Skip to content
Home Creative Processing The Work Product How This Place Works FAQ About Join Refugees →
Monday Guides Welcome Letter The Book Music Roots About Refugees Constitution Launch Archive
NeuroSpicy Refugees

What I Meant

Nuance, text, and misreading each other inside our own spaces.

Why tone drops out of text, why the same words can hit an idea or a person, why autistic spaces aren't automatically smooth, and what repair looks like when they aren't.

Monday, November 16 4PM Pacific / 7PM Eastern Join on Zoom →

Last week I left a comment on an essay that got shared into an autistic Facebook group.

The essay was about thinking in systems instead of steps, and about being told your whole life that you're overcomplicating things. It was long. It also read, to me anyway, like something that had come out of a chatbot and gone onto the page without an edit. So I said so:

"That's a lot of words on a page for something that can be said in far fewer. And if the OP didn't have so many hallmarks of being drafted with AI, I'd be much more inclined to forgive its verbosity. But it was one prompt away from a proper edit… 🥺"

For the record, I build AI tools for a living. I use them every day, including to help put these guides together. My gripe was with skipping the edit. That's not how it read.

The group's host hid the comment. Her reasoning, as best I can paraphrase it: a lot of people in that group build meaning through context and elaboration, and the essay was describing how they think. I'd knocked the way they communicate in a thread about how they communicate.

I explained what I meant. She said she understood, and that impact doesn't stay inside intent. I asked her to put the line in the group's agreements, in general terms, so the next person would know where it was before they stepped over it. She said she would. Not long after, a post about AI went up, addressed to everyone, and I couldn't tell if it was about me. I still can't.

Nobody in that thread was cruel. I meant one thing, she heard another, and we were each right about our own half. That's what we're getting into tonight.

Part One

Text drops the tone, for everyone

We all think our tone comes through. It mostly doesn't.

In a 2005 study, people sent short messages to a partner, some sincere, some sarcastic. Everybody predicted their partner would read the tone right about 78% of the time. By voice, partners got it about 75% of the time. By email, 56%, not much better than a coin flip. The readers were sure they'd gotten it right about 90% of the time. (Kruger, Epley, Parker & Ng, J Pers Soc Psychol, 2005)

Small marks carry weight nobody intended. In a 2016 study, one-word text replies like "Okay." read as less sincere with a period than without. The same period on a handwritten note changed nothing. (Gunraj et al., Computers in Human Behavior, 2016)

Emoji don't fix it. When 304 adults rated the exact same emoji image, they disagreed about whether it was positive, neutral or negative a quarter of the time. I ended my comment with 🥺. I know what I meant by it. I have no idea what anyone else saw. (Miller et al., ICWSM, 2016)

Part Two

Same words, two targets

The same sentence can carry an argument and a judgment at the same time.

Text has been good to a lot of us. In a 2014 survey of 291 autistic people, online communication gave them more comprehension, more control, access to people like them, and "the opportunity to express their true selves." (Gillespie-Lynch et al., 2014)

It also costs. In a 2025 study of 34 autistic social media users, some had been misread so often that they started padding everything with words like "seem" and "not sure," just to stay out of trouble. A participant in another study said emoji were "totally confusing," and proposed a fix: show your draft to a couple of trusted people before you post. (Koteyko et al., Autism in Adulthood, 2025; Barros Pena et al., CHI, 2023)

Getting the meaning and feeling the weight are two separate skills. In an eye-tracking study, autistic adults understood ironic criticism fine. What they didn't do was register it as softer or harsher than the same criticism said straight. You can know exactly what you meant and still misjudge how much it stung. (Barzy, Filik, Williams & Ferguson, Autism Research, 2020)

Tone tags like /s and /gen feel like they help. In the only study of them with neurodivergent users, 18 people chatted with and without tags. With tags, people felt more expressive and rated their conversations better. They got no better at reading each other's tone. (Xiao, Iftikhar & Huang, CSCW, 2025)

Part Three

Inside our own spaces

We came to autistic spaces to stop being misread. Then it happens anyway.

