IQ and Autism: Why the Test You Take Changes the Answer

There is no autistic IQ. Autistic people are found across the entire distribution, from intellectual disability to the far upper tail, and any article offering you a single number is describing a sample rather than a population.

What the research does show, and show unusually clearly, is that which test you sit changes the answer far more for autistic people than for anyone else. That finding is the most useful thing on this page.

The 30-percentile-point finding

In 2007, Michelle Dawson, Isabelle Soulières, Morton Ann Gernsbacher and Laurent Mottron published a study in Psychological Science that changed how a lot of researchers think about this question.

They gave 38 autistic children two different measures: the Wechsler scales, the standard clinical instrument, and Raven’s Progressive Matrices, a test of abstract pattern reasoning with no verbal content and minimal instruction.

The autistic children scored on average 30 percentile points higher on Raven’s than on the Wechsler. In some individual cases the gap exceeded 70 percentile points.

The control condition is what makes this meaningful. Typically developing children showed no such discrepancy. Their scores on the two tests agreed, as you would expect from two measures of the same underlying ability. A comparable contrast appeared when autistic adults were compared with non-autistic adults.

That control group turns the result from a flattering claim into a technical one. If both groups had shown a gap, it would say something about the tests in general. Only one group showing it says something about how these particular tests interact with autistic cognition, and it points in a specific direction: the standard instrument appears to understate autistic reasoning ability, sometimes dramatically.

Why the two tests diverge

The two instruments differ in ways that are easy to overlook if you think of an IQ test as a neutral window onto ability.

• Verbal instruction load. Wechsler subtests are administered orally with spoken explanations and prompts. Raven’s is almost entirely non-verbal: you are shown a pattern with a piece missing and asked to pick the piece.
• Social demands of administration. A Wechsler assessment is a sustained one-to-one interaction with an unfamiliar adult who is watching you closely. That is a different task from doing puzzles, and it is not the task the score claims to measure.
• Timed, speeded subtests. Several Wechsler subtests reward fast simple output. Processing speed is a common relative weakness in autistic profiles, so those subtests pull the composite down independently of reasoning.
• Breadth versus focus. The Wechsler composite averages several quite different abilities. Raven’s measures one thing, abstract reasoning, which in many autistic profiles happens to be a relative strength.

Subsequent work by the same group found that enhanced visual processing contributes to matrix reasoning performance in autism, which suggests the Raven’s advantage reflects a genuine cognitive strength rather than simply the absence of obstacles.

What the distribution actually looks like

People search for the average IQ of autistic people and there is no honest single answer, for two reasons worth understanding.

The range is the whole range. Autistic people are found at every level of measured cognitive ability. Any figure you see is the mean of a particular sample, and samples differ enormously depending on where they were recruited.

The proportion with intellectual disability has fallen sharply. CDC surveillance of eight-year-olds in 2020 put co-occurring intellectual disability, defined as an IQ of 70 or below, at roughly 38% of autistic children. Older figures were far higher. Longitudinal work covering 2000 to 2016 found autism without intellectual disability rising around five-fold over that period while autism with intellectual disability roughly doubled.

That divergence is not autism changing. It is identification changing. Broadened diagnostic criteria and better recognition found large numbers of autistic people who were previously missed entirely, and those people were disproportionately the ones without intellectual disability, because they were the ones whose support needs had not forced the issue.

Which means older statistics describe who used to get diagnosed rather than who is autistic, and any figure quoted without a date and a source should be treated with suspicion. Identification rates also vary substantially between demographic groups, which reflects differences in access to assessment rather than differences in autism itself.

Uneven profiles and what they mean

Autistic cognitive profiles are frequently spiky. Index scores vary more, often considerably more, than in non-autistic profiles, with relative strengths in perceptual and visual-spatial reasoning and relative weaknesses in processing speed and sometimes verbal comprehension.

This matters for the same reason it matters in ADHD, covered in our guide to IQ and ADHD. A Full Scale IQ averages the indexes together. When they are far apart, the average describes none of them.

Clinical guidance is explicit about this: where the gap between the highest and lowest index exceeds roughly 23 points, about one and a half standard deviations, the Full Scale IQ should not be interpreted as a single summary of ability. For a markedly spiky profile the headline number is arithmetically valid and descriptively useless.

The practical consequence is that the index breakdown carries almost all the useful information. A profile showing strong perceptual reasoning and weak processing speed tells you something actionable about how someone works best. The composite that averages them tells you almost nothing.

What to ask for in an assessment

If cognitive testing is part of an assessment for you or your child, four things are worth raising:

• Which instrument is being used, and whether a non-verbal measure will be included. Given the Dawson finding, a Wechsler-only assessment may not capture reasoning ability well.
• The index scores individually, not just the Full Scale figure. This is where the information is.
• Accommodations for the administration itself. Sensory environment, breaks, familiarity with the assessor and reduced time pressure all affect performance without affecting ability.
• Confidence intervals. Every index carries error and a good report gives ranges rather than points.

And one thing to hold lightly: a spiky profile is characteristic of autistic cognition but is not evidence of autism. Plenty of non-autistic people have uneven profiles. The pattern is consistent with, which is a much weaker claim than indicative of.

Frequently asked questions

The bottom line

Autistic people span the full IQ range and there is no meaningful single average. What there is, is strong evidence that the standard clinical instrument understates autistic reasoning: 30 percentile points on average in the Dawson study, more than 70 in some cases, with no comparable gap in non-autistic controls.

If you take one thing from this page, make it that a Full Scale IQ from a Wechsler assessment is a weaker description of an autistic person than of almost anyone else, and that the index breakdown and a non-verbal measure will tell you considerably more.