Module 9 — TAS-20 — Alexithymia, When You Can't Name What You Feel

 

What alexithymia actually is, why the questionnaire's three parts are not equally trustworthy, why roughly half of autistic adults score above the line and half do not, and what a total score of 61 does and does not settle.

 

Autistic Self-Discovery · Part Four — Feeling, Naming, Connecting · 13 min read, about 26 min with the workbook

 

The big idea

 

If someone has ever asked how you are and you genuinely did not know — not that you did not want to say, but that the information was not there — this module is about that gap. It has a name, a twenty-item questionnaire behind it, and thirty years of argument about what the number means.

 

Two different methods — one statistical, one judgment-based — independently flagged the same subscale and the same item. If your total is pushed up by the externally-oriented thinking items, it is telling you less than it appears to. If it is pushed up by the two feelings subscales, it is solid.

 

Step 1 — The lesson

 

A. The four-hour delay

 

Something happens at eleven in the morning. A colleague says an ordinary sentence in a tone you cannot place. You carry on with the day. At two you are shorter than usual with someone who did not deserve it. At four your jaw aches and you cannot settle. At nine that evening, washing up, the answer arrives fully formed: you were angry, and you have been angry since eleven.

 

The emotion was not missing. It ran the whole day. What was missing was the label, and without the label you could not say it, act on it, or let it go.

 

That gap has a clinical name: alexithymia. The word is built from Greek and means, almost exactly, no words for feelings. It is not a diagnosis. It does not appear in any diagnostic manual as a condition. It is a trait that runs on a dimension, like height, and everybody sits somewhere on it.

 

Autistic people describing it from the inside reach for physical language rather than emotional language. One autistic-led resource collects descriptions like surges of electricity rushing through me and all fuzzy and floaty, and names the specific trap of not recognizing you are in distress until you are a long way into it.¹ The English feeling words do not fit what is actually arriving, so people build their own.

 

Diagram — A · What it looks like, what it is. The left column is the reputation you get. The right column is the mechanism. Someone who takes four hours to identify anger looks, for those four hours, exactly like someone who is not angry. That is how a processing delay gets read as a personality.

 

This is measurable outside a clinic. A study fitted 29 autistic adults and 28 non-autistic adults with smartphone prompts for a week and asked them, repeatedly and in the moment, what they were feeling. The autistic group reported significantly more moments of having an emotion they could not name, and more moments of holding a positive and a negative feeling at the same time.² The gap is not a retrospective story people tell about themselves. It shows up live.

 

It is also not rare, and not only autistic. In a representative sample of 1,859 German adults aged 20 to 69, ten per cent scored at or above the standard alexithymia threshold, with average scores just under fifty in both men and women.³ One person in ten, in the general population, walking around with the same delay.

 

B. Three parts, and one of them wobbles

 

The instrument that measures this is the Toronto Alexithymia Scale, twenty items, usually written TAS-20. In its published form it is twenty statements, each rated from one to five, giving a total somewhere between 20 and 100. Statements like I am often confused about what emotion I am feeling, or it is difficult for me to find the right words for my feelings.

 

The twenty items split into three subscales, and knowing which one is high tells you far more than the total does. Difficulty identifying feelings is seven items and covers the four-hour delay: something is happening in you and you cannot tell what. Difficulty describing feelings is five items and is a different problem — you know what it is and cannot get it out of your mouth. Externally-oriented thinking is eight items and is meant to capture a habit of attention: looking outward at events rather than inward at what they did to you.

 

Diagram — B · Three parts, one number. Seven items, five items, eight items, one total. The first two subscales describe genuinely different problems — not knowing, and not being able to say — with different practical answers. The third is the one the research keeps arguing about.

 

The scale was built in 1994 by Bagby, Parker and Taylor, who took a longer pool of items, kept the ones that held together, and cross-validated the three-factor structure in clinical and non-clinical samples alike. Those three factors have been the standard reading of it ever since.⁴ A companion paper tested the scale against personality measures and against an observer's rating of alexithymia, and found the pattern of correlations the theory predicted.⁵

 

The largest test of that structure came much later. A meta-analytic factor analysis pooled 88 samples from 62 studies — 69,722 people, 25 countries, 16 languages — and found the original three-factor solution fitted better than eight rival models. It also found something less comfortable. Reliability for difficulty identifying feelings was 0.84 and for difficulty describing feelings 0.75, which is respectable. For externally-oriented thinking it was 0.62. The authors recommended reading the three subscale scores rather than the total.⁶

 

Half of the externally-oriented thinking items are reverse-worded, and in that analysis the reverse-worded ones loaded more strongly on a statistical artefact of being reversed than on the thing they were supposed to measure. In plain terms: part of what that subscale records is how you handle a sentence written backwards. That is a real and long-standing weakness, and it is one third of the questionnaire.

