Tuesday 21 June 2011

Autism diagnosis and hyper-systemizing parents: Nottingham vs. Eindhoven

Family life in Eindhoven?
© cartoonstock.com

With an insensitivity one might have expected from a Sun headline writer, the New Scientist summarised the findings from a recent study thus: Childhood autism spikes in geek heartlands.  The bottom line is that autism is diagnosed more than twice as often in Eindhoven, a city in the Netherlands where many are employed in the IT industry, than in two other cities of comparable size, Haarlem and Utrecht. The explanation favoured by the researchers is that autism is characterised by a ‘systemizing’ style of thinking, which is part of normal human variation, and which is common in those in technical and mathematical occupations. Cities with a strong IT presence will attract an unusually high number of high-systemizer residents, and these people will be more likely to have offspring with autism. This is, then, a genetic explanation, and it’s given some plausibility by a large body of research demonstrating that parents of children with autism are more likely than other parents to show indicators of mild autistic-like characteristics, the so-called ‘broad autism phenotype’. (Although, on an admittedly quick search, the evidence that parents of children with autism tend to be high systemizers seems rather weak: mild symptoms of social and communicative problems seem a more pronounced feature of the broad autism phenotype than hyper-systemizing behaviours (Bishop et al., 2004; Wheelwright et al., 2010; Windham et al., 2010)).

As the researchers themselves note, the hyper-systemizing account is not the only possible explanation for their result. Crucially, the study relied on counting diagnoses from school records, rather than screening the population in a standard fashion. Although it’s not hard to recognise a case of classic Kanner autism, there’s far more disagreement about diagnoses for children with milder symptoms. As I argued on a Guardian blog, unless we have clear objective criteria for diagnosis, it’s hard to compare one prevalence rate with another. The different numbers could just reflect local expertise, policy or practice in diagnosing autism.

One limitation of the published study is that the researchers are quoted as saying that their study was prompted by anecdotal reports that autism was abnormally common in Eindhoven. While it is worth checking out if the anecdote is accurate, this makes Eindhoven less than ideal for testing the hyper-systemizing hypothesis, as it potentially capitalises on a chance blip. What would be better would be a study with clear a priori predictions, based solely on levels of IT industries in different cities. In theory, it should be possible to do this using publicly-available data from the UK published by the Department of Children, Schools and Families. This dataset has the advantage of being comprehensive, unlike the school report data from the Eindhoven study, which relied on schools providing the data (- the response rate was 75% for Eindhoven but only 50% for Haarlem and 46% for Utrecht). A recent report by Lindsay (2011)  presented some data from this UK database on numbers of children with Special Educational Needs categorised as having Speech, Language and Communication Needs (SLCN) or Autistic Spectrum Disorder (ASD). The numbers of children with these labels varied massively from place to place, as shown in Table 1.

Table 1: Percentages of children with SEN diagnosed with SLCN or ASD; 
Data from local authorities with the highest or lowest % of either diagnosis

I have no idea whether the number of IT experts is higher in Nottingham than Leeds, but it’s noteworthy that places where you might expect high levels of hyper-systemizing, such as the university towns of Oxford and Cambridge, don’t feature among the places with very high rates of ASD diagnoses. Lindsay also points out that “the two neighbouring authorities of Nottingham and Nottinghamshire have almost identical prevalence rates for SLCN and ASD, despite one being a large city, and the other a shire county” (p. 143). Clearly, my intuitions are no substitute for real data, and it’s also important to note that the data in Table 1 are not population prevalence figures, but instead are proportions of children who have already been identified as having Special Educational Needs. One would need to use the frequencies of ASD in the population as a whole to test the hypothesis of a correlation between level of IT industry and rates of autism. The data are available, and this might seem like a nice project for someone to do, except for a major problem. As Lindsay emphasised, it is impossible to conclude from the UK data whether prevalence really do vary across the country, because the definitions of disorders are inconsistent from one area to another. It’s possible that criteria for autism are more standardised in the Netherlands than in the UK, but the UK data make me suspect that it’s just not possible to draw meaningful conclusions about prevalence from data based on educational records.


Bishop, D. V. M., Maybery, M., Maley, A., Wong, D., Hill, W., & Hallmayer, J. (2004). Using self-report to identify the broad phenotype in parents of children with autistic spectrum disorders: a study using the Autism-Spectrum Quotient. Journal of Child Psychology and Psychiatry, 45, 1431-1436.

Lindsay, G. (2011). The collection and analysis of data on children with speech, language and communication needs: The challenge to education and health services. Child Language Teaching & Therapy, 27(2), 135-150.

Roelfsema, M., Hoekstra, R., Allison, C., Wheelwright, S., Brayne, C., Matthews, F., & Baron-Cohen, S. (2011). Are Autism Spectrum Conditions More Prevalent in an Information-Technology Region? A School-Based Study of Three Regions in the Netherlands Journal of Autism and Developmental Disorders DOI: 10.1007/s10803-011-1302-1 
Wheelwright, S., Auyeung, B., Allison, C., & Baron-Cohen, S. (2010). Defining the broader, medium and narrow autism phenotype among parents using the Autism Spectrum Quotient (AQ). Molecular Autism, 1(1), 1-9. 
Windham, G. C., Fessel, K., & Grether, J. K. (2009). Autism spectrum disorders in relation to parental occupation in technical fields. Autism Research, 2(4), 183-191.


  1. Neuroskeptic on the same study


    If only epidemiology really was this simple.

