Predictive Processing in Autism: How Expectations Shape Perception, Uncertainty & Surprise

You walk into a familiar room and immediately notice that something has changed. Someone says, “We’ll see how it goes,” and the lack of a clear plan feels much more demanding than it seems to feel for everyone else. A sound you expected is manageable, while the same sound arriving unexpectedly makes you jump.

One way researchers are trying to understand experiences like these is through predictive processing: the idea that perception involves combining incoming information with expectations built from previous experience. Predictive-processing theories of autism ask whether autistic brains may sometimes learn, weight or update expectations differently — but the evidence is much more nuanced than the popular claim that “autistic people cannot predict.”

🧠 expectations helping interpret what happens next
👀 incoming information confirming or contradicting those expectations
⚡ unexpected events creating a prediction error
🔄 expectations changing as new information arrives
🎧 sensory input sometimes remaining unusually important
❓ uncertainty making prediction harder
📅 routines making parts of the environment easier to anticipate

🧠 What Is Predictive Processing?

Your brain does not simply wait for information to arrive and then analyse everything from scratch. It uses previous experience and current context to anticipate what is likely to happen next.

If someone reaches toward a light switch, you expect the light to change. If you hear the beginning of a familiar sentence, you already have ideas about how it might continue. When you walk downstairs, your body predicts roughly where the next step will be before your foot reaches it.

Predictive-processing theories describe perception as an ongoing interaction between:

🔮 Prediction — what the brain expects
👀 Sensory evidence — what actually arrives
⚡ Prediction error — the difference between the two
🔄 Updating — changing the prediction when the difference matters

This is a broad framework in cognitive neuroscience, not an autism-specific theory. Autism researchers have become interested in whether differences somewhere in this cycle could help explain certain sensory, learning or uncertainty-related experiences.

🍎 Your Brain Rarely Starts From Zero

Imagine seeing a partly hidden object on your kitchen counter.

You can only see part of its shape, but context helps your brain interpret it. The counter, colour, shape and your previous experiences all contribute to what you perceive.

The same thing happens with language. If someone says:

“Can you pass me the salt and…”

you probably already expect pepper before the final word arrives.

That expectation speeds up processing because your brain does not need to treat every possible word as equally likely.

Prediction therefore does not mean consciously guessing the future. Much of it happens automatically and continuously.

⚡ What Is a Prediction Error?

A prediction error occurs when reality differs from what was expected.

You expect the lift doors to open.

They do not.

You expect your colleague to be at the meeting.

Someone else appears.

You expect the supermarket aisle to contain bread.

The entire shop has been rearranged.

The brain now has information that its existing model did not fully predict. It needs to decide whether the mismatch matters and whether the internal expectation should change.

⚡ Prediction: bread is in aisle four
👀 Reality: aisle four now contains cleaning products
🔄 Update: bread must be somewhere else

Prediction errors are therefore not mistakes in the ordinary sense. They are information.

They tell the brain:

something about the current model may need updating.

⚖️ The Brain Also Needs to Decide What to Trust

Predictive processing becomes more interesting when expectation and sensory input disagree.

Suppose you expect someone to say your name but hear something ambiguous.

Should your brain trust:

🧠 the expectation?

or:

👂 the incoming sound?

Predictive-processing models describe this partly in terms of precision. In simplified language, the brain needs to estimate how trustworthy different sources of information are.

A strong, reliable expectation may influence perception more.

Clear sensory evidence may override the expectation.

The balance constantly changes depending on context.

This idea is central to several predictive-processing theories of autism, but researchers disagree about exactly where any autistic difference might occur.

♾️ How Did Predictive Processing Become an Autism Theory?

Researchers noticed that several autistic experiences might, in principle, relate to prediction.

These included sensory differences, responses to uncertainty, preference for predictability, repetitive behaviours and difficulties when environments change unexpectedly.

Different theories then proposed somewhat different mechanisms. Some suggested autistic perception might be influenced less strongly by previous expectations. Others proposed that sensory evidence or prediction errors might sometimes receive relatively greater weight. Another influential proposal suggested that the issue may concern how flexibly the brain estimates which prediction errors are important.

