Friday, February 14, 2014

Autism and the average

Stephen Jay Gould. Source:
Around this time every year, my almost-namesake, John Brockman, publishes on his website, the Edge, a selection of answers to an Annual Question. What have you changed your mind about? What is your favourite deep or beautiful explanation? What should we be worried about?

This year's question was particularly intriguing. What scientific idea is ready for retirement? Diverse answers include: big data, universal grammar, entropy, string theory, radical behaviourism, common sense, carbon footprints, neuroplasticity, the continuity of time, the idea that science is self-correcting.

My favourite response, however, came from Nicholas Christasis, a physician and social scientist at Yale University. His answer: the Average.
"Ever since the landmark invention of diverse statistical techniques 100 years ago that allow us to properly compare the difference between the averages of two groups, we have deluded ourselves into thinking that it is such differences that are the salient—and often the only—important difference between groups. We have spent a century observing and interpreting such differences. We've become almost obsessed, and we should stop." 
As an autism researcher, this is an issue particularly close to my heart. Despite the incessant talk about the "heterogeneity problem" in autism research, most autism studies begin and end with the question of whether or not on average people with autism are different on some measure to people who don't have a diagnosis.

This carries with it the assumption that it is the average (and not the variation around the average) that matters; that it captures some essential property of autism. It also invariably leads to contradictory findings, where one study finds some interesting difference on average between autistic and non-autistic people - and another finds no difference or even the opposite effect.

My latest article for the Simons Foundation Autism Research Initiative elaborates on this point. It was inspired initially by a wonderful essay by Stephen Jay Gould, discussing life-expectancy statistics in the context of his own background in evolutionary biology. As Gould wrote:
"All evolutionary biologists know that variation itself is nature’s only irreducible essence. Variation is the hard reality, not a set of imperfect measures for a central tendency. Means and medians are the abstractions."
This, I argue, is an important lesson for autism research.
"We have to take heterogeneity seriously as the object of investigation, rather than an excuse for inconsistent results or an inconvenience in our quest to understand the essence of autism."
The search for autism's core essence is something that has driven research ever since Kanner first described autism in 1943. However, seventy years later, we're still not much closer to an answer - as evidenced by the controversies surrounding the latest attempt to redefine autism in DSM-5.

There's no denying that "autism" is a useful meme in terms of raising awareness and in communicating in very broad terms the kinds of difficulties a person may face. "Autism" also serves an important purpose in bringing together likeminded people for support, advocacy, and research.

Clearly, the symptoms that affect people diagnosed with autism are real and for many present huge challenges. However, autism is not an explanation for those challenges - it is those challenges. And given the incredible diversity within autism, it's not clear that autism is an actual thing to be explained either.

So, inspired by John Brockman, my question to you is this: Are we now at the point where, as a scientific concept, "autism" is ready for retirement?

Related posts:

Further reading:
  • Intellectualizing: Review of Rethinking Autism: Variation and Complexity by Lynn Waterhouse


  1. The thing that always strikes me in the incredible diversity of externally observable autistic behaviors you can see in identical twins who both share an autism diagnosis. If you have two people with the almost identical genes, the same prenatal environment, and the same early childhood environment - or all of the places we know there are "risk factors" - then how can the disorder manifest itself differently?

    I think the obvious answer is it doesn't and we are defining "autism" using the wrong thing. We are looking at manifestations of an underlying condition and calling that "autism" even though the manifestations are essentially random. The specific stims, rigidities, sensory issues, and even the social/communication behaviors that a person has are essentially random and unique in the same way that a person's personality is random and unique.

    But just because identical twins can have vastly different autistic behaviors doesn't mean that they don't have the same underlying condition. It just means that we don't have a more specific way of labeling where the actual problem is at.

    So while I agree with you that it is likely that the autism label includes many different related conditions, I don't think that the response is to throw out the entire label. We need to get better at picking out the sub groups in the autism population based on other measurable biological imbalances because it is these biological imbalances that are causing the behaviors.

    1. Thanks MJ. This is a really important point. It shows just how tricky (impossible perhaps) it is to define "autism".

      As you say, the case of identical twins suggests that the same underlying genetic factors can have very different manifestations. But does that then mean that, even if they're very different "on the surface", they have the same thing?

      With MZ twins concordant for autism diagnosis that sounds like a reasonable argument. But does that then mean that two completely unrelated people (to your twins or to each other) have the same thing because one looks like twin A and one looks like twin B?

