether context changes the effect, or whether the task is sensitive to details we haven't identified yet. We shouldn't choose our favorite explanation and call it proven.

8. Try a version of the method

This is a learning activity, not a replication. Don't sort people by diagnosis, ask classmates to disclose private information, or infer anybody's neurotype from their performance.

Story-passing activity (15–25 minutes)
  1. Ask for volunteers and explain the task. Offer a non-speaking or observation role.
  2. Use an original, neutral story with 12 numbered details. Only the first person sees it.
  3. Pass the story verbally or in writing, one person at a time, without referring back.
  4. After each handoff, note how many original details remained. Don't rate participants as better or worse communicators.
  5. Compare the outcome with a second round that allows clarifying questions or written notes. Keep the groups comparable if you want a cleaner comparison.
  6. Discuss what changed: memory load, uncertainty, checking for understanding, and the value of repair.

Our companion packet contains a ready-to-use twelve-detail story and scoring grid. All participants may opt out without explanation.

Be careful with the conclusion. If your second round preserves more details, you have demonstrated something about your own activity and procedure. You have not proved a neurotype difference, replicated Crompton, or measured autistic communication.

9. Let's argue about the interpretation

These are questions that work well in a classroom, faculty discussion, or family reading group:

  1. What did the 2020 paper actually establish that still seems defensible after 2025?
  2. Why does it matter that the second study was larger and preregistered?
  3. What does "no significant difference" tell us, and what does it leave open?
  4. If participants feel more comfortable with certain partners, does that necessarily predict how much information they can share?
  5. How might knowing someone's diagnosis change the interaction? Why isn't that a blanket argument for disclosure?
  6. Could a different task, one with back-and-forth questions, produce a different result? How would you test it?
  7. What would a useful college study of communication need to measure beyond the number of story details remembered?
  8. Who should help design that study, and why?

10. What should we do differently tomorrow?

We shouldn't use these papers to prescribe one communication style. We can use them to challenge the reflex to treat misunderstandings as deficits located entirely in the autistic participant. In a classroom, an office, or a job, try making expectations clear, offering ways to check understanding, and allowing communication repair. Then ask the people involved whether those changes actually helped.

If you want to push the question further, return to the double empathy discussion or browse the Research Library. And if you want to learn more about reading papers in general, How to Read Research starts with the basics.

Vocabulary without the headache

Term Plain-English meaning
Hypothesis A prediction the researchers can test
Diffusion chain A sequence in which information passes from one person to the next
Neurotype A way of describing neurological or cognitive variation; here, the studies grouped people as autistic or non-autistic
Operational definition Exactly how researchers measured an idea, such as counting story details as a measure of information transfer
Rapport How comfortable or connected someone reports feeling during an interaction
Sample The people or cases actually included in the study
Replication Repeating or extending a study to see whether a finding holds up
Preregistration Recording the research plan before seeing the results
Registered Report A publication format in which methods and analysis plans receive review before results are known
Statistically significant A result meeting the threshold used by the statistical model; not a synonym for important or proven
Generalizability How reasonably findings can apply beyond the people and setting studied

The papers, in full

  • Crompton, C. J., Ropar, D., Evans-Williams, C. V. M., Flynn, E. G., & Fletcher-Watson, S. (2020). Autistic peer-to-peer information transfer is highly effective. Autism, 24(7), 1704–1712. Open full text. DOI.
  • Crompton, C. J., Foster, S. J., Wilks, C. E. H., Dodd, M., Efthimiou, T. N., Ropar, D., Sasson, N. J., Lages, M., & Fletcher-Watson, S. (2025). Information transfer within and between autistic and non-autistic people. Nature Human Behaviour, 9, 1488–1500. Open full text. DOI.

The numbers and design descriptions above come from the published articles. The diagrams, prompts, and teaching activity are TTW Nest's own explanations, not materials supplied or endorsed by the study authors. This guide does not diagnose communication abilities or make recommendations about individual disclosure.

More to explore

Looking at the same problem from a different perspective?

Understanding neurodivergenceResearch and evidence