About a month ago, the eminent developmental psychologist and autism cognitive researcher Uta Frith shared a number of opinions publicly in short succession.  First, Frith senior-authored (along with other scholars, including Simon Baron-Cohen as lead author) a paper defending the idea that autistic people sometimes struggle with theory of mind.  I was somewhat disappointed the paper did not more fully engage with the relevant issues, but I don’t necessarily disagree with the conclusion.¹ However, almost immediately afterward, an interview Frith did for Tes Magazine appeared online, rapidly followed by an appearance on a podcast hosted by Naomi Fisher.  In these more wide-ranging discussions, Frith did not hesitate to offer frank and candid comments on numerous issues.  They were then repeated in other outlets.  Now, there have been some great responses to Frith’s remarks already, but I’m chiming in a bit belatedly to comment with my takes on a couple of issues:

1. Frith’s comments about the diversity of autism – specifically, her argument that early- and late-diagnosed autistic people are two different populations, and that only the former should be considered genuinely autistic
2. Frith’s dismissal of sensory issues, her suggestion they are specific to late-diagnosed people, and her skeptical remarks about crucial protections and accommodations

That’s not to say that I necessarily accept Frith’s other points – I am quite concerned about her remarks on masking and social anxiety, for one – but this is already a long post, so I thought I should try to keep it at least somewhat focused.

Also, while I do disagree strongly with some of Frith’s views here, and I’m concerned about the way some outlets are amplifying Frith’s views without providing counterpoints, I am nevertheless glad that Frith shared her views on these questions.  It’s important to have such discussions so that our field can address disagreements openly and navigate to a consensus based on evidence and reason.² While there has been a lot of discussion recently about the diversity of autism, the sensory issues are less openly debated.  While I definitely see sensory issues being dismissed by people in the community, it seems like those who think sensory issues are fake or unimportant are usually just quietly ignoring them, rather than actively debating their existence.  Because Frith spoke openly about her doubts regarding sensory issues in autism, a dialogue becomes possible.

Furthermore, in the podcast, Frith also directs attention to the lack of research that is inclusive of autistic people with co-occurring intellectual disabilities.  At risk of being accused of hypocrisy (since much of my research focuses on people without intellectual disabilities), I do agree and believe we need to develop mechanisms to prioritise funding for meaningful, community-partnered, serious, methodologically-careful research regarding autistic people with varying degrees of co-occurring intellectual disability.

Frith also highlights the importance of transdiagnostic research and comparing different neurodivergent and disability groups with each other, rather than just with the general or neurotypical population.  I completely agree with that.

Diversity and subgroups in autism

Nevertheless, I strongly disagree with Frith’s take on the diversity of autism.  Like many others, Frith is concerned about the ongoing rise in autism prevalence and the increasing diversity of autism.  Indeed, many orders of magnitude more people are receiving autism diagnoses than was the case at the beginning of Frith’s career.

As Frith says, this is almost certainly due to cultural factors.  I’ve previously written about how I think the closure of many institutions in Western countries during the 1970s and 1980s played a key role – once autism came with special education supports in mainstream settings, rather than a one-way ticket to an institution, the diagnosis would have become much less threatening. Furthermore, I’ve suggested that there’s a number of reasons why autistic traits may be becoming increasingly disabling, making such diagnoses more important as well.  Our modern world is extremely competitive in many ways, including finding friends in settings where one’s social world isn’t pre-given, which tends to magnify the importance of individual differences like neurodivergent traits.  I also noted that our cities are large and overwhelming, making individual differences in susceptibility to overload more crucial.  It probably doesn’t help that industrial jobs in the West have been vanishing for decades, being lost to either automation or globalisation.  These jobs weren’t particularly demanding in terms of interpersonal or technical skills, but especially in the postwar period, they were a reliable source of middle-class wages.  Even if they might have involved sensory stresses for some people, the existence of these jobs probably helped to benefit everyone by mitigating some of the modern world’s complexity and competitive pressures. However, in their absence, the economy is increasingly divided between low-wage unskilled³ service jobs versus higher-wage knowledge economy jobs.  We’ll see how generative AI changes things, but for the moment and the foreseeable future, competing over the well-paid jobs is difficult and stressful, and that probably again heightens the importance of neurodivergent traits.  Even when children are growing up, parents need to worry about their ability to succeed in this competitive knowledge economy – not only in terms of technical skills and education, but social dynamics of workplaces.

