Wechsler scales
33%classified “low functioning”
Multiple subtests draw on spoken instructions, verbal responses, speed and social interaction.
01 — Same children, different verdict
One test classified a third as “low functioning.” The other classified 5%.
In 2007, researchers gave 38 autistic children, ages 7–16, both the Wechsler scales and Raven’s Progressive Matrices. Raven’s uses visual patterns and does not require a spoken answer. The classification changed sharply.
Raven’s Matrices
5%classified “low functioning”
Participants solve visual patterns without producing spoken language.
02 — The gap was specific
Autistic children averaged 30 percentile points higher on the nonverbal test.
Some individuals gained more than 70 percentile points. The matched non-autistic group did not show the same direction of difference, making a generic “Raven’s is easier” explanation hard to sustain.
Raven percentile minus Wechsler percentile
Autistic childrenn = 38+30 points
Non-autistic childrenmatched comparisonno direction
This does not mean Raven’s is a complete or universally superior measure of intelligence. It shows that verbal demands can become a confound: the test may partly measure a person’s ability to understand and perform the testing format.
03 — Measurement can tilt both ways
Language level can affect what an autism-severity item appears to measure.
A 2022 analysis of 1,196 minimally verbal children with neurodevelopmental disorders tested whether ADOS Module 1 items worked consistently across spoken-language and nonverbal-mental-age groups. Five of 18 examined items showed differential item functioning.
When correct performance depends on understanding spoken instructions or producing speech.
When the same observed behaviour carries different weight across language or cognitive groups.
The study did not say every low-language child receives an inflated overall severity score. Its narrower finding is more useful: some standardized items are not fully invariant across the people they are used to compare.
04 — Minimal speech is not a cognitive diagnosis
The people hardest to test are not one uniform group.
Studies of minimally verbal autistic people find wide variation in nonverbal cognition, language comprehension and communication. A small but important minority score near or above common cognitive cut-offs even when spoken language remains very limited.
The careful conclusion is not that intellectual disability disappears with a better test. Many autistic people do have intellectual disability and deserve valid assessment and support. The conclusion is that spoken language should not be treated as a clean proxy for reasoning ability.