The Double Empathy Problem in Autism: Theory and Evidence

Author: Angela Bogdanova · Published: October 18, 2026 · Editorial Policy

Research cluster: Cluster Autism and Adult Neurodiversity · Written by Artificial Sapiens

The double empathy problem in autism is a theory of reciprocal difficulty in understanding across people whose experiences, communication practices, and expectations differ. Originally articulated by Damian Milton in 2012, it proposes that some misunderstandings between autistic and non-autistic people arise within their interaction, through a mismatch between perspectives, rather than being explained entirely by one participant's abilities. It directs scientific attention to both partners, the history of their relationship, the social setting, and the standards used to judge a successful conversation. The central claim is relational: understanding another person is an achievement involving two perspectives, and each participant may misinterpret the other. Milton (2012).

The empirical record has developed significantly. Some studies find more efficient information transfer, stronger rapport, or more positive interaction ratings when autistic participants communicate with other autistic participants than when neurotypes are mixed. Others find that matched participants do not perform better on carefully measured communication efficiency, and recent methodological criticism shows that rapport and communication quality are often mistakenly treated as direct measures of empathy. The scientific conclusion is that mutual social understanding is a consequential research target, while the scope, mechanism, and generality of the double empathy hypothesis remain under active investigation. Crompton et al. (2020); Fellowes (2026); De Laet et al. (2026).

What the Double Empathy Problem Means

The expression double empathy problem refers to difficulties in reciprocal understanding between people with substantially different ways of perceiving, interpreting, or participating in social situations. In the autism literature, the principal comparison is between autistic and non-autistic people. An autistic adult may infer an intention that their non-autistic partner did not mean; the non-autistic partner may likewise mistake the autistic adult's words, expression, pauses, interests, or body movement for something else. Successful exchange requires both participants to establish a workable shared meaning.

Milton's original formulation was a conceptual and social-theoretical intervention in autism research. It asked investigators to analyze social interaction as a relationship between participants with potentially distinct experiential worlds, rather than inferring all causes of a conversational breakdown from an autistic person's diagnosis. The theory also questions whether the social conventions of the majority should define the only valid form of interpersonal understanding. These are important questions about the object of scientific explanation, including how observers decide whose behavior is considered unusual or unsuccessful. Milton (2012).

A precise formulation has several parts. First, conversational coordination is influenced by both partners. Second, differences in expectations and interpretive habits can make coordination harder. Third, a partner who is highly skilled in one social community may encounter difficulties interpreting another community's conventions. Fourth, the quality of an interaction can depend on how well the partners learn each other's individual signals. These claims generate testable predictions but do not require that all autistic people communicate in the same way or that all difficulties arise from neurotype mismatch.

Why the Theory Is Called Double

The word double describes the possibility that misinterpretation operates in both directions. Consider an autistic employee who provides a detailed, literal account of a technical problem and a non-autistic colleague who interprets the directness as interpersonal criticism. The employee may regard the colleague's indirect request as too ambiguous to act on. Both have received information and both have assigned meaning, yet their interpretations diverge. Whether this constitutes reduced empathy, pragmatic communication mismatch, a problem of workplace norms, or simply unfamiliarity requires evidence about what each person intended and understood.

An observation of mutual misunderstanding does not automatically demonstrate equal responsibility, equivalent social power, or identical communication capacity. A clinician has obligations to make health information accessible; a manager exercises authority over a worker; a partner may have more knowledge of the context or greater power to determine the consequences of a conflict. The theory invites investigation of each perspective while preserving these contextual differences. A balanced scientific analysis asks whose assumptions shaped the interaction, how error was assessed, and what changes would improve understanding.

A Relational Hypothesis, Not a Diagnostic Category

The double empathy problem is not a DSM-5-TR or ICD-11 diagnosis, autism subtype, screening instrument, or clinical symptom. Autism remains a neurodevelopmental condition assessed using developmental history and the relevant criteria for social communication and restricted or repetitive behaviors, interests, or sensory features. The double empathy framework may inform understanding of particular interactions but does not establish or rule out autism. NICE CG142.

A person may be autistic and experience substantial difficulty with language, interpreting others' intentions, and everyday adaptation even in a sympathetic environment. Another autistic person may communicate fluently in many settings while expending substantial effort to coordinate with unfamiliar partners. Intellectual disability, language development, co-occurring ADHD, anxiety, sensory conditions, social history, and environmental pressures can also affect the interaction. A useful relational account remains compatible with evaluating individual support needs.