These ideas are intellectually attractive because they could potentially connect several apparently separate autistic experiences.

But attractive theories are not automatically correct.

And this is where the research has become much more interesting.

🚫 Autism Does Not Mean “Weak Predictions”

One of the earliest popular versions of the theory was sometimes translated into:

“Autistic people have weak priors.”

A prior is essentially an expectation based on what came before.

Taken literally, the weak-prior idea suggests that previous experience has less influence over autistic perception, leaving incoming sensory information relatively more influential.

But empirical research has not supported that as a universal explanation.

In a 2023 fMRI study, autistic and non-autistic adults both learned predictions and both had their perceptions influenced by those expectations. The researchers found differences in some neural signals, but behaviourally the autistic participants’ predictive abilities were intact.

Another 2023 study using statistical learning likewise found that autistic adults successfully learned predictive regularities, with no evidence for a broad impairment in predictive processing.

So:

autistic brains clearly form and use expectations.

🔄 The Difference May Sometimes Be in Updating

A more interesting possibility is that differences emerge not when expectations are first created, but when circumstances change.

A study of autistic adults learning associations under uncertainty found that participants successfully learned the relationship between cues and outcomes. However, they updated their existing expectations more slowly when conditions changed.

A 2025 study adds another layer. Autistic participants could use previous information, but the dynamics of how their expectations changed across repeated trials differed from the comparison group. Over time, their behaviour became more similar, leading the researchers to propose an atypical iterative updating account rather than a general inability to form priors.

That distinction is important.

The emerging picture is less:

“Autistic people cannot predict.”

and more:

“Under some conditions, predictions may be updated differently.”

📊 One Study Can Show a Difference and Another Can Show None

This is one reason predictive-processing research can seem contradictory.

Different experiments test very different kinds of prediction.

Researchers may study:

🎵 expected sounds
👀 visual patterns
⏳ timing
🧩 statistical sequences
🗣️ language
🤝 social expectations
🏆 rewards
🎯 learned cue–outcome relationships

Someone can perform typically in one domain and differently in another.

The 2021 systematic review of 47 autism prediction studies found evidence for differences in some forms of predictive learning and low-level processing, but results were inconsistent across tasks. The review did not support the idea of one global autism-wide prediction deficit.

That has become even clearer as newer studies have accumulated.

🎵 Prediction Can Be Intact in Complex Domains Too

A 2024 study examined prediction in both music and language. It found that performance depended strongly on individual musical, linguistic and cognitive abilities rather than showing one general autistic failure to predict complex sequences.

This matters because predictive-processing theories are sometimes stretched too far.

If an autistic person experiences difficulty in a social situation, it is tempting to explain it through prediction.

If they experience sensory overload, prediction.

If plans change, prediction.

If communication is difficult, prediction.

At that point, the theory risks becoming impossible to disprove.

A useful theory needs to identify when and how predictive processing differs, not merely attach the word “prediction” to every autistic experience.

🎧 Could Predictive Processing Help Explain Sensory Differences?

This is one of the most compelling applications of the theory — but it remains a hypothesis rather than a settled explanation.

Imagine hearing the hum of an air-conditioning unit.

If the brain expects the sound and predicts it accurately, repeated input may become less informative. It can fade into the background.

If incoming sensory information remains comparatively important or prediction errors are not reduced as strongly, the sound could remain more noticeable.

That provides a possible way of thinking about why some repeated sensory information may remain difficult to ignore.

But autistic sensory processing is highly heterogeneous, and current evidence does not justify saying:

“Sensory sensitivity happens because autistic brains have prediction errors.”

That would turn a developing theoretical framework into a proven biological explanation.

For the established everyday picture, see Sensory Processing in Neurodivergent Adults.

🔊 Expected and Unexpected Sensations Can Feel Different

Prediction may nevertheless help explain an important everyday observation: control and predictability can change how sensory input feels.

Consider loud music.

