What Is the Double Empathy Problem in Autism?

The double empathy problem is the idea that when autistic and non-autistic people struggle to understand each other, the difficulty runs in both directions. Rather than treating communication breakdowns as a deficit located inside the autistic person, this framework holds that both parties face a gap in mutual understanding because they process social information differently. The concept, introduced by autistic researcher Damian Milton in 2012, has reshaped autism research over the past decade, and a growing body of experimental evidence now supports it in concrete, measurable ways.

How Information Degrades Between Neurotypes

Some of the most compelling evidence for the double empathy problem comes from “diffusion chain” experiments, which work like the telephone game. A piece of information is passed from one person to the next down a chain, and researchers measure how much detail survives. In one widely cited study, chains made up entirely of autistic participants retained just as much information as chains of entirely non-autistic participants. But when the chains were mixed, with autistic and non-autistic people alternating, information degraded significantly faster. Autistic-only chains averaged about 13 recalled details and non-autistic chains about 12, with no meaningful difference between them. Mixed chains dropped to roughly 9 details on average, and the decline from one person to the next was steeper.

Rapport ratings told the same story. Participants in mixed chains reported feeling less connected to their partner than those in same-neurotype chains. The researchers concluded that neither group lacked communication skill in any absolute sense; the penalty was specific to the mismatch.

A follow-up study used a visuospatial task instead of verbal storytelling, asking participants to reproduce a drawing and pass it along the chain. Here, the picture was more nuanced: the three chain types did not differ significantly in objective performance. Information still decayed down the chain as expected, but the mixed chains did not show the selective penalty seen in the verbal study.

That inconsistency is actually informative. It suggests the cross-neurotype communication barrier is not a blanket problem affecting every kind of exchange. The breakdown may be most pronounced in tasks that rely heavily on narrative style, social cues, and implicit understanding of what details matter to the listener. When the task strips away those social layers and becomes more spatial or procedural, the gap narrows or disappears. This is consistent with the idea that the issue is not raw cognitive ability but rather a difference in communication conventions between neurotypes.

Rapport in Groups, Not Just Pairs

Most early double empathy research looked at pairs. A recent study expanded the lens to small groups of four people, assigned to one of four compositions: all autistic, all non-autistic, three non-autistic with one autistic member, or three autistic with one non-autistic member. After a collaborative building activity, participants rated how the interaction felt.

All-autistic groups did not differ from all-non-autistic groups on overall rapport. On two specific items, enjoying the interaction and perceiving friendliness from others, the all-autistic groups actually scored higher than both types of mixed groups. At the individual level, autistic participants felt progressively less at ease as the proportion of non-autistic people in their group increased. Non-autistic participants, by contrast, reported roughly consistent rapport regardless of who else was in the group.

That asymmetry is worth sitting with. Non-autistic people did not report feeling uncomfortable in mixed groups, but autistic people did. One plausible reading: non-autistic social norms are the assumed default in most group settings, so non-autistic people feel at home regardless of composition, while autistic people feel the friction of navigating a communication style that does not match their own. The double empathy problem is not just about two sides failing equally; it plays out against a backdrop where one set of norms is treated as standard and the other as deviant.

The Speed of Snap Judgments

If the double empathy problem were purely about mismatched communication conventions, you might expect it to resolve once people spend enough time together. But research on first impressions suggests the barrier can form almost instantly, and that the non-autistic side contributes heavily. In one set of three studies, non-autistic observers watched brief clips of autistic individuals in social situations and rated them less favorably on a range of traits. The observers also expressed less willingness to interact with them. These negative judgments formed within seconds, did not improve with longer exposure, and persisted across child and adult age groups.

⁴

A follow-up investigation flipped the question: what actually drives those first impressions? The answer was surprising. Variability in how autistic adults were rated was explained more by characteristics of the non-autistic raters than by characteristics of the autistic adults being judged. Raters who held more stigmatizing views toward autism gave harsher ratings, while raters with more autism knowledge gave more favorable ones. Only ratings of “awkwardness” were driven primarily by the autistic person’s behavior. When raters were told the person they were watching was autistic, impressions generally improved, but for raters with high stigma, knowing about the diagnosis actually made their ratings worse.

