Interoception

Interoception is the sense of what is happening inside the body: signals such as heartbeat, hunger, thirst, temperature, tiredness and fullness, and the bodily side of emotions.^1,2,3 One model, set out in 2015, splits it into three parts.¹ Interoceptive accuracy is how well a person does on an objective test, usually a heartbeat task. Interoceptive sensibility is how a person rates their own body sense on a questionnaire. Interoceptive awareness is whether a person's confidence tracks their actual accuracy.

In the study that set out this model, the three measures were distinct in 80 non-autistic adults, and accuracy was only partly predicted by the other two.¹ A meta-analysis of 22 studies found that the two main heartbeat tasks agree only weakly on accuracy, and that the awareness scores they produce do not agree at all.⁴ Studies that use different measures can therefore reach different answers, and the autism findings below often do. A 2019 paper argues that the model needs revising, to separate how interoception is measured (a task or a questionnaire) from what is measured (accuracy or attention); "awareness" has been used in more than one sense since.³¹ Newer questionnaires also ask about attention (how much a person notices body signals) and evaluation (whether they read those signals as good or bad).⁵

Most lab studies use the heartbeat. In heartbeat counting, people silently count their heartbeats over set periods without feeling for a pulse. In heartbeat discrimination, they judge whether a series of tones is in time with their heartbeat.^6,7

A 2023 meta-analysis pooled 15 studies with 467 autistic and 478 non-autistic participants.⁷ Autistic groups scored a little lower on heartbeat counting (g = −0.33, a small difference). They did not differ on heartbeat discrimination, though only four samples had used that task. Autistic participants were also more confident in their counting than non-autistic participants (g = 0.43) despite scoring lower, an estimate that also rests on four samples and that the authors call imprecise. A 2025 meta-analysis of five studies of adults, about three-quarters of them men, found a slightly smaller difference in the same direction (−0.21) that did not reach statistical significance (p = .06).⁸ The two meta-analyses draw on overlapping studies, so they are not independent checks. All of these tasks use the heart; a 2018 paper noted that evidence linking interoception to alexithymia came mainly from such cardiac tasks.³²

Individual studies point in different directions. One found lower accuracy in autistic participants.⁹ Another, of 46 autistic adults, found no difference, and its Bayesian analyses favored no difference.¹⁰ A later study from that group found lower accuracy in 21 autistic children but not in 21 autistic adults.¹¹ The 2023 meta-analysis found evidence against age group explaining why studies differ. How well the groups were matched on IQ explained about 38% of the variation between studies, short of statistical significance, and the authors recommend adjusting for IQ.⁷

Heartbeat counting is the most used task, and it has drawn strong criticism. In 572 people, more than 95% of counting scores reflected undercounting, and the number of beats people reported matched their actual heartbeats only weakly (r = .16).¹² When the same group told 123 people to report only the beats they actually felt, accuracy scores fell by about half. In the standard task, people described several strategies, one of which was estimating their heart rate.¹³ In 134 people, knowing roughly what a resting heart rate is statistically explained the link between intelligence and counting scores.¹⁴ In 48 people, counting scores were unrelated to a task that measured when heartbeat sensations were felt,⁶ and across 22 studies, counting and discrimination accuracy were only weakly related.⁴

This bears on autism research in two ways. Most of the evidence for a group difference comes from heartbeat counting, and groups that differ in IQ or in what they know about heart rates could score differently for reasons other than body sense.^7,14

When asked about their bodies, many autistic adults report difficulty. In one study, 74 autistic adults reported much lower body awareness and thirst awareness than 228 non-autistic adults, large differences (d = −1.26 and −1.02); the comparison group was mostly women.³ On a later questionnaire from the same group, the Interoception Sensory Questionnaire, 74% of autistic adults reported confusion about internal body states unless those states were extreme.¹⁷ A later check of that questionnaire in a large autistic sample found that its single-scale structure fitted the data poorly, and proposed a shorter 8-item version; that study had no non-autistic comparison group.³³