Thus, there are good reasons why more and more people are being diagnosed with or identifying with autism.  Again, Frith doesn’t deny that – she admits that all of those included in the broad constellation of autism “really do have problems,” as she puts it.  But her argument is that autism has become so broad as to be incoherent. 

Now, there’s a couple of different ways one could deem autism incoherent.  One would be to say that the range of supports and services different people need is so broad that the category has no practical value, and the other is that the category itself lacks coherence.  Frith seems to be trying to apply both of these arguments, but I’ll address the supports and services one first.

Coherence of supports and services

Frith, in the Tes Magazine article, points particularly to her belief that the groups require different educational supports.  Here, she actually distinguishes three categories of people:

1. Students with co-occurring intellectual disabilities and “challenging behaviour” that “cannot very easily be accommodated in the mainstream.”
2. Students with good language who would have fitted into the old Asperger’s category, who can “really benefit from mainstream education.”
3. The “extended” spectrum or constellation of autism, with the people who Frith does not consider genuinely autistic.  Children in this category, Frith says, need to work on “building resilience,” which I presume means continuing to attend mainstream school and persisting through challenges there.

However, I disagree with Frith’s views about appropriate educational settings.  I’m by no means attached to the idea that everyone should be in mainstream schools, but I don’t like the way this framework seems to match what is convenient for the system rather than the needs of the students.  Perhaps students in the first category are difficult for the system to accommodate in the mainstream, but they might still benefit from being in the mainstream.  Conversely, some – but not all – students in either⁴ of the other two categories might find the mainstream highly stressful and benefit considerably from access to alternatives if they want them.  We want children to enter adulthood with resilience, yes, but if they’ve burnt themselves out trying to cope with overwhelm and arbitrary social expectations even when that was unnecessary, they may end up with a lot of mental health damage and worse resilience…

In general, though, I would concede that the autism constellation has become too broad for a simple autism label to be particularly useful in directing people to supports and services, except in early childhood when developmental and (more controversially) naturalistic-developmental behaviour interventions may be useful for a broad range of autistic people.  In older childhood and adulthood, certainly, service needs are highly variable depending on an individual’s stable traits, fluctuations in things like mental health and burnout, the contexts around them, and their goals.

The feasibility of Frith’s proposal to subdivide autism

However, that very diversity of support needs – and note again it depends on contexts, goals, and fluctuating characteristics, not just stable traits – means that I’m not convinced we’re in a position to carve autism up into a set of more useful subgroups.  Specifying co-occurring conditions can be helpful, but I’ve yet to see a proposal to subdivide autism itself that I consider practical.⁵

Frith proposal’s is no exception.  In general, she distinguishes between two groups:

1. Those with early diagnoses – who are, in Frith’s view, genuinely autistic
2. Those with late diagnoses – who, in Frith’s view, have been improperly added to the autism constellation

Frith’s idea here aligns somewhat with a recent paper suggesting there might tend to be some genetic differences between people with early and later autism diagnoses.  However, this by no means implies that we are dealing with two categorically discrete populations, or that they can be practically disentangled in any sort of diagnostically useful way.  Notably, twins with identical genetics, even if they are both autistic, can have wildly different levels of autism traits.

Furthermore, while I’ve always in the past tended to dismiss optimistic studies claiming that low average ages of autism diagnosis are being achieved in the community (e.g., suggestions that the average age of diagnosis is 4 or 5 years),⁶ it seems like today a very large and diverse range of autistic children finally are getting diagnosed in the preschool years.  For example, recent American data suggests 3% of 4-year-olds can be identified as autistic through searching their records, and 70% of them had already received an autism test.  Realistically, in such a large group, we must be capturing some people in the supposedly “later-diagnosed” group at early ages these days – and it’s notoriously difficult to predict future developmental outcomes at such ages.  Thus, at least in those Western countries with high autism awareness levels, age of diagnosis hardly seems like a practical or useful dividing point today.

To be fair, I’m not sure that’s exactly what Frith was suggesting – she also seems to be pointing to other features she believes distinguish the two groups.  But much of this seems confused or wrong to me.  I don’t mean any disrespect, but Frith for example argues that being able “to get irony and humour” would in her view be a “contraindication” for autism, and I find this comment quite bewildering – I’ve met plenty of early-diagnosed people and people with more obvious autism profiles who are extremely humorous!  Granted, the content of the humour may not always be conventional, and the timing and delivery can sometimes be unusual, but it is certainly often there.