You choose the music, know when it starts and can stop it.

Now compare that with someone unexpectedly dropping a metal tray behind you.

Both may be loud.

But the second event creates much more surprise.

🎵 chosen: expected, controllable, meaningful
💥 unexpected: sudden, uncontrolled, unpredictable

This does not prove a specific autism predictive-processing mechanism. Human brains generally respond differently to expected and unexpected events.

But for autistic people who already experience strong sensory responses, predictability may meaningfully change the total demand.

That is why sensory processing and predictive-processing theory are related without being the same thing.

📅 Why Routines Can Reduce Processing Demand

Predictive-processing theories also offer an interesting perspective on routines.

A routine makes part of the future easier to anticipate.

Morning happens in roughly the same order.

The route to work is familiar.

Objects remain in predictable locations.

The same food tastes approximately the same.

Fewer things need to be newly interpreted or decided.

That can reduce several forms of demand at once:

❓ uncertainty
🧠 decision-making
🔄 switching
🎧 unexpected sensory input
📋 planning

This does not mean autistic routines exist because of predictive processing.

Routines can have many functions, including enjoyment, efficiency, sensory regulation and reducing executive load.

Predictive processing provides one possible framework for understanding why predictability itself may sometimes be useful.

🔄 What Happens When the Expected Pattern Changes?

Imagine that your usual train leaves from platform three.

You arrive and discover:

PLATFORM CHANGED TO 11

Many people find this mildly annoying.

For someone already carrying sensory, executive or social load, the change can create a much larger cascade.

The brain now needs to:

⚡ register the mismatch
🔄 abandon the old expectation
🧭 understand the new situation
🎯 create a new action
🚶 navigate somewhere unfamiliar
⏳ recalculate timing

This is where predictive processing overlaps conceptually with Cognitive Flexibility in Autism, ADHD & AuDHD.

The prediction changed.

But cognitive flexibility determines how easily behaviour can change with it.

These are related ideas, not identical mechanisms.

❓ Predictive Processing and Uncertainty

Prediction becomes harder when the environment itself is inconsistent.

Suppose someone behaves warmly one day and coldly the next.

There is no stable pattern.

Your brain cannot easily decide:

“What should I expect from this person?”

The uncertainty remains active because each new interaction may force another update.

Research into autism has found strong relationships between intolerance of uncertainty and anxiety, but intolerance of uncertainty is a psychological construct in its own right. It should not be presented as proof of predictive-processing differences.

Still, the concepts fit together intuitively.

If a situation is difficult to predict, it remains cognitively unresolved.

Read Intolerance of Uncertainty in ADHD & Autism for the practical anxiety side of this pattern. (sensoryoverload.info)

🧠 Predictability Is Not the Same as Certainty

This distinction is useful.

Predictability means having enough structure to make a reasonable expectation.

Certainty means knowing exactly what will happen.

Very little in life offers certainty.

You may know:

“The meeting lasts around an hour and we will discuss these three topics.”

That is predictable enough to prepare for.

You do not need to know every sentence people will say.

This can be a useful practical goal for autistic support:

increase usable predictability without pretending uncertainty can be eliminated.

💬 Social Situations Are Prediction-Heavy

Human interaction is extraordinarily dynamic.

During conversation you are constantly interpreting:

🗣️ language
🎵 tone
👀 expression
🤝 context
⏳ timing
💬 previous interactions
🧠 what the person probably means

Then the person responds and you update again.

Social prediction is therefore difficult for everyone because other human beings are noisy, context-dependent systems.

Some predictive-processing theories have suggested that differences in building or updating social expectations could contribute to autistic social experience. Neuroimaging research has found differences in neural processing of certain prediction errors, including social domains, but the evidence is nowhere near strong enough to reduce autistic communication differences to one predictive mechanism.

Social communication also involves language, attention, processing speed, sensory demands and differences between communicative styles.

🤝 The Double Empathy Problem Still Matters

Predictive-processing explanations can become overly individualistic if we are not careful.

Imagine an autistic person failing to predict what a non-autistic person means.