Intellectual Origins: From Theory of Mind to Interaction

Traditional autism research often investigated social understanding through individual tasks involving mental-state inference, facial expressions, nonliteral language, joint attention, and theory of mind. These methods addressed important developmental and cognitive questions, but the everyday act of understanding another person is embedded in a dynamic exchange. Milton's 2012 paper brought attention to the limits of explaining reciprocal encounters by measuring only one participant's cognitive performance or by using non-autistic behavior as the default standard of correctness. Milton (2012).

Theory of mind describes the capacity to represent another person's beliefs, desires, intentions, and knowledge. Empathy can include recognizing a person's feelings or perspectives and responding emotionally. Pragmatic communication concerns how participants use language, context, implications, turn-taking, and shared reference. These constructs overlap in real conversations but pose different research questions. A theory-of-mind test asks whether a participant understands a target mental state under specified conditions; a relational experiment asks how two people construct shared meaning across an interaction; a measure of interpersonal liking asks yet another question.

The double empathy theory gained momentum as autistic advocates and researchers challenged assumptions that atypical outward behavior reveals a universal lack of caring or interest. It also aligns with work on the social model of disability, which emphasizes how an environment's arrangements can create or intensify barriers. These intellectual links do not settle empirical disagreements. A model can illuminate neglected dimensions of experience and still require rigorous testing of its strongest causal predictions.

What Counts as Empathy in This Research?

The ordinary word empathy covers several different capacities and processes. Cognitive empathy commonly refers to representing or understanding another person's perspective, thoughts, or feelings. Affective empathy concerns emotional resonance or an emotional response to another person's experience. Empathic concern, personal distress, compassion, agreement, rapport, and friendliness each describe further dimensions. A pleasant conversation may show good rapport without requiring a researcher to infer accurate knowledge of another person's internal experience.

This distinction is especially important in autism research because the results depend on the measure used. A 2025 systematic review and meta-analysis by Cusson and colleagues examined 226 studies. The authors found larger group differences on measures of cognitive or undifferentiated empathy than on measures of affective empathy. The small average affective-empathy difference became nonsignificant when the analysis was restricted to higher-quality studies. Different subscales could point in opposing directions, including lower average empathic concern but greater personal distress. Such group estimates do not describe every autistic individual, and measurement differences substantially change the apparent picture. Cusson et al. (2025).

A 2026 critical methodology review by Sam Fellowes examined 20 empirical papers presented as relevant to double empathy. Fellowes concluded that many studies used outcomes such as rapport, warmth, ease, communication flow, or signaling rather than valid direct measures of cognitive or affective empathy. Some experimental tasks and brief encounters also offered limited opportunities for either form of empathy to arise. The review therefore distinguishes evidence of better interaction in a particular condition from evidence establishing a particular theory about empathic capacities. Fellowes (2026).

Interaction Quality, Empathic Accuracy, and Emotional Concern

Empathic accuracy can be examined by comparing a participant's inference about another person's reported thoughts or feelings against that person's own reports, recognizing that self-report is itself incomplete. Affective response may require validated emotion ratings, physiological or behavioral measures, or qualitative descriptions appropriate to the concept. Interaction quality may be assessed by conversational cooperation, factual accuracy, ease, perceived friendliness, or success on a shared task. Studies should name the outcome rather than using empathy as a universal synonym.

Two partners may feel connected while conveying inaccurate technical instructions; another pair may achieve an objective task with little warmth. One participant may understand another person's feelings but respond in an atypical manner, or be emotionally moved while misreading the person's intention. These configurations are entirely plausible. The research question becomes which component neurotype matching affects, in which settings, through what pathway, and whether its effects survive changes in task and participant composition.

How Researchers Test the Double Empathy Hypothesis

The most direct designs compare interactions between two autistic participants, two non-autistic participants, and one autistic plus one non-autistic participant. Researchers may examine information retained in a communication chain, subjective rapport, observer-rated rapport, conversational accuracy, interpersonal perception, or collaborative efficiency. A strong study states its specific prediction in advance, recruits adequately characterized participants, explains how partners are assigned, reports both successes and null findings, and distinguishes a participant's diagnosis from the characteristics of a particular relationship.

The phrase same-neurotype pair is a useful shorthand in these experiments. It does not guarantee cognitive or communicative similarity. Autistic people vary widely in expressive language, sensory preferences, interests, culture, gender, co-occurring conditions, and communication experience. Two non-autistic people can also differ dramatically. The autism-status category represents one variable among many that can shape compatibility. Evidence of an average difference between experimental groups should not be converted into a rule about whom any individual will understand best.

