The 6 Stages of Crisis Through the Predictive Processing Theory Lens
Working with SEN students you will know about the 6 stages of crisis. This blog gives a deep insight into this by using the latest cognitive science thinking - Predictive Processing Theory.
You see it in real time: a student’s breathing quickens, their body tenses, and their responses spiral out of control. As adults, we often respond to what we can see — behaviour, tone, language — but underneath that is a deeper story about how their brain is trying to cope. Predictive Processing Theory gives us a powerful lens to understand what is happening and why and how our responses can help.
This post has been inspired by the substack posting 🧠Understanding the 6 Stages of Crisis🌋 by the excellent SEMH Education.
Quick Intro to Predictive Processing
The brain is not a passive sponge soaking up the world. Instead, it’s a prediction machine. It continuously generates internal predictions about what’s going to happen next — based on past experiences, current context, and what it’s learned over time.
This predictive processing theory perspective represents a major shift in our understanding of the brain and is now widely regarded as a dominant theory of cognition among leading neuroscientists and philosophers of mind. Pioneers like Karl Friston (Free Energy Principle), Andy Clark (Experience Machine), Anil Seth (Being You), and Jakob Hohwy (The Predictive Mind) have contributed to this model, which reframes everything from sensory processing to emotion, memory, and mental health through the lens of prediction error minimisation.
Image: From predictablycorrect.substack.com
Crucially, the brain does not construct a sense of reality by first processing raw sensory input and then building a model of the world. Instead, it starts with a generative model — a generative simulation of what it expects stored in long-term memory — and uses sensory input only to correct errors in that model. In this view, perception itself is a controlled hallucination, constrained by the world but shaped by prior beliefs.
Then, it compares these predictions to what’s actually coming in from the senses. Any mismatch is called a prediction error. To manage the noisy sensory inputs from the world, it applies a precision weighting( confidence level) between the predictions it makes and the sensory input it receives.
When things are calm and predictable, the brain’s generative models (its internal map of how the world works) are doing a great job — the errors are small, and the brain adjusts gently.
But when the world feels unpredictable or threatening, or when previous trauma hijacks the system, prediction errors start to spike. The brain adjusts the precision — the confidence it gives to different sources of information — to try and make sense of the chaos. Often, this spirals into a dysregulated state.
But predictive processing is only part of the story. Active inference builds on this by proposing that the brain doesn't just passively update its internal model based on errors — it can also take action to minimise those errors.
This can be small and automatic, like an eye movement to bring the next word into focus, or turning your head to catch someone’s tone of voice more clearly. But it can also be large and effortful: avoiding certain people, shutting down emotionally, or even lashing out physically — all as attempts to reduce the gap between what the brain expects and what it’s sensing.
In short, we don't just react to the world — we act on it to make it more predictable.
You can learn more about predictive processing theory in my Substack blog: PredicablyCorrect
The Brain Predicts More Than We Think
When we talk about prediction in the brain, we often think in terms of the five senses and navigating the physical world. But in reality, the brain is constantly generating and correcting predictions across a vast range of systems — all at once.
It’s not just sight, sound, smell, touch, and taste. The brain is also tracking 30+ interoceptive channels — predicting things like heart rate, hunger, temperature, bladder fullness, and breathing — in order to maintain homeostasis. It’s managing proprioception to track the body’s position in space. It’s processing the 3D environment and navigating obstacles and motion. And, crucially, it’s constantly simulating the social environment — predicting the words, facial expressions, emotional states, and intentions of everyone nearby.
When a student becomes dysregulated, all of these predictive systems are affected. That’s why the experience can feel so globally overwhelming — it's not just emotional distress, but a collapse in the brain’s ability to manage everything it normally tracks automatically.
This also explains why strategies like removing the audience, having only one trusted adult interacting, and reducing social unpredictability are so effective. They aren’t just de-escalation tools — they’re ways of unburdening the brain’s prediction system so it can begin to recover.
6 Stages of Crisis
Image from: Headteacher Update
Let’s explore each stage of the crisis model through this lens — and understand what strategies actually support the brain at each point. I’ve added stage 0-Regulated to the team teach 6 stages of crisis model to illustrate what a non-crisis state is like through the predictive processing lens.
Not in Crisis - Regulated
What it looks like:
The student is calm and alert. They are coping with classroom demands, engaging in learning, and responding to social cues and guidance.
Predictive Processing Lens:
The student’s generative models are accurately predicting what’s going to happen — socially, physically, and emotionally. Sensory input matches what the brain expects, so prediction errors are low and easily resolved. Precision (the balance between trusting sensory input vs internal models) is well-regulated.