That could be interpreted as:

“The autistic person’s social model is inaccurate.”

But the non-autistic person may also fail to predict what the autistic person means.

That is why The Double Empathy Problem remains important.

Human prediction is shaped by experience.

People become better at interpreting communication styles they encounter repeatedly.

A predictive-processing framework should therefore not quietly return us to:

autistic person = defective predictor, non-autistic person = correct model of social reality.

Social reality is created between people.

👀 Detail-Focused Perception: An Interesting but Unsettled Connection

Predictive-processing theories have sometimes been linked with the idea that autistic perception may be influenced more strongly by immediate sensory evidence and less by broad contextual expectations.

That could, theoretically, help explain attention to local detail.

But this is another area where simple stories are dangerous.

The evidence for predictive differences is mixed, and autistic detail-focused processing itself varies between people and tasks. A systematic review found that predictive performance depended substantially on the kind of regularity being learned and the conditions under which it appeared.

So it is safer to say:

predictive-processing theories offer one possible way to investigate detail-versus-context differences

rather than:

autistic people see details because their brains lack priors.

The second claim goes far beyond the evidence.

🧠 What Does “Precision” Mean?

This is one of the most confusing terms in predictive-processing discussions.

Precision does not mean intelligence or accuracy.

It roughly refers to how much confidence the system places in a piece of information.

Imagine two weather forecasts.

Forecast A comes from a highly reliable system.

Forecast B comes from someone guessing.

Both say it will rain.

You naturally give more weight to A.

Predictive-processing models propose that brains also need to estimate which signals deserve more weight.

If incoming sensory information seems highly reliable, the brain should listen to it.

If an expectation is based on a very stable pattern, it may deserve more weight.

One autism theory — often referred to as HIPPEA, for high and inflexible precision of prediction errors — proposes that autistic processing may sometimes assign prediction errors unusually high or inflexible importance. But this remains a theoretical proposal rather than an established property of autistic brains.

🧩 The Same Person May Predict Differently in Different Contexts

This may be one of the most important conclusions from the research.

An autistic adult might learn a stable pattern perfectly well.

But prediction may become more difficult when:

🌪️ the environment changes frequently
❓ cues are ambiguous
🎯 relevant signals are weak
👥 social information is highly variable
🔄 expectations need rapid updating

The 2023 Nature Communications study found intact predictive ability under its experimental conditions, while discussing previous evidence showing differences under more uncertain conditions.

That points away from a simple trait such as:

“poor prediction.”

It points toward an interaction:

person × type of information × uncertainty × context.

🧠 The 2025 Updating Study Makes This Even More Interesting

The 2025 study by Shi and colleagues examined how autistic and non-autistic participants combined previous expectations with incoming information across repeated trials.

The autistic group did not simply ignore prior information. Instead, the groups showed different updating dynamics, and the differences reduced over repeated experience. The researchers proposed that autistic predictive processing may involve atypical iterative prior updating rather than an inability to use prior knowledge.

This is a much more nuanced model.

A new situation might initially be weighted differently.

With repeated exposure, the internal model may gradually converge.

That possibility fits better with the evidence than claiming that autistic people permanently experience the world without expectations.

🧪 Predictive Processing Is a Framework, Not a Diagnostic Test

There is no brain scan or behavioural task that can currently tell you:

“You are autistic because your prediction errors work like this.”

Predictive-processing research is trying to understand mechanisms at a group level.

It does not currently provide an individual diagnostic profile.

An autistic person might show typical performance on one prediction experiment and differences on another. A non-autistic person may show similar predictive patterns under certain conditions.

So predictive processing should not become another online self-test.

It is a research framework.

⚠️ Do Not Turn Theory Into Biology Too Quickly

Popular neuroscience often moves too fast from:

“Researchers have proposed…”

to:

“This is how the autistic brain works.”

Those are very different claims.

At present, predictive-processing theories of autism are plausible and actively researched, but major questions remain unresolved. The 2021 systematic review found a mixed pattern; several 2023–2025 studies have shown intact predictions alongside more subtle differences in updating or neural processing.

