I usually keep a pretty close eye on the autism news, and it seems like a day can hardly go by without me seeing some new, hysterical reference to the rising rates of autism diagnosis.  It’s an epidemic, we’re told.  It’s a crisis, we’re told.

I not only find it tiresome to hear this idea repeated endlessly, but I’m also concerned that this sort of fearmongering is dangerous.

It’s true I have some biases here.  All communities want to imagine themselves as eternal.  We do that when we get nationalistic about countries – we reimagine history in terms of the national communities and nationalism that exist today (Anderson, 1983/2006).  I identify as part of an autistic community, and I therefore have a similar interest in seeing that autism is permanent.  It’s also worth noting that some of the debate around neurodiversity seems to hinge on the question of whether autism is natural or unnatural, with the idea being that a natural neurotype would be acceptable whereas an unnatural one should be eliminated.  I think my personal idea of neurodiversity doesn’t really depend on this distinction,[1] but it is another reason why I’m probably biased against the idea of a true increase in autism.

But even if I have my biases, at least I’m aware of them and taking them into account.  That’s more than most people can say.  Furthermore, I think I have a whole bunch of real, good reasons not to believe that we’ve seen much of a true increase in autism.  Let’s run through them.

The Burden of Proof

Whose view is right by default?  Is there an autism epidemic unless proven otherwise, or is there no autism epidemic unless we prove that there is?

I say there’s no epidemic unless proven otherwise.  We know that the diagnostic criteria for autism have broadened.  We know that there’s more awareness of autism today than before.  We also know that there’s tons of undiagnosed autistic adults out there – I keep meeting people in their twenties, thirties, forties, and fifties who are only now getting diagnosed for the first time.  We know that we’ve gotten better at diagnosing autism.  We know that there will have been an increase in autism diagnosis, regardless of whether there’s a real autism epidemic.

To say that there is a true increase in levels of autism symptoms, on top of better diagnosis, we need additional evidence.  As far as I’m concerned, the burden of proof is on anyone who argues in favour of the autism epidemic.

Is There a Biological Mechanism?

Those who believe in an autism epidemic have advanced all sorts of ideas as to what sort of environmental factors might be causing autism.  These deserve to be taken quite seriously, because if there is a clear biological mechanism causing an increase in incidence of autism, with no plausible mechanism to cause a corresponding decrease, that would be enough to prove a true increase in incidence.

I will admit that one certain mechanism out there is parental age.  We know that older mothers and fathers are more likely to have an autistic kid (Sandin et al., 2016).  Modern people wait longer to have kids.  Ergo, modern people are more likely to have an autistic kid.  This is absolutely right.  Basic logic.  I confess it – parental age is likely causing an increase in the true incidence of autism.

But is it enough to matter all that much?  Most likely not.  Mothers and fathers in their forties are about 15% and 30%, respectively, more likely to have autistic kids than parents in their twenties.  It’s not that big of a difference, really, not when you measure it against the ~3000-3500% increase in autism prevalence we’ve observed since the 1960s.  Certainly not enough to deserve the label of an “epidemic.”  Besides, younger parents are also more likely to have autistic kids, and rates of teen pregnancies have been dropping.  We also have this cool modern healthcare system, which helps to avoid birth complications, and avoiding birth complications will also make autism less likely.

That’s the funny thing about many of the explanations for an increase in autism rates.  People look for an explanation for an increase, but there might actually be more reasons to believe in a decrease.

Let’s take pollution.  Some people have argued that exposure to air pollution increases autism risk.  Quick question: has pollution increased in Western urban centres over the last few decades?  Not really – it’s been declining (Laden et al., 2006).  We’ve exported pollution to the Global South, of course, so the environmental impact is still there, but air quality has not gotten worse in Western urban centres.  Therefore, if air pollution causes autism, we ought to expect autism incidence to be decreasing in Western countries.

I suppose there might be an intergenerational epigenetic process of some kind that would make the effects of past pollution manifest in the present, but even if such a process exists and substantially increases rates of autism, that still leaves us with plenty of other biological reasons why autism rates might be decreasing.  For example, there used to be all sorts of childhood diseases we don’t have today.

What about diet and nutrition?  Well, diet is kind of funny.  We get lots of strange ideas about diet, then we start believing them as though they’re indubitable facts.  There’s a cool study showing that physicians agree more about diet with laypeople from their own countries than they do with foreign physicians – even in relatively similar, industrialized, Western countries (Leeman et al., 2011)!  Beliefs about diet vary so much across time and culture that I can’t really take them too seriously.  I think the only safe conclusion is that a balanced diet with adequate nutrition is healthy, and I’m pretty sure we eat better today than we have for most of human history, so I’m not inclined to see modern food as an explanation for autism.

