The Double Empathy Problem: Why Cross-Neurotype Misunderstanding Is Nobody's Fault

Two people are in the same conversation. One is autistic. One isn't. Both are trying. Both are paying attention. Both leave the interaction feeling vaguely misread, slightly off, like something didn't quite land — and neither one can name exactly what.

This is the double empathy problem, and it's one of the most practically useful reframes in contemporary autism research.

For most of the twentieth century, the dominant explanation for autistic social difficulty was deficit-based and one-directional: autistic people have impaired theory of mind, which means they struggle to infer what others are thinking and feeling, which means social interaction is hard for them.

This framing wasn't invented maliciously. It emerged from a real observation — autistic people do often struggle in social contexts — and it produced a generation of interventions aimed at teaching autistic people to read neurotypical social cues more accurately.

The problem is that it located the difficulty entirely inside the autistic person. The neurotypical person in the interaction was treated as the standard, the baseline, the one who was reading the situation correctly. The autistic person was the one with the impairment.

In 2012, autistic sociologist Damian Milton published a paper that reframed this. His argument was structurally simple and empirically well-supported: the difficulty in autistic–neurotypical interaction isn't one-directional. It runs both ways.

Neurotypical people are also poor at reading autistic people. They misinterpret autistic communication styles, misread autistic emotional expression, and frequently attribute negative intent to behavior that has a completely different meaning inside autistic social logic. The empathy gap isn't a deficit located in one party — it's a mismatch between two different systems, neither of which is broken.

Milton called this the double empathy problem: the difficulty is doubled because it belongs to both sides of the interaction, and it arises from the gap between two genuinely different ways of being in the world.

It removes the burden from one side. If cross-neurotype misunderstanding is bidirectional, then the autistic person isn't failing at something the neurotypical person is succeeding at. They're both failing to translate across a gap that neither of them created. That's a different kind of problem, and it calls for a different kind of response.

It reframes the goal. If the problem is a translation gap rather than a one-sided deficit, then the goal isn't to make autistic people better at performing neurotypical social scripts. It's to build mutual legibility — which requires both parties to do work, not just one.

It explains a lot of relationship friction. A non-autistic partner who reads their autistic partner's reduced expressiveness as coldness, or their directness as aggression, or their story swapping as one-upping, isn't doing anything wrong in the narrow sense. They're applying their own conversational protocol to behavior that was generated by a different one. The autistic partner isn't doing anything wrong either. Two coherent systems are producing a mismatch that neither one is solely responsible for.

It has direct implications for misdiagnosis. A significant portion of autistic people — particularly those who are late-diagnosed, or who have spent years masking — have been told at some point that their social difficulties reflect a character problem rather than a neurological difference. The double empathy problem provides a structural explanation for why those difficulties exist that doesn't require pathologizing the person.

The double empathy problem shows up in small, concrete ways that are easy to miss until you know what you're looking at.

Tone misreads. An autistic person delivers information in a flat, direct tone that they experience as neutral. A neurotypical listener reads it as cold, dismissive, or rude. Neither reading is wrong — they're accurate to two different sets of norms for what neutral sounds like.

Silence misreads. An autistic person goes quiet during a conversation because they're processing. A neurotypical partner reads the silence as withdrawal, anger, or disengagement. The autistic person wasn't withdrawing — they were thinking. The neurotypical person wasn't misreading maliciously — silence means something specific in their protocol.

Conversational structure misreads. An autistic person answers a story with an adjacent story of their own — story swapping — as a signal of connection and resonance. A neurotypical listener reads it as one-upping or making it about themselves. Both readings are internally coherent. They're just reading from different grammars.

Emotional expression misreads. An autistic person's face doesn't track their internal state the way a neurotypical face typically does. A neurotypical observer reads the reduced expressiveness as absence of feeling, or as deception. The autistic person is feeling — the signal just isn't being transmitted on the frequency the observer is tuned to receive.

In each of these cases, the neurotypical read gets treated as the objective read by default, simply because it's the more common frame of reference. The double empathy problem names why that default is wrong.

Milton's original paper has held up well. A 2019 study by Catherine Crompton and colleagues found that autistic people communicate more effectively with other autistic people than with non-autistic people — and that non-autistic people communicate more effectively with other non-autistic people than with autistic people. The information transfer was better within each group than across the neurotype divide. This is exactly what the double empathy problem would predict: the difficulty is in the translation, not in either party's individual capacity.

Subsequent research has continued to support the bidirectional framing. The field has moved, slowly, away from the deficit model and toward a difference model — not because autistic people don't experience real difficulty, but because locating that difficulty entirely inside the autistic person was never the most accurate account of what was actually happening.

If you're in a relationship with someone of a different neurotype — romantic, familial, professional — the double empathy problem offers a reframe that's practically useful.

The friction you're experiencing probably isn't about bad faith. It probably isn't about one person being more empathetic than the other. It's about two people running different protocols that don't natively translate, and both of them interpreting the other's behavior through their own protocol's lens.

The fix isn't for one person to become more like the other. It's for both people to develop enough literacy in each other's protocol to stop assuming that their own read is the objective one.

That's harder than it sounds, and it doesn't happen automatically. It requires naming the gap explicitly, which requires knowing the gap exists. That's what the double empathy problem gives you: a name for something that was already happening, and a frame that makes it possible to address without assigning blame.

• The double empathy problem (Milton, 2012) reframes autistic social difficulty as bidirectional: neurotypical people are also poor at reading autistic people, and the empathy gap belongs to both sides of the interaction.
• The difficulty arises from a mismatch between two different systems, neither of which is broken.
• Neurotypical reads of autistic behavior get treated as the objective read by default, simply because they're the more common frame of reference — the double empathy problem names why that default is wrong.
• A 2019 study (Crompton et al.) found that autistic people communicate more effectively with other autistic people, and non-autistic people with other non-autistic people — consistent with the translation-gap model.
• The fix isn't for one party to become more like the other; it's for both to develop literacy in each other's protocol.

Milton, D. E. M. (2012). On the ontological status of autism: The 'double empathy problem.' Disability & Society, 27(6), 883–887.

Crompton, C. J., Ropar, D., Evans-Williams, C. V., Flynn, E. G., & Fletcher-Watson, S. (2020). Autistic peer-to-peer information transfer is highly effective. Autism, 24(7), 1704–1712.

This article is for educational and informational purposes only.

Well wishes. 🙏

L.C. Dialogos Buddhist chaplain, Plum Village tradition

© 2026 LC Publishing Company. All rights reserved.