⁵

This body of work complicates the traditional framing considerably. If the social difficulties autistic people face were primarily about skill deficits on their end, you would expect the autistic person’s behavior to be the main predictor of social outcomes. Instead, the non-autistic observer’s attitudes and biases play a larger role. The double empathy problem is not just a mismatch in communication style; it is amplified by prejudice and lack of familiarity on the non-autistic side.

Nonverbal Misunderstandings Run Both Ways

Much of the clinical literature on autism focuses on difficulties autistic people have reading non-autistic nonverbal cues, like facial expressions, eye contact, or tone of voice. The double empathy framework predicts that the reverse should also be true: non-autistic people should struggle to read autistic nonverbal cues.

Qualitative research with autistic adults bears this out. In one study, autistic participants described frequently having their emotional states misread by others. They also reported misinterpreting other people’s emotions, but the crucial point was the bidirectional pattern. Both sides experienced the confusion; it was not a one-way deficit.

⁶

Experimental work on empathic accuracy adds a quantitative dimension. When non-autistic participants watched video clips of people recounting emotional events, they were significantly less accurate at identifying the emotions of autistic narrators compared to non-autistic narrators, especially for happiness and sadness. Interestingly, the same participants reported feeling more intense physical sensations in their own bodies while watching autistic narrators, particularly during accounts of anger and fear. They were picking up that something emotional was happening, but they could not correctly identify what it was.

⁷

Differences in prosody and gaze also contribute to the friction. Autistic individuals often have atypical intonation patterns and may avoid direct eye contact, both of which have been shown to create understanding problems during conversation. But calling these “deficits” in the autistic person misses something important. Atypical prosody is only a problem relative to the listener’s expectations. Among autistic conversational partners who share similar patterns, these features may not create confusion at all, which aligns neatly with the diffusion chain and rapport findings above.

⁸

Rethinking the “Empathy Deficit” in Autism

The traditional clinical view held that autistic people lack empathy, full stop. More recent research has broken empathy into components and found a far messier picture. One study comparing autistic and non-autistic young adults found that the autistic group had lower accuracy at identifying anger specifically, but did not differ from the non-autistic group on broader measures of cognitive or affective empathy when tested using a behavioral task rather than self-report.

⁹

Meanwhile, a large qualitative study asked autistic adults directly about their empathic experiences. A striking pattern emerged: a high proportion reported not too little empathy but too much. Many described their empathic responses as overwhelming or distressing. Some felt others’ emotions so intensely that it became difficult to function. These hyper-empathic experiences do not fit the “empathy deficit” narrative at all. They suggest instead that autistic people may experience empathy differently, sometimes more intensely, and that the apparent gap in empathy is really a gap in how that empathy is expressed or recognized by non-autistic observers.

¹⁰

This is where the double empathy problem intersects with how we measure empathy in the first place. Most empathy assessments were designed by and normed on non-autistic populations. They tend to reward empathy that looks a particular way: making eye contact, mirroring facial expressions, using expected vocal tones to signal understanding. If you feel someone’s pain intensely but express it through withdrawal rather than a concerned face, standard measures may record that as an empathy failure. The tool is measuring conformity to a social convention, not the internal experience itself.

When Alexithymia Is the Real Culprit

An important complication in the empathy research is alexithymia, a trait characterized by difficulty identifying and describing your own emotions. Alexithymia is not unique to autism, but it occurs at much higher rates in autistic populations. And it turns out to be a powerful confound.

In a study of preadolescent children, alexithymia predicted empathy difficulties significantly better than autistic traits did. The children who struggled most with empathy were those with high alexithymia, regardless of whether they were autistic.

¹¹

Neuroimaging research tells a similar story. One study compared brain activation patterns between autistic and non-autistic adults while viewing images of people in pain. Initial analysis showed group differences. But when the researchers statistically controlled for alexithymia and anxiety, both of which were elevated in the autistic group, the group differences in brain activation disappeared. The apparent empathy-related brain differences were being driven not by autism itself but by co-occurring alexithymia and anxiety.