 

C. The half that is not autism

 

In 2013 Geoffrey Bird and Richard Cook published an argument that has reorganised this whole corner of the field. Autistic people, they pointed out, are described in the literature as having impaired emotion recognition and reduced empathy — and yet the autistic population is wildly heterogeneous on exactly those measures. Their proposal, which they called the alexithymia hypothesis, is that where those impairments are found they belong to co-occurring alexithymia rather than to autism itself.⁷

 

The evidence they assembled is specific. Alexithymia is present in roughly ten per cent of the general population and, by their reading, in 40 to 65 per cent of autistic adults. In studies of recognizing emotion in the voice, correlations with alexithymia exceeded 0.60 while autism severity predicted nothing. In brain imaging, empathic activity in the anterior insula tracked alexithymia in autistic and non-autistic participants alike, with no independent effect of autism.⁷

 

Diagram — C · About half, not all. Fifty dots in a hundred. Common rather than universal is the most useful phrase in this literature: a general claim about how autistic people process emotion is a claim about roughly half of them, and describes the other half wrongly.

 

A systematic review and meta-analysis put a number on it. Across 15 studies comparing autistic and non-autistic groups on the Toronto Alexithymia Scale — 366 autistic and 348 non-autistic participants — 49.93 per cent of the autistic group scored above the alexithymia threshold, against 4.89 per cent of comparison groups. The authors' conclusion is worth quoting in shape if not in words: alexithymia is common, rather than universal, in autism.⁸

 

The figure climbs in help-seeking samples. Among 190 autistic adults seen by a specialist adult assessment service in London, 66.3 per cent scored at or above 61, and difficulty identifying feelings was associated with the severity of both depression and anxiety.⁹ People who go looking for an assessment are not a random slice of autistic adults, and a number from a clinic is not a population rate.

 

The mechanism usually proposed is interoception — the sense that reports on the inside of your body. If the signals about heart rate, muscle tension and gut arrive faint or garbled, you have less raw material from which to build a feeling. It is a good story and the evidence is thinner than the story. A 2025 review and meta-analysis of 31 interoception studies found that across the five adult studies using comparable cardiac accuracy tasks there was no significant difference between autistic and non-autistic adults — a pooled estimate of −0.21, p = 0.06.¹⁰ Hold the interoception explanation loosely.

 

One correction is needed, because the hypothesis is sometimes flattened into it was never autism, it was always alexithymia. A study of 179 autistic and 158 non-autistic adolescents and adults, followed up 12 to 24 months later, found difficulty describing feelings was the strongest predictor of self-reported social communication difficulty, and difficulty identifying feelings predicted anxiety at follow-up.¹¹ Alexithymia is not a nuisance variable to be subtracted from a research design. It has consequences of its own.

 

D. What the number is, and what it is not

 

Here are the bands the screener returns, exactly as the live page states them: 0 to 51, no to low presence of alexithymia traits; 52 to 60, moderate presence of traits; 61 to 100, high presence of traits.

 

Diagram — D · The bands on the live screener. These three bands are the ones you will see, reproduced without alteration. They come from a convention set by the scale's authors rather than from a study that tested where the line should fall in autistic adults — which is the difference between a threshold and a tradition.

 

The first honest thing to say is about the range. The live screener page states a total range of 0 to 100. The published TAS-20 cannot produce a score below 20: twenty items, each scored a minimum of one, so the floor is 20. We have reproduced the live bands unchanged because those are the bands you will be shown, and flagged the discrepancy for correction. The practical effect is small but real — a low score is low against a floor of 20, not 0, so the bottom band is narrower than it looks.

 

The second is about where 61 came from. It is the developers' convention, applied across three decades and hundreds of studies, which is what gives it weight. It is not a line drawn against a criterion in autistic adults. The German population study judged it possibly too restrictive and proposed a percentile alternative falling at 53 for men and 52 for women.³ Scoring 60 and scoring 62 is not a difference in kind.

