Autismology AutismologyAutistic studies

Autism and ASD: Current State of Knowledge and Main Theories Under Discussion

Other approaches

B — Critical examination

v1.22026-10-09
0:00 / 0:00

Object of study, levels of explanation, causality, and contradictions

B.1 — A clinical category is not automatically a natural essence

An ASD diagnosis can be useful for identifying needs, organizing services, forming research cohorts, and facilitating access to rights. That usefulness does not, by itself, demonstrate that the category corresponds to a homogeneous natural entity with a single mechanism.

The issue becomes methodological when one moves, without saying so, from:

“these people meet the same diagnostic criteria”

to:

“they necessarily have the same thing at the same causal level.”

That transition is not logically required.

B.2 — Heterogeneity can be both a scientific finding and a signal about how the object is constructed

The literature rightly emphasizes the heterogeneity of diagnosed populations. But two possible interpretations must be distinguished.

The first is that autism itself is extraordinarily heterogeneous.

The second is that the category used to construct the groups may bring together several different phenomena, which may share some manifestations without sharing all of their mechanisms.

These two hypotheses are not equivalent.

When successive studies find very different genetic, neurobiological, cognitive, and developmental profiles, heterogeneity should not only be statistically “corrected”; it may also provide information about the relevance of the initial category.

B.3 — Precisely measuring a category does not guarantee that the category is correctly constructed

Research can be technically excellent and produce accurate measurements while studying an object that was defined too broadly or too compositely.

This is a general methodological point: precision of measurement does not compensate for uncertainty about what has been grouped together.

If a category combines autistic functioning, adaptive difficulties, associated conditions, intellectual disability, reactions to the environment, and very different levels of support needs, then the search for one gene, biomarker, or unique mechanism for that category will inevitably encounter a great diversity of results.

That does not mean the data are false. It means their interpretation depends on the quality of the object defined at the outset.

B.4 — Observation, interpretation, and explanatory hypothesis

A large part of the controversies can be clarified by separating three levels.

  1. Observation: a result was measured in a task or a group.
  2. Interpretation: the result is described as a deficit, preference, strategy, compensation, increased precision, rigidity, and so on.
  3. Explanatory hypothesis: a broader mechanism is proposed to account for several observations.

For example, better performance on a local perceptual task is an observation. Describing it as “superior local processing” is already an interpretation. Making it the consequence of a particular neurodevelopmental organization is an additional hypothesis.

The distinction does not prohibit any theory; it only prevents one level from being presented as though it automatically contained the next ones.

B.5 — Correlation, association, and causality

Genetic associations illustrate this problem particularly well.

A variant may be statistically more frequent in a diagnosed population without being a unique cause of the diagnosis. A variant associated with several different developmental presentations does not justify concluding that it specifically produces “autism” as a single-mechanism disease.

The same principle applies to imaging, EEG, metabolism, or the microbiome: an average difference between groups does not automatically establish a complete causal chain.

The proper question is therefore not only “is there a difference?”, but also:

  • in which population?
  • with what effect size?
  • does this difference precede the phenomenon being studied?
  • is it specific?
  • is it necessary or sufficient?
  • could it be a consequence, an adaptation, or an indirect correlation?

B.6 — Several levels of analysis are often placed on the same plane

The source document juxtaposes genetics, neurophysiology, perception, categorization, social interaction, identity politics, and support needs. All of these domains are relevant, but they do not answer the same question.

A genetic variant, an EEG latency, a perceptual preference, a categorization strategy, a conversational difficulty, and an administrative category are objects at different levels.

A theory becomes fragile when it claims to move directly from one level to another without a sufficiently established intermediate mechanism.

Saying that a phenomenon is “biological” does not replace an explanation of how it becomes an experience, a behaviour, or a concrete difficulty.

B.7 — Vocabulary can introduce a conclusion before the analysis

Apparently technical words sometimes carry an implicit interpretation.

“Autism risk” presents autistic existence itself as the undesirable event whose occurrence should be prevented. When a study actually measures a statistical association with the probability of receiving an ASD diagnosis, it is more precise to describe that association.

“Severity” may summarize a useful operational score, but it becomes ambiguous unless one specifies what is severe: support need, communication difficulties, epilepsy, intellectual disability, anxiety, functional impact, or deviation from a behavioural norm.

“Deficit” may be legitimate when a measured mechanism or capacity is genuinely reduced. But the same result can sometimes be described as a different priority, a processing style, or a strategy. The word should not decide among those possibilities in advance.

B.8 — Subcategories can move the problem without solving it

“Syndromic / non-syndromic,” “idiopathic / secondary,” “essential / complex,” “prototypical,” “profound”: each of these distinctions addresses a real problem.

But multiplying subcategories does not automatically guarantee better understanding.

A subcategory may be useful:

  • for identifying a genetic etiology;
  • for comparing developmental trajectories;
  • for describing support needs;
  • for constructing a more homogeneous research protocol.

