In 1996, a group of notable American scientists asserted that “most, if not all” people with autism also suffered from some degree of what they were then calling “mental retardation.”[1]  Nine years later, in 2005, a Canadian researcher reviewed the existing literature on autism and intelligence and concluded that 55-70% of those with autism had an intellectual disability.[2]  Another nine years later, in 2014, the U.S. Centers for Disease Control reported that 31% of autistic children studied in their most recent survey had an intellectual disability.[3]  How can we explain these rapidly changing numbers?  Why has the proportion of autistic people considered intellectually disabled dropped from “most, if not all” to less than one-third over the course of the last twenty years?

One significant factor is change in the population being studied.  Recent research generally includes not only those with what might be called “classic autism” (who tend to do poorly on standard intelligence tests), but also people with Asperger Syndrome.  Because Aspergians, by definition, have been assessed on standardized tests as having average or above-average intelligence, including them in research populations automatically raises the average IQ found.  But an equally important part of the explanation has to do with the methods by which autistic intelligence is studied.  From the 1960s to the 1990s, scientists regularly used intelligence tests (Stanford-Binet, Wechsler, etc.) designed for neurotypicals in their autism research, without questioning whether those tests were really appropriate for the population being studied. Around the turn of the millenium, however, scientists began to raise serious questions about these tests and the validity of their results.

One very basic issue is the use of language-based tests in studying a population that famously “thinks in pictures.”[4]  A 2007 study conducted by a team of scientists based in Montreal—a team led by Michelle Dawson, who is herself autistic—elegantly demonstrated the unreliability of conventional tests in assessing the intelligence of those who think in non-conventional ways.  The Montreal researchers looked at what happened when autistic children and adults, along with a group of neurotypical controls, took two different intelligence tests.  The first was the very widely used Wechsler test, which includes both verbal and non-verbal sections.  The second test was Raven’s Progressive Matrices (RPM), which is entirely non-verbal (it requires the completion of visual patterns) and which is generally considered to assess high-level analytical skills.  What they found was striking: as expected, the autistics did worse than the neurotypical controls on the Wechsler test—especially on the verbal sections.  However, on the RPM test, autistic scores rose dramatically (in some cases by as much as 70 percentile points) into what would be considered the “normal” range for a neurotypical population, while the neurotypical scores remained basically unchanged. [5]  The Montreal team concluded that because of the use of inappropriate tests, autistic intelligence has generally been greatly underestimated.  This point is confirmed by the personal experiences of many autistics.  A contributor to the Wrong Planet website, for example, writes:  “When I was 8 years old, my standard IQ scored in the moderately mental retardation range, but I scored 135 on a nonverbal IQ test at that same time.”[6]

A further problem, well-known to most people with autism, is that their ability to concentrate and communicate may vary drastically from day to day, for a wide variety of reasons, including how much sleep they’ve had, how much sensory overload they are facing, how much powerful emotion they are processing, how tiring they find the test itself, etc., etc.  An intelligence test given on any single day is, thus, much more a snapshot of their functioning on that particular day, than a reliable indicator of their overall intelligence.  Testing results may vary wildly depending on when they are given.  A middle-aged man who uses the moniker “Horus” online notes that “My own IQ scores have changed dramatically and I’ve taken seven professionally administered tests in my life. I scored 120 on the first one I took in 9th grade. I only scored 94 on the next one I took at 23. On the five tests I’ve taken since….I’ve scored as high as 143 and as low as 104.”[7]  If autistic intelligence is to be tested at all, it must clearly be done through a series of tests, given over the course of different days.

And a final fundamental problem has to do with the “global” scores derived from intelligence tests—the single number that most of us think of as someone’s “I.Q.”  For decades, an I.Q. of 100 has been defined as “average,” while one below 70 indicates intellectual disability, and one above 160 signifies “genius.”  But where do these single numbers come from?  They are calculated, using a variety of formulas (depending on the test being used) from the sub-scores on the different parts of the test, so that the global score is what might be considered an “average” of all the sub-scores.  Specialists in testing regularly warn us that global scores are not terribly significant in and of themselves.  It is more important to look at the sub-scores, to see each individual’s areas of intellectual strength and weakness.  Nevertheless, global scores continue to be widely used in research, to make claims about the intelligence of different groups in society (including autistics), as well as in everyday practice, to make decisions about what classes to place children in or what special services to offer them.