A scientifically responsible summary is:

some aspects of predictive processing may differ in autism under some conditions.

That is less dramatic.

It is also much closer to what the evidence actually supports.

🌪️ Predictive Processing Is Not Sensory Overload

The terms can sound related because both involve incoming information.

But sensory overload is an experiential state in which sensory and related demands become too much to manage effectively.

Predictive processing is a theoretical framework describing how expectations and incoming information may interact.

Someone can experience Sensory Overload without needing a predictive-processing explanation.

Predictive theory may eventually help explain part of why certain sensory situations are demanding for some autistic people, but it should not replace the practical concepts readers already use.

🧱 Predictive Processing Is Not Cognitive Rigidity

Likewise, difficulty changing plans does not automatically prove that expectations are “too strong.”

Cognitive flexibility includes the ability to switch strategies, perspectives or actions when conditions change.

Predictive processing asks how the brain represents and updates what is expected.

These may interact.

But a person can accurately recognise:

“The plan changed.”

while still finding the behavioural transition difficult.

The prediction has updated.

The transition system is struggling.

That is why Cognitive Flexibility remains its own useful concept.

😰 Predictive Processing Is Not Anxiety Either

Anxiety often involves anticipation.

What if this happens?

What if something goes wrong?

What if I cannot cope?

Predictive-processing frameworks are sometimes used in theories of anxiety too, but anxiety cannot be reduced to prediction error.

Likewise, an autistic person’s preference for knowing what will happen does not automatically mean they are anxious.

Predictability can reduce:

🧠 cognitive load
🔄 switching demand
🎧 sensory surprise
📋 planning demand
❓ ambiguity

It can simply make the environment easier to process.

Anxiety may be present as well, but the two should not be collapsed into one explanation.

📅 Practical Meaning: Make Important Things More Predictable

You do not need to believe a predictive-processing theory of autism to see the practical value of predictability.

If someone finds unexpected change difficult, useful support might include:

📅 giving advance notice of changes
📝 explaining what will happen
🗺️ showing the sequence of an unfamiliar situation
⏳ giving transition time
❓ identifying what is known and what remains uncertain
🔄 providing a backup option when possible

This reduces unnecessary ambiguity.

The goal is not to create a life where nothing changes.

It is to avoid making every change a complete surprise.

🧩 Prepare for the Type of Change, Not Every Possible Outcome

Perfect prediction is impossible.

Trying to anticipate every scenario can become exhausting.

A more sustainable approach is often:

“I do not know exactly what will happen, but I know what I can do if something changes.”

For example:

“The train may change platforms. If it does, I will check the board and give myself ten extra minutes.”

or:

“The meeting agenda may change. If an unexpected decision is needed, I can ask to respond afterward.”

That creates predictability around coping, even when the event itself remains uncertain.

This connects closely with Intolerance of Uncertainty in ADHD & Autism.

🎧 Reduce Unnecessary Surprise in Sensory Environments

You cannot make the sensory world completely predictable.

But some surprises can be reduced.

For example:

🔔 warning before a loud sound
💡 explaining that lights will be dimmed
🩺 describing what a medical procedure will feel like
🚪 allowing someone to see a room before an event
🎧 giving control over sound where possible

That can reduce both sensory uncertainty and the executive demand of adapting without warning.

This is especially useful in healthcare, schools and workplaces, where many “small” surprises are completely avoidable.

💼 Predictability at Work

Work becomes much easier to model when expectations are explicit.

Compare:

“We need this soon.”

with:

“Please send the first draft by Thursday at 15:00. It only needs sections one to three.”

The second gives the brain something stable to organise around.

Useful workplace predictability can include:

📅 clear deadlines
🎯 explicit priorities
📝 written instructions
🔄 advance notice of changes where possible
👥 clear meeting agendas
❓ clarity about what is still undecided

These are not only autism accommodations.

They are generally good communication.

But they can be especially valuable when uncertainty itself adds substantial processing demand.