Information Transfer: The Diffusion-Chain Experiment

A widely discussed study by Crompton and colleagues used diffusion chains in which successive adults passed a story from one participant to another. The sample included 72 participants arranged into chains of autistic participants, non-autistic participants, or mixed-neurotype participants. Information was lost more quickly in the mixed chains than in the matched chains, while autistic-only chains were not significantly worse than non-autistic-only chains. The design directly tests transmission of story details across social exchanges and demonstrates that effective communication can occur among autistic adults. Crompton et al. (2020), Autism.

The outcome concerns detail retention under a particular sequential retelling procedure. It does not directly measure whether each participant accurately identified another person's emotions, nor whether the same result would appear during therapy, intimate conversation, disagreement, or high-stakes shared decision-making. Chains also accumulate the effects of earlier transmissions, so the group-level result is a property of the entire sequence. Its contribution is a concrete challenge to any blanket prediction that autistic-autistic communication must necessarily transmit information less successfully.

Rapport in Dyadic Interactions

In a separate 2020 paper, Crompton and colleagues compared autistic-autistic, non-autistic–non-autistic, and mixed pairs using participant and observer judgments of interpersonal rapport. Neurotype matching influenced some ratings of rapport, whereas autism status by itself was not a simple explanation of interaction quality. The findings support the importance of partner composition and challenge an automatic equation between autistic status and poor social connection. Crompton et al. (2020), Frontiers in Psychology.

Rapport is a valuable outcome because social participation involves experiences of ease, acceptance, reciprocity, and enjoyment. Yet judgments of rapport can be shaped by individual preference, self-presentation, expectations, evaluator bias, conversational topic, and prior experience. Rapport is neither identical to empathic accuracy nor a direct test of the developmental mechanisms underlying autism. A nuanced synthesis treats subjective connection as a meaningful endpoint in its own right rather than automatically elevating it into an omnibus test of empathy.

Two-Sided Understanding in Families

Heasman and Gillespie examined misunderstandings in 22 pairs involving people described at the time using the diagnosis Asperger's syndrome and their family members. Each side estimated how the other perceived them across a set of domains. The study found that misunderstandings could involve inaccurate assumptions by family members about the autistic participant's perspective. It also demonstrated a practical method for analyzing both participants' representations of one another rather than treating only the autistic person's answers as the outcome. Heasman and Gillespie (2018).

This research is valuable for showing that interpretation of another person's intentions can be studied within established relationships. Its participants had shared histories, expectations, and potentially longstanding disagreements. Findings from these family dyads should not be presented as a population-wide prevalence estimate or proof that every disagreement between relatives follows the same mechanism. The diagnostic terminology in the original study reflects historical classification; contemporary discussion generally uses autism spectrum disorder or autism.

Reading Mental States From Movement

Edey and colleagues studied whether participants interpreted simple social animations differently depending on who had produced them. The animations were generated through manual movement of shapes. Non-autistic observers were more successful in interpreting mental states represented by animations made by non-autistic rather than autistic producers, while autistic observers did not show the same in-group advantage. The work illustrates an asymmetry: non-autistic perceivers may have difficulty reading expressions generated in an unfamiliar style. Edey et al. (2016).

This is a specific experimental finding about interpreting stylized motion, not a demonstration that all participants reciprocally misunderstood one another to the same degree. The lack of a matched-group advantage among autistic viewers also complicates a simplified prediction that autistic people must always interpret autistic signals more accurately. Evidence can support one part of a relational account while leaving a stronger symmetric claim unresolved.

What the 2024 and 2025 Studies Added

Later investigations broadened the outcome measures beyond information-transfer chains and one-to-one ratings. These studies help explain why the field now distinguishes ease of interaction from objective accuracy and distinguishes direct participants' reports from observers' judgments. They also show that a familiar pattern in one type of task may coexist with a different pattern in another.

How Non-Autistic Observers Judge Interactions

Jones and colleagues asked 102 non-autistic observers to evaluate recorded interactions involving autistic, non-autistic, and mixed dyads. Observers rated mixed interactions as less successful overall and often rated participants more favorably when they interacted with non-autistic partners. The observers also underestimated some of the autistic participants' own judgments of their partners, suggesting that outside evaluations can diverge from the interpretations held by people participating in a conversation. Jones et al. (2024).

The study helps identify how evaluation and social stigma may operate around an interaction, extending the discussion beyond actual conversational performance. However, the observers were non-autistic, the filmed interacting participants were men, and the ratings involved brief recorded encounters. The study therefore cannot establish how all communities would judge all social exchanges. An outside observer's perception is one social outcome, not an error-free representation of the intentions of both partners.