One study found the reverse: autistic adults rated themselves as more sensitive to body signals than non-autistic adults. A later study of autistic children and adolescents did not find this.^9,18 Across eight samples, the 2023 meta-analysis found no average difference in self-reported attention to body signals. The large spread between studies came mostly from that one study, which the meta-analysis treated as an outlier.⁷ One commentary argues that questionnaires showing higher sensitivity in autism mostly capture anxious attention to bodily symptoms.¹⁹

In posts written by adults who identified as autistic, many described limited awareness of hunger, fullness or thirst. Others described trouble noticing pain or illness, or telling harmless body signals from ones that need a doctor.²⁰

Alexithymia is difficulty identifying and describing one's own emotions.²¹ Across 15 studies using the Toronto Alexithymia Scale, autistic participants scored higher than non-autistic participants. In the 9 studies that reported how many people passed the scale's cut-off, about half of autistic participants (49.93%, ranging from 33% to 63% between studies) scored in the alexithymic range, against about 5% (4.89%) of non-autistic participants.²² The authors conclude that alexithymia is common in autism but not universal.²² A later study of 743 autistic and 721 general-population adults found that this questionnaire did not fit the data well in either group, and that a shorter 8-item score worked better; in the autistic sample, mostly women diagnosed in adulthood, its scores also tracked autistic traits.³⁴

Alexithymia overlaps with body sense more widely. In one study of two mostly female samples, one from the general population and one of people reporting a psychiatric diagnosis, people with more alexithymia also reported more trouble sensing non-emotional signals such as hunger and temperature. All the measures were self-report, and the questionnaires were new.²

The alexithymia hypothesis proposes that the emotional differences seen in some autistic people come from co-occurring alexithymia rather than from autism itself.²¹ Applied to interoception, it predicts that body-sense differences track alexithymia more closely than autism. In two small experiments, one of 38 non-autistic adults and one of 19 autistic adults with comparison adults matched on alexithymia, alexithymia predicted heartbeat-counting accuracy and autism did not.²³ The same group later ran three small experiments in non-clinical young adults: people with more alexithymic traits used breathing cues less and judged muscle effort and taste less accurately, and autistic-trait, depression and anxiety scores did not explain this. No autistic group was tested.³² A 2026 study from the same group tested expectation rather than accuracy, and used pain instead of the heartbeat. In 96 adults (28 of them autistic, 25 with marked alexithymia, and 55 with neither; the first two groups overlap), a sham treatment led people to expect less pain. More alexithymia went with a larger effect of that expectation on the pain they reported, and more trait anxiety with a smaller one; any effect of autistic traits was weaker and unreliable.³⁵ A study of 26 autistic and 26 comparison adults gives a mixed result: the autistic group scored lower on heartbeat accuracy and on a measure the authors call interoceptive awareness, and alexithymia was linked to awareness but not to accuracy. Autistic participants with alexithymia had lower empathy, and the authors suggest an alexithymic subgroup.²⁴

Other findings do not fit. A study of 137 non-autistic adults found no link between alexithymia and heartbeat accuracy, and its second experiment found no difference in 46 autistic adults, 27 of them with marked alexithymia, from non-autistic adults, with or without alexithymia.¹⁰ The 2023 meta-analysis tested whether studies whose groups differed more in alexithymia found bigger counting differences. Its text reports evidence against this, but its own table shows an inconclusive result.⁷ Across 66 samples, alexithymia was only weakly linked to interoception overall (r = −.16); the direction depended on how interoception was measured, and the link was clearer in clinical samples than in typically developing ones.²⁵ Much of the evidence on both sides uses heartbeat counting, with the problems described above; a few studies use other signals instead, such as breathing, muscle effort, taste and pain.^32,35