 

The third is about autistic respondents specifically, and it is the important one. The largest psychometric study of the TAS-20 in autistic adults compared 743 autistic adults with 721 from the general population. The full scale fitted poorly in both groups. Externally-oriented thinking had a reliability of 0.451 in the autistic sample, with seven of its eight items loading poorly on the general alexithymia factor. And one item — I have physical sensations that even doctors don't understand — behaved measurably differently for autistic and non-autistic respondents at the same underlying level of alexithymia. The authors recommend an eight-item score instead, correlating 0.910 with the full total and behaving consistently across both groups.¹²

 

A separate line of work reached the same place from the other direction. Sixty-nine judges were asked what each TAS-20 item actually measures, rated against eight defined constructs. Difficulty identifying feelings and difficulty describing feelings came out measuring what they claim to. Externally-oriented thinking did not: only two of its eight items were endorsed at all. The same item about physical sensations doctors cannot explain came out measuring health anxiety better than it measured alexithymia.¹³

 

The scale's original authors do not accept the criticism. Their twenty-five-year review concludes that the preponderance of evidence supports the reliability and validity of the TAS-20 and that it measures the construct as originally defined.¹⁴ This is a live disagreement between serious people. What it means for you is narrower than it sounds: read your two feelings subscales with confidence, read the third with a raised eyebrow, and do not treat 61 as a door that opens or closes.

 

INSIDE THE INSTRUMENT

 

TAS-20 — Twenty-Item Toronto Alexithymia Scale

 

Where it came from

 

Developed by Bagby, Parker and Taylor and published in the Journal of Psychosomatic Research in 1994 as a revision of the earlier 26-item Toronto scale. Item selection and cross-validation of the three-factor structure appeared in the first paper; convergent, discriminant and concurrent validity against the NEO Personality Inventory, psychological mindedness, need for cognition and an observer-rated alexithymia measure appeared in the second.⁴⁵ The construct comes from psychosomatic medicine, not from autism research, and predates the scale by two decades.

 

What it is made of

 

Twenty self-report items on a five-point scale, five of them reverse-keyed, yielding a total of 20 to 100 and three subscales: difficulty identifying feelings (7 items), difficulty describing feelings (5 items), and externally-oriented thinking (8 items). The conventional bands are 51 or below for non-alexithymia, 52 to 60 for possible alexithymia and 61 or above for alexithymia. Note the discrepancy documented in section D: the live screener page states a range of 0 to 100 rather than 20 to 100.

 

How well it performs

 

A meta-analytic confirmatory factor analysis of 88 samples from 62 studies (N = 69,722, 25 countries, 16 languages) supported the original three-factor solution over eight competing models, with omega of 0.84 for difficulty identifying feelings, 0.75 for difficulty describing feelings and 0.62 for externally-oriented thinking. The authors recommend subscale scores over the total.⁶ Population data are available: in a representative German sample of 1,859 adults, 10% exceeded the threshold of 61, with means of 49.5 for men and 48.2 for women.³

 

Where it was validated

 

Originally in Canadian student and clinical samples, then across 25 countries and 16 languages. Autistic samples came late and separately. A meta-analysis of 15 comparison studies (366 autistic and 348 non-autistic participants) found 49.93% above threshold against 4.89% of comparison groups.⁸ In a specialist adult assessment service (N = 190), 66.3% scored 61 or above, with difficulty identifying feelings associated with depression severity (beta 0.30 to 0.38) and anxiety severity (beta 0.36).⁹

 

What it cannot do

 

It does not measure alexithymia well in autistic adults without modification. In 743 autistic adults compared with 721 general-population adults, the full scale showed subpar fit (CFI 0.924, RMSEA 0.072); externally-oriented thinking had omega 0.451 in the autistic sample and seven of its eight items loaded poorly on the general factor. Item 3 showed practically significant differential item functioning between groups. The authors recommend an eight-item general alexithymia factor score, correlating 0.910 with the full total, with negligible between-group differential functioning.¹²

 

One of its three subscales does not survive content-validity testing. With 69 judges rating every item against eight defined constructs, only difficulty identifying feelings and difficulty describing feelings represented their intended constructs; only two of eight externally-oriented thinking items were endorsed as measuring externally-oriented thinking, and the four reverse-scored items performed worst. Two items overlapped substantially with health anxiety.¹³

 

The cut-off is a convention, not a calibrated threshold for this population. No study has established sensitivity or specificity for 61 against a criterion measure in autistic adults, and no autistic normative table exists. The German population study proposed a 66th-percentile alternative of 53 for men and 52 for women on the grounds that 61 may be too restrictive.³ Elevated totals are also not specific to autism, occurring across eating disorders, schizophrenia, substance use, Parkinson's disease, multiple sclerosis and anxiety disorders.⁷

 

Self-report of a self-report deficit is a structural problem. The instrument asks people who have difficulty identifying and describing their internal states to identify and describe their internal states. It has also been criticised for omitting the imaginal and fantasy components of the original construct entirely.⁸

 

The developers dispute much of this. Their 25-year review concludes that the preponderance of evidence supports the reliability and validity of the TAS-20 and that it measures the construct as originally defined.¹⁴ Present the disagreement rather than resolving it for the client.

