Show the full module text
Module 4 — Reflective Listening
The listening technique taught at the end of your first session: ask for a time, say it back, and check that it landed.
Most hard conversations at home start in a doorway, with one of you halfway out and the other saying “by the way.” The last part of your first session teaches a different way to have them. You ask for a time, one of you talks, the other says it back and checks, and you take the whole thing home on one sheet.
What this is about
Here is a conversation that happens in a lot of houses. One of you says something. The other answers the thing they think was said. The first one hears an answer to a question they did not ask, and says it again, louder or flatter. A while later both of you are angry, and neither of you can say what it was about.
Nobody did anything wrong. Something was dropped in the handover between you. Neither of you could see it drop, so each of you decided something about the other person instead.
What each of you decided. You were not listening.
You did not care enough to follow.
This session teaches a way of talking where the handover is visible. One of you sends a message. The other says it back. The first one confirms it arrived, or says what was missed.
That check is the whole technique.
We build it in on purpose. When two differently wired people talk, the sense that something landed is exactly what tends to go missing.
Most people already do the second step by accident; the third step is the one that gets skipped, and it is the one doing the work.
It also starts before the conversation does. Instead of the doorway version, you ask for an appointment. The whole of that is one sentence: I have something I would like to talk to you about. When would be a good time? Then it goes on the calendar like anything else.
What happens in the session
Your therapist teaches this in the last stretch of the first session, and you run it once before you leave, on a real but small topic. Your therapist will stop it early if it starts to escalate. Being stopped is not a mark against you. The shape you learn is the same shape on the handout you take home.
Every step ends with a question to the person talking, and you do not move on until it has been answered.
Reflective Listening
1 Make an appointment. A time and a place, an agreed length, phones off. Pick a small topic. Before you start, each of you says how wound up you are, on a scale of one to ten.
2 One of you talks, the other says it back. In their exact words or your own. Then two questions: right? and more? Go around that loop until the answer to more? is no. The listener adds nothing of their own.
3 Summarize, then feelings, then validate. The main message, in your own words this time. Right? More? Then the talker names up to three feelings, the listener says them back, and adds: you make sense to me.
4 Switch, and the other one does the whole thing. Say the switch out loud. The second person can answer what they heard or bring something new. If a step does not come off, that is fine.
The two questions do different jobs. Right? checks whether you got it. More? says you are still interested. The second one is the one people leave out, and it is the one that brings out the sentence somebody has never said out loud. Most of us say part of a thing and stop.
You make sense to me is not agreement.
You are not saying your partner is right. You are saying you can see how they got there. Your therapist says that distinction out loud more than once.
Setting it up. We will do this once here so it is not new when you try it at home. It will feel artificial. That is fine. Say one thing. Not the biggest thing. This is your whole turn, not a sentence.
The two questions. Now: did I get that right? And then the other one, and mean it: is there more?
Into a long silence. He is working. Let us both leave it. There is no time limit on this, and I am not going to rescue anybody from a pause.
Validating. Say it back to her, and then: you make sense to me. You are not agreeing. You are saying you can see how she got there.
Right? More? is what you take home. Everything else is scaffolding around those two words.
The way this goes wrong. The listener says the words back and adds a few of their own. “So you felt ignored when I was late, which I have already apologized for.” That is not a reflection. It is a defense dressed up as one, and the talker hears the difference at once. A sincere apology in the first step does the same thing: it ends the turn before the talker has finished. If you catch yourself adding a clause, stop at the period. You get your whole turn afterward.
Four things we change
Word for word is a real answer. A listener who mirrors exactly can be told you are just parroting me, after doing the most careful thing they know. For many autistic listeners, precision is the care. Bringing the words back intact says I did not decide what you really meant. So your therapist asks each of you which you prefer, in front of each other, before the first round.