  2. Good post. Personally I think there must be a link but I doubt it would come out at the level of cities. E.g. Oxford and Cambridge may have lots of intellectuals in the city centre, but as cities they are much bigger and the great majority of people who live there, and in particular those who raise kids there (which junior academics don't tend to!), are not academics.

  3. Thanks for the brilliant analysis of the limitations of any comparison of prevalence of ASD diagnosis. Replication is essential taking into account the precautions mentioned. Thus, start with HighTec Areas and not with pockets of raised incidence. Indeed, here is a nice project for somebody.

    The Table on figures of prevalence of Language impairments and Autism spectrum disorder in different localities in the UK is striking. Epidemiologists would know what the unit size should be for comparisons. I guess that units much larger than average local authority size are needed to overcome noise/measurement error, with an estimated 1% base rate. Still, we have to worry about local differences in clinical and educational practice.

  4. Semantic-Pragmatic disorder, PDD, adhd, PDD-NOS Educational Autism, Aspergers, BD ...In order, by a neurologist, pediatrician, pediatrician, psychiatrist, Occupational Therapist, psychiatrist, psychologist, all labels for my son. You also have to take into account that school systems with more astute pricipals/superintendents are going to go for the more highly compensated disability here in America. Ask any school secretary.

    I give my son 2 labels with my education degree which are the only ones that make sense to me, and the only ones that are not questionable in the eyes of the public because of the "good" versus "scary" P.R. that's out there for all things autism, as well as the scarier answer (ABA) that "science" serves.

    My son is dyslexic, in 2 areas, calculations and writing. Those were the things that interfered with his schooling. These are things that most people realize are beyond anyone's ability to change, and that take really hard work to be overcome. These are the things that he needed heavy duty help in that the school couldn't give him in a special ed classroom.

    I am a systemizer... we all are a little autistic. Science, if you get right down to it, is overrun with autistic-type people. Very insensitive at times.... ;)

  5. You mention only in parentheses the other important issue - does "hypersystemizing" characterise the broader phenotype? Your thoughts on this in a blog would be very welcome, as I fear we are all rather confused and in need of some clear thinking. As you point out, this study is suggesting a genetic explanation. But it would be irresponsible of us to allow expensive genetic studies to be based on an unclear notion of the behavioural and cognitive features of any broader autistic phenotype.

  6. Very interesting analysis.
    I agree that if you look at cities, there seem many factors that could play a role. However, I do not see why you could not analyse at a city-level, as Neuroskeptic suggests. Isn't your point (the majority of people in the cities are not academics / IT professionals) the same for all the cities and balanced out?

  7. The problem with these kind of studies is that the parents are not handicapped nor do they meet diagnostic criteria for 'autism' using any of the gold standard diagnostic checklists (DSM-IV, ICD-10, ADOS-G, ARI-R). Diagnostic criteria for developmental disorders have been so broadened that there are no definitive boundaries between disorders (autism, ADHD, schizophrenia, language disorders) or between disorder and normality (autism, broad autism phenotype).


    In the last decade the most unexpected and surprising result is that the human genome is far more complex, diverse and even unstable than has been thought to exist. There are any number of common genetic polymorphisms associated with autism risk (serotonin transporter gene variations, COMT, MET) but these genetic influences cluster not only within families with a history of autism, but they extend very broadly throughout the general population and they are also associated with risk for schizophrenia, ADHD, depression, bi-polar disorder, etc.

    Is the BAP associated with technical occupations, perhaps, but the association may not be with autism per se but rather with what has been commonly called the BAP which are common, normal human characteristic traits within the complex diversity of all normal human characteristic traits.

    The association between technical occupations (I/T, engineering) may exist but the association is not necessarily with autism per se but rather with what has been called the BAP.

    That is entirely consistent with what Rutter has called the 'two-hit' hypothesis which states that the genetic influences underlying the BAP may not be the same as the genetic and environmental factors involved in the transition to the handicapping disorder.


    Within the framework of the two-hit hypothesis the genetic influences underlying the BAP is one of two independant component parts with genetic (Downs Syndrome, Rhetts Sydrome) and environmental (Congenital Rubella Syndrome, Valproate Acid Syndrome and Thalidomide Embryopathy) factors associated with disruption of early brain development and the neuroanatomical alterations that distinguishes affected from unaffected family members who may share the genetic influences underlying the BAP independant component part. The synergy between the BAP component part and the biological factors involved in the early disruption of brain development may follow a developmental trajectory towards an autism diagnosis.

    The autism twin studies are routinely referenced in studies of genetic factors in autism etiology and it is always stated that autism is the most heritable of the developmental disorders with a heritability of >90%. That interpretation is a contradiction to the concept of autism being a complex multifactorial disorder.

    What exactly is it that heritable that is being reflected in autism twin studies? Within the framework of the two-hit hypothesis it is assumed that the BAP component part is always present in accurately diagnosed cases therefore it may be the independant BAP component part, but not autism per se, that is heritable and that is being reflected in the autism twin studies.

  8. Autism is the fastest-growing serious developmental disability in the U.S. You can show support and raise awareness by wearing one of many autism awareness products .

  9. Hello, I really enjoy reading your blog. The things that people struggle with the most are those basic daily activities. I found out about these magnetic pictograms which have helped to reinforce those daily tasks and develop important life skills whilst still having fun!


    I also now put this star chart on my fridge so all the kids know when they have done a good job!


    Enjoy, and good luck!!