 

What a clinician does with it

 

Read the two feelings subscales as separate clinical targets: not knowing is an interoception and labeling problem, not being able to say is an expression problem. Discount the externally-oriented thinking subscale unless the item-level pattern is clinically meaningful. Use the total as a rough index, or the published eight-item score if the client is autistic. Note the therapy implication: in 269 autistic adults, higher alexithymia predicted finding cognitive strategies harder to use but was unrelated to how helpful people found them.¹⁵ Difficulty is not futility.

 

Validity tier: 1 — validated and published, calibrated by independent labs. Thirty years of independent replication across 25 countries and 69,722 pooled participants earns the tier for the total and the two feelings subscales; the externally-oriented thinking subscale and the 61 cut-off in autistic adults would not earn it on their own.

 

Diagram — E · What is reported, what is not. The ticks are what the literature reports. The blanks are what nobody has measured: there is no autistic normative table and no study establishing what 61 catches and misses in autistic adults. A blank is a limit on what the score can be asked to do, not a criticism of your score.

 

STRENGTHS LENS

 

You have been making decisions on incomplete data, accurately, for years.

 

Most people are handed their feelings pre-labeled and act on them immediately, which is efficient and also how a great many bad decisions get made. If your emotional information arrives late, you have spent your life compensating: reading situations from evidence rather than from gut, checking your reasoning, waiting before you reply. People routinely describe you as steady in a crisis. That steadiness is not fake — it is what you built.

 

The delay is also not as permanent as it feels. What is missing is a mapping between a physical state and a word, and mappings can be learned in adulthood. People who work at this do not become different people; they get a four-hour delay down to forty minutes, which is the difference between saying the thing and swallowing it. Section F is about how.

 

F. What helps

 

Almost every popular suggestion for this — sit with the feeling, name it to tame it — assumes the label is available and you are avoiding it. If it is not available, that advice is an instruction to read a page in the dark. What follows starts from the body.

 

1. Start with the body, not with the word.

 

Four times a day, do a ten-second inventory of physical questions only. Jaw: tight or loose. Shoulders: up or down. Stomach: heavy, light, nothing. Temperature. Heart rate. Then guess a feeling, and let the guess be wrong. You are building a lookup table between a physical state and a word, and the physical column has to be filled in first.

 

2. Get very good at five feelings before you try thirty.

 

The emotion wheels with ninety words on them are counterproductive here. Choose five: angry, sad, frightened, ashamed, glad. Learn what each does in your body specifically — not in general, in you. Most people find two of the five are already reliable and three are guesswork, and knowing which is which beats a larger vocabulary you cannot apply.

 

3. Buy time out loud, with a sentence you have already written.

 

The damage usually happens in the gap between being asked and knowing. Have one prepared line and use it without embarrassment: "Something is off and I do not know what yet. I will tell you by tomorrow evening." This converts what reads as stonewalling into an appointment. Say the second half; the commitment to come back is the part that protects the relationship.

 

4. Track the delay, not just the feeling.

 

Keep two timestamps for a fortnight: when the thing happened, and when you worked out what it was. The number you care about is the gap, because that is the number that moves. Watching it fall from six hours to two beats trying to score yourself on how well you feel things, which is not a thing you can score.

 

5. Tell a therapist what is hard to use, not what fails to help.

 

In a study of 269 autistic adults, higher alexithymia predicted finding cognitive strategies harder to use but had no relationship to how helpful people found them.¹⁵ Hand that distinction over directly. "Asking me what I am feeling stalls the session; asking me what my shoulders are doing does not" usually gets the approach adapted rather than abandoned.

 

Step 2 — Take the screener

 

Twenty questions, five to seven minutes, free and confidential. You get a total score with a band, plus three component readings: difficulty identifying feelings, difficulty describing feelings, and externally-oriented thinking. One honest steer before you start — the first two components are the well-evidenced part of this questionnaire and the third is not, so read the components rather than fixing on the total.

