For a long time, we’ve used the metaphor of a spectrum to describe autism.  This is a metaphor that has worked fairly well for us in many ways.  It alerted us to the fact that instead of being a discrete category, there’s a lot of heterogeneity and variability within autism.

Indeed, I would argue that autism lacks biological reality – it’s a social construct, a category that we made up, and its boundaries have shifted over space and time.  I’m not saying that autism doesn’t exist – we have diagnostic instruments that can measure it fairly well (albeit not wholly objectively), so clearly it is a measurable thing.  It’s also quite useful as far as eligibility for supports and services is concerned, and it’s a positive identity for many people, including myself.  My point is that the thing we’re measuring is a thing we defined ourselves, and not necessarily based on any underlying neurobiological or etiological coherence.  As far as etiology is concerned, there are many, many different genetic variants that are all associated with autism, while lots of other people have polygenetic autisms that are so etiologically complex as to be, for all intents and purposes, idiopathic.  The environment can influence autism – I suspect much of this environmental influence is prenatal – but there again it’s unusual for there to be a single environmental factor we can unambiguously point to as a primary cause.  Meanwhile, as far as heterogeneity in neurobiology is concerned, studies aiming to find reproducible patterns in autistic people’s brains across large samples usually don’t succeed (e.g., King et al., 2019).

In any case, the spectrum metaphor helped to alert us to the fact that autism wasn’t a single, simple categorical entity and instead helped us see the need to take account of dimensional heterogeneity in autism.  Unfortunately, the spectrum metaphor implies that there’s only one dimension in autism: a dimension going from greater to lesser “severity” of autism features.  The spectrum metaphor reduces the heterogeneity of autism to a linear continuum from the “highest functioning” autistic person to the “lowest functioning” autistic person. This idea of a simple linear gradient of severity is an idea which deeply bothers a lot of self-advocates, who point out that they might struggle more or less at different times, in different domains, and in different contexts.

To be fair, this idea of a linear spectrum of severity was not the original intent behind the idea of an “autism continuum,” the precursor to the autism spectrum.  When people like Lorna Wing and Judith Gould first introduced the notion of an autism continuum, they were imagining multiple dimensions, including different dimension of autism features as well as other things like cognitive ability.  Wing (1988, in Diagnosis and Assessment in Autism, pp. 91-110) even mentioned anxiety and depression, albeit in passing.  (There wasn’t really any research on autistic people’s mental health back in the 1980s.)  But unfortunately, people ignored this multidimensional aspect to the continuum idea and instead started talking about a continuum simply from greater to lesser severity.

We need an alternative metaphor that will still recognize the existence dimensional heterogeneity in autism, without losing sight of the reality that there are multiple dimensions that matter. For a while I was speaking of autism as a “multidimensional mush,” which was hardly the most attractive metaphor ever, but a while ago I had the pleasure of reading Sue Fletcher-Watson and Francesca Happé’s update of the book Autism.

[Update – 2021-08-28: I was just reminded that Fletcher-Watson and Happé did not invent this idea, and that Caroline Hearst had already designed a physical model of an autism constellation and presented it at events. Moreover I’ve been informed that the idea was already circulating through word of mouth before Hearst’s model. I’m told that Larry Arnold was an early proponent of the idea, and that Damian Milton and the late Dinah Murray were discussing similar concepts.]

They present the idea of autism as a multidimensional “constellation.”  Instead of a thinking of a single dimension of autism severity, we could be free to imagine a multidimensional space.  We could pick any three dimensions we want – for example, I simulated data labelled with the dimensions of cognitive ability, anxiety, and discomfort to loudness in the figure above.¹ That’s a constellation based on those three dimensions.  We could even imagine higher multidimensional spaces with constellations of four, five, or more dimensions – however many we want.  We could pick dimensions of the characteristics we’ve chosen to associate with autism, or we could pick any other dimensions along which individuals might vary.

Steven Kapp also mentioned something similar to this in a 2018 article – he spoke of autism as a “cloudy constellation” where individuals might have uneven skills and abilities across different dimensions.

There is much that is attractive about this metaphor.  Besides allowing us to talk about heterogeneity in more than one dimension, it also highlights the important role of unique individuals in our constellation – each single “star” is a person.  It also allows us to speak of changes over time, for just as individuals’ states and traits change over time and across contexts, the celestial disc is not fixed.  Planets wander rapidly across the skies, and even stars slowly drift in their positions.

In contrast, I can’t see any particular advantage to the idea of a “spectrum” other than habit and familiarity.  Perhaps it is time we moved on and embraced the multidimensional constellation metaphor as a way of talking about heterogeneity in autism?

1. Of course, I don’t mean to imply that we can necessarily measure these constructs perfectly.  Exhibit 1: there are serious discrepancies between different measures of cognitive ability in autism.  Exhibit 2: measures of anxiety designed for typical development don’t necessarily work very well in autism, partly because autistic people can have forms of anxiety not generally found in neurotypicals, and partly because it’s sometimes hard to tell whether we ought to treat something as a feature of anxiety or autism.  Exhibit 3: discomfort to loudness is a complete and total mess.  The questionnaires currently used to index sensory features in autism don’t even distinguish discomfort to loudness – hyperacusis – from other forms of hypersensitivity like misophonia (the frustration and anxiety caused by exposure to soft but aggravating sounds).  It’s also unclear whether questionnaire reports of hypersensitivity map well onto loudness discomfort levels, or the intensity/loudness of sound people have to be exposed to before they say they are uncomfortable.  When we’re trying to measure these sensory features, we’re trying to measure an internal experience, and that’s not easy.  Fortunately, people like Zachary Williams at Vanderbilt are doing important work trying to tackle problems of measuring these sensory features and experiences.  Hopefully we’ll continue to improve our measures so we can make our image of the autistic constellation more accurate.