It becomes problematic when it is treated as a new “species” without independent evidence that the boundary corresponds to a distinct mechanism.

B.9 — Why apparently contradictory theories can all find results

Weak central coherence, EPF, Bayesian models, predictive processing, categorization models, and “enhanced rationality” can appear incompatible if each is read as a general theory of autism.

But many of these studies actually concern different tasks, different levels, and different subgroups.

One person may simultaneously show:

  • strong perceptual discrimination in one task;
  • different use of context in another;
  • good explicit learning ability;
  • difficulty in a particular probabilistic-learning task;
  • lower susceptibility to certain decision biases;
  • and substantial difficulties in real-life social interaction.

These observations are not necessarily contradictory. The conflict appears when each is converted into a total explanation.

B.10 — Group studies do not automatically describe an individual

A statistically significant mean difference can coexist with substantial overlap between groups.

If the mean of an autistic group differs from that of a non-autistic group, this does not mean that every individual in the first group has the characteristic, nor that no individual in the second group has it.

Research must therefore avoid turning a probabilistic group difference into an essentialist individual description.

This is particularly important in public communication, where modest findings may be simplified into claims such as “the autistic brain does X.”

B.11 — Research instruments may incorporate the frame of reference they claim to measure

A social or cognitive task is never entirely neutral. It depends on instructions, conventions, a response format, the time allowed, a definition of the correct answer, and a laboratory context.

A test built from majority functioning may measure both the capacity of interest and familiarity with the format or with non-autistic expectations.

Research on Raven/Wechsler profiles and on double empathy shows precisely that the result may change when the tool or the interaction partner changes.

This does not make tests useless. It requires analysis of what they actually measure.

B.12 — Environment and interaction have long been secondary variables

Much clinical research begins with characteristics observed in the individual. That is logical for an individual diagnosis, but less sufficient for understanding phenomena that arise within an interaction or an environment.

A communication difficulty may depend on:

  • the autistic person;
  • the interaction partner;
  • the channel used;
  • noise or lighting;
  • the processing time allowed;
  • predictability;
  • implicit expectations;
  • reciprocal errors of interpretation.

Double empathy made this issue visible in the social domain. A similar approach may be necessary in other domains.

B.13 — The non-autistic frame of reference often remains implicit

Many expressions compare the autistic person with a norm without studying that norm with equal intensity.

“Weak coherence,” “restricted interest,” “rigidity,” “social atypicality,” “lack of flexibility,” or “abnormality of contact” describe a deviation. But a deviation only has meaning relative to a point of comparison.

Majority functioning is often treated as the zero point of the instrument: it does not need to be explained, while every deviation requires an explanation.

Scientifically, this asymmetry deserves to be made explicit. A rigorous comparison should be able to describe the properties of both forms of functioning and their interactions, rather than analyzing only one side of the difference.

B.14 — What can be retained without accepting every interpretation

Critical examination does not lead to rejecting existing research. On the contrary, it allows its findings to be used more effectively.

For each theory or result, one can distinguish:

  1. the probably robust observation;
  2. the interpretation proposed by the authors;
  3. assumptions that may be unnecessary or insufficiently demonstrated.

For example, a difference in susceptibility to certain perceptual illusions may be a robust result, while its interpretation as a deficit or as a superiority may remain debatable.

Likewise, a genetic association may be highly robust without providing a definition of autism.

This separation avoids two opposite errors: mechanically accepting all discourse produced within a medical framework, or discarding useful scientific data simply because an interpretation appears inadequate.

B.15 — Conclusion of the critical examination

The main problem in the current state of knowledge is not a lack of data. The data are numerous and often sophisticated.

The problem is determining which data concern which object, at what level, and within which frame of reference.

The diversity of findings may reflect the real complexity of human development. It may also reveal that several phenomena have been brought together under the same clinical category and that different levels of explanation are then compared as though they answered a single question.

Part C examines what this landscape becomes when the method of Autismology is explicitly taken as the starting point.



B.15bis — Detailed methodological precautions preserved from V1.2

What group studies can and cannot say

Most research compares groups.

If an autistic group obtains, on average, a different result from a non-autistic group, this does not mean that every person in the first group has that characteristic.

Two distributions can overlap extensively while still having statistically different means.

This is especially important for concepts such as:

  • sensory sensitivity;
  • central coherence;
  • Bayesian priors;
  • rationality;
  • intelligence;
  • categorisation;
  • empathy.

A serious theory of autism should therefore avoid turning an average difference into a universal individual essence.

Association, correlation and causality

Another frequent pitfall is to confuse three levels.

Association: a characteristic appears more often with autism.

Correlation: two measures vary together.

Causality: one characteristic directly produces the other.

For example, finding a genetic variant that is more frequent in autistic people does not necessarily mean that the variant “causes autism” in a simple deterministic way.