 

Take the TAS-20

 

Before you start

 

The TAS-20 is a published, widely validated self-report measure of alexithymia — difficulty identifying and describing your own feelings — and it has been tested in 25 countries and tens of thousands of people. It is a screen and not a diagnosis, and alexithymia is not itself a diagnosable condition in any case: it is a trait that runs on a dimension. Its externally-oriented thinking subscale has persistently poor reliability, no normative data exist for autistic adults, and no study has established what the cut-off of 61 catches or misses in an autistic population. A high score is also not specific to autism, occurring across depression, anxiety, eating disorders and several other conditions. Only a clinician can diagnose.

 

Step 3 — Your workbook

 

Your answers save to this device only — we cannot see a word of what you write. This module lets you record your total and your three component scores, work out which of the two feelings problems is actually yours, and build the body-first inventory and the time-buying sentence that do most of the practical work.

 

Your TAS-20 results

 

Took the screener? Put your numbers in below. The subscales matter more than the total here — not knowing what you feel and not being able to say it are two different problems with two different answers. Entirely optional — skip it if you would rather just read.

 

Score bands: 0–51 = No to Low Presence of Alexithymia Traits; 52–60 = Moderate Presence of Traits; 61–100 = High Presence of Traits

 

Fields: TAS-20 · Toronto Alexithymia Scale, 20 items; Total score (0–100); My total (enter 0–100); The three components (the screener names these but publishes no ranges or bands for them, so write down whatever it gives you); Difficulty identifying feelings; Difficulty describing feelings; Externally-oriented thinking

 

Which of the two is yours

 

Section B. Not knowing and not being able to say are separate problems. Most people are much worse at one than the other, and the useful work is different for each.

 

Fields: The harder one for me: Working out what I feel / Saying it out loud to someone / Both, about equally / Neither is especially hard; A recent time this cost me something — what happened, and what it cost

 

Your delay, timed

 

Section F, item 4. Two timestamps, not a feelings diary. The gap between them is the number that actually moves.

 

Fields: Most recent example: when the thing happened; When I worked out what it was; What finally gave it away — a physical sign, something someone said, or a thought; Tick: I will keep both timestamps for the next fortnight

 

The body inventory

 

Section F, item 1. Physical answers only. You are filling in the left-hand column of a lookup table, and the emotion word can wait.

 

Fields: What my body does when I am angry — specifically, in me; What my body does when I am frightened; What my body does when I am sad, or ashamed; Which physical signal reaches me earliest and most reliably: Jaw or face / Shoulders or neck / Chest or breathing / Stomach or gut / Temperature / Restlessness / I am not sure yet

 

Five feelings, learned properly

 

Section F, item 2. Five words, not ninety. Mark honestly which ones you can already spot and which are guesswork.

 

Fields: Feeling one, and how I know; Feeling two, and how I know; Feeling three, and how I know; Feeling four, and how I know; Feeling five, and how I know; Which of the five are still guesswork

 

The sentence that buys time

 

Section F, item 3. Write it once, so it is available when you need it and you are not composing under pressure.

 

Fields: Who I most need it for; My exact sentence, including when I will come back to them

 

What to hand a therapist

 

Section F, item 5. Difficulty using a strategy is not the same as the strategy not helping, and saying which one you mean changes what you get offered.

 

Fields: What stalls me in a session, and what does not; The adaptation I would ask for

 

Appendix — Research companion

 

Peer-reviewed research

 

4. Bagby RM, Parker JD, Taylor GJ (1994). The twenty-item Toronto Alexithymia Scale--I. Item selection and cross-validation of the factor structure. Journal of Psychosomatic Research, 38(1), 23-32. DOI 10.1016/0022-3999(94)90005-1. View the paper The originating paper for the TAS-20, written to address limitations in the earlier 26-item Toronto scale. Two studies produced a twenty-item scale with good internal consistency and test-retest reliability and a three-factor structure - difficulty identifying feelings, difficulty describing feelings and externally-oriented thinking - theoretically congruent with the alexithymia construct, with the structure then replicated by confirmatory factor analysis in both clinical and non-clinical samples. It is the reference for the scale's content, its three subscales and its conventional interpretation. Limitation: developed in Canadian student and clinical samples in the early 1990s, with no autistic participants and no receiver-operating-characteristic analysis establishing the sensitivity or specificity of any cut-off score.