      What if the twins aren't concordant for autism diagnosis? Do we then broaden the diagnostic criteria yet further. Twin A has autism so Twin B must have autism too (even though B doesn't meet our diagnostic criteria). So now anyone who looks like Twin B also has autism.

      So then we have to come back to your twins and say, yes they're very different, but do they really both "have" the same thing? Does it make sense to say that they both have autism?

      I don't know the answer (in general and certainly not in the case of your actual twins). I just think we need to be careful saying "These two people are genetically identical so the answer must be yes"!

    2. You only have a real conundrum if you assume that autism is mostly "genetic" (i.e. caused by one or more permanent changes to dna). But if recent studies of twins have shown us anything it is that "autism" is in large part caused by something other than stable genetic mutations. Under that sort of model it is quite possible to have one identical twin who have autism while the other does not. The diagnostic criteria don't broaden because it is not automatically assumed that the non-diagnosed twin must have autism as well.

      On the flip side, if both identical twins have autism then it should be a safe assumption that there is some genetic or environmental factor that is common to both of them that causes the behaviors of autism. Even if that commonality is environmental such as something that they were exposed to in the prenatal environment. I think it would be a real stretch for identical twins to have disparate causes because they have so much genetically and environmentally in common. Although I have heard of cases where a second event, such as illness, can make one twin more severe than the other.

      I think the real point is that you can't assume that the behaviors necessarily reflect the underlying condition because the specific behaviors are random and individual. You have to look at the general pattern in the behaviors as well as at more concrete (and measurable) biological disruptions. And when you look at that level you start seeing patterns that might be more useful for grouping like people together for study and treatment.

      Take my family for example - all three of my daughters have a diagnosis of autism but all three are quite different from each other with respect to their autistic behaviors. We will see some behaviors in only one of them, sometimes two - although not only just the twins, but only very rarely will all three have the same behavior.

      However, if you look behind the specific behaviors and at general patterns you can see a lot of commonalities. All three are very visual in their stims even if the specific way they stim is quite different. All three are sensitive to sound although one twin and the youngest are overly sensitive to sound while the other twin likes constant loud sounds. The twins have significant problems with the functional aspects of communication while the youngest has only minor problems in that area.

      On the biological side they have even more in common. All three have chronic zinc deficiency even after taking supplements. All three have issues with being borderline anemic even though they have a diet that is high in iron. All three have no detectable level of lithium in their blood (someone needs to explain that one to me). All three have a similar skewing of their immune system across several different tests. All three have issues with certain dietary proteins - two have benefited from a restricted diet while the third has benefited from another approach. The two on the diet have shown signs of high levels of IgG antibodies against specific proteins.

      The question is then should we look at just the different behaviors in all three and say that that they might not even have the same condition or should we look beyond their individual behaviors to the commonality behind the behaviors?

      Based on factors like the ones I listed about plus years upon years of experience I think it is a safe bet to say that all three of my children have the same condition even if they look different on the surface. I don't think my family is unique in this regards.

    3. Classical twin study design is long past its retirement date. It is folly to think that you can calculate a 'heritablity' estimate based on the concurrence rates between MZ and DZ twin pairs. According to classical twin study design, Down Syndrome, not autism, would be the most 'heritable' of the developmental disorders. Almost all MZ twins are concordant for Down Syndrome and almost all DZ twin are discordant for autism. MZ twins share the same genetic architecture but 1/3rd of MZ twins do not share the exact same prenatal environment. 1/3rd of MZ twins develop in seperate placentas while 2/3rds of MZ twins share the same prenatal environmental since they develop in MC (monochorionic). That may be why many classical twin studies have reported a 60% concordance rate in MZ twins. Classical twin study has fostered more misunderstanding of autsm inheritance (ie autism is the most heritable of the developmental disorders).

  2. This comment has been removed by the author.

  3. As a scientific concept, 'autism' is narrow, limiting and fails to express the complexity of human nature. If your experience of the material world is overwhelming and nonsensical because the reality that makes sense to you is within a transcendent 'home', wouldn't you have a tough time adapting to what some people are trying to convince you is real?

    I offer a different perspective to mainstream understanding of autism. To understand autism means that we might consider that some people exist in an expanded state of consciousness all the time. Being in such a state for prolonged periods also enhances and amplifies all of the sounds, sights, noises, densities, tastes, emotions and thoughts of people and things around the person.

    Unlike a spiritual aspirant or master who has learned to transcend to higher states of consciousness gradually to eventually stably exist in an expanded state permanently, while managing the input of all things at lower, ground states, children, adolescents and adults with autism might not have had the training required to help them understand and manage what's happening on the ground, so to speak. Nor have the people who work with them, unless they're people like Donna Williams.