So, at this point, I don’t think Frith has advanced a credible set of criteria or indicators for distinguishing the two putative groups she is trying to separate.

Coherence of autism itself

But Frith does nevertheless reject “the idea that there is something that unites all the people who are diagnosed as autistic.”  This may seem like a contradiction in terms.  After all, by definition, doesn’t everyone who gets an autism diagnosis need to meet criteria from a list of behavioural features?  And as Frith says, our interpretation of the list has changed much more than the list itself has changed, but surely the fact people do in some way share the traits on the list means they do have something in common?

Be that as it may, I think I can understand what Frith is getting at – that even if people might seem to in some way meet these criteria, the criteria might be vague enough that there are starkly different ways of meeting them.

Frith seems to think that the late-diagnosed individuals who, in her view, are not really autistic are different from other autistic people in a couple of broader ways than specific indicators like humour levels.  First, she dismisses their social difficulties as simply social anxiety – I don’t agree, but I’m not going to address that point here because this is long enough already.  But second, in the Tes Magazine article, she suggests that the late-diagnosed, not-genuinely-autistic group is “characterised mainly by a sort of hypersensitivity.”  In the podcast, she explicitly adds that sensory issues were “not originally” being discussed.  All of this seems to suggest that Frith believes sensory differences are not characteristic of those early-diagnosed individuals she does believe are autistic, and that they are characteristic of the late-diagnosed individuals she believes are not autistic.

Sensory responses and experiences

However, I found Frith’s remarks about sensory hypersensitivity being associated with later-diagnosed individuals puzzling and inconsistent with evidence.  In fact, sensory issues have long been apparent in autism, since well before its expansion.  For example, according to Frith’s own translation, Asperger commented on hypersensitivity and hyposensitivity to noise in his 1943 paper.  Admittedly, in general, early literature tended not to dwell on sensory differences, but there were exceptions clearly demonstrating that unusual sensory reactions were present.  For example, one 1947 paper focuses largely on sensory sensitivities, and in the 1960s, Ornitz and Ritvo published exhaustive descriptions of sensory differences, including heightened awareness and responsiveness to stimuli.  Furthermore, when autistic people first began to describe their experiences, even before the autism spectrum had expanded to anything like its broad range today, sensory issues occupied a prominent part of the narrative.  Temple Grandin, in particular, was a pivotal early leader in emphasising the centrality of the sensory differences in many autistic people’s everyday experience (and Grandin continues to be a powerful voice on that issue today).

It is no coincidence, I think, that Frith notes in the podcast that people seeking autism diagnoses or identities today “are talking about their sensory issues” [emphasis added].  These are fundamentally internal experiences, so if people can articulate their internal sensory experiences – which not only requires language but also quite a bit of self-insight – it is much more likely that these experiences will be recognised by others.  Sure, there are behavioural reactions, but it’s easy to ignore those reactions when other things are more salient to outside observers, like a seeming-indifference to social contact or dangerous self-injurious behaviours.  That’s probably why so few of the early autism researchers noticed the sensory reactions, and of those who did, fewer grasped how important they might be.⁷ Thus, later-diagnosed people with more verbal fluency are inherently going to have an easier time pointing out and advocating for their sensory needs – but that in no way implies that the early-diagnosed people don’t have them!

Indeed, while direct comparison is very hard, I would not be surprised if the sensory experiences of the autistic people with the earliest diagnoses and highest support needs might sometimes be more atypical than those of later-diagnosed individuals.  For one thing, those with generally-higher support needs are more likely to have genetic syndromes that can directly affect peripheral sensory signalling and that might have more extensive effects on central processing of stimuli.  For another, insofar as particularly extensive sensory differences (e.g., difficulties integrating different percepts into one whole, diminished awareness of and hyporesponsivity to sounds) might disrupt language learning, we’d expect the people with the highest support needs to be more likely to have such experiences.

Thus, we can be pretty confident that unusual sensory experiences and reactions are common in early-diagnosed autistic people.  Once again, we are seeing that Frith’s proposal to subdivide autism rests on weak foundations.

Sensory protection and accommodations

However, Frith not only restricts sensory experiences to late-diagnosed individuals, but also questions the usefulness of accommodations and protections, such as ear defenders and dimmed lights.  She claims these adjustments have “very little scientific grounding” and expresses skepticism about their value.

