Alexithymia
Literally “no words for feelings” — a stable trait of reduced ability to identify and describe one’s own emotional states, and to distinguish emotions from the bodily sensations that accompany them. Quadt et al. (2018) give it a load-bearing role: it is repeatedly the variable that, once controlled, explains away interoceptive deficits credited to other conditions.
The confound that keeps reappearing
The pattern the review documents:
- Autism. Autistic adults show reduced heartbeat-tracking accuracy — but at least one study (Shah et al. 2016) finds the deficit tracks alexithymia, which is highly comorbid with autism, rather than autism per se. On this reading the emotion-processing deficits of autism, characterized by high alexithymia, may be the principal driver of its interoceptive impairment. A recent study links impaired interoceptive awareness (not sensitivity) to autistic traits, alexithymia, and empathy together (Mul et al. 2018).
- Non-autistic populations. High alexithymia is associated with impaired interoceptive accuracy independent of autism (Brewer, Cook & Bird 2016), whose title states the strong form: alexithymia as “a general deficit of interoception.”
- Eating disorders. Alexithymia (with depression and anxiety) is a comorbidity that ED interoception studies “often do not take into account”; one bulimia study found the apparent accuracy deficit vanished once comorbid alexithymia, depression and anxiety were covaried out.
The methodological lesson the review draws is that a great deal of “condition X impairs interoception” may be “alexithymia impairs interoception, and alexithymia rides along with condition X.” Whether the interoceptive deficit is primary to alexithymia or a shared cause is the open form of the claim.
Why it matters theoretically
If emotions are, per construction and interoceptive inference, categorizations built on interoceptive signals, then a person who cannot read their bodily signals well should struggle to identify their emotions — which is exactly the alexithymic profile. Alexithymia is thus a natural test case for the interoception→emotion link: the Brewer/Bird claim that alexithymia is fundamentally an interoceptive deficit is the constructionist prediction stated as a personality trait. It also connects to empathy: reduced capacity to read one’s own affect co-occurs with reduced reading of others’, consistent with the perception-action model in which empathy re-uses self-directed interoceptive machinery.
Acquired alexithymia: the trait produced by a lesion
Bonaz et al. (2021) add a form of the construct this page did not have, and it bears directly on the causal question above. Among the specific emotional and motivational deficits produced by acquired focal lesions to the insula, they list acquired alexithymia — alongside disgust insensitivity, psychopathy, and loss of drug-craving.
Alexithymia is otherwise treated throughout this page as a stable personality trait of unknown origin, comorbid with autism and eating disorders, and possibly identical to a general interoceptive deficit. The lesion cases say the phenotype can also be produced, in a previously unaffected adult, by damage to the cortex that re-represents bodily state. Two consequences:
- It is the strongest available support for the Brewer/Bird direction of causation — that alexithymia is fundamentally an interoceptive deficit rather than a correlate of one. Damage the interoceptive cortex; get the alexithymic profile. The developmental trait and the acquired syndrome need not share a mechanism, but the acquired case shows the interoceptive route is sufficient.
- It gives Bonaz et al.’s neurological section its most direct relevance to the wiki’s constructionist material: if emotion words require interoceptive content (emotion-prototypicality, ferre-2024-emotion-prototypicality) and categorization over interoceptive signals (theory-of-constructed-emotion), then removing the signal’s cortical representation should cost a person their emotion vocabulary. That is what “no words for feelings” describes.
Held at one remove, as with everything else on this page: Bonaz et al. give this a clause, and the primary insular-lesion literature is not in raw/.
The trait seen as a continuum in ordinary talk
Carter & Ogden (2023) add a qualitative sighting: interviewing twelve adults about their bodily states, they find insight into how emotion colours the body varies from acute (“I’m fully aware of the extent to which depression can make you feel physically ill”) to minimal, and read this range explicitly as reflecting alexithymia (citing Bird/Bagby). Their own prior work supplies a health-psychology construct for the same territory — “Interoceptive Crossover” (Carter & Ogden 2020, not in raw/), individual differences in how far a person perceives emotions and physical symptoms as similar. That is the alexithymic difficulty-separating-feelings-from-bodily-states factor (§ above) approached from symptom perception rather than from the TAS-20, and a plausible mediator of both the alexithymia/interoception link and the gut-centred emotion localisation those interviews report. Held loosely: the 2020 quantitative source is unread here, and the 2023 evidence is interview talk, not a measure.
The construct measured first-hand at last — and it does not separate from bodily self-report
Everything above is carried by review. Ventura-Bort, Wendt & Weymar (2021) is the wiki’s first source to administer the TAS-20 (α = 0.83 in 109 adults), and to do so alongside dedicated interoceptive-sensibility scales — the IAS, the ICQ, and the maia.
The zero-order correlations are large and run the way the Brewer/Bird claim predicts: Difficulty Identifying Feelings ↔ ICQ (difficulty interpreting non-affective bodily states) r = .55, and DIF ↔ IAS (believed interoceptive accuracy) r = −.55. In the factor analysis, all three TAS-20 subscales load negatively on the same component as the IAS and MAIA Attention regulation/Trusting — a single dimension of “I can tell what is going on inside me,” on which emotional and bodily items are not distinguishable.
Two readings, and the difference between them is the whole causal question this page carries.
The substantive reading. This is “alexithymia is a general deficit of interoception” (Brewer, Cook & Bird 2016) demonstrated psychometrically at trait level: difficulty identifying feelings and difficulty identifying bodily states are one factor, not two correlated ones.
