Interoception in autism (ASD)

Atypical sensory processing is a DSM-5 diagnostic criterion of autism — but the well-known examples (hyper/hyposensitivity to touch, sound, pain) are all exteroceptive. Whether the internal sensory channel is also atypical, and in which direction, is a smaller and much less settled literature. This page holds what the wiki knows about it.

The evidence base is thin, and the earliest map of it is a census

DuBois et al. (2016), the field’s first dedicated review, ran a systematic search across three databases and 35 years and found five original studies measuring interoception in ASD. All had samples under 50; all were male-skewed, normal-IQ, verbal; and every one excluded comorbid developmental disability, though >50% of autistic individuals have one. So the whole literature describes a slice of the spectrum — high-functioning, verbal adults and children — and says nothing about the majority.

The two things that are genuinely contested

Accuracy splits by sample. Cardiac accuracy (tracking) was intact in autistic children (Schauder et al. 2014 — no group difference, and ASD children sustained attention to the heartbeat more constantly across durations) but impaired in autistic adults (Garfinkel et al. 2016). Quadt et al. (2018) report the same divergence and attribute it partly to comorbid alexithymia rather than autism per se — the standard confound on this literature (autism and alexithymia co-occur, and at least one cited study found atypical interoception tracked alexithymia, not autism).

Sensibility runs both ways — this is the load-bearing tension. The wiki’s ITPE material carries a “low accuracy + elevated sensibility” signature, from Garfinkel et al. (2016), who found autistic adults over-report their interoceptive ability (Porges BPQ) while performing poorly. But two of the three sensibility-relevant studies in DuBois’s review find the opposite: Fiene & Brownlow (2015, n=74 vs 228) found sharply decreased body and thirst awareness (large effects, d≈−1.0 to −1.3), and Elwin et al. (2012, qualitative) found lifelong hyposensitivity to interoceptive cues and pain. DuBois’s own overall read — a “slight tendency toward hyporeactivity in interoceptive awareness” — sides with the majority, against the elevated-sensibility premise the ITPE story needs. The wiki should carry both: the ITPE signature is real and reported, but it is one of two directions, and the review that names the most autism studies leans the other way.

The ITPE: a discrepancy, not a level

What partly dissolves the sensibility conflict is Garfinkel’s move to treat the accuracy–sensibility gap as the quantity of interest. The interoceptive trait prediction error (objective accuracy minus subjective sensibility) was enlarged in ASD in her study and predicted trait anxiety in both autistic and control samples — so the clinically relevant signal is the mismatch, not the sensibility level, which is exactly why studies disagreeing on the level can still be reconciled. Whether the ITPE survives when sensibility is low (Fiene & Brownlow) rather than high is untested: a large negative sensibility with low accuracy could shrink the discrepancy, which the ITPE framework has not confronted.

The neural picture: connectivity, not lesion

Autism supplies the wiki a rare lesion-free insula dissociation. Barttfeld et al. (2012) found that during an interoceptive state (focus on breathing/heart) functional connectivity increased in ASD relative to controls — the reverse of the exteroceptive task — with more compact coupling in cingulo-opercular, fronto-parietal and DMN systems, and the degree of interoceptive-state hyperconnectivity scaled with ADOS symptom severity. Resting-state work more broadly reports insula hypo-/hyperconnectivity — a reduced posterior→anterior insula response, salience- and DMN-network alterations — but the direction (hypo vs hyper) is unresolved and appears to depend on whether the task targets socioemotional or internal-sensory processing.

The framing: autism as a disorder of prediction

DuBois reads ASD interoception through predictive coding (Seth 2013; Seth, Suzuki & Critchley 2012): interoceptive prediction errors as a candidate substrate of the condition. This is the same “autism as a disorder of prediction” lens the wiki meets on the exteroceptive side in sensory-over-responsivity (Schwarzlose et al.), and it is developed formally in Seth & Friston (2016), who extend active interoceptive inference to autism. The through-line: if autistic sensory atypicality is a single precision/prediction failure, it should show on both sides of the skin — which raises the specificity worry that “interoceptive dysfunction in autism” may be one face of a domain-general sensory-precision difference, not an interoception-specific one.

What the literature has never done

• Measured a non-cardiac channel with any weight. Both physiological ASD studies used heartbeat paradigms, yet the review argues the evolutionarily salient cues — hunger, thirst, bladder urgency, GI discomfort, all elevated in autism — are where the effect likely lives. Fiene & Brownlow’s finding that autistic thirst awareness dissociated from actual fluid intake is the sharpest hint. See is-interoception-domain-general.
• Measured the non-verbal majority. Every construct is a verbal report of a counted or rated quantity (the field’s measurement bottleneck); the spectrum’s non-verbal, developmentally-disabled majority is unmeasured.
• Separated interoception from alexithymia cleanly. The confound is named everywhere and controlled almost nowhere.
• Run an intervention through interoception. Biofeedback, TMS/DBS and wearables are proposed; none has been tested as working via interoception.