❤️ Predictability in Relationships

Relationships contain huge amounts of implicit prediction.

What does silence mean?

Is the other person angry?

Are plans still happening?

When will they be home?

What happens after an argument?

Unclear patterns can produce substantial cognitive work.

Useful agreements might include:

💬 “If I need space, I will tell you rather than disappearing.”

📅 “If plans change, we message each other.”

❤️ “Needing time to process does not mean the relationship is ending.”

These agreements reduce ambiguity without eliminating spontaneity.

Predictability in relationships does not mean scripting every interaction.

It means making important patterns trustworthy enough that the other person does not have to constantly guess.

🌱 Prediction Can Also Be a Strength

Predictive-processing discussions often focus only on difficulty.

But recognising stable patterns can also become a major strength.

An autistic adult may become exceptionally knowledgeable within an area of interest and notice regularities that other people overlook. Deep familiarity with a system can make predictions precise because the person has accumulated enormous amounts of relevant information.

That does not prove that autism produces superior prediction.

It illustrates a broader principle:

prediction depends on experience, context and the kind of information being predicted.

The same person may find one environment highly predictable and another almost impossible to model.

🧠 Familiarity Changes the Problem

This may explain why the first experience of something can feel radically different from the tenth.

The first visit to a new clinic contains:

❓ unknown route
🚪 unknown building
👥 unknown people
💡 unknown sensory environment
🗣️ unknown procedure

The tenth visit contains far fewer unknowns.

The world has become more predictable because you have built a model of it.

Again, this is not uniquely autistic. All human learning works partly through familiarity.

But when uncertainty and sensory load already carry high costs, the difference between first time and known routine can become unusually important.

🧪 What Can You Learn From Your Own Prediction Patterns?

Rather than asking:

“Does my autistic brain have weak priors?”

which is not something you can meaningfully determine from daily experience, ask more useful questions:

❓ Which kinds of uncertainty are especially difficult?
🔄 Is unexpected change harder than expected change?
🎧 Does sensory input become easier when I control or predict it?
📅 How much advance information actually helps?
👥 Are social situations harder when the rules are ambiguous?
🧠 Does repeated exposure make unfamiliar situations much easier?
⚡ Do I notice every mismatch or only particular kinds?
🪫 Does unpredictability become harder when capacity is already low?

Those questions produce practical information regardless of which predictive-processing theory ultimately survives.

🌱 A Better Way to Think About the “Predictive Autistic Brain”

The most interesting idea in predictive-processing research is not that autistic people somehow live without expectations.

That picture is increasingly inconsistent with the evidence.

Autistic adults clearly learn statistical patterns, form expectations and use previous information. Some studies find remarkably typical predictive behaviour. Other studies find differences in updating, uncertainty or neural signals under particular conditions.

So instead of:

“Autistic brains cannot predict.”

a better question is:

“Under what conditions does an autistic brain rely on, update or respond to expectations differently?”

That question leaves room for heterogeneity.

It also gives science something precise enough to test.

🌱 Conclusion

Predictive processing is an influential framework for understanding how brains combine what they expect with what actually happens.

It has inspired several theories of autism because sensory differences, responses to unpredictability, routines and adaptation to change all seem potentially relevant to prediction. But twenty years of theory have not produced a simple finding that autistic people have weak expectations or cannot predict effectively.

Instead, the evidence is increasingly nuanced. Autistic adults can clearly form and use predictions. Some studies find intact predictive learning, while others find subtler differences in how expectations are updated or how prediction errors are represented, especially as uncertainty or task demands change.

Predictive processing therefore should not become a new master explanation for everything autistic people experience.

It is better understood as a promising research framework that may eventually explain part of the picture.

And the practical lesson does not require the theory to be completely settled.

Predictability can reduce unnecessary work. Advance information can make change easier. Familiarity can change how demanding an environment feels. Clear expectations can leave more capacity for everything else.

The world does not need to become perfectly predictable.

But sometimes knowing enough about what comes next makes it much easier to meet it.