 

14. Bagby RM, Parker JDA, Taylor GJ (2020). Twenty-five years with the 20-item Toronto Alexithymia Scale. Journal of Psychosomatic Research, 131, 109940. DOI 10.1016/j.jpsychores.2020.109940. View the paper Twenty-five-year review by the scale's original authors of the accumulated reliability, validity and structural evidence, written as a direct response to the controversies about the instrument. They conclude that the preponderance of evidence supports the reliability and validity of the TAS-20 and that it accurately reflects and measures the construct as it was originally defined, including the operational thinking component that the externally-oriented thinking subscale is meant to capture. It is the strongest statement of the case for the scale and is included here so the disagreement is presented in both directions. Limitation: a narrative review by the instrument's own developers rather than an independent systematic review, with no meta-analytic pooling and no separate treatment of autistic samples.

 

5. Bagby RM, Taylor GJ, Parker JD (1994). The twenty-item Toronto Alexithymia Scale--II. Convergent, discriminant, and concurrent validity. Journal of Psychosomatic Research, 38(1), 33-40. DOI 10.1016/0022-3999(94)90006-X. View the paper The companion validity paper published immediately after the development paper in the same issue. In university student samples the TAS-20 showed the pattern of correlations with the NEO Personality Inventory scales and with separate measures of psychological mindedness and need for cognition that the alexithymia theory predicted, supporting convergent and discriminant validity. Concurrent validity was supported by positive correlations with observer ratings of alexithymia in a sample of behavioural medicine outpatients. Limitation: validity was established against personality questionnaires and observer ratings rather than against any diagnostic criterion, in student and medical outpatient samples that included no autistic participants.

 

7. Bird G, Cook R (2013). Mixed emotions: the contribution of alexithymia to the emotional symptoms of autism. Translational Psychiatry, 3(7), e285. DOI 10.1038/tp.2013.61. View the paper The paper that set out the alexithymia hypothesis: that impaired emotion recognition and reduced empathy in autistic people, where they are found, are attributable to co-occurring alexithymia rather than to autism itself. It reports alexithymia in roughly 10% of the general population against 40 to 65% of autistic adults, correlations above 0.60 between alexithymia and vocal affect recognition accuracy with no independent contribution of autism severity, and anterior insula empathic responses that tracked alexithymia in autistic and non-autistic participants alike. It also notes elevated alexithymia in eating disorders, schizophrenia, substance use, Parkinson's disease, multiple sclerosis and anxiety disorders. Limitation: a theoretical review synthesising the authors' own earlier studies rather than new data, with small experimental samples behind several of the effects it summarises.

 

2. Costache ME, Gioia F, Vanello N, Greco A, Lefebvre F, Capobianco A, Weibel S, Weiner L (2026). Exploring emotion control and alexithymia in autistic adults: an ecological momentary assessment study. Journal of Autism and Developmental Disorders, 56(2), 587-601. DOI 10.1007/s10803-024-06551-8. View the paper Twenty-nine autistic adults and 28 non-autistic adults matched on age, sex and education were prompted by smartphone to report emotions, arousal and perceived emotional control repeatedly over seven days. The autistic group reported significantly more instances of having an emotion they could not name, and when they could label an emotion they reported more negative and more mixed positive-negative states. A link between reporting no emotion at all and impaired emotional control appeared only in the autistic group. Limitation: 57 participants in total, self-report only, a single week of sampling and a cognitively able adult sample, so the effect sizes should not be treated as population estimates.

 

3. Franz M, Popp K, Schaefer R, Sitte W, Schneider C, Hardt J, Decker O, Braehler E (2008). Alexithymia in the German general population. Social Psychiatry and Psychiatric Epidemiology, 43(1), 54-62. DOI 10.1007/s00127-007-0265-1. View the paper Standardisation study of the German TAS-20 in a representative random sample of 1,859 adults aged 20 to 69 drawn from the general population. Ten per cent of the sample exceeded the conventional threshold of 61, with mean total scores of 49.5 (SD 9.3) in men and 48.2 (SD 9.2) in women and only minimal sex differences. The authors judged the threshold of 61 possibly too restrictive for identifying highly alexithymic people and proposed a 66th-percentile alternative of 53 for men and 52 for women. Limitation: a single German sample, so the norms and the proposed alternative cut-offs do not automatically transfer to other countries, to clinical groups or to autistic adults.

 

8. Kinnaird E, Stewart C, Tchanturia K (2019). Investigating alexithymia in autism: a systematic review and meta-analysis. European Psychiatry, 55, 80-89. DOI 10.1016/j.eurpsy.2018.09.004. View the paper Systematic review and meta-analysis of 15 studies using the Toronto Alexithymia Scale to compare autistic and neurotypical groups, covering 366 autistic and 348 non-autistic participants. Autistic participants scored significantly higher on the total and on every subscale, and 49.93% met the alexithymia threshold against 4.89% of comparison participants, giving a significantly increased risk. The authors conclude that alexithymia is common rather than universal in autism and that autistic people with alexithymia may represent a distinct subgroup with distinct clinical needs. Limitation: only 15 studies with small individual samples, all relying on self-report in a population whose defining difficulty is self-report of emotion, and the review notes the TAS omits the fantasy and emotionalising components of the alexithymia construct.