Rapport in Groups Rather Than Pairs

Foster and colleagues tested rapport in groups containing four participants, with all-autistic, all-non-autistic, and mixed compositions. In this 2025 study of 143 adults, autistic-only groups were capable of experiencing strong rapport; some enjoyment and friendliness ratings were higher in those groups than in mixed compositions. Overall rapport in autistic-only groups did not differ significantly from non-autistic-only groups. This expands the evidence that interaction outcomes depend on social context and group composition. Foster et al. (2025), Autism.

The activity was short, and the experiment involved a collaborative building task. Participants were informed of group diagnostic composition, which may affect expectations. The study illuminates immediate, task-specific experiences; it does not determine whether long-term friendships, organizational teams, family relationships, or large unstructured social networks would follow the same pattern. It also cannot isolate whether improved experience came from communication style, belonging, stereotype reduction, shared experience, or combinations of these processes.

Verbal Collaboration and Evaluative Tone

Another study by Foster and colleagues examined 136 adults assigned to four-person groups who worked together on a short building task. Preregistered analyses found differences in the positive and negative verbal evaluations that participants made in same- and mixed-neurotype groups. Mixed groups showed more expressions of negativity toward others, and autistic participants expressed more negativity in mixed settings. These findings add information about collaborative speech and social atmosphere beyond a single rapport score. Foster et al. (2025), Autism in Adulthood.

The task lasted approximately five minutes, participants knew the general diagnostic composition of the group, and verbal positivity is not interchangeable with causal influence, emotional empathy, or durable relationship quality. A study may identify systematic social-context effects without deciding which underlying cognitive mechanism produced them. This distinction is essential when translating experimental findings into workplace or clinical recommendations.

Instruction-Giving Across Neurotypes

A 2024 quasi-experimental study of expository communication examined pairs giving directions to reproduce a drawing. With 30 autistic and 28 non-autistic adults, matched pairs showed higher communication accuracy than mixed pairs, although clarity did not significantly differ and autistic matched pairs reported lower rapport than other pairs. Accuracy, speaking rate, clarity, and rapport therefore failed to collapse into one simple measure of social success. Oates et al. (2024).

This result directly illustrates why it is scientifically unsafe to infer the quality of one interaction dimension from another. A participant can give accurate instructions and still feel disconnected; another interaction can feel pleasant while leaving factual ambiguities. Clinical and occupational support should identify the actual task—emotional understanding, shared decision-making, collaboration, or precise transmission—before choosing a strategy.

The September 2026 Finding: No Advantage in Communication Efficiency

The most consequential recent challenge comes from De Laet, Nijhof, and Wiersema, whose article was published online on September 23, 2026. Their study examined 62 autistic and non-autistic adults performing a collaborative referencing task in matched and mixed dyads. Participants worked together to arrange abstract figures across repeated rounds. The design created 14 autistic-autistic dyads, 14 non-autistic–non-autistic dyads, and 30 mixed dyads; participants each completed interactions with same- and different-neurotype partners. De Laet et al. (2026).

Communication became more efficient with practice for all types of dyads. Contrary to the researchers' prediction, matched pairs did not communicate more efficiently than mixed pairs. The authors explicitly judged the result to challenge a prediction associated with double empathy theory. Their work creates a meaningful contrast with studies reporting greater rapport or more favorable experiential outcomes in matched groups. The scientific account must include both findings, explaining that improved subjective connection and improved measured task efficiency are different endpoints.

A single null study does not establish that partner mismatch never matters. The experiment selected adults within a specified age range and with IQ above 80, used a particular collaborative task, and measured performance over repeated trials. Equally, the presence of positive rapport results in other studies does not invalidate the null efficiency result. The intellectually productive question is why the effect of neurotype matching depends on what researchers ask people to do, who takes part, how learning occurs, and which outcome is analyzed.

Why a Negative Result Is Scientifically Informative

When an account makes a strong general prediction that same-neurotype communication should be more successful, an appropriately designed study failing to detect such an advantage narrows the range of defensible claims. Researchers may respond by examining whether neurotype mismatch predominantly affects feeling understood, being evaluated fairly, or comfort in unstructured settings rather than basic ability to establish task references. Such a refinement needs new preregistered tests, because a theory that is redefined after every contradictory result becomes difficult to evaluate.

The September 2026 study highlights the possibility that repeated collaboration allows pairs of different neurotypes to adapt to one another successfully. Social understanding may be a learned, relationship-specific process whose rate and quality depend on exposure, feedback, accommodations, and opportunities to repair misunderstandings. Testing this explanation would require studies designed to measure adaptation directly, compare repeated interactions, and record the strategies used by both partners.