Nobody fills the pause. For many autistic adults a long silence is where the answer is being put together, not hesitation. I am processing is a complete sentence that holds the floor. A step can also come back tomorrow, in writing, and count in full.
Look wherever helps you listen. You do not have to face each other, and side by side is allowed. The talker is told what averted eyes mean here, so that relieving one of you does not simply move the injury to the other.
When the feeling will not come, there is a menu: hurt, angry, worried, or something else. A body report, like something going tight in your chest, is a real answer.
Getting a feeling roughly right and being corrected is the exercise working, not failing.
Why we do it this way
Very little of this is about being nice. Almost all of it is about stopping one person from doing several things at once. Holding a gaze, following the words, watching your own face and building a reply is four jobs on one channel. Each of the changes above takes a job away.
The facts usually get across. What goes missing is the sense of having been understood.
That is what thousands of neurodiverse couples have told us, and research points the same way. When a story is passed between autistic and non-autistic people, the detail crosses about as well as it ever does, but both people report feeling less in tune than pairs where neither is autistic.1 Saying it back and asking right? puts back the missing proof that it landed.
The facts arrive about as well either way. What is thinner between an autistic and a non-autistic partner is the sense of having been understood, and that is the gap the two questions are built for.
That gap was measured on both people in the pair, not found inside one of them. It does not say either of you communicates badly. If you have spent years being told you are the one who does not listen, or the one who is impossible to follow, that is worth knowing.
There is one experiment on the technique itself. People whose words were said back to them, and who were asked to say more, felt more understood than people who only got nods.2 It is a small study, and we say more about its limits at the foot of the page. What we can tell you from our own rooms is simpler. The check takes a few seconds, and the failure it catches is expensive.
After the session
The homework is appointments, not conversations. One of you makes one in the two days after the session. The other makes one on day three or four, and makes it even if the first one did not happen. Each is an hour, split: twenty minutes one of you talks and the other listens, twenty minutes the other way, then twenty minutes of open discussion if you want it. Use the full twenty, and do not go past it.
Asking for a time does not change the subject. What changes is that both of you arrive at it with the same amount of notice.
This is not a problem-solving tool. For now the whole job is that each of you gets heard. If you are not used to being listened to, let the silences sit.
If one of you escalates. Agree a line in advance: “I love you too much to fight. Let us try again at…” Name a time and a place, then walk away. Naming the next appointment is what makes walking away different from storming off.
Afterward, write down three things in the workbook below: what went well, where you struggled, and what you felt. Expect some of these to go badly at first. That is the part everybody works through.
The one thing, if that is all you have. On a week with no capacity in it, keep the appointment and shorten it. Ten minutes each way, said back and checked, is a real appointment. Cancelling breaks the rhythm; shortening does not.
Your workbook
Your answers save to this device only - we cannot see a word of what you write. Fill in the first group before your appointment and the second one after it.
Your appointment
The homework is an appointment, not a conversation. One of you takes the first two days after the session; the other takes day three or four, and makes it even if the first one did not happen.
Which slot are you taking? — The first two days, Day three or four, Not agreed yet
When you are the one being heard, which lands better? — My exact words back, In their own words, Either is fine, I do not know yet
One small, low-stakes thing you could practice on this week, such as what to do about the car
After the appointment
These are the three things your therapist will ask about, so write them down while they are fresh. Fill this in afterward, not before.
Did the appointment happen? — Yes, both of ours, Yes, one of them, We started and stopped, Not yet
What went well? Even if it is small.
Where did you struggle?
What did you feel? Up to three words is plenty.
If it broke down, where? — Talking too long or too little, Adding a clause to the reflection, We skipped is there more, It turned into the real argument, We never started, It did not break down
Where this comes from
Every source below was checked against the published record. They are grouped by the kind of evidence they are, so the numbers may not run straight down the page — a number points back to where the source is used in the lesson.