 

10. Klein M, Witthoft M, Jungmann SM (2025). Interoception in individuals with autism spectrum disorder: a systematic literature review and meta-analysis. Frontiers in Psychiatry, 16, 1573263. DOI 10.3389/fpsyt.2025.1573263. View the paper Systematic review and meta-analysis of 31 interoception studies across the lifespan in autism, comprising 9 child or adolescent studies, 15 adult studies and 7 lifespan studies. Across the five adult studies using comparable cardiac interoceptive accuracy tasks there was no significant difference between autistic and non-autistic adults (pooled estimate -0.21, SE 0.11, p = 0.06), and findings for interoceptive sensibility and awareness were inconsistent in both directions. The review notes that roughly half of autistic people have co-occurring alexithymia and that interoceptive confusion is the mechanism usually proposed to link the two. Limitation: only five studies entered the adult accuracy meta-analysis, the tasks differed widely between studies, and alexithymia was discussed but never tested as a moderator.

 

15. Mazurek MO, Pappagianopoulos J, Brunt S, Menezes M, Smith JV, Howard M (2026). Alexithymia, inner thinking patterns, and perceptions of mental health therapy strategies among autistic adults. Journal of Autism and Developmental Disorders, 56(4), 1370-1381. DOI 10.1007/s10803-024-06643-5. View the paper Survey of 269 autistic adults aged 21 to 77 (mean 37.5), of whom 59.9% were diagnosed in adulthood, 80.3% had a lifetime anxiety disorder and 81.8% a lifetime depressive disorder, examining how alexithymia and inner thinking style relate to experience of seven common therapy strategies. Autistic adults with greater alexithymia found cognitive strategies harder to use, and those with more frequent self-talk found them easier, while greater visual thinking made guided imagery easier. Crucially, alexithymia and inner thinking patterns were unrelated to how helpful participants found the strategies, only to how difficult they were to use. Limitation: a cross-sectional online survey of a predominantly white, largely female, cognitively able volunteer sample, relying on retrospective recall of therapy rather than measured outcomes.

 

11. Oakley BFM, Jones EJH, Crawley D, Charman T, Buitelaar J, Tillmann J, Murphy DG, Loth E (2022). Alexithymia in autism: cross-sectional and longitudinal associations with social-communication difficulties, anxiety and depression symptoms. Psychological Medicine, 52(8), 1458-1470. DOI 10.1017/S0033291720003244. View the paper Analysis of 337 adolescents and adults (179 autistic, 158 non-autistic) from the EU-AIMS LEAP cohort, with 135 of them followed up 12 to 24 months later. Autistic participants showed significantly elevated alexithymia (r = 0.48); difficulty describing feelings was the strongest correlate of self-reported social-communication difficulty (beta 0.57) and continued to predict it longitudinally after accounting for baseline severity, while difficulty identifying feelings predicted anxiety symptoms at follow-up (beta 0.31, p = 0.01), with anxiety partially mediating the path to depression. This complicates any reading of alexithymia as a mere confound to be partialled out. Limitation: the longitudinal subsample fell to 135 participants, all measures were self-report questionnaires, and the cohort is cognitively able and predominantly white European.

 

6. Schroeders U, Kubera F, Gnambs T (2022). The structure of the Toronto Alexithymia Scale (TAS-20): a meta-analytic confirmatory factor analysis. Assessment, 29(8), 1806-1823. DOI 10.1177/10731911211033894. View the paper Meta-analytic structural equation modelling of 88 samples nested in 62 studies, totalling 69,722 participants from 25 countries in 16 languages, comparing nine competing measurement models. The original three-factor solution fitted best (CFI .936, RMSEA .027, SRMR .041), but reliability differed sharply across subscales: omega was .84 for difficulty identifying feelings, .75 for difficulty describing feelings and .62 for externally-oriented thinking, and every negatively keyed externally-oriented thinking item loaded more strongly on a method factor than on its content factor. The authors recommend reporting subscale scores rather than a single total. Limitation: 75% of the pooled samples were non-clinical and mostly university students, no autistic samples were analysed separately, and the pooled correlation matrix cannot detect problems specific to any one population.

