Alexithymia is common in autism—not universal

Alexithymia—difficulty identifying or describing feelings—is often discussed as though it were a defining autistic trait. Our own alexithymia pages cite this review because it offers a stronger answer: the co-occurrence appears common, the average group difference is large, and a substantial share of autistic participants did not cross the measure’s cut-off.

The note also matters because the headline number is easy to detach from its method. “Half” sounds like a clean prevalence finding. It came from selected case-control studies using one family of self-report scales, with considerable variation between studies.

The reviewers asked whether autistic and non-autistic groups differed on the Toronto Alexithymia Scale, including its subscales for difficulty identifying feelings, difficulty describing feelings, and externally oriented thinking. They also asked what proportion in each group crossed the TAS-20 cut-off used in the included studies.

The paper situated those questions inside a larger hypothesis: some emotion-processing differences attributed to autism may instead be associated with co-occurring alexithymia. The meta-analysis tests co-occurrence and group difference more directly than it tests that broader explanation.

The team searched four databases through January 2018. Eligible studies compared an autistic group with a non-autistic group, used the 20- or 26-item Toronto Alexithymia Scale, and reported the means and standard deviations needed for analysis. Fifteen non-overlapping studies remained, covering 366 autistic and 348 non-autistic participants.

The reviewers used random-effects models because the samples and measures differed. They assessed study quality, inspected publication bias, and tested whether age helped explain variation. Nine studies reported how many participants crossed the TAS-20 cut-off and contributed to the weighted proportion and risk-ratio analysis.

Most studies matched groups on at least age or gender. Far fewer measured or adjusted for anxiety and depression, even though both can affect alexithymia scores. Only one study adjusted for both in its group comparison.

The pooled difference on total alexithymia scores was large: Cohen’s d = 1.51, with a 95% confidence interval from 1.21 to 1.81. Differences were also large for identifying feelings (d = 1.28) and describing feelings (d = 1.29), and moderate for externally oriented thinking (d = 0.50).

Among the nine studies using the TAS-20 cut-off, estimates in autistic groups ranged from 33.3% to 63%. The weighted figure was 49.93%, compared with 4.89% in non-autistic groups. The pooled risk of crossing the cut-off was 6.50 times as high in autistic participants, with a wide 95% confidence interval from 3.26 to 12.93.

The direction of the finding was clear. Its exact size was less stable. Overall scores varied significantly between studies, and mean age was related to the estimated difference. The available reports did not let the reviewers examine gender, anxiety, depression, or other possible explanations in a consistent way.

Not a population prevalence estimate. These were case-control studies gathered for research, not a representative survey of autistic people. Weighting the included proportions does not remove selection bias.

Not every form of emotional awareness. The Toronto scale leaves out parts of the broader alexithymia concept, including fantasy or imagination. Its externally oriented thinking subscale has also drawn reliability concerns.

Not every autistic person. The self-report format requires language, reflection, and access to the questionnaire. Included studies frequently focused on people with average or higher IQ and no major communication difficulty.

Not a settled explanation. The review shows frequent co-occurrence. It does not prove that alexithymia explains every emotion-recognition, empathy, interoception, or regulation difference in autism. Anxiety and depression were rarely accounted for.

The paper gives us a useful correction in both directions. Difficulty finding words for feelings should not be dismissed as unrelated to autism: it appeared far more often in the autistic samples. But it should not be treated as an autistic universal either. Nearly half the participants did not cross the chosen cut-off, even within samples likely to underrepresent people who could not complete the scale.

That makes alexithymia a question, not an assumption. “Do you know what you feel?” may be too broad. It can help to separate noticing a body change, distinguishing one state from another, finding a word, deciding whether to share it, and having someone understand the words used. One questionnaire score compresses those steps.

It is also worth resisting a circular use of the measure. Asking people to self-report difficulty identifying their internal states is practical, but the result should not be treated as a perfect window into that same difficulty.

The analysis represents 714 participants from 15 published studies that used one of two closely related Toronto scales. The studies generally included autistic people able to complete self-report measures and often described participants using outdated functioning language.

It tells us much less about people with intellectual disability, people with limited spoken or written communication, autistic people outside research clinics, and how culture or language changes the naming of feelings. Those are not small exceptions to a claim about autism as a whole.

Future studies should combine self-report with accessible interviews, observer information chosen by the participant, and tasks that separate bodily sensation, identification, language, and expression. Representative samples should report communication needs, intellectual disability, race, gender, anxiety, depression, and age rather than treating those factors as background noise.

Longitudinal work could also test whether alexithymia scores change with safety, therapy, language, interoceptive support, or mental health. That would tell us whether a score reflects a fixed trait, a learned vocabulary, current distress, or several things at once.