Research discussion
The technique is reflective listening. The structured version we use is closest to the Imago tradition, where the published evidence is a small pilot trial rather than a body of outcome research. The one direct experiment had people talk to a stranger who either paraphrased and asked for more, nodded, or gave advice. Paraphrasing beat nodding on feeling understood. It did not beat advice, and it was a single conversation between undergraduates and a stranger.2
A larger study followed newlywed couples for about three years. How well they communicated barely predicted how satisfied they were later. The authors wrote that this raises “important doubts about theories and interventions that prioritize couple communication skills.”3 We would rather quote that than hide it.
There is no research on reflective listening between autistic and non-autistic partners. We looked. The nearest evidence is a pair of story-passing studies. A small 2020 study found that mixed chains of autistic and non-autistic people lost detail faster. A much larger 2025 study, registered before the data existed, did not find that. What held was rapport: mixed chains felt less in tune, and both speaker and listener said so.1 So this module’s claim is about the sense of being understood, not accuracy, and it is still not a test of the technique.
Those chain studies are the main evidence behind what is called the double empathy problem: the proposal that breakdowns between autistic and non-autistic people are mutual, not a deficit in one party.4 You may run into arguments about it. A 2025 review argues the idea is too loosely defined, is being applied faster than it has been pinned down, and should not yet be carried into clinical work.5
Researchers in the field, including the people who proposed it, have replied that it is a sociological account of interaction being judged by the wrong yardstick. They say the evidence favors a probabilistic version: autistic-autistic interactions can go as well as non-autistic ones, and mixed interactions tend to be the hardest.6 We think the caution is fair, and we are not treating a theory. What we work on is the actual gap between how the two of you send and receive, which we can watch happen in front of us.
Who this research was done with. The couples studies behind most of this module drew heavily on white, comparatively well-off, English-speaking participants. If your household carries pressures those samples did not — money, immigration, racism, disability, unsafe housing — the practice still applies, but the room you are practicing in is harder. That is the room, not you.
Peer-reviewed research
1. Crompton CJ, Foster SJ, Wilks CEH, Dodd M, Efthimiou TN, Ropar D, Sasson NJ, Lages M, Fletcher-Watson S (2025) Information transfer within and between autistic and non-autistic people. Nature Human Behaviour, 9(7), 1488-1500. https://doi.org/10.1038/s41562-025-02163-z Stage 2 Registered Report -- the analysis plan was accepted by the journal before the data existed, so the result could not be shopped for. 324 adults recruited and 311 tested (154 autistic, 157 non-autistic) at three sites: Edinburgh, Nottingham and UT Dallas. Six-person diffusion chains, all-autistic, all-non-autistic or alternating, passing a 30-detail fictional passage and a 30-detail factual one. Detail was lost steeply along every chain (fictional b = -5.29, t(51.3) = -9.07, p < 0.001; factual b = -4.57, t(50.9) = -8.29, p < 0.001) but chain type did not predict the rate of loss, with Bayes factors of 7.91 and 125.0 favoring the model without it. In other words the 2020 mixed-chain information-transfer effect did not replicate. Rapport did differ: all-non-autistic chains rated rapport higher than mixed chains, among speakers (b = -35.8, p = 0.0012) and listeners (b = -22.3, p = 0.04); speakers in all-non-autistic chains also rated it higher than those in all-autistic chains (p = 0.002). Participants told the neurotype of the others in their chain reported higher rapport as listeners (p = 0.04), with the same comparison falling just short among speakers (p = 0.054). Limitation: all participants were from the USA and UK, mean IQ was high, everyone communicated by speech, and this is strangers passing on a story rather than a couple in a disagreement. It is also a study of rapport ratings, not of any listening technique.