Calculation of a global IQ score is, however, based on the assumption that an “average” actually has some meaning—in other words, that the sub-scores on the different parts of the intelligence test will be roughly comparable with one another.  For most neurotypicals, there tends to be overall consistency across the sub-scores.  Some will be a bit higher, some will be a bit lower, but they mostly fall within the same general range, and so the global score can give us some indication of what that range is.  People with autism, however, tend to end up with very different scores on different parts of intelligence tests. It has been repeatedly shown that autistics perform much better on the non-verbal than on the verbal sections of intelligence tests, but it is also the case that they score much better on some of the subtests within these two large categories than on others.  In other words, if you look at an autistic person’s scores across all the different sub-sections of an intelligence test, you will very often see a spikey pattern of highs and lows, with some of the highs being very much higher than the lowest low.  (It is not unusual for someone with autism to achieve a “genius” level score on one part of a test, and a “sub-normal” score on another.)  But how should such varied results be factored into the determination of a global IQ score?   How useful is an average, when the numbers being averaged are so dissimilar?  Based on these discrepancies in sub-scores, a recent article in the Journal of Autism and Developmental Disorders has declared global IQ scores for autistics essentially meaningless.[8]

What has emerged, then, over the course of the last two decades, is a scientific consensus that people with autism are much more intelligent than used to be thought.  While far too many service providers, school administrators, teachers, therapists and other professionals still assume that a person with autism must be intellectually disabled, not only dry scientific data, but a growing number of compelling anecdotes illustrate the inaccuracy of that assumption.  It has also become clear that we still don’t know how to measure autistic intelligence accurately.  And until that problem is resolved—if it ever can be—we would do better to give autistic people the benefit of the doubt, or “presume competence,” than to deny them a proper education, jobs, or the opportunity to live independently, on the assumption that they are just not “smart enough.”

[1]  Marie Bristol, Donald Cohen, E. Jane Costello, et al., “State of the Science in Autism:  Report to the National Institutes of Health,” Journal of Autism and Developmental Disorders 26: 2 (April, 1996), p. 124.

[2] Eric Fombonne E (2005) “Epidemiology of autistic disorder and other pervasive developmental disorders,” Journal of Clinical Psychiatry 66 (2005):  3-8.

[3]  Jon Baio, “Prevalence of Autism Spectrum Disorder Among Children Aged 8 Years,” Centers for Disease Control and Prevention, Surveillance Summaries, March 28, 2014: http://www.cdc.gov/mmwr/preview/mmwrhtml/ss6302a1.htm?s_cid=ss6302a1_w

[4] Temple Grandin, Thinking in Pictures, and Other Reports from My Life with Autism (New York, 1995).

[5] Michelle Dawson, Isabelle Soulieres, Morton Ann Gernsbacher and Laurent Mottron, “The Level and Nature of Autistic Intelligence,” Psychological Science 18 (2007), 657-62.  See also Isabelle Soulieres, Michelle Dawson, Morton Ann Gernsbacher and Laurent Mottron, “The Level and Nature of Autistic Intelligence II: What About Asperger Syndrome?” PLOS/One September 28, 2011:  http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0025372; Valerie Courchesne, Andree-Anne Meilleur, Marie-Pier Poulin-Lord, Michelle Dawson, and Isabelle Soulieres, “Autistic Children at Risk of Being Underestimated:  School Based Pilot Study of a Strength-Informed Assessment,” Molecular Autism 6 (2015): https://molecularautism.biomedcentral.com/articles/10.1186/s13229-015-0006-3.  Dawson et al’s 2007 study is discussed Temple Grandin and Richard Panek, The Autistic Brain:  Helping Different Kinds of Mind Succeed (New York, 2013).

[6] Post dated September 13, 2010.  Discussion topic “Your IQ as a Child vs. Your IQ as an Adult”:  http://wrongplanet.net/forums/viewtopic.php?t=137652

[7] Post dated September 13, 2010.  Discussion topic “Your IQ as a Child vs. Your IQ as an Adult”:  http://wrongplanet.net/forums/viewtopic.php?t=137652

[8] Kerri Nowell; G. Schanding; Stephen Kanne; Robin Goin-Kochel, “Cognitive Profiles in Yourth with Autism Spectrum Disorder:  An Investigation of Base Rate Discrepancies using the Differential Ability Scales—Second Edition,”Journal of Autism and Developmental Disorders, 45 (2015), 1978-1988.