The deflationary reading, which the wiki should weight at least as heavily. These are all self-report instruments completed in one sitting by the same people, so shared method variance is a complete alternative explanation. A person disposed to endorse “I don’t know what’s happening in me” will endorse it about hunger and about sadness. Nothing here measures whether either judgment is true — no heartbeat task, no physiological recording, no objective emotion-identification criterion. That is exactly the gap this page names in its final paragraph, and this study does not close it; it makes it more visible.
The useful residue: the alexithymia/interoception relationship is now known to hold at the level of beliefs, which was not previously demonstrated first-hand here, and which is a weaker claim than the one the reviews above report. It also puts alexithymia in a specific place in the interoception vocabulary — reversed, the TAS-20 is a sensibility measure in the believed-accuracy sense, and on the same factor as the well-being correlations, not on the attention factor. See emotional-granularity for the one behavioural (non-questionnaire) outcome the factor predicted.
And the objective-task version of the claim is null
The belief-level result above has an exact complement on the performance side. Desmedt et al. (2022), meta-analysing 23 studies, find heartbeat-counting accuracy and alexithymia share 0.01% of variance (r = −0.01, ns) — the weakest of the seven associations in that meta-analysis. So the Brewer/Bird “alexithymia is a general deficit of interoception” claim splits cleanly by measurement level: it holds where interoception is believed (self-report factors together with the TAS-20, as Ventura-Bort shows) and vanishes where interoception is performed (the objective cardiac task). That is exactly the pattern the deflationary reading predicts — shared method variance among questionnaires, nothing in the behaviour — and it is why the causal question on this page stays open. The strong claim survives only if the HCT is too invalid to test it (is-the-heartbeat-counting-task-valid), which is not a comfortable place for the claim to have to stand.
The earliest form of the claim, and its embodiment framing (Herbert & Pollatos 2012)
The Brewer/Bird “general deficit of interoception” direction has a Tübingen predecessor the wiki can now name. Herbert & Pollatos (2012) reframe alexithymia not as an emotion-classification deficit but as a disturbance in translating bodily signals to conscious awareness — a disturbance of embodiment in Fuchs & Schlimme’s sense (see embodied-selfhood). They cite their own finding (Herbert et al. 2010a/2011) that high alexithymia goes with low interoceptive awareness on a heartbeat-perception task, and read it through Lane et al.’s (1997) “affective agnosia” — a present emotional bodily response that is not experienced as a feeling.
This lands squarely in the measurement-level split the two sections above establish, and on the wrong side of the objective one: it is a heartbeat-perception (counting-type) association, and the later meta-analysis (Desmedt et al. 2022) puts the alexithymia ↔ heartbeat-counting-accuracy relationship at essentially zero. So the Herbert result is best read as the belief/attention-flavoured version of the claim stated early and confidently, of a piece with the Ventura-Bort questionnaire factor rather than with any objective deficit. Its durable contribution is the framing — alexithymia as failed translation of body into feeling — not a new data point; and it is held at one remove, the 2010a/2011 primaries being outside raw/.
Which half of self-report it is a deficit in — and it is the belief half, not the attention half (Murphy et al. 2020)
The two sections above establish a clean measurement-level split: the Brewer/Bird claim holds where interoception is believed and vanishes where it is performed. Murphy et al. (2020) — the Bird/Catmur lab itself — narrow the surviving half considerably, and the narrowing is not in the claim’s favour rhetorically but is in its favour evidentially.
Self-reported interoception divides into believed accuracy and deployed attention (see interoceptive-taxonomy). The TAS-20 tracks the first and not the second:
| correlation with TAS-20 | |
|---|---|
| IAS (believed accuracy) | r = −.430 (n = 76), −.572 (n = 35) |
| ICQ (believed accuracy, reversed) | r = .648 |
| BPQ (deployed attention) | r = .079, .067 — ns |
Three things this adds.
It is not merely “self-report correlates with self-report.” The deflationary shared-method-variance reading recorded above predicts the TAS-20 to correlate with any body questionnaire completed in the same sitting. It correlated with two and not with a third. Whatever produces the association is specific to believed accuracy, which is a stronger result than Ventura-Bort’s one-factor loading on its own.
It survives the obvious third variables. Partialling self-esteem left the TAS-20↔IAS relationship intact (partial r = −.255, p = .027) — the control included because a general tendency to describe oneself as impaired would produce the raw correlation. In regression across age, gender, depression and anxiety, alexithymia was one of only two significant predictors of IAS scores (b = −.606, p = .002; age was the other), and it predicted ICQ scores too. Meanwhile the BPQ was predicted by anxiety alone (b = .588, p = .022), and that effect did not survive robust regression.
And the objective null is reproduced in the same paper. The TAS-20 was unrelated to heartbeat-counting accuracy in these samples (all ps > .13), consistent with Desmedt et al.’s r = −.01. So the same study that sharpens the belief-level claim confirms the performance-level null — and does so while demonstrating that the heartbeat task can correlate with a believed-accuracy questionnaire (the IAS, r = .27–.34). That last detail is the one that costs the strong claim most: on this evidence the objective task is not too blunt to detect a believed-accuracy relationship. It detected one. It just did not detect alexithymia.
Placement in the wiki
Alexithymia had been mentioned unhosted on interoceptive-trait-prediction-error (as a complication — atypical interoception tracking alexithymia rather than autism, without directly testing ITPE) and on interoceptive-psychopathology (as an uncontrolled comorbidity). It now has a home as a transdiagnostic interoceptive-deficit construct. It is held at one remove: the primary alexithymia sources (Shah, Brewer, Bird; Berthoz; the Toronto Alexithymia Scale literature) are not in raw/, so the wiki carries the construct through Quadt et al.’s review, not first-hand. The unresolved question — is alexithymia the cause of these interoceptive deficits, a label for them, or a correlated third thing — is the same cause-or-consequence problem that runs through interoceptive-psychopathology.