Likewise, observing a brain difference in autistic adults does not always make it possible to know whether that difference is an early cause, a developmental consequence, an adaptation, the effect of different experiences, or a combination of these factors.

Why theories of autism often contradict one another

Several factors explain the apparent contradictions.

The groups studied are not the same.

The tasks do not measure exactly the same thing.

Children and adults may present different phenomena.

People with and without intellectual disability are sometimes mixed together or, conversely, excluded.

Samples have historically included far more boys and men.

Diagnostic criteria have changed over time.

Researchers use different theoretical models to interpret the same result.

Negative studies are sometimes published less often.

Sample sizes are often small.

Some theories are formulated at a very general level but tested with very narrow tasks.

Research from 2020–2026 increasingly tends to acknowledge these problems of heterogeneity and reproducibility explicitly.

What is relatively robust in 2026

Without claiming absolute consensus, several propositions are now relatively robust.

  • ASD diagnostic criteria concern characteristics that emerge during development, even if their identification may occur late.
  • Current ASD diagnosis is clinical and developmental, not based on a single biomarker.
  • In populations diagnosed with ASD, the genetic contributions identified are substantial but extremely heterogeneous.
  • Some presentations meeting ASD criteria are observed in the context of identifiable syndromes or genetic variants.
  • Sensory, cognitive, language and adaptive profiles are highly variable.
  • Some perceptual tasks show superior performance in some autistic groups; others show difficulties.
  • Differences in context processing, categorisation, prediction or decision-making are real in some tasks, but they do not yet form a universal theory.
  • Social-interaction difficulties can no longer be studied solely as isolated properties of the autistic person; the role of the interaction partner and context is now a legitimate research object.
  • Support needs vary enormously and cannot be inferred from a single measure.

Important but non-definitive models

Enhanced Perceptual Functioning: an influential model supported by several perceptual and imaging findings, but not universal.

Veridical mapping: a productive proposal for some forms of learning and talent, but not a general theory.

Weak central coherence: a historical model progressively reformulated as a cognitive style; useful but insufficient on its own.

Bayesian / predictive processing: an important family of models, with partial and highly heterogeneous empirical support.

Enhanced rationality: a set of interesting findings in some decision-making tasks, but not a general superiority.

Double empathy problem: a major relational correction to the one-way social-deficit model, with growing empirical evidence but mechanisms still under discussion.

Prototypical autism: a recent and structured proposal, but not validated as a new consensual clinical category.

Asymmetric developmental bifurcation: a recent theory of developmental variation; highly speculative at this stage.

Scientific questions that remain open

Several fundamental questions remain.

Is there one or are there several mechanisms common to all people currently diagnosed as autistic?

Does the current diagnostic category group together several different developmental realities?

Can reproducible subgroups be identified without creating arbitrary boundaries?

How much of perceptual difference comes from the sensory level itself, and how much from contextual weighting, attention, learning or experience?

How do perceptual characteristics develop into differences in language, behaviour and social interaction?

Are differences in priors global or specific to certain domains and types of learning?

Which characteristics are primary and which result from adaptation to the environment?

How can people with little spoken language be studied scientifically without confusing lack of response, lack of understanding, and motor or communication difficulties?

What is the best level of analysis: genes, neural networks, perception, cognition, development, interactions, or a combination of these levels?

An important precaution concerning the word “explanation”

No theory presented in this document should be described as “the explanation of autism”.

A theory may explain one phenomenon rather than autism as a whole.

EPF explains some perceptual findings.

Predictive processing attempts to explain some relationships among perception, expectations and uncertainty.

Double empathy explains part of relational difficulties.

Genetics explains an important part of the statistical variation in the probability of an ASD diagnosis and some identifiable aetiologies.

“Profound autism” organises functional needs.

These levels are not interchangeable.


Sources and cited documents

S1 — WHO Clinical Descriptions and Diagnostic Requirements for ICD-11 Mental, Behavioural or Neurodevelopmental DisordersWorld Health Organization2024Open source ↗
S5 — Genetic contributions to autism spectrum disorderGenovese A; Butler MG2020/2021Open source ↗
S14 — Enhanced perceptual functioning in autism: an update, and eight principles of autistic perceptionMottron L; Dawson M; Soulières I; Hubert B; Burack J2006Open source ↗
S16 — The level and nature of autistic intelligenceDawson M et al.2007Open source ↗
S22 — When the world becomes too real: a Bayesian explanation of autistic perceptionPellicano E; Burr D2012Open source ↗
S26 — Category learning in autistic individuals: A meta-analysisWimmer L et al.2024Open source ↗
S30 — Autism and the double empathy problem: Implications for development and mental healthMitchell P; Sheppard E; Cassidy S2021Open source ↗
S31 — Do you feel me? Autism, empathic accuracy and the double empathy problemCheang RTS et al.2024Open source ↗
S38 — State of the Field: Differentiating Intellectual Disability From Autism Spectrum DisorderThurm A et al.2019Open source ↗