Being with other autistic people often feels easier. Twelve autistic adults described it that way in 2020, and one gave the study its title: "I never realised everybody felt as happy as I do when I am around autistic people." That expectation is real. It's also why a blowup inside our own space hurts more than one outside it. (Crompton et al., Autism, 2020)

Researchers filmed 30 members of an autistic adults' charity playing video games together. They gave each other a lot of credit, assuming a shared background and dropping references without explaining them. When the other person caught the reference, rapport came fast. When they didn't, the same remark came across as disruptive or self-centred. They were also quick to shrug off small misunderstandings and move on. (Heasman & Gillespie, Autism, 2019)

Autistic-to-autistic isn't automatically smooth. In a 2025 study of 43 pairs, the autistic pairs' rapport varied more than anyone's, from easy to rough. In another study, autistic adults preferred talking with other autistic adults, and also rated them as more awkward. And a 2025 survey of 504 people in the autism community found real disagreement among neurodiversity supporters themselves. The authors warned that "oversimplified rhetoric may cause confusion regarding advocates' views." (Efthimiou et al., 2025; Morrison et al., 2020; Dwyer et al., Autism, 2025)

The aha

We don't all use the same dictionary

We share a lot in spaces like that. We don't share definitions. When I wrote "verbosity," I was talking about a tool. Somebody who's been told their whole life that they talk too much read it as being about them, and I can't say they were wrong to. Inside a group that's supposed to understand you, people stop checking which meaning the other person is using. So when it turns out to be a different one, it lands like a betrayal.

Part Four

The heat

Some of us feel it faster, and some of it has a name.

Emotional dysregulation runs high in ADHD. A 2020 meta-analysis of 13 studies with 2,535 adults found it much higher than in controls, and called it "a core feature" of ADHD. In other words, the reply goes out before the thinking catches up. (Beheshti, Chavanon & Christiansen, BMC Psychiatry, 2020)

Rejection sensitive dysphoria, the sting a lot of us know, isn't in the diagnostic manual. The research so far is small interview studies, and there's no reliable way to measure it yet. Plenty of us know the feeling anyway. The research just hasn't caught up. (Ginapp et al., PLOS ONE, 2023)

Autistic people are often called justice-sensitive, and plenty of us recognize ourselves in that. Nobody has measured it in autistic adults with a validated scale yet. The closest study found more sensitivity to unfair treatment among young people with ADHD symptoms. Black-and-white thinking has better evidence: higher autistic traits went with more all-or-nothing thinking, mostly through a harder time tolerating uncertainty. A comment you can't read is a kind of uncertainty, and all-or-nothing is one way to make it go away. (Bondü & Esser, 2015; Shi & Hirai, Scientific Reports, 2024)

Part Five

Intent, impact, and the line

Both are real, and last week I was on both ends of it.

"I didn't mean it that way" is true and doesn't undo how it landed. "It hurt" is true and doesn't prove someone meant to hurt. A lot of fights in our spaces happen because each side holds one half of that and demands the other give theirs up.

The group I was in had already written the first move down. Its proposed agreements, shared here with the host's permission, include two that would have helped us both:

  • "Ask before assuming." "If something lands badly, ask 'what did you mean?' first. Different minds misread each other in both directions."
  • "Disagree with care." "Disagree with ideas as much as you like. No ridicule, contempt, slurs or pile-ons."

The hard part is the space between those two. Was my comment about an idea, a tool, or a person? I'd have said a tool. A lot of readers heard a person. The fix I asked for was a written line, so the next person knows where it is before they cross it. That's still the fix I'd ask for.

Part Six

Repair, before and after

Some of these have research behind them. The rest are lived wisdom, and the guide says which.

Before you post

Name your target out loud

"I'm talking about the tool, not the writer." One clause would have changed how my comment read. Mine didn't have it.

Lived wisdom
Before you post

Use a human tone-checker

Show the draft to someone you trust before it goes up. An autistic participant in a design study came up with this one.

Barros Pena et al., 2023
When it lands badly

Ask what they meant

Readers in the 2005 study were 90% sure they'd read the tone right and got it right 56% of the time. Check the part you feel surest about.

Kruger et al., 2005
When it lands badly

Move it to voice

Hearing someone explain a view made them seem more thoughtful and human than reading the same words. Voice built closer bonds than text without the awkwardness people expected. Offer it. Don't force it on someone who needs text.

Schroeder, Kardas & Epley, 2017; Kumar & Epley, 2021
When it lands badly

Wait one cycle before replying

If your heart's racing, you're answering the feeling. Draft it, close it, come back. There's no direct study on this one. It's what most of us learned the hard way.

Lived wisdom
For the space

Write the line down

When a host draws a line, ask for it in the agreements, in general terms. That way the next person doesn't have to find it by stepping on it, and nobody has to wonder who the rule was written for.

Lived wisdom
For the space

Say what you want up front

Autscape, an autistic-run conference, uses colored badges: red for "please don't start a conversation," yellow for people you already know, green for "please talk to me, I struggle to start." Online, a line like "looking for feedback" or "just venting" does the same job.