It’s true that parental age would cause an increase in autism incidence, and maybe there are some other mechanisms that would cause an increase, but given everything else – better healthcare, better nutrition, etc. – I don’t know that we can actually conclude autism incidence has increased.  One could just as well argue that there’s probably more biological mechanisms that would be causing a decrease.  At the very least, we need a lot more research comparing different biological mechanisms before we can draw the conclusion that true incidence has increased.

What About a Cultural Mechanism?

Strangely, cultural explanations haven’t attracted too much attention of late.  There are some fringe people out there in France and South Korea and other places who still believe that bad parenting causes autism, but since that theory was debunked, we’ve not really advanced many cultural explanations.  Right now, the Zeitgeist is all biology, so people are looking for biological explanations.  There’s more money in it, and it fits in better with the current research culture.  Biology is in fashion, and culture is out of fashion.

I actually think a cultural argument would be more compelling than a biological one.  I’ve written elsewhere about how I think our society has gotten more complex and competitive, and I think that that makes it more difficult for autistic people – leaving co-occurring conditions aside – to function, at least relative to neurotypicals.  While I don’t have any bizarre beliefs about autistic people’s lives being perfect in the past,[2] I think it’s possible that autism could be more noticeable in the modern world because it is more impairing.

But this argument, interesting as it is, wouldn’t actually be about a true increase in autism incidence, at least not as I see things.[3]  It might make people’s lives more miserable and cause an increase in the rate of co-occurring mental health problems, and it might create more barriers and make autism more impairing and noticeable, but it wouldn’t make neurotypical people suddenly turn into autistics.[4]

Controlling For X…

What else could we do to argue a true increase in autism?  Well, we could always go back to the trusty old argument from personal incredulity.  It goes something like this:

I cannot imagine that anything except X, Y, or Z could be driving an increase in autism diagnosis, independently of a true increase in incidence.

Autism rates still increase after controlling for X, Y, and Z.

Therefore, there has been a true increase in autism incidence.

That’s not a terribly good argument, because it depends on the author successfully imagining all possible reasons why we might be getting better at diagnosing autism.  Authors don’t usually do this, because there’s lots of reasons.  Not just obvious things like broadening of diagnostic criteria, or even diagnostic substitution, but things like people’s awareness of autism and the salience of the autism diagnosis (see, e.g., Liu & Bearman, 2015).[5]

Also, statistical control only works properly if your measurement is perfectly reliable.  That’s why ice cream causes pool deaths, even after controlling for subjective temperature (because subjective temperature ratings aren’t perfectly reliable; see Westfall & Yarkoni, 2016).  This is actually a problem with a lot of the studies trying to find biological causes of autism – inadequate statistical control for confounds.

Is an Answer Possible?

So far, we’ve wandered all over the landscape of autism.  We’ve seen that there are biological mechanisms that could cause increases in true incidence, but also others that might cause decreases.  We’ve discussed the possibility of a cultural mechanism, but dismissed that as being not exactly the same as a true increase.  We’ve also seen that an argument based on statistical control is inadmissible, both because of failures of imagination and failures of statistics.

So what sort of evidence would I accept, without any counterargument?

Well, I can think of one thing.

This is a bit tricky, because it’s possible that our reports about autism symptoms might change over time as people become more aware of the existence of autism.  However, if there was a way to get a good, reliable sense of the true level of autism symptomatology independent of diagnosis in successive representative cohorts of children over time, that would allow one to say that autism incidence has increased.  If you found a true increase in autism symptoms, you’d be able to say that, yes, autism has truly increased.  In this type of cross-temporal analysis, there would be no need for statistical control or any of that.

Unfortunately for those who believe in the true increase in incidence, when we do a cross-temporal analysis, that’s exactly the opposite of what we find.

Lundström et al. (2015) asked all Swedish parents of twins (which is representative enough of the general population for our purposes) to report their kids’ autism symptoms (and those kinds of reports are probably not perfectly reliable over time, because we know they’re affected by culture,[6] but they’re a good start) in successive years with successive cohorts.  The researchers also asked them to report whether their kids were diagnosed with autism.  Guess what?  Autism diagnosis increased; autism symptoms did not.[7]

The burden of proof might be on those arguing for a true increase, but right now, it looks rather like we have some good evidence that there is no true increase.  At the very least, if there is a true increase, it can’t be a particularly large one.  So perhaps we should think about shifting some of the resources going into studying the environmental and biomedical causes of autism and instead using them to study things like employment interventions, school placement, and mental health?

If you have thoughts, please comment below!  I am interested in your reactions and feedback.

