An autistic child sits two intelligence tests a few weeks apart. On one the result places them near the middle of the population. On the other they land near the top. Nothing about the child changed. Only the test did.
This is one of the better documented findings in the research on autism and cognition, and it has a practical consequence for any parent about to read a score. Which test was used matters more for an autistic child than it does for a non-autistic one.
What the research found
In 2007 a team led by Michelle Dawson at the University of Montreal gave two different intelligence tests to autistic children and adults. One was a Wechsler scale, the standard clinical instrument, which mixes verbal questions, general knowledge, timed hands-on tasks and spoken answers. The other was Raven’s Progressive Matrices, a nonverbal test made entirely of visual patterns with a missing piece.
The autistic participants scored substantially higher on Raven’s. The average difference was in the region of thirty percentile points, and for some individuals it was far larger. A child placed around the middle of the population by one test was placed near the top by the other. Non-autistic participants in the same study showed no comparable gap (Dawson et al., 2007).
The authors argued that the conventional figure had been measuring something other than autistic reasoning ability. Published estimates of intellectual disability in autism rested largely on Wechsler-style testing, and that choice of instrument had quietly shaped what the field believed.
Why the two tests diverge
Look at what each test actually requires and the gap stops being mysterious.
| What the test asks of the child | Wechsler scales | Raven’s style matrix test |
|---|---|---|
| Understand spoken instructions from a stranger | Throughout | Once, at the start |
| Answer out loud, in sentences | Several subtests | No |
| Work against a stopwatch | Several subtests | Usually not |
| Read social cues from the examiner | Continuously | Not at all |
| Infer an unstated rule from a pattern | One subtest | Every item |
| General knowledge learned from other people | Several subtests | None |
A Wechsler assessment is a social event. A stranger sits across a table, gives instructions out loud, waits for spoken answers, times some tasks with a stopwatch and makes notes. For a child who finds spoken language effortful, who needs longer to formulate an answer, or who finds being watched by an unfamiliar adult genuinely uncomfortable, a large part of what the session measures is the session itself.
A matrix test removes nearly all of that. The instructions are given once. Every item afterwards is a visual pattern with a rule hidden in it, and the task is to work out the rule. There is no spoken answer, no general knowledge, and usually no stopwatch. What is left is reasoning, which is what everyone wanted to measure in the first place. This is the broader argument for nonverbal and culture-fair testing, and autistic test takers are the clearest case for it.
The spiky profile
The other thing that shows up repeatedly in autistic test results is scatter. Most people score at a fairly similar level across the different parts of a test. Autistic profiles are frequently uneven, sometimes dramatically so, with a very high score on pattern reasoning sitting next to a much lower one on a timed or verbal task.
A single overall IQ figure averages that spikiness away. It produces a number that describes a child who does not exist: one who is moderately good at everything, rather than one who is exceptional at some things and finds others hard. When you see a full scale score for an autistic child, the section scores underneath it carry more information than the headline does.
As a mother, I got this wrong first
A friend’s son, who is autistic and was nine at the time, did one of the matrix-style tests with my daughter one afternoon, mostly because he wanted to and she was already doing it. He worked through it without a word, fast, and scored higher than anyone else in the house.
What I remember is my own reaction, which was surprise. He had been assessed at school the year before and the number from that assessment was ordinary. I had absorbed that number without ever asking what had produced it, and I had quietly adjusted my expectations of a nine year old on the strength of it. The second test did not make him cleverer. It made visible something the first format had been getting in the way of.
That is the mistake I would warn other parents off. A score is a measurement taken through a particular instrument, and instruments have blind spots that fall differently on different children.
What this means in practice
If your child is autistic and has been assessed with a Wechsler scale, the resulting figure is likely to understate their reasoning ability, and the amount it understates by is not small. That does not make the assessment worthless. A clinical assessment is answering questions about support needs as well as about ability, and it does things no online test can do.
If you want a second view that leans on reasoning rather than on language, speed and social interaction, a nonverbal test gives you one. Our free IQ test for kids is built from exactly this kind of matrix item, with no reading required beyond the instructions and no stopwatch on most age bands. There is an adult version too, which autistic adults who were assessed as children sometimes find clarifying.
Two cautions. A high matrix score does not mean support needs disappear, and a score above 130 in an autistic child often sits alongside real difficulties that the number says nothing about. And a low score on any single test, for any child, means considerably less than people assume.
If attention rather than autism is the question in your mind, the pattern there is different again, and worth reading separately.
References
- Dawson, M., Soulieres, I., Gernsbacher, M. A., & Mottron, L. (2007). The level and nature of autistic intelligence. Psychological Science, 18(8), 657-662. doi:10.1111/j.1467-9280.2007.01954.x