 

13. Veirman E, Van Ryckeghem DML, Verleysen G, De Paepe AL, Crombez G (2021). What do alexithymia items measure? A discriminant content validity study of the Toronto-alexithymia-scale-20. PeerJ, 9, e11639. DOI 10.7717/peerj.11639. View the paper Discriminant content validity study in which 69 judges rated every TAS-20 item for relevance to eight defined constructs, including the four alexithymia components and anxiety, depression and health anxiety. Difficulty identifying feelings and difficulty describing feelings were endorsed as measuring their intended constructs, but only two of the eight externally-oriented thinking items were endorsed as measuring externally-oriented thinking, with the four reverse-scored items performing worst. Item 3 was judged to measure health anxiety better than alexithymia and item 7 measured both equally, and the total score was not endorsed as measuring the broader alexithymia construct. Limitation: the judges were psychology students without expertise in alexithymia research rating item wording against dictionary-based definitions, so this establishes what items appear to measure, not how they perform against outcomes.

 

12. Williams ZJ, Gotham KO (2021). Improving the measurement of alexithymia in autistic adults: a psychometric investigation of the 20-item Toronto Alexithymia Scale and generation of a general alexithymia factor score using item response theory. Molecular Autism, 12, 56. DOI 10.1186/s13229-021-00463-5. View the paper Psychometric study comparing 743 autistic adults from the SPARK cohort with 721 general-population adults matched on mean age. The full TAS-20 fitted poorly in both groups (CFI 0.924, TLI 0.900, RMSEA 0.072 in the autistic sample); the externally-oriented thinking subscale had omega of 0.451 with seven of its eight items loading poorly on the general alexithymia factor, and item 3, about physical sensations doctors do not understand, showed practically significant differential item functioning between autistic and non-autistic respondents. The authors derived an eight-item general alexithymia factor score correlating 0.910 with the full total, with negligible between-group differential item functioning and no loss of nomological validity in autistic participants. Limitation: an online, largely white and highly educated self-report sample from a research register, and this article is a revised replacement for an earlier version retracted over copyright concerns.

 

Clinical frameworks and position statements

 

9. Josyfon E, Spain D, Blackmore C, Murphy D, Oakley B (2023). Alexithymia in adult autism clinic service-users: relationships with sensory processing differences and mental health. Healthcare, 11(24), 3114. DOI 10.3390/healthcare11243114. View the source Cross-sectional study of 190 adults aged 18 to 68 (median 33) seen by a tertiary autism diagnostic service in the UK, of whom 153 received a diagnosis. 66.3% scored 61 or above on the TAS-20. Total alexithymia and difficulty identifying feelings were associated with depression severity (beta 0.30 to 0.38, p < 0.002) and difficulty identifying feelings with anxiety severity (beta 0.36, p < 0.001), and the two feelings subscales partly mediated the relationship between sensory processing differences and both depression and anxiety. Limitation: a help-seeking tertiary clinic sample, so the 66.3% figure is not a population prevalence, and the cross-sectional design cannot establish that alexithymia causes the mental health outcomes it predicts.

 

Lived experience

 

1. Williams Z, Gilham B (2026). Alexithymia. Autism Understood, an autistic-led resource by Spectrum Gaming. View the source Co-produced explanatory resource written for and by autistic people, authored by an autistic writer and a neurodivergent contributor from the Spectrum Gaming community. It describes alexithymia in first-person terms rather than clinical ones, collecting descriptions such as feeling surges of electricity rushing through the body or being all fuzzy and floaty, and notes that single emotion words often feel insufficient for what is actually arriving. It also flags the practical trap of not recognising distress until you are already a long way into it, and links the experience to interoception. Limitation: a community education resource written primarily for autistic young people, with no sample, no data and no peer review, so it illustrates the experience rather than evidencing any prevalence or mechanism claim.

 

Peer-reviewed = checked by independent experts before publication. Clinical model = an established professional framework, not a single study.

 

Up next

 

Module 10 - OAQ-G2 — A Longer Look at Alexithymia

 

All modules in Autistic Self-Discovery

 

Not knowing what you feel is workable. This course was built by clinicians who are part of the New Path Family. If therapy has previously stalled on the question "and how did that make you feel", that is a solvable problem rather than evidence that therapy is not for you — the approach adapts, and we have adapted it often. Therapy for clients in California and coaching worldwide, all by telehealth, are offered by our sister company New Path Family of Therapy Centers, Inc. A conversation costs nothing and there is no pressure. Saving this for later counts too. Talk with the New Path team