First of all, I want to emphasise that the claim about scientific evidence is not quite true.  While the autism sensory literature has indeed tended to focus on how individuals process stimuli and neglect questions of greater applied relevance regarding sensory accommodations and sensory-inclusive design, there is still some evidence addressing it.  For example, Beth Pfeiffer and colleagues found that noise-attenuating headphones can decrease physiological arousal and improve community participation.

Second, Frith in the podcast says she “would really like to know what the people with the ear defenders and the silent spaces actually think about it.”  I don’t mean to come off as rude, but it’s not like we’ve never made our thoughts known.  There’s a reason so many of us seek out these adjustments and escape spaces.  Sometimes, blocking access to these adjustments – putting people in an overwhelming, distressing setting they have no control over and preventing them from escaping or putting any barrier between themselves and the chaos – will lead straight to a meltdown.  In other cases, it will simply cause preventable and pointless distress and discomfort.

Admittedly, there are some reasons to sometimes be concerned about people overprotecting themselves from sensory distress.  There is a great deal of concern in the hyperacusis community about people avoiding sound in ways that could result in oversensitisation and that could exacerbate anxiety.⁸ However, let me contextualise that concern.  One might, for example, be concerned about an autonomous adult who uses noise-cancelling headphones constantly without pause, who structures their life to as much as possible avoid leaving the home and exposing themselves to stimulation, and so on.  But this is obviously starkly different from a situation where a child, with far less autonomy, is subjected to the constant sensory stresses and overloads of a busy, chaotic school environment.  There are severe inherent limits to what one can do to protect oneself when one is in these chaotic settings, and I think the last thing we need to be doing is chipping away at what little protection people can create for themselves there.  Forcing people to ignore their needs and burn themselves out dealing with these overloads does not promote resilience – it is a huge and completely unnecessary waste.

In my own experience, leaving the mainstream school and doing distance education had huge benefits.  I’m not saying that would be the best option for everyone, but when I was no longer burning myself out trying to constantly cope with overwhelming stimulation at school (and that was even after I had finally been granted some minimal accommodations, like the option to flee overwhelming cafeterias and gymnasiums at school when they became too much), I actually found that my sensory coping ability was much improved.  Indeed, pretty much everything improved – my mental health, grades, etc.  Facing impossible and pointless tests every day did not improve my resilience.  I achieved resilience when my family and I found the right balance between the Scylla of overprotecting myself and the Charybdis of overloading myself.  This requires care and precision, not sweeping generalisations.

The benefits of unity

Finally, I believe there is one last issue with Frith’s ideas.  Even if we could develop a scientifically compelling proposal to subdivide autism itself, should we follow it?  We would need to carefully assess the benefits of any greater precision offered by the new subcategories – in terms of access to services, public understanding, and self-understanding – against the cost of breaking up the autism community.

At present, I do genuinely believe there are common themes and experiences that run through the broad autistic population, so we would need to consider the cost of some populations potentially losing access to concepts and ideas with broad explanatory value.  Moreover, surely there are common advocacy goals we ought to be able to agree and work collectively on, like promoting autism acceptance and improving inadequate disability benefits.  Moreover, neurodiversity advocacy emphasises how many of the political and social issues facing autistic people apply at least as much to those with the highest needs as to those with lower needs.  For example, Holly Sutherland is doing excellent work drawing attention to how problematic power dynamics and “double empathy”-type misunderstandings of autistic people’s behaviours can affect autistic people with high needs receiving community living supports.  If we break up autism, do we send the message that those with the highest needs are not deserving of such consideration?  It is not a logically inevitable consequence, but it seems a disturbingly realistic danger.