    Science is still limited in this area and attempting to understand 'autism' will require an expansion of the human paradigm used to develop any scientific theories.

  4. Hi Jon,

    Very interesting piece, thanks! I hear your call for caution in lumping together heterogeneous pathologies. But I also just read Hobson ( and think he has a few good arguments for the possibility of something like autism with a common core. I'd like to hear your take on them.

    - We know other diseases that are very heterogeneous yet coherent. Hobson gives the example of diabetes. Some symptoms (eg heart vs eye failures) may not be expressed at all for any particular patient. Still, "that individual is vulnerable to the full set of physical complications, some of which may exist in subclinical form." There is no reason to think that ​autism is any simpler than that model. We don't know as much about the level of coherence between mental faculties as we know about 'coherence' in biological organs.
    - As you rightly say, autism is not an explanation but a description of the challenges. The same largely holds for existing cognitive theories, I would say. Hobson again: "one might question how far they render the pathogenesis of the syndrome intelligible in terms of processes that we understand better than the phenomena we are seeking to explain."​ Can we discard unity in autism before we understand underlying processes better? How do we know we are searching at the right level of description?

    Science is about taking phenomenology seriously but going beyond, finding the right concepts to understand phenomenology. For autism we haven't gone much further than phenomenology, one could argue. Maybe this means there just is no such thing as autism, maybe we don't understand the symptom domains (and normal functioning) enough to see that a limited set of processes could in fact cause the different deficits in differing degrees. With which position lies the burden of explanation? To me it seems more fundamental research is especially needed (maybe by reorienting some of the funding for autism research proper?).

    Clearly, 'autism' was and still is very useful 'politically' (in getting resources, both for researchers and for patients/caregivers), possibly to the extent that these politics were partly driving the lumping together of different pathologies? Put differently, is a middle position possible that acknowledges that the autism category has expanded too much to be scientifically useful, but at the same time still finds enough commonality (in a subset of cases) to warrant a term (why not 'autism')? This is akin to what commenter MJ above describes, though I don't think it will be as simple as finding the right 'biological imbalances'. I do agree with your thesis that we should take individual differences seriously in all their messiness, because they can probably teach us important things about the underlying processes (my labmates wrote this interesting chapter about that:

  5. It is good to see this issue being aired; can I just add the perspective of someone trying to understand for the purpose of education and support? I take it as obvious that there are different 'autisms' and that behavioural diagnosis is crazy - there are (as the comments make clear) many disparate ways in which an autism may be expressed behaviourally and, conversely, many other conditions (notably, anxiety states) that can lead to what are mislabelled as 'autistic' behaviour. The trouble with many diagnostic systems (especially those approved of as 'scientific' by the journals) is that they depend on increasingly narrow behaviour-based definitions that exclude all the 'messiness' of real -world cases; as a result we have come to know more and more about less and less i.e. mythical cases of 'pure' autism. No wonder there is such poor overlap between scientific findings and practical understanding and support.

    I would like to make the case articulated by Hobson (and before that by Lorna Wing) that what unites those with ASD and distinguishes them from others is that they do not seem to have the basic instinctual make up that enables them to lock onto, and benefit from, the social world. If they are intellectually bright and are lucky enough to receive appropriate education and support, they can often acquire the behaviours that enable them to fit into the social world but they always remain at a disadvantage (and thus vulnerable to disruptions in development) since cognitively acquired understanding and skills are always less fluent and easy than those that come as a birthright, through instinctual processes. It is important that parents, teachers and the individuals themselves recognise this difference because that is the only way we will develop the explicit teaching and nurturing that are needed. Of course, not being socially obsessed, as the rest of us are, also carries some advantages and can lead to high levels of talent and ability as the many achievements of those with autism are testament to.

    Yet being asocial in a social world is undoubtedly a disadvantage and we need a way of recognising and compensating for that, where possible. Most of us have social guidance to help us interpret the world, understand others and know what is dangerous or safe; only those with autism are left trying to work things out for themselves. It is hard enough trying to train professionals to understand these differences and develop helpful strategies. Without a label that identifies this asocial development, I fear we will be left with us NTs making assumptions about behaviour, based on our understanding of the typical, and be back in the situation of ignorance where people on the spectrum are 'blamed' for their differences - neither understood nor supported. We are beginning to understand what it is like to exist without the capacity to benefit from the social world and to develop strategies that are supportive. If science has a warped view of autism, because of its obsession with replicability rather than ecological validity, then surely it is science that needs to change.