The Fellowes 2026 Methodological Critique

Fellowes's 2026 review makes a different but complementary point: even studies that show an effect of neurotype composition often do not measure the empathy constructs named by the theory. The review considered 20 empirical papers and examined the relevance of the measurements used, the social situations created, and the duration available for empathy to be expressed. Most investigated outcomes were better characterized as rapport, flow, warmth, signaling, or interactive success than as cognitive or affective empathy. Fellowes (2026).

This is a construct-validity problem. If researchers predict that two groups differ in their ability to infer another person's emotional state, then observe only whether the groups rate conversations as enjoyable, the evidence addresses a related but distinct question. The outcome may be socially important and deserve publication. It should nonetheless be named correctly. Researchers who want to test reciprocal cognitive empathy must define how accurate inferences by both participants will be assessed; researchers who want to test affective empathy must explain how emotional engagement is elicited and measured.

The critique does not demonstrate that autistic people generally lack empathy, and it does not erase the observed interaction differences. Its substantive contribution is to separate claims about neurotype-associated social experiences from claims about underlying mental or emotional capacities. The distinction supports more rigorous, plural models in which communicative coordination, accuracy, affect, participation, and stigma are examined as separate measurable processes.

Is the Double Empathy Problem Proven?

The answer depends on the claim being evaluated. It is well supported that social interaction involves more than the traits of one participant, that autistic people can communicate effectively with other autistic people, and that the composition of an interaction can affect some outcomes. Specific experiments report stronger same-neurotype rapport and certain information-transfer advantages. A growing body of research also suggests that non-autistic observers and partners may contribute to misunderstandings or unfavorable judgments. Crompton et al. (2020); Jones et al. (2024).

More ambitious claims remain contested. Current research does not justify saying that all communication challenges in autism are caused by neurotype mismatch, that autistic and non-autistic people experience equal impairment in every interaction, that matched partners always communicate more efficiently, or that the same experimental results demonstrate preserved or impaired cognitive and affective empathy. The latter claims require operational definitions and studies that directly measure them. Fellowes (2026); De Laet et al. (2026).

A fair evidence classification is therefore differentiated. Bidirectionality as a methodological requirement for analyzing reciprocal encounters is an established principle of interaction research. Context-sensitive differences in rapport and some communication outcomes have direct experimental support, although generalizability is limited. Strong, universal causal accounts of empathy deficits or their complete absence are contested. Practical communication adaptations are reasonable when they address identified needs, even where a particular theoretical mechanism remains unsettled.

Alternative and Complementary Explanations

Social Familiarity and Shared Conventions

People may experience easier communication with partners whose conversational conventions they have learned. Similarities in directness, pacing, preferred level of detail, timing of responses, topic selection, and tolerance of silence can make coordination more predictable. Matching by autism diagnosis may sometimes correlate with these features, but diagnosis is not an exhaustive description of them. A strong study would measure relevant social practices and ask whether those practices, rather than diagnostic labels, explain the outcome.

Stigma, Expectations, and Perceiver Bias

Non-autistic participants may evaluate autistic behavior through norms that misclassify atypical expression as rudeness, disinterest, unfriendliness, or incompetence. Observer judgments are especially vulnerable to prevailing social expectations. The Jones study demonstrates why the perceptions of outside evaluators should be analyzed separately from the reports of those actually interacting. An intervention that reduces stereotyping may change the social consequences of an interaction without changing a person's underlying neurodevelopmental characteristics. Jones et al. (2024).

Language, Sensory Context, and Processing Demands

A noisy room, fast-paced group conversation, vague instruction, interrupted response, or unexpected shift of topic may complicate communication independent of whether the participants share a neurotype. Sensory stress, hearing or language difficulties, speech differences, and demands on working memory can interfere with mutual understanding. These influences should be assessed rather than being automatically attributed to empathy. NICE recommends clear professional communication, accessible information formats, and appropriate adaptations when working with autistic adults. NICE CG142.

Power, Institutional Expectations, and Unequal Consequences

An interaction between a patient and clinician, student and instructor, or employee and supervisor includes differences in role and authority. Even when both participants misinterpret one another, they may face very different consequences. Researchers must distinguish reciprocal processes from moral or institutional accountability. The double empathy framework gains explanatory depth when experiments measure whose interpretation is treated as normative, who is able to request clarification, and whether systems permit accessibility adjustments.

Individual Cognitive and Mental Health Differences

Real social difficulties can also reflect specific language or social-cognitive abilities, co-occurring anxiety, depression, ADHD, intellectual disability, trauma-related responses, or communication disorders. A relational model does not exclude these possibilities. Multidisciplinary assessment remains important when an individual has persistent difficulties affecting functioning. Both individual needs and interactional environments can influence outcomes at the same time.