1. Baron-Cohen, Frith, and the other authors were essentially arguing that, on average, autistic people have performed more poorly on theory of mind tasks than non-autistic people.  This initially sounds like an authoritative and decisive scientific response, until you consider that Frith and her colleagues were replying to a paper by Travis LaCroix that, among other things, criticizes the tasks.  Thus, I felt like the response kind of misses the point…
   As the international popularity of hockey seems to have increased since the release of a certain Canadian TV show, I think I might try a hockey analogy to illustrate this.  Let’s say that one team scores a goal, but is accused of having too many players on the ice.  Now, that probably wouldn’t be malicious: having too many players on the ice is usually accidental.  But the question would be whether the rules of hockey were followed, not whether a goal was scored.  Frith and colleagues, in their response, were effectively focusing mostly on whether the puck went in the net, not whether it did so legitimately.
   Now, I should say that Baron-Cohen, Frith, and the other authors do briefly suggest that peer reviewers wouldn’t have allowed theory of mind papers to be published if the methodology wasn’t rigorous.  But on the other hand, LaCroix’s paper pointing out the issues had been peer-reviewed too…  And while science has ways of self-correcting, the implication that peer review offers a foolproof guarantee against methodological flaws or limitations certainly seems bold to me.  If we extend our hockey analogy, I guess the argument that peer reviewers would have prevented issues is kind of like saying that the referees would have caught any violation in the moment, so there’s no need to examine video replays.
   There are also comments in the Baron-Cohen et al. article about “sensationalist activism,” and I guess these remarks could be considered the academic equivalent of hockey chirping?
   Furthermore, Baron-Cohen, Frith, and the others point to evidence of relationships with social functioning and autistic traits as evidence of validity, but that’s kind of a circular argument, and LaCroix might not necessarily agree that a valid indicator of theory of mind would be related to autism…  Admittedly, I’m struggling to fit that bit into the same hockey analogy, but hopefully you get the idea.
   So again, I do agree with Frith and Baron-Cohen insofar as I don’t think all autism theory of mind research should be dismissed as necessarily pseudoscientific, nor do I deny that autistic people sometimes have difficulties with perspective-taking.  But realistically, I think it’s hard to deny that theory of mind research today is affected by methodological concerns (not least of which being the flaws of the Reading the Mind in the Eyes task) and conceptual questions (e.g., are autistic people’s difficulties to do with representing or inferring mental states?).  Thus, I wish the response from Baron-Cohen, Firth, and others had engaged more extensively with the core scientific issues.  I think we need to be very careful about invoking science and meta-analysis outcomes as a form of authority – “The science says this, which means I’m right.”  To me, science is more of an ideal, an approach that we strive to realise as best as we can, and part of that is the need to carefully justify the validity of the methods and empirical data on which we base our theories and paradigms. Of course their past empirical theory of mind research does discuss the tasks, but I lament the lack of an updated integrated reflection and discussion of how to continually improve the rigour of the science…

2. Of course, I am taking a broad view of what counts as evidence: rigorous and methodologically careful quantitative scientific work, qualitative interviews, experiential narratives, and more.  There are many kinds of evidence that all have their own uses, if they are considered critically and thoughtfully.

3. At least, I mean that they don’t require particularly strong technical skills or formal education – they may still require interpersonal skills in ways that can disadvantage autistic people!  Or they can impose sensory stresses, etc…

4. If anyone thinks the mainstream can’t be stressful for people meeting a 1990s type understanding of autism/Asperger’s, please read Clare Sainsbury’s brilliant book, Martian in the Playground.

5. The recent proposal to create a “profound autism” category seems to be effectively defined based on the presence of something like a so-called “severe” to “profound” level of intellectual disability, so in my view, it is effectively about specifying the presence of a co-occurring condition.  This seems especially clear in the new definition I heard about last year, which makes clear that the category will not include nonspeaking autistic people who can communicate complex ideas via typing – only those whose minimal language would imply the presence of an intellectual disability.
   The reason for describing it as an “autism thing” rather than an “intellectual disability thing” seems, as far as I can tell, to be representational politics and strategy rather than science.  Calling it an “autism thing” sets up a fight only with the neurodiversity movement, whereas focusing on the co-occurrence between autism and intellectual disability would also lead to conflict with self-advocates with intellectual disabilities.  (Many of the advocates for profound autism have views that are very much out of line with those of self-advocates with intellectual disability: for example, on integrated versus specialised/segregated living facilities for adults.)  Champions of the profound autism concept may believe a dispute with neurodiversity advocates alone would be easier for them to win, since neurodiversity advocates can be more easily denigrated as an out-of-touch elite.

6. It simply hasn’t been credible – the studies often make no adjustment for the fact that including people who have been diagnosed at early ages, but not the unknown ones who have not, tends to bias things downwards!

7. Also, I wonder if some of the early researchers and clinicians, because they were looking for cases of childhood schizophrenia, may have misinterpreted sensory differences as evidence of hallucinations.

8. On this issue, I recommend Rob Littwin’s chapter, Hyperacusis Management: A Patient’s Perspective, in the edited volume Hyperacusis and Disorders of Sound Intolerance: Clinical and Research Perspectives.