Autscape
For the space

Repair in private first

A public correction reads like a verdict to everyone watching, even when it's meant gently. A private check-in leaves room for "oh, I meant something else."

Lived wisdom
Bring these to the room

Both halves

Answer one, or all of them, or just listen. Chat counts.

When were you sure you knew someone's tone, and wrong?

Either direction. You misread them, or they misread you.

What word means one thing to you and another to someone else?

"Fine." "Verbose." "Okay." "Interesting." Name yours.

Has a neurodivergent space ever hurt more than a regular one?

No names needed. What made it land harder?

What line do you wish had been written down before you crossed it?

Or before someone crossed it with you.

The lines you've heard

What people say

"It's obvious what I meant."
To you. People expected their tone to come through by email almost 80% of the time. It came through 56%.
"We're all autistic here, we get each other."
Often, yes. Not automatically. Autistic pairs' rapport varied more than anyone's, and shared references only work when the other person shares them.
"Intent doesn't matter, only impact."
Both matter. The hurt needs repair either way, and knowing what someone meant tells you whether you're dealing with a slip or a habit.
The receipts

Where all of this comes from

Text and tone

Kruger, Epley, Parker & Ng (2005). "Egocentrism over e-mail: Can we communicate as well as we think?" J Pers Soc Psychol 89(5):925–936. Figures as reported by the APA Monitor (2006).
Gunraj, Drumm-Hewitt, Dashow, Upadhyay & Klin (2016). "Texting insincerely: the role of the period in text messaging." Computers in Human Behavior 55:1067–1075.
Miller, Thebault-Spieker, Chang, Johnson, Terveen & Hecht (2016). "'Blissfully happy' or 'ready to fight': Varying interpretations of emoji." Proceedings of ICWSM 2016.

Autistic people online

Gillespie-Lynch, Kapp, Shane-Simpson, Smith & Hutman (2014). Intellectual and Developmental Disabilities 52(6):456–469.
Koteyko, Van Driel, Billan, Barros Pena & Vines (2025). "Stigma management strategies of autistic social media users." Autism in Adulthood 7:273–282.
Barros Pena, Koteyko, van Driel, Delgado & Vines (2023). "'My Perfect Platform Would Be Telepathy'." CHI '23.
Barzy, Filik, Williams & Ferguson (2020). Autism Research 13:563–578. Xiao, Iftikhar & Huang (2025), "Tone indicators: Designing accessible CMC cues for neurodiverse users," Proc. ACM HCI 9(7), CSCW496.

Inside our own spaces

Crompton, Hallett, Ropar, Flynn & Fletcher-Watson (2020). Autism 24(6):1438–1448. Heasman & Gillespie (2019), "Neurodivergent intersubjectivity," Autism 23(4):910–921.
Efthimiou et al. (2025). "Social motor synchrony and interactive rapport in autistic, non-autistic, and mixed-neurotype dyads." Autism. Morrison et al. (2020), Autism 24(5):1067–1080.
Dwyer et al. (2025). "Community views of neurodiversity, models of disability and autism intervention." Autism 29:2297–2314.

The heat

Beheshti, Chavanon & Christiansen (2020). BMC Psychiatry 20:120. Ginapp et al. (2023), PLOS ONE 18:e0292721.
Bondü & Esser (2015), European Child & Adolescent Psychiatry 24:185–198. Shi & Hirai (2024), Scientific Reports 14:23334.

Repair

Schroeder, Kardas & Epley (2017). "The humanizing voice." Psychological Science 28:1745–1762. Kumar & Epley (2021), J Exp Psychol: General 150:595–607.
Autscape, "Autistic needs" and interaction badges. Proposed group agreements from an autistic-led research community, quoted with the host's permission.

Several of these studies are small, and most of the repair tools are community practice, not tested. The guide marks which is which. The story at the top is told from my side. The people on the other side would tell it differently, and their version would be just as true.

How we hold this

  • You can pass on anything. Pass is a complete sentence.
  • No names. We're talking about patterns, not putting anyone on trial, including the people in my story.
  • If something said tonight lands badly, ask what they meant. We'll practice right here.
  • Long answers and short answers are both welcome. That one's not a joke.

Getting misread by your own people can sting worse than anything outside. If you're carrying one of those right now, it's welcome in the room.

Say what you meant.

Then ask what they heard. Bring the one you're still chewing on, and we'll read it from both ends.