Umwelten

Predictive processing and autism

🔬ResearchPeer-reviewed

6 min read

Summary

Predictive processing (PP) is a computational theory of the brain that says: the brain is not a passive receiver of sensory information. It is a prediction machine. It continuously generates models of the world, compares those predictions to incoming sensory data, and updates the models based on prediction errors — the difference between what was expected and what arrived. Under this framework, what we experience as “perception” is not the raw sensory signal; it is the brain’s best guess, shaped by the tug-of-war between prior expectations and sensory evidence.

Several researchers have proposed that autism involves a shift in this balance: the brain assigns more weight to sensory evidence and less to prior expectations, or equivalently, prediction errors are treated as more precise (more trustworthy) than they are in neurotypical brains. This single idea unifies a striking number of autistic experiences: heightened sensory sensitivity, attention to detail, discomfort with unpredictability, resistance to change, and the literal, concrete quality of autistic cognition.

This page introduces the framework and its application to autism. It does not endorse PP as the single correct theory — the evidence is actively debated — but it is currently the most comprehensive mechanistic account available.

The core idea

Priors, sensory evidence, and precision

In PP, every level of the brain has a model of what it expects to happen next. When the model’s prediction does not match the incoming sensory data, the mismatch — the prediction error — is passed up the hierarchy. The brain then either updates its model (learning) or suppresses the error (filtering). The decision about which to do depends on precision: the brain’s confidence in its prediction versus its confidence in the incoming data.

  • High prior precision → the brain trusts its expectations, filters out surprises, and experiences the world as predictable. Routine feels natural. Detail is smoothed over.
  • High sensory precision → the brain trusts the incoming data, passes errors up, and experiences the world as vivid, detailed, and potentially overwhelming. Surprises register strongly. Detail is foregrounded.

The autistic shift

The key proposal (Pellicano & Burr 2012, Van de Cruys et al. 2014, Lawson et al. 2014): in autism, sensory precision is set high. Prediction errors are treated as important. This means:

  • More detail reaches awareness. Things that neurotypical brains smooth over — background sounds, texture, the flicker of a fluorescent light — remain salient. This maps onto hyperresponsivity.
  • Surprises are harder to absorb. If every mismatch between expectation and reality triggers a strong error signal, the world feels less predictable. Routines and sameness reduce the load by making the world match its predictions.
  • Learning from regularities may differ. Building strong priors requires many prediction errors being absorbed and averaged over time. If each individual error is heavily weighted, the averaging process may be slower or noisier, leading to the slower habit-formation and the “every time is the first time” quality of some autistic experience.
  • Context modulates everything. The precision assigned to both priors and sensory evidence is not fixed — it is modulated by attention, arousal, safety, and environment. This is why the same autistic person can be overwhelmed by a supermarket and utterly focused at a darts match (see the SGL synthesis example in Hypo- and hyperresponsivity). PP predicts exactly this context-dependence.

The REBUS connection — psychedelics as the other end

Carhart-Harris & Friston’s REBUS model (Relaxed Beliefs Under Psychedelics, 2019) proposes that psychedelics reduce the precision of high-level priors — particularly self-referential models held in the default mode network — while leaving sensory-level processing relatively intact. This produces ego dissolution, enhanced flexibility, and perceptual vividness.

Roseby & Osborn Moar 2025 — Predictive processing as a lens on psychedelics and autism draws the comparison: both autism and psychedelic states involve a shift towards sensory information, but at different levels of the cortical hierarchy. In autism, the alteration is primarily at lower (sensory) levels; under psychedelics, primarily at higher (self-referential) levels. The comparison illuminates both without conflating them.

Why this matters for practice

For carers, parents, and teachers, the predictive processing framework reframes several everyday realities:

  1. Sensory sensitivity is not a volume knob turned too high. It is the brain treating its own predictions as less certain. This means the intervention is not “reduce the signal” but “make the environment more predictable” — which reduces prediction error at source.
  2. Resistance to change is rational. If your brain treats every deviation from expectation as a high-precision error, routines are not rigid preferences — they are load-management strategies.
  3. Context sensitivity is not inconsistency. A person who copes well in a familiar setting and falls apart in a new one is not malingering or choosing to be difficult. Their system assigns different precision to different contexts, and unfamiliar contexts produce more errors.
  4. Prikkelbalans is a PP concept in practical clothing. The four-zone framework in Prikkelbalans — Stimulus balance — green (matched capacity), orange (errors accumulating), red (overwhelm), blue (under-stimulation) — maps naturally onto a PP account of how prediction-error load fluctuates.

Open questions

  • Whether PP applies in the same way to autistic people with intellectual disability, where cognitive resources for generating and maintaining predictions are structurally different. Most PP-autism research is in adults without ID.
  • Whether the “high sensory precision” account explains hyporesponsivity as well as hyperresponsivity — or whether hyporesponsivity requires a different mechanism (e.g. system shutdown when prediction error load exceeds capacity).
  • The role of interoception in PP: if internal bodily signals are also subject to aberrant precision weighting, this could explain alexithymia, difficulty with emotional regulation, and poor interoceptive awareness in autism.
  • Whether the REBUS model’s predictions about 5HT₂A pathways will be confirmed by the PSILAUT trial (see Whelan et al. 2024 — PSILAUT).

Key sources

  • Pellicano, E., & Burr, D. (2012). When the world becomes “too real”: a Bayesian explanation of autistic perception. Trends in Cognitive Sciences, 16(10), 504–510.
  • Van de Cruys, S., et al. (2014). Precise minds in uncertain worlds: predictive coding in autism. Psychological Review, 121(4), 649–675.
  • Lawson, R.P., Rees, G., & Friston, K.J. (2014). An aberrant precision account of autism. Frontiers in Human Neuroscience, 8, 302.
  • Carhart-Harris, R.L., & Friston, K.J. (2019). REBUS and the anarchic brain: toward a unified model of the brain action of psychedelics. Pharmacological Reviews, 71(3), 316–344.
  • Roseby & Osborn Moar 2025 — Predictive processing as a lens on psychedelics and autism

Graph View

Backlinks