The Promise of Predictive Processing

I managed to spend a decade researching schizophrenia without seriously encountering the predictive processing literature. It was not until I put forward, on X, the controversial idea that we may one day see autism and schizophrenia recombined diagnostically that I began to look into it.

I was understandably chastised by X’s autism community, which includes many fierce advocates of autism as an identity rather than a diagnosis, as well as several self-diagnosed individuals.

My theory is not strongly supported by the evidence, and I am aware that the decision to split autism and schizophrenia in DSM-III carries a great deal of cultural baggage among those operating in the area of neurodiversity—not to be confused with neurodivergence, as it often is.

My hunch does not come from a desire to be edgy or to achieve fame or notoriety on X. It is instead the natural landing point following a decade of schizophrenia research that has inevitably required engagement with the autism literature. It is also informed by my clinical and personal experiences.

Many individuals with schizophrenia whom I have supported as a healthcare assistant or interviewed as a researcher have been referred, at some stage following their schizophrenia diagnosis, for autism assessments. This also happened to me. I was assessed for autism in 2023 and felt not to meet the criteria: many years after my schizophrenia diagnosis in 2015.

These observations lead me to question the idea that autism and schizophrenia are always sufficiently distinct in clinical presentation that one can confidently diagnose one while excluding the other. Many of my research participants appeared to speak in ways I had heuristically associated more closely with autism than with the schizophrenia and formal thought disorder literature. Many were quite concrete, had restricted or repetitive interests, and perseverated on topics that I sensed we had exhausted.

While processing the backlash to my post on X, I began looking into some of the criticisms coming my way. One man called me an idiot because autism and schizophrenia are ‘literally the complete opposite’.

I took this to mean that he was gesturing towards the diametric opposition literature, with which I am familiar. In some theoretical accounts, autism and schizophrenia are not simply situated at different points along a continuum. They appear more closely to resemble mirror images of one another.

My own suspicion is that they may nevertheless represent different expressions of some shared underlying processes, with developmental and life-course factors helping to mediate the direction in which predictive processing differences manifest. Concrete, bottom-up processing, with greater weighting of sensory information, is often associated with autism. More abstract, top-down processing, involving greater weighting of prior beliefs, is often invoked in accounts of schizophrenia.

This is obviously a gross oversimplification of a complex and contested literature, but that is about as far as my expertise on this particular issue extends.

I do not expect my suspicion—that decoupling autism and schizophrenia in DSM-III may have been premature or mistaken—to be empirically supported any time soon, if at all. Nor do I expect contemporary autism communities to suddenly become comfortable with the idea.

Schizophrenia remains heavily stigmatised and sits firmly within the category of severe mental illness. Autism is increasingly situated within a framework of neurological difference, with less overt stigma in at least some communities and contexts. Bringing these conditions together, even if it were eventually judged academically sensible, would therefore be ideologically charged.

But in looking into the predictive processing literature, I began to develop a much deeper appreciation for its explanatory potential within what appears to be a significant current in psychiatric thinking.

From my own experience, particularly of ideas of reference, I have noticed that my symptoms are much more severe when I encounter the speech of strangers or people with whom I have a complex and largely negative interpersonal history.

In interactions with my daughter, for example, I do not generally experience ideas of reference. There is no historical data to suggest that she is exploiting pragmatic communication to convey some unspoken insult. She is a six-year-old who says six-year-old things because she sees the world in six-year-old ways.

My experiences with strangers are different. Other people have historically been more hostile, more unpredictable, and more difficult for me to interpret. In those contexts, ambiguity carries a different weight. A predictive processing account therefore offers some explanatory potential for why my ideas of reference fluctuate in severity depending on environmental and interactional context.

The prior is not the same.

That, to me, is where the framework becomes particularly interesting. My response to an ambiguous utterance may depend not only on the words themselves, but on what previous experience has taught me to expect from the person speaking, from people like them, or from the interactional context more broadly. Under conditions of uncertainty, those expectations may shape what I perceive an utterance to mean.

For this reason, I think predictive processing accounts of schizophrenia are well worth keeping an eye on. I am particularly interested to see the area develop beyond broad accounts of psychosis as a state and turn, with greater specificity, towards phenomena such as auditory verbal hallucinations, delusions, and ideas of reference.

For now, I remain uncertain about whether autism and schizophrenia will ever be brought closer together diagnostically. I may simply be wrong.

But I am increasingly persuaded that understanding how people weight sensory evidence, prior expectations, and uncertainty may tell us something important about why the same world can become so radically different from one mind to another.