Understanding Autism through Predictive Processing
By: Julia Daley
Autism Spectrum Disorder (ASD) is a developmental disability that impacts perception, communication, and learning in profound ways.1 Well-known hallmarks of Autism include struggles with communication, restricted interests, and repetitive behaviors (including stimming), and hyper- (and hypo-) sensitivity to sensory stimuli. When their brains get overwhelmed, autistic people can experience meltdowns (sudden loss of control) and shutdowns (withdrawing inwards); recovery from these episodes can take a few hours in a non-stimulating space.
Autistic people require support to thrive in a neurotypical world, but how much support they need will vary depending on the level of their Autism diagnosis; Level 1 equates to low support needs; Level 2 has moderate support needs; and Level 3 requires substantial support. When there’s a chronic mismatch between an autistic person’s support needs and their daily life, there is a risk of developing autistic burnout (sometimes called autistic regression); this is a debilitating condition that is characterized by intense exhaustion, increased sensitivity to sensory stimuli, and can make previously-possible daily tasks suddenly impossible (Raymaker et al., 2020).
In short, autistic brains are quite different from neurotypical brains, which neuroscience is still working to understand. Predictive Processing may help with painting a fuller picture of the autistic brain.
1 For a child-friendly look at the many complex behaviors of Autism, I recommend watching this Sesame Street clip that introduces the autistic character Julia to the cast of the show. Her character’s autistic traits were thoughtfully developed in collaboration with the autistic Self-Advocacy Network (ASAN), and while Julia continues to appear on Sesame Street, the partnership with ASAN has been discontinued.
Predictive Processing and Learning
First, let’s fill in one of the gaps from the Kurzgesagt video anchoring this issue and take a brief look at how learning works via the Predictive Processing model of the brain. In short, your brain predicts what reality should be, based on a large database of prior experiences, and then checks in with incoming sensory information to ensure that the prediction is accurate. Any discrepancy between the prediction and reality produces an error signal. Your brain analyses the error signal for relevance, and if it’s deemed applicable, updates its internal models to generate better future predictions.
How much “weight,” or importance, the brain places on these error signals is known as precision. Depending on the situation, your brain adjusts its precision dial; a lower precision level means you are more tolerant of ambiguity, whereas a higher precision level means you are more exacting in interpreting sensory information. Subjectively, errors in our predictive models create a feeling of surprise–how much surprise we feel will depend on the size of the error (the bigger the error signal, the more surprised you may feel).
Let’s say you’ve made a new acquaintance, perhaps at work. You both have dogs and are eager to share pictures of your beloved pooches. Your new friend excitedly flips through their phone for the perfect photo of their dog, while you wait in anticipation. As the friend reaches out their arm and shows their phone screen to you, your brain has already pre-emptively generated possible mental images of what this dog could look like. Your body has prepared a social response, based on past experiences, of how to appropriately react to this sharing-of-cute-dog-photos scenario–smile, lean forward for a closer look, compliment the dog’s cuteness. You’re already smiling and leaning in before you pause and go “huh” as the reality of what your eyes are actually perceiving is finally processed by your brain—a loud error signal has you hitting the brakes, figuratively.
“Isn’t that a wolf?” you ask, perplexed. “No, he’s a Czechoslovakian Wolfdog! Isn’t he so cute?” your new friend doesn’t miss a beat, as your surprised reaction is one they’ve encountered many-a-time before. After another few moments, you resume your what-a-cute-dog script and compliment the wolfdog photo before you. Your mental model of “what cute dogs look like” has adapted to accommodate this previously-unknown breed of dog. Error signal, resolved! Learning achieved. The next time you see your friend’s–or anyone else’s–wolfdog, you’ll be ready.2
2 This fictional scenario was influenced by Julia’s previous encounters with Czechoslovakian Wolfdogs in Europe. These dogs are mind-boggingly wolf-like in appearance and make people everywhere stop and stare and update their predictive models of the world.
Predictive Processing and Autism
Part of the struggle to fully understand Autism is that its key features—sensory overwhelm (and underwhelm), intolerance of the unexpected, and difficulty interpreting the meanings and intentions of the people around them—are all quite different from each other, neurologically speaking. It’s been hard for researchers to figure out how these aspects of Autism are connected to each other in the brain. Predictive Processing might hold the key to finally improving our understanding of ASD… it’s just the how that’s still very much up for debate among neuroscientists.