📌 Key Takeaways

🧠 Predictive processing describes how the brain combines expectations with incoming information
⚡ A prediction error is the mismatch between what was expected and what occurred
🔄 Prediction errors can help the brain update its internal model
⚖️ Predictive-processing models also consider how much weight is given to expectations versus sensory evidence
♾️ Several predictive-processing theories have been proposed to explain aspects of autism
🚫 Current evidence does not support the simple claim that autistic people cannot predict or universally have “weak priors”
🧪 A 2021 systematic review found mixed results across different kinds of prediction.
🧠 Several 2023 studies found intact predictive behaviour in autistic adults.
🔄 Other research suggests that updating expectations may differ under some conditions, particularly when uncertainty changes.
🎧 Predictive processing may eventually help explain some sensory experiences, but it is not a proven cause of autistic sensory sensitivity
📅 Routines can reduce uncertainty and processing demand without being reducible to predictive processing
❓ Intolerance of uncertainty and predictive processing are related ideas but not the same construct
🔄 Cognitive flexibility and predictive updating are also different mechanisms
🤝 Predictive-processing theories should not replace relational explanations such as the Double Empathy Problem
🛠️ Advance information, clearer expectations and predictable transitions can make everyday environments easier to navigate
🌱 Predictive processing is a research framework, not a diagnostic test or a complete theory of autism

🔗 Read Next

🎧 Sensory Processing in Neurodivergent Adults
🔄 Cognitive Flexibility in Autism, ADHD & AuDHD
❓ Intolerance of Uncertainty in ADHD & Autism
🌪️ What Is Sensory Overload?
🧠 Cognitive Load in Neurodivergent Adults
⏱️ Processing Speed in Autism, ADHD & AuDHD
🤝 The Double Empathy Problem
🧊 Autistic Shutdowns in Adults
♾️ Autism Learning Hub
🛠️ Autism Coping Skills & Tools

📚 Sources & Research

Cannon, J. et al. (2021). Prediction in Autism Spectrum Disorder: A Systematic Review of Empirical Evidence.
Read the full paper
Reviews 47 studies and finds a mixed picture: differences appear in some forms of predictive learning and low-level processing, but the evidence does not support one general prediction deficit across autism.

Sapey-Triomphe, L-A., Weilnhammer, V. A., & Wagemans, J. (2022). Associative Learning Under Uncertainty in Adults With Autism: Intact Learning of the Cue–Outcome Contingency, but Slower Updating of Priors.
View the PubMed record
Autistic adults successfully learned predictive associations, but updating of prior expectations was slower when the environment changed.

Sapey-Triomphe, L-A. et al. (2023). Neural Correlates of Hierarchical Predictive Processes in Autistic Adults.
Read the full paper
Both autistic and non-autistic adults made accurate predictions and were influenced by prior expectations. Differences appeared in some neural signals, particularly at higher predictive levels, arguing against a universal weak-prior account.

Pesthy, O. et al. (2023). Intact Predictive Processing in Autistic Adults: Evidence From Statistical Learning.
Read the full paper
Found intact learning of statistical regularities in autistic adults and no evidence for a broad predictive-processing impairment on the task used.

Arthur, T. et al. (2023). Testing Predictive Coding Theories of Autism Spectrum Disorder Using Models of Sensorimotor Behaviour.
Read the full paper
Directly tests competing predictive-coding accounts and illustrates why different autism theories can make subtly different predictions rather than supporting one simple “predictive brain” explanation.

Zhao, C. et al. (2024). Predictive Processing of Music and Language in Autism Spectrum Disorder.
Read the paper
Across music and language tasks, predictive performance depended substantially on individual linguistic, musical and cognitive abilities rather than showing one general autistic prediction impairment.

Shi, Z. et al. (2025). Predictive Processing in Autism Spectrum Disorder: The Atypical Iterative Prior Updating Account.
View the PubMed record
Found that autistic participants could use prior information but showed different updating dynamics across repeated trials. The authors propose atypical iterative updating rather than an inability to form or use priors.