Anxiety is common in autistic adults: pooled estimates put current anxiety disorders at 27% and lifetime at 42%.²⁶ In a study of 20 autistic and 20 non-autistic adults, the gap between heartbeat-task accuracy and self-rated sensitivity, which the authors call a "trait prediction error", correlated with anxiety.^9,7 A study of 75 children and adolescents, half of them autistic, found the same link, though self-rated sensitivity on its own predicted anxiety best.¹⁸ Both studies measured everything once, so they cannot show which comes first; anxiety could also draw attention to the body. Across 55 studies, not specific to autism, heartbeat-task accuracy by itself was not linked to anxiety.²⁷ In general-population adults measured once, alexithymia statistically overlapped with part of the link between self-rated body sensitivity and anxiety.²⁸

In a trial at one center, 121 autistic adults were assigned to six sessions of heartbeat-feedback training (ADIE) or to a control therapy. Three months later, trait anxiety was lower in the training group (d = 0.30), and 31% met the trial's recovery criteria against 16% of controls. Only about half of participants gave data at three months, more in the training group (59%) than the control group (42%), and the missing data were estimated statistically. No other team has repeated the trial as of October 2026.²⁹

Autistic writers often describe missing the body's everyday signals. Megan Anna Neff, an autistic psychologist, writes that "Many Autistic people grapple with interoception" and names it, with masking, as one of two main barriers to self-care.^c3 Bianca Toeps describes autistic people who barely seem to feel pain or cold, who put off urinating until they have a bladder infection, or who forget to eat or drink for a whole day; for her it happens when her mind is busy with a photo shoot or with code.^c1 Megan Griffith links it to hyperfocus: "We might forget to use the restroom until it becomes an absolute emergency, or forget to eat until we're lightheaded, or forget about our other very important obligations."^c2

Devon Price gives the example of a non-autistic person who sees coworkers leave for lunch and checks whether they are hungry too, a link an autistic person absorbed in thought may not make. Price adds: "Many of us become quite adept at ignoring pain, just as we neglect our own hunger or thirst."^c4 Neff and Price both mention the wish to be "a mind in a jar" or a "floating brain in a jar".^c3,c4

Writers also describe what missed signals cost in health care. Pete Wharmby never seems to feel thirst until it is extreme, and finds pain scales hard to answer: without knowing what 10 would feel like, "the safest and most logical thing to do is hang around the average and say '5'", which a doctor may read as not serious.^c6 Carly Jones, an autistic advocate for autistic girls, writes that interoception differences can affect whether a girl knows "she is pregnant, if she is in pain, if she needs the toilet", and that some autistic girls avoid drinking at school because they cannot tell in time when they need the toilet.^c7,c8

Not every account is of too little signal. Wharmby also describes the reverse: a mind that fixes on every heartbeat, twinge and ache, which Wharmby calls "not quite the same as interoception" because it runs the other way, and links with worry about illness.^c9

Writers differ on where this comes from. In Unmasking Autism, Price says it is unclear how much is neurological and how much is a by-product of masking.^c4 In the sequel, Price goes further and argues that "our detachment from our bodies is a learned trauma response, not an innate disability."^c5 That is Price's argument; the research sections above do not test it.

The practical advice these writers give is similar: outside cues in place of inside ones. Neff suggests timers, visual schedules and a water bottle to finish by the end of the day, because "your body may not always register thirst and hunger cues"; Toeps has an assistant remind her to drink and keeps a refillable bottle within reach.^c3,c1 Neff also describes body scans and writes that interoception "is like a muscle in that the more you practice using it, the stronger you'll get!"^c3 Daniel Burford disagrees with that kind of exercise: a technique that begins "tune into your body" asks for "a signal that may not be reporting reliably", and Burford reports that most people on a self-compassion course built for autistic people found its body exercises brought distress up rather than settling it.^c10 Burford's alternative is an outside measure: "Buy a thermometer and believe it instead of your body."^c11 Luke Beardon, an autism educator, suggests planned meals for people who cannot feel hunger or fullness.^c12