Quite a few possible Predictive Processing-based frameworks have now been proposed to help researchers better define the root neurological causes of Autism. All of them have the benefit of helping us study the autistic brain in a more precise way, and yet, as of this writing, no particular model has become the clear “winner.” This is all very new science, and it will likely be many years before we have a good understanding of all the quirks of the autistic brain. Still, let’s take a brief look at a few of the main theories that researchers are currently developing, each one building on its predecessor.
Weak Priors
In 2012, Pellicano and Burr were the first researchers to propose that Autism is best understood within the Predictive Processing Framework. In their paper, they suggest that at the core of all aspects of ASD might be priors—or the things we’ve learned in the past that influence our predictive models. They argue that autistic brains either struggle to form these priors properly, or else are not heavily influenced by them. This could then lead autistic people to rely heavily on sensory information to make sense of the world—in other words, whereas neurotypical people see the world as it should be, autistic people instead see the world as it actually is. 3
In short, Pellicano and Burr theorized that the root difference at the core of autistic brains has to do with the atypical development of priors, which they refer to as either “attenuated priors” or “hypo-priors,” and thus a heavy reliance on real-time sensory processing. For instance, autistic people generally aren’t fooled by visual illusions; this may be due to the fact that the success of these illusions often depends on tricking the predictive brain, which autistic people might not rely on. autistic people are regularly overwhelmed by the sensory stimuli around them, sometimes to the point of perceiving loud sounds or bright lights as painful. Neurotypical brains are better able to tune-out the sensory noise and focus just on the input that’s most salient (e.g. the information that’s most likely to confirm or disprove their predictions). If autistic brains struggle to accurately predict how the world should be, they would have to spend more time constructing reality via their senses, and thus try to pay attention to everything around them.
In their initial proposal, Pellicano & Burr (2012) focused just on the perceptual differences of autistic brains. They wondered if this sensory overload could be the impetus behind other autistic challenges, like navigating a conversation or craving routine.
3 It’s still unclear if this processing difference extends to all facets of the autistic brain or only parts of it. In addition, there could well be a lot of individual variation to this effect. With time and more research, we’ll hopefully get clearer answers!
High Precision
The next big theory of Autism comes from Van de Cruys et al. in 2014 and builds off of Pellicano & Burr (2012). In their view, the focus on priors alone could not fully account for all the quirks of the autistic brain. They had a different idea: what if autistic brains’ “precision dials” are permanently set on high? Perhaps autistic brains are uniquely inflexible, and thus autistic people can’t adjust their precision levels to different situations.
High precision means lots of error signals, so the autistic brain might be spending a lot of time sifting through the salience of these internal alarm bells to determine where its predictive models need adjusting. This always-highly-precise mode of perceiving the world could also explain a lot about Autism. It implies that autistic people can learn from prior experiences, but that they may struggle to generalize their predictive models to similar situations that may differ from their original priors. For a classic example (Silberman, 2016), an autistic child first introduced to chocolate in square shapes refused to eat round chocolates; this might be a case of the autistic brain unable to flexibly adapt their prior (chocolate is square) with new information (chocolate is also round). Perhaps this always-high level of precision is the root cause for autistic people’s often rigid behavioral patterns. The more reality conforms to their predictions, the fewer alarm bells they have ringing in their heads.
Aberrant Precision
Another possibility that Van de Cruys and his co-authors consider is that autistic brains might struggle instead with interpreting error signals, and thus are less able to self-regulate their precision dials. Rather than an always-precise way of perceiving the world, autistic brains don’t learn to recognize when different situations require more flexibility or more precision. This latter idea is more of an afterthought, as they focus primarily on the effects of high precision, but it gets picked up and expanded upon by Lawson et al. in 2014.
They argue that there might be a mismatch between autistic brains’ predictive models and the sensory input, and that mismatch means their precision dials are often on the wrong setting—sometimes too precise, sometimes too flexible. In uncertain contexts, where there’s lots of novelty, autistic people might have a hard time because they can’t adequately draw upon their prior experiences to determine how much weight to give to the sensory signals they are receiving. If they’re highly precise, they might generate lots of error signals, many of which are not relevant or necessary; if they’re less precise, however, autistic people might miss an important cue that their predictions are inaccurate.