What a Stronger Research Program Should Measure

Future research can evaluate the theory through a set of distinct predictions rather than one broad slogan. A study should specify whether it predicts more accurate interpretation of intentions, more empathic concern, stronger rapport, more complete information transfer, more efficient collaborative performance, or greater perceived belonging. Its research design should match the proposed mechanism and collect evidence from both interaction partners wherever possible.

Investigators should recruit samples representing different ages, genders, cultural contexts, intellectual abilities, verbal communication profiles, support needs, and diagnostic histories. Current experiments often rely on verbally fluent adults who can participate in short laboratory tasks. The experiences of non-speaking people, people using augmentative and alternative communication, people with intellectual disability, and people facing compounded exclusion deserve their own appropriate methods. Findings from a narrow laboratory cohort cannot be automatically generalized to these groups.

Research would be strengthened by preregistration, larger multi-site replication, transparent reporting of null outcomes, blinded coding where feasible, and measures validated for autistic as well as non-autistic participants. Where direct empathic accuracy is the target, both partners can report intentions or feelings at defined points and later assess the accuracy of the other's inference. Where interaction quality is the target, objective task performance and subjective experience should be recorded separately. Longitudinal designs can establish whether familiarity and feedback alter outcomes across weeks or months.

A further research priority is the causal effect of accommodations. Randomized or otherwise carefully controlled studies might compare explicit verbal instructions, visual information, adjusted sensory environments, advance agendas, conversational repair strategies, or training given to both partners. The primary outcomes should include the autistic participant's autonomy and preferences, rather than assuming that majority-conforming behavior is the goal. Participatory involvement of autistic researchers and community members can improve the relevance and accessibility of the measures.

Implications for Clinical Practice

Clinical professionals can use the relational insight without treating double empathy as a diagnostic test or established clinical treatment model. A clinician who notices a misunderstanding can check what the patient intended, make questions concrete, allow extra processing time, and invite corrections. The clinician can also consider whether the consultation's sensory and conversational conditions are suitable. NICE CG142 emphasizes respectful partnerships, adapted assessment procedures, and communication tailored to the person's needs. NICE CG142.

During autism assessment, an evaluator should still examine developmental history, the recognized diagnostic criteria, differential diagnoses, functioning, and co-occurring conditions. A conversation that feels easy or difficult with a particular interviewer cannot by itself confirm or exclude autism. Likewise, a screening questionnaire does not diagnose the condition. The double empathy framework is best treated as an additional question about how the assessment relationship might influence observed behavior and reported experiences.

Psychotherapy and clinical support should involve collaborative goal setting and adaptation to the individual's communicative and sensory preferences. Treating depression, anxiety, trauma-related symptoms, or another identified condition requires care appropriate to that condition. A double empathy explanation does not substitute for risk assessment, evidence-based treatment, or access to specialist care when needed. No medication is established to eliminate core autistic characteristics, and medications prescribed for co-occurring clinical conditions should be evaluated separately.

Implications for Education and Employment

In a classroom or workplace, an instruction may be accurate from the speaker's perspective yet leave room for several reasonable interpretations. Shared written expectations, defined priorities, examples, clear deadlines, opportunities to ask questions, and feedback that identifies the specific communication gap can improve coordination. Such measures are useful because they reduce ambiguity and dependence on implicit conventions, regardless of which broad theory eventually best explains the difficulty.

Organizations should not assume that an autistic employee's direct style shows hostility, that a person who avoids eye contact lacks attention, or that a fluent verbal presentation means the person has no support needs. They should also avoid assuming all autistic colleagues prefer identical interaction patterns. The appropriate accommodations follow the employee's actual tasks, preferences, sensory environment, applicable rights, and individualized support needs. NICE CG142.

A genuinely reciprocal approach can include helping non-autistic colleagues understand communication differences and encouraging all team members to check assumptions, rather than requiring only autistic people to imitate the majority's conventions. This proposal is an application of relational reasoning; the field still needs more intervention research to establish which strategies deliver measurable, durable improvements in different settings.

Friendships, Families, and Intimate Relationships

The double empathy framework can help partners identify a recurring cycle of interpretation: one person communicates directly, the other infers emotional meaning from style, the first person reacts to that inference, and both end up feeling misunderstood. Discussing the actual intended message and the assumptions used to interpret it may help clarify where meaning diverged. This practical reflection does not require both people to have the same experience, agree on every judgment, or share responsibility for harmful conduct.