2. Weger H Jr, Castle Bell G, Minei EM, Robinson MC (2014) The relative effectiveness of active listening in initial interactions. International Journal of Listening, 28(1), 13-31. https://doi.org/10.1080/10904018.2013.813234 Experiment with 115 undergraduates (70.9 per cent female, mean age 20.0) talking to a confederate who responded with active listening (paraphrase plus requests to elaborate), simple acknowledgement (nods and back-channels), or advice. Active listening produced significantly greater feeling understood and higher conversational satisfaction than simple acknowledgement, but did NOT beat advice-giving on conversational satisfaction. Limitation: strangers in a single lab conversation, undergraduate communication students, no nonverbal variables measured, and nothing about couples, conflict, or repeated use over time.
3. Lavner JA, Karney BR, Bradbury TN (2016) Does couples' communication predict marital satisfaction, or does marital satisfaction predict communication? Journal of Marriage and Family, 78(3), 680-694. https://doi.org/10.1111/jomf.12301 431 low-income, ethnically diverse newlywed couples (76 per cent Hispanic) in Los Angeles County, observed four times at 9-month intervals across about 3 years. Cross-sectionally, satisfied couples communicated more positively; longitudinally the cross-lagged effects were very small, with only 7 of 36 communication-to-satisfaction effects significant (median absolute beta .02 to .06). Satisfaction predicted later communication in 6 of 12 lags (median absolute beta .09 to .10). The authors state the results raise important doubts about interventions that prioritize couple communication skills. Limitation: first-married low-income newlyweds in the first 3 years only, structured lab discussions coded for positivity, negativity and effectiveness rather than for reflective listening specifically. It does not show communication training is useless, only that observed communication is a weak forward predictor.
Clinical sources and public data
4. Milton D, Gurbuz E, Lopez B (2022) The 'double empathy problem': Ten years on. Autism, 26(8), 1901-1903. https://doi.org/10.1177/13623613221129123 Editorial revisiting Milton's 2012 proposal that breakdowns in understanding between autistic and non-autistic people are bidirectional rather than a deficit in one party. The authors state plainly that the original was developed from personal experience, anecdotal accounts and limited qualitative data, and that empirical support has accumulated since. Limitation: an editorial framing a research program, not itself a study; cited here for the concept, with the empirical weight carried by Crompton and colleagues above.
6. Kilgallon E, Botha M, Dwyer P, Bottema-Beutel K, Milton D, Sasson NJ, Crompton CJ (2026) What the double empathy problem is (and is not). Autism in Adulthood, OnlineFirst. https://doi.org/10.1177/25739581261456653 The reply from within the field, including Milton and Crompton themselves. It reframes the double empathy problem as a sociological account of interaction rather than a cognitive deficit theory, and states that there is growing evidence in favor of a probabilistic version: autistic-autistic interactions can be as effective as non-autistic ones, while mixed interactions tend to be the most challenging. Limitation: a position and framing paper rather than new data, written partly by the researchers whose work is under discussion.
Still being argued about
5. Livingston LA, Hargitai LD, Shah P (2025) The double empathy problem: A derivation chain analysis and cautionary note. Psychological Review, 132(3), 744-757. https://doi.org/10.1037/rev0000468 A critical analysis arguing that the double empathy problem is poorly conceptualized and is being conflated with other constructs, which the authors call a striking example of a weak derivation chain in psychological science. They caution against translating double-empathy research into applied settings until the concept is better specified. Note what this paper does NOT say: it does not report failed replications of the interaction findings. Its objection is conceptual, about how loosely the theory is defined and how quickly it is being applied. Limitation: a theoretical critique rather than new data, and one position in a live disagreement -- see Kilgallon and colleagues (2026), which argues the theory should not be judged solely inside the framework this critique applies to it.
Neither of you is the one who communicates badly
The Neurodiverse Couples Counseling Center works with couples where one or both partners are autistic, ADHD or AuDHD. We do not treat either of you as the one who communicates badly, because the research does not support that and neither does the room. Therapy for clients in California, coaching worldwide, all by telehealth. A first conversation costs nothing.
Up next
Module 5 — When We Don't Talk About It