In practice, this could mean that autistic brains learn to operate under the assumption that life is unpredictable. This means, somewhat counterintuitively, that autistic people are not as surprised when something surprising happens, but they’re conversely surprised when something ordinary happens. Early research from Lawson et al. in 2017, where they measured the pupils of autistic and neurotypical people to measure their surprise during different tasks (our pupils grow bigger when we are surprised, and the size change correlates to how surprised we are). As they put it:
The surprise experienced on finding a pineapple in your sock drawer depends on the strength of your prior expectation to see only socks. The results of this study imply that adults with autism show a tendency to over-estimate the volatility of the sensory environment… In other words, adults with autism may be mildly surprised by both the pineapple and the socks. (Lawson et al., 2017, p. 11)
This Aberrant Precision theory, then, is another possible explanation for why autistic people crave routine—they know what to expect from their daily rituals, and thus how precise their brains should be. Through experience, neurotypical brains will learn when attention to detail and high accuracy is important, and when it’s OK to move through life on auto-pilot; they learn when certain situations are likely to be more volatile and surprising, and which situations are mundane and ordinary. With this theory, it may be that in autistic brains, this kind of meta-learning is weakened.
Active Inference
While the preceding theories focused on how autistic brains perceive the world, this next one instead tries to explain autistic behavior—why do autistic people do what they do? For Palmer et al. (2017), the secret to understanding Autism lies in understanding the balance between perception and action.
Let’s say you’re thirsty, feeling that tell-tale dryness in your throat. There’s a glass of water beside you; all you have to do is reach out your arm, grasp it, and bring it to your mouth. Your body predicts where your hand will be and how the glass will feel within it, and then takes action to fulfill this vision. You drink the water and perceive cool relief as it moves down your throat. It’ll take a bit of time before that water goes where it needs to within your body, but you no longer perceive thirstiness as your body has already predicted its future state of re-hydration.
That balance of perception (feeling thirsty) and action (grabbing the glass) of drinking water could well be impaired in autistic brains. Perhaps they struggle to perceive their thirst, due to poor interoception (the ability to recognize your internal state of being). Maybe there is a gap between reality (where their arm currently is) and the predicted future (where their arm needs to go to grab the glass), and due to the mismatch, the autistic person ends up knocking the glass over. Perhaps the autistic brain is so focused on interpreting real-time sensory input that it forgets to rely on its prior input, like the location of body parts. We know that autistic people generally struggle with interoception (Seth & Friston, 2016), and they are also a lot clumsier than neurotypical people. This theory might help us understand these two perplexing facets of Autism.
This theory, then, tries to connect the previous theories’ understanding of autistic perception to how autistic people interact with the world. If autistic brains are in a constant state of uncertainty—unsure of how they are feeling or where they exist in relation to the world around them—that could explain why they engage in a lot of repetitive behaviors. For autistic people, it’s better to focus on the known—like their favorite stim—than drown in the unknown of the world around them.
No Clear Answers… Yet
Research connecting Autism to Predictive Processing is still in its early days, and there’s a lot we don’t know yet. Many studies so far have tried to get quantitative results to validate one theory or another, but most have had mixed results. Sometimes their autistic participants demonstrate differences that match Predictive Processing hypotheses, but other times there’s no major difference between the autistic and neurotypical participants. Plus, designing experiments that accurately measure priors or precision is, in itself, a challenge as Predictive Processing itself is still a relatively new part of neuroscience.
It turns out that studying autistic people can be challenging! How questions are designed (closed vs. open-ended), or how many trials are conducted, or how long participants’ outcomes are tracked can all have huge impacts on the results. Some of this can be ameliorated by consulting with autistic researchers. The neuroscientific community is slowly starting to organize a series of best practices, as well, especially for studies with autistic participants (Angeletos Chrysaitis & Seriès, 2023). There’s also the ongoing problem of representation in Autism studies, and the lack of female autistic participants can make it hard to extrapolate a study’s results to the broader autistic population (Rippon, 2023).