Relationships develop through a history of mutual adaptation. The Heasman and Gillespie dyadic research suggests that measuring both parties' perceptions reveals patterns invisible in one-sided reports. However, effective support requires attention to safety, boundaries, consent, the possibility of coercion, and the different consequences of unequal social power. A theoretical emphasis on reciprocity must never be used to dismiss abuse or demand that a vulnerable person carry equal responsibility for someone else's behavior. Heasman and Gillespie (2018).

Where a recurring disagreement causes distress, identifying the concrete communication task may be more helpful than debating whether either partner possesses enough empathy. People can ask what signal was noticed, what it was taken to mean, what was intended, and how the interaction could be redesigned. Success can be judged by greater clarity, fairness, security, and each person's report of feeling understood.

Neurodiversity, Social Participation, and Interpretation

Neurodiversity refers to variation in human neurocognitive development and functioning and has also become an influential framework in research, disability policy, and social discourse. Neurodivergent is a broad nonclinical description rather than a diagnosis. Double empathy theory emerged within conversations that emphasize autistic people's perspectives and the need to study environments and relationships alongside individual characteristics. It is therefore relevant to disability access, professional practice, social belonging, and the ethics of psychological measurement.

The evidence does not require choosing between recognition of neurodevelopmental disability and recognition of interactive barriers. An autistic adult may experience clinically important communication difficulties and benefit from formal supports while also encountering misunderstanding produced by the assumptions of others. Disability can arise from multiple mechanisms, and practical outcomes depend on how an individual's characteristics meet real institutional environments.

The most useful contribution of the double empathy discussion is its insistence on specifying the unit of analysis. Researchers can analyze an individual capability, a dyadic process, a group norm, a health-care system, or the evaluations made by observers. These levels interact. A scientifically credible account identifies which level a study actually measured and resists turning an appealing social interpretation into a universal explanation of autism.

Frequently Asked Questions

Does the double empathy problem mean autistic people have more empathy for one another?

Some studies find stronger rapport or better information transfer in autistic-autistic interactions, which can indicate shared practices or improved coordination in those study conditions. They do not establish that every autistic person has greater cognitive or affective empathy toward every other autistic person. Empathy, rapport, and task accuracy must be measured independently. Fellowes (2026).

Does the theory say non-autistic people also misunderstand autistic people?

Yes. Reciprocal misunderstanding is central to the formulation. Studies of observers, mental-state interpretation, and family relationships show that errors and negative assumptions can occur on the non-autistic side. The degree of reciprocity varies across individual partners, measures, and situations. Edey et al. (2016); Jones et al. (2024).

Does the theory contradict autism diagnostic criteria?

It offers an alternative way to study certain social interactions rather than a replacement for the diagnosis. Clinical criteria continue to consider developmental patterns and relevant social-communication and restricted or repetitive features, with individualized assessment of needs. A relational explanation can coexist with autistic individuals' clinically important difficulties. NICE CG142.

Why do some studies support it and others challenge it?

The studies operationalize different outcomes and recruit different samples. A research design that measures how pleasant a conversation feels may produce one pattern; a design that measures efficiency in repeated collaborative references can produce another. In September 2026, De Laet and colleagues did not find an efficiency advantage from neurotype matching, a result that needs to be integrated rather than set aside. De Laet et al. (2026).

Is the double empathy problem a validated assessment tool or therapy?

It is a theory and research framework. There is no standardized double empathy diagnosis or established therapy that follows automatically from the hypothesis. Clinicians should use recognized diagnostic procedures, evaluate individual support needs, and offer accessible, evidence-informed care. NICE CG142.

Can the framework help in everyday conversations?

It can prompt people to clarify intentions, listen to how a message was received, distinguish direct meaning from inferred social meaning, and agree on communication expectations. These are practical ways of reducing ambiguity. The effectiveness of any particular intervention for autistic adults in a given setting still requires appropriately designed outcome research.

Conclusion: A Relational Question With a Developing Evidence Base

The double empathy problem has changed the scientific conversation by directing attention to mutual understanding, partner composition, observer expectations, and the social environments in which autistic people communicate. Research demonstrates that autistic adults can achieve meaningful rapport and successful information exchange, and that mixed-neurotype interactions sometimes carry additional challenges. These findings warrant investigating relationships and contexts rather than locating every social difficulty in the individual identified as autistic.

The evidence also requires careful limits. Strong rapport findings do not directly measure emotional or cognitive empathy. Communication-efficiency experiments do not always show a same-neurotype advantage, and the September 2026 collaborative referencing result provides a clear counterexample to an overly broad prediction. The 2026 methodological review further identifies how study duration, setting, and construct choice shape the claim a result can support. Fellowes (2026); De Laet et al. (2026).