¹²

This matters for how we interpret the double empathy problem. If empathy difficulties in autism are substantially explained by alexithymia rather than by autism per se, then the “empathy gap” between neurotypes may be less about fundamentally different wiring and more about the challenge of reading and conveying emotional states when you struggle to label your own. It also suggests that interventions aimed at helping autistic people “learn empathy” are targeting the wrong thing. Addressing alexithymia, or simply building awareness that it exists as a distinct trait, might be far more productive.

Masking and the Cost of Bridging the Gap

Many autistic people do not simply accept the communication gap. They work to close it unilaterally, through masking, sometimes called camouflaging. Masking involves consciously suppressing autistic behaviors and performing non-autistic social scripts: forcing eye contact, rehearsing small talk, mimicking facial expressions, suppressing stimming. It can be effective enough that others do not realize the person is autistic. But it comes at a steep price.

Research has linked increased masking behaviors to higher rates of depression, anxiety, burnout, and exhaustion in autistic people.

¹³

The workplace is where this dynamic becomes especially visible. A study of neurodivergent employees examined their experiences through the lens of the double empathy problem and found that stigma around neurodivergence, combined with the mutual misunderstanding the theory describes, pushed many employees into camouflaging their natural communication styles. That camouflaging was associated with reduced well-being and a sense that their organizations did not support them.

¹⁴

The double empathy problem reframes masking as a logical but harmful adaptation to a one-sided social contract. If both neurotypes struggle equally to understand each other, but only one is expected to do the work of bridging that gap, then the burden falls disproportionately on autistic people. The mental health consequences of masking are, in this view, partly a symptom of a society that treats non-autistic communication norms as the only valid ones and asks autistic people to conform rather than meeting in the middle.

What Happens in the Brain During Cross-Neurotype Interaction

Neuroscience is beginning to map the double empathy problem at the level of brain activity. One study used EEG hyperscanning, which records the brain activity of two people simultaneously during a live interaction, to compare pairs of two non-autistic participants with mixed pairs of one autistic and one non-autistic participant. During social imitation tasks, the mixed pairs showed different patterns of inter-brain synchrony, specifically in low-alpha frequency bands, compared to same-neurotype pairs. Autistic participants in the mixed dyads also showed a preference for the follower role rather than the lead role during imitation.

¹⁵

Inter-brain synchrony is a rough index of how attuned two brains are to each other during an interaction. Lower synchrony in mixed pairs is consistent with the behavioral findings: something about the cross-neurotype exchange is less coordinated at a neural level, not because either brain is broken but because the two brains are running on slightly different rhythms. The preference for the follower role may reflect a compensatory strategy: when you are uncertain how to lead an exchange with someone whose social timing differs from yours, following is safer.

When the Framework Gets Stretched Too Far

The double empathy problem has become enormously influential in autism advocacy, and for good reason. It shifts the locus of difficulty from a broken individual to a relational mismatch. But it is worth noting where the evidence is strong and where it gets thinner.

The diffusion chain studies, the rapport research, and the first impression experiments offer solid support for the core claim: communication breakdowns between autistic and non-autistic people are mutual, and same-neurotype pairs perform comparably to each other. The visuospatial study, however, did not replicate the selective penalty for mixed pairs, suggesting the effect is context-dependent rather than universal. And most of the evidence comes from autistic adults without intellectual disability who can participate in lab studies. Whether the framework applies equally well to autistic people with higher support needs is an open question that the current literature does not resolve.

There is also a risk of overcorrecting. Saying that communication difficulties are bidirectional does not mean they are symmetrical in consequence. In a world built around non-autistic social norms, the practical fallout of being misunderstood falls much harder on the autistic person: missed job opportunities, social exclusion, misdiagnosis. Framing the problem as mutual is accurate at the level of mechanism, but it can accidentally minimize the unequal real-world impact if taken as the whole story. The strongest version of the double empathy problem acknowledges both the bidirectionality and the asymmetry, recognizing that mutual misunderstanding plays out on an uneven field.