Still, given time, this is a very promising area of research that seems sure to help us better understand Autism and its neurological underpinnings.
Ideas for Teachers
While this research is still very new, the ideas being discussed might still have some relevance to what teachers do in the classroom. All of the competing theories do seem to agree that autistic brains are more heavily focused on interpreting real-time sensory information. If most brains are living simultaneously in the past and the future, autistic brains seem to be stuck in the present. How then, should we manage our classrooms when our students’ brains are, figuratively, operating in different timezones?
As a teacher, my first thought goes to wait-time, or the time you spend waiting for a response from students. In my experience, my autistic students have significantly longer wait-times than my other students. With a lot of patience, if I wait quietly for long enough, I can get great responses from my autistic students. Whereas other students might need a 5 or 10 second wait-time, it seems like 30 to 60 seconds is more appropriate for autistic students. The trick is, of course, being quiet for the necessary amount of time for my students to process what I’ve asked and formulate their responses. If I fill the gap too soon—perhaps by repeating myself or suggesting answers—I reset the processing timer for my autistic students.
Classroom management seems to be another area that could make a difference for autistic students. Sometimes classes are noisy, but there’s a difference between managed mayhem and uncontrolled havoc. Keeping students on-task and preventing disruptions are important for good learning for any brain, neurotypical or neurodivergent. Consistency will be key—whatever behavior norms you set for your class need to be consistently maintained. We should try to create clear expectations for how students should ask questions, interact with their peers, and transition between tasks. Autistic brains crave routine—which makes intuitive sense if they’re trying to navigate a chaotic world full of conflicting sensory information; the more routines we can build into our classroom, the better able they will be to focus on the content we are trying to teach.
References
Angeletos Chrysaitis, N. & Seriès, P. (2023). 10 years of Bayesian theories of autism: A comprehensive review. Neuroscience & Biobehavioral Reviews, 145. https://doi.org/10.1016/j.neubiorev.2022.105022
- Lawson, R., Geraint, R., & Friston, K. (2014). An aberrant precision account of autism. Frontiers in Human Neuroscience, 8. https://doi.org/10.3389/fnhum.2014.00302
- Lawson, R., Mathys, C., & Rees, G. (2017). Adults with autism overestimate the volatility of the sensory environment. Nat. Neurosci, 20(9): 1293-1299. https://doi.org/10.1038/nn.4615
- Palmer, C. J., Lawson, R. P., & Hohwy, J. (2017). Bayesian approaches to autism: Towards volatility, action, and behavior. Psychological Bulletin, 143(5), 521-542. https://doi.org/10.1037/bul0000097
Pellicano, E., & Burr, D. (2012). When the world becomes too real: a Bayesian explanation of autistic perception. Trends in Cognitive Sciences. https://doi.org/10.1016/j.tics.2012.08.009
Raymaker, D. M., Teo, A. R., Steckler, N., Lentz, B., Scharer, M., Delos Santos, A., Kapp, S., Hunter, M., Joyce, A., & Nicolaidis, C. (2020) “Having all of your internal resources exhausted beyond measure and being left with no clean-up crew”: Defining autistic burnout. Autism in Adulthood, 2(2). https://doi.org/10.1089/aut.2019.0079
Rippon, G. (2023). Off the spectrum: Why the science of Autism has failed women and girls. Seal Press.
Seth, A. K., & Friston, K. J. (2016). Active interoceptive inference and the emotional brain. Phil. Trans. R. Soc. B 371: 20160007. http://dx.doi.org/10.1098/rstb.2016.0007
- Silberman, S. (2015). Neurotribes: The legacy of autism and the future of neurodiversity. Avery.
- Van de Cruys, S., Evers, K., Van der Hallen, R., Van Eylen, L., Boets, B., de-Wit, L., & Wagemans, J. (2014). Precise minds in uncertain worlds: Predictive coding in autism. Psychological Review, 121(4), 649-75. https://doi.org/10.1037/a0037665
Julia Daley is a senior lecturer and Assessment Coordinator at Hiroshima Bunkyo University, and she received her Masters in TESOL from Northern Arizona University. She has taught English writing and conversation in many classrooms in the US and Japan. Her brain’s knack for real-time detail makes her excellent at spotting all sorts of wildlife while hiking.