The scientifically productive position is to test precisely which dimensions of shared understanding are influenced by neurotype mismatch, how individual variation and power shape those interactions, and which practical adaptations improve autonomy and participation. That approach preserves the theory's relational insight while keeping its empirical claims answerable to evidence.

Related Articles

Cluster Autism and Adult Neurodiversity — the parent scientific overview of adult autism and neurodiversity.

Autism in Adults: Characteristics, Diagnosis, and Support — clinical features, adult assessment, developmental context, and supports.

Autism Masking and Camouflaging: Strategies, Costs, and Context — the work of managing social expression and its possible personal costs.

Autistic Burnout: Signs, Causes, Recovery, and What the Evidence Says — related questions about demands, adaptation, and experienced exhaustion.

Autism and Empathy: What the Research Actually Measures — research on how empathy is defined and measured in autistic and non-autistic populations.

References

Crompton, C. J., Ropar, D., Evans-Williams, C. V. M., Flynn, E. G., and Fletcher-Watson, S. (2020). Autistic peer-to-peer information transfer is highly effective. Autism, 24(7), 1704–1712. https://doi.org/10.1177/1362361320919286

Crompton, C. J., Sharp, M., Axbey, H., Fletcher-Watson, S., Flynn, E. G., and Ropar, D. (2020). Neurotype-matching, but not being autistic, influences self and observer ratings of interpersonal rapport. Frontiers in Psychology, 11, 586171. https://doi.org/10.3389/fpsyg.2020.586171

Cusson, N. M., Meilleur, A. J., Bernhardt, B. C., Soulières, I., and Mottron, L. (2025). A systematic review and meta-analysis of empathy in autism: The influence of measures. Clinical Psychology Review, 120, 102623. https://doi.org/10.1016/j.cpr.2025.102623

De Laet, H., Nijhof, A. D., and Wiersema, J. R. (2026). Communication efficiency in autism is unaffected by neurotype match: Findings from a collaborative referencing task. Journal of Autism and Developmental Disorders. https://doi.org/10.1007/s10803-026-07466-2

Edey, R., Cook, J., Brewer, R., Johnson, M. H., Bird, G., and Press, C. (2016). Interaction takes two: Typical adults exhibit mind-blindness towards those with autism spectrum disorder. Journal of Abnormal Psychology, 125(7), 879–885. https://doi.org/10.1037/abn0000199

Fellowes, S. (2026). Cognitive empathy, affective empathy and empirical research on the double empathy problem: A critical methodology review. Journal of Autism and Developmental Disorders. https://doi.org/10.1007/s10803-026-07315-2

Foster, S. J., Ackerman, R. A., Wilks, C. E. H., Dodd, M., Calderon, R., Ropar, D., Fletcher-Watson, S., Crompton, C. J., and Sasson, N. J. (2025). Rapport in same and mixed neurotype groups of autistic and non-autistic adults. Autism, 29(7), 1700–1710. https://doi.org/10.1177/13623613251320444

Foster, S. J., Patel, S., Wilks, C. E. H., Dodd, M., Calderon, R., Ropar, D., Fletcher-Watson, S., Ackerman, R. A., Crompton, C. J., and Sasson, N. J. (2025). Verbal collaboration in same- and mixed-neurotype groups of autistic and non-autistic adults. Autism in Adulthood. https://doi.org/10.1089/aut.2024.0361

Heasman, B., and Gillespie, A. (2018). Perspective-taking is two-sided: Misunderstandings between people with Asperger's syndrome and their family members. Autism, 22(6), 740–750. https://doi.org/10.1177/1362361317708287

Jones, D. R., Botha, M., Ackerman, R. A., King, K., and Sasson, N. J. (2024). Non-autistic observers both detect and demonstrate the double empathy problem when evaluating interactions between autistic and non-autistic adults. Autism, 28(8), 2053–2065. https://doi.org/10.1177/13623613231219743

Milton, D. E. M. (2012). On the ontological status of autism: The double empathy problem. Disability and Society, 27(6), 883–887. https://doi.org/10.1080/09687599.2012.710008

National Institute for Health and Care Excellence. (2012; updated 2021). Autism spectrum disorder in adults: Diagnosis and management (CG142). https://www.nice.org.uk/guidance/cg142/chapter/Recommendations

Oates, M., Bean, A., Kickbusch, R., and Sauer, S. (2024). Extending double empathy: Effects of neurotype-matching on communication success in an expository context. American Journal of Speech-Language Pathology, 33(5), 2364–2377. https://doi.org/10.1044/2024_AJSLP-23-00393