Footnotes

[1] When I say that we have to take special care to protect “innate” mental variation, that’s really contrasting someone who has been a certain way since very early in development from somebody who got TBI or something later on.  I’m using “innate” very loosely there – better psychologists would remind me that innateness is super-complicated and we have experience-dependent and experience-expectant mechanisms, blah blah.  But I think it conveys the idea of something that has pretty much always been part of the person, which is all I mean.  Doesn’t matter whether it is “natural” or not.

[2] For one thing, “relatively” better does not mean better in absolute terms.

For another, those with severe co-occurring disabilities would be at great risk of abuse.  This was true even before the horrors of institutional psychiatry – it’s easy to forget, but early campaigners for institutionalization like Dorothea Dix were thought to be humanitarians, because people in community settings had terrible lives at the time.

[3] I suppose someone bound to a pathology paradigm who sees impairment as a required feature of autism might disagree with me.

[4] One intriguing cultural hypothesis I read the other day is the idea that cellphones cause autism because they keep parents away from eye contact with their kids (Davidovitch et al., 2018).  There’s no firm evidence just yet, but it’s certainly an interesting idea.  However, we should be cautious about leaping to conclusions.  The increase in autism began well prior to the emergence of the smartphone.  We should also bear in mind that Western families have fewer children today than they used to, and the idea of insufficient parental attention causing autism rather contradicts everything we’ve been hearing about helicopter parenting.  In fact, there’s a lot of cross-cultural variation in the amount of time parents spend in face-to-face contact with their infants (Keller et al., 2009).

[5] For the record, I most certainly did not make the error of arguing from personal incredulity in the biological mechanisms section.  I admitted that there is at least one cause of a true increase in autism incidence – parental age – and I admitted there might be others.  My point was that there are also biological reasons why true autism incidence would decrease.  Besides, if we’re arguing it’s an epidemic, we need more than a minor increase: we need a big increase!

[6] See, e.g., Neuhaus et al., 2018.

[7] [Added January 1, 2020: The Swedes later also found that fewer characteristics of autism became necessary for a diagnosis over time.  This supports the idea that factors like an expanded definition of autism, as well as greater awareness and salience of the category, were responsible for the apparent increase in autism rates.  See Arvidsson et al., 2018.]

References

Anderson, B. (2006). Imagined communities (New ed.). London, UK: Verso. Original work published 1983.

Davidovitch, M., Shrem, M., Golovaty, N., Assaf, N., & Koren, G. (2018). The role of cellular phone usage by parents in the increase in ASD occurrence: A hypothetical framework. Medical Hypotheses, 117(January), 33–36. https://doi.org/10.1016/j.mehy.2018.06.007

Keller, H., Borke, J., Staufenbiel, T., Yosvi, R. D., Abels, M., Papaligoura, Z., … Su, Y. (2009). Distal and proximal parenting as alternative parenting strategies during infants’ early months of life: A cross-cultural study. International Journal of Behavioral Development, 33(5), 412-420. https://doi.org/10.1177/0165025409338441

Laden, F., Schwartz, J., Speizer, F. E., & Dockery, D. W. (2006). Reduction in fine particulate air pollution and mortality. American Journal of Respiratory and Critical Care Medicine 173(6): 667-672. https://doi.org/10.1164/rccm.200503-443OC

Leeman, R. F., Fischler, C., & Rozin, P. (2011). Medical doctors’ attitudes and beliefs about diet and health are more like those of their lay countrymen (France, Germany, Italy, UK and USA) than those of doctors in other countries. Appetite 56(3), 558-563. https://doi.org/10.1016/j.appet.2011.01.022

Liu, K., King, M., & Bearman, P. S. (2010). Social influence and the autism epidemic. American Journal of Sociology, 115(5), 1387–1434. https://doi.org/10.1086/651448

Lundström, S., Reichenberg, A., Anckarsäter, H., Lichtenstein, P., & Gillberg, C. (2015). Autism phenotype versus registered diagnosis in Swedish children: Prevalence trends over 10 years in general population samples. BMJ, 350, h1961. https://doi.org/10.1136/bmj.h1961

Neuhaus, E., Beauchaine, T. P., Bernier, R. A., & Webb, S. J. (2018). Child and family characteristics moderate agreement between caregiver and clinician report of autism symptoms. Autism Research, 11(3), 476–487. https://doi.org/10.1002/aur.1907

Sandin, S., Schendel, D., Magnusson, P., Hultman, C., Surén, P., Susser, E., … Reichenberg, A. (2016). Autism risk associated with parental age and with increasing difference in age between the parents. Molecular Psychiatry, 21(5), 693–700. https://doi.org/10.1038/mp.2015.70

Westfall, J., & Yarkoni, T. (2016). Statistically controlling for confounding constructs is harder than you think. PLoS One, 11(3): e0152719. https://doi.org/10.1371/journal.pone.0152719