The everyday account of sensory overload is a volume metaphor. Lights are too bright, sound is too loud, a label in a collar is unbearable — as though the gain has been turned up across the board. It is a useful shorthand and the research does not support it in that form. What the studies find is both more specific and harder to summarise, and the disagreement between two ways of measuring it is the most interesting thing in the literature.

What Questionnaires Describe

Most of what is known at scale comes from behavioural report — parents, teachers and autistic people themselves describing responses to everyday sensory environments.

A 2026 synthesis of research across the lifespan finds that autistic youth show elevated and variable sensory responsivity. The variability is not noise: hypersensitivity predicted internalising symptoms, while sensory seeking was linked to externalising behaviours. Those are different profiles with different consequences, not degrees of one thing.

It organises this using an existing sensory taxonomy rather than inventing categories, which matters for comparing findings across studies that would otherwise describe overlapping things in incompatible language.

It also reports that affective reactivity is consistently elevated across cultures and associated with anxiety and with caregiver stress — and, notably, that sensory seeking may function as a coping mechanism rather than a symptom (Current Psychiatry Reports, 2026).

What the Laboratory Finds Instead

Psychophysics — measuring perceptual thresholds directly under controlled conditions — returns a different picture.

The same review reports domain-specific differences: reduced tactile adaptation, altered motion noise exclusion, and enhanced pitch discrimination.

Read those three together. One is a difference in how quickly a constant touch stops being noticed. One concerns separating signal from visual noise. And one is a measurable advantage in hearing fine differences in pitch.

That last one does not fit an overload model at all. It is better performance on a perceptual task, and any account of autistic sensory processing that cannot accommodate it is incomplete.

The Tension, Stated Plainly

The review is explicit that these perceptual findings often show limited correspondence with questionnaire measures.

That is an uncomfortable result and it is important. The measure that scales — asking people — and the measure that is precise — testing thresholds — are not tracking the same underlying quantity.

There are several ways that could be true at once. Questionnaires may capture how manageable an environment is rather than how it is detected. Laboratory tasks isolate single modalities under controlled conditions, which is nothing like a supermarket. And what makes a sensory environment difficult may be less about detection thresholds than about accumulation, unpredictability, or the inability to leave.

None of those is established. The honest statement is that two methods disagree and the field has not resolved why.

What the Neurophysiology Adds

A 2026 meta-analysis went looking for the underlying signals, pooling early event-related potentials and event-related fields measured with EEG and MEG across studies published from 1980 onwards.

It included 145 studies — 3,778 autistic and 3,484 non-autistic participants — with moderator analyses for technique, sensory modality, age group, sex and language impairment (European Child and Adolescent Psychiatry, 2026).

The moderator analyses are the part worth noting. The reviewers tested whether results varied by recording technique, by sensory modality, by age group, by sex and by language impairment — because a pooled average across studies that differ on all five would be close to meaningless. Asking which of those actually moves the result is how a meta-analysis earns its conclusion rather than merely stacking studies.

Its framing is candid: while sensory processing alterations in autism are well documented, the neurophysiological correlates remain unclear. A meta-analysis of that size opening with that sentence is telling you where the field actually is.

Why the Distinction Matters Practically

Because ‘oversensitive’ invites the wrong response. If the model is a volume knob, the implied fix is reduction — quieter, dimmer, less — and the implied explanation is that the person is fragile.

Domain-specific differences point elsewhere. Reduced tactile adaptation means a sensation that others stop noticing keeps being noticed, which is a difference in habituation rather than in threshold. Altered motion noise exclusion is a difference in filtering signal from clutter. Enhanced pitch discrimination is not a deficit in any direction.

The caregiver finding deserves its own line. Affective reactivity was associated not only with anxiety in the autistic person but with caregiver stress — meaning the sensory environment is a shared problem in a household rather than an individual one, and that support aimed only at the person misses part of what is happening.

And the finding that sensory seeking may serve as coping reframes a behaviour often treated as a problem to be reduced. A description that fits the evidence is more useful than one that fits the metaphor, which is the same reason filtering deserves its own account.

What This Evidence Does Not Establish

It does not describe any individual. Elevated and variable is the finding: profiles differ between people, and group averages are a poor guide to one person’s experience.

It does not settle mechanism. The largest meta-analysis available says the neurophysiological correlates remain unclear, and that should be taken at face value rather than filled in.

It does not adjudicate between the two measures. Questionnaires and psychophysics disagreeing is a documented problem, not a resolved one, and neither has been shown to be the more valid account of lived difficulty.

And it is not diagnostic. Nothing here identifies autism, and no sensory measure is used to.

What it does support is a better description: sensory differences in autism are real, measurable, specific rather than global, sometimes advantageous, and consequential for mental health and for the people around an autistic person.