How to Support (and Why it Helps):
• Maintain routines and predictable environments.
• Signal transitions clearly.
• Provide consistent adult responses.
These strategies keep prediction error low. They reinforce the student’s generative models by confirming that their predictions are valid, which keeps the brain feeling safe.
1- Anxiety / Trigger
What it looks like:
Small but noticeable changes: fidgeting, withdrawal, heightened sensitivity, or frustration. The child may become quieter or more irritable. Something has been triggered.
Predictive Processing Lens:
A trauma-linked generative model may have been activated by a familiar pattern of sensory inputs — even a partial match. Prediction errors are increasing across multiple channels. The brain begins reallocating precision — often tipping toward hypervigilance — and this increased load pushes the system closer to its bandwidth limit. Adrenaline is likely rising.
How to Support (and Why it Helps):
• Acknowledge their emotions early and calmly.
• Reduce sensory load where possible.
• Use simple, straightforward language.
• Offer structure, predictability, and choice.
These interventions help reduce incoming prediction errors and encourage the brain to trust sensory input again. They also ease the overall prediction load, giving the system space to regain balance.
2-Defensive / Escalation
What it looks like:
More obvious signs of dysregulation: verbal outbursts, pacing, challenging staff, refusal. Language processing starts to break down.
Predictive Processing Lens:
A trauma-linked generative model may have been activated by a familiar pattern of sensory inputs — even a partial match. Prediction errors are increasing across multiple channels. The brain begins reallocating precision — often tipping toward hypervigilance — and this increased load pushes the system closer to its bandwidth limit. Adrenaline is likely rising.
How to Support (and Why it Helps):
• Acknowledge their emotions early and calmly.
• Reduce sensory load where possible.
• Reduce social load where possible.
• Use only 1-2 simple words.
• Offer structure, predictability, and choice.
These interventions help reduce incoming prediction errors and encourage the brain to trust sensory input again. They also ease the overall prediction load, giving the system space to regain balance.
3- Crisis
What it looks like:
Overwhelm. Shouting, aggression, self-harm, or complete shutdown. The child is no longer responsive to language or reasoning.
Predictive Processing Lens:
Prediction errors are now far beyond bandwidth. The system can't resolve them through perception alone — so it detunes sensory input altogether. The brain falls back fully on internal predictions. At this point, active inference takes over. The student’s brain is now attempting to bring the world into line with its predictions, even if those predictions are shaped by trauma.
This is why aggressive behaviour in crisis is not just impulsive or reactive — it may be a desperate attempt to make the world match a generative model that says, “I’m not safe,” or “people will hurt me,” or even “I am bad.”
In this state, fight, flight, or freeze aren’t just survival reflexes — they are strategic actions to reduce intolerable prediction errors by changing either the input (sensory experience) or the environment.
How to Support (and Why it Helps):
• Prioritise safety: reduce stimuli and move slowly.
• Be physically and emotionally calm.
• Minimise social interaction, preferably just one known/safe adult.
• Avoid verbal interaction if possible.
These approaches reduce the need for the brain to process new inputs, and allow it to slowly lower prediction error levels. Consistent presence reassures the system without adding further sensory or social complexity.
4 - Recovery
What it looks like:
Breathing slows, eye contact may return. The child appears calmer but remains fragile — easily tipped back into crisis.
Predictive Processing Lens:
Prediction errors are now reducing, but the brain still doesn’t trust sensory input. Precision is still biased toward internal models. The system is cautiously testing the waters again.
How to Support (and Why it Helps):
• Keep everything simple, familiar, and low-demand.
• Avoid conversations or requests.
• Offer presence, predictability, and sensory grounding.
These strategies support the gradual reweighting of precision — encouraging the brain to start trusting the present environment again, at a manageable pace.
5 - Depression
What it looks like:
Withdrawn, exhausted, teary or ashamed. The student may be quiet but emotionally vulnerable.
Predictive Processing Lens:
The prediction system is coming back online, and precision is balancing out. However, the student’s experience of the crisis was shaped by their internal generative models — so what they think happened may differ significantly from reality.
How to Support (and Why it Helps):
• Be available without demanding interaction.
• Re-engage gently with very simple tasks, preferably physical in nature 1st.
• Avoid factual reviews of the crisis yet.
Low-stakes engagement helps the system continue to recalibrate safely.
6 - Restoration
What it looks like:
The child is calm and able to talk. They may reflect on the incident and show readiness to rejoin the group or lesson.
Predictive Processing Lens:
The system is now fully regulated. Prediction errors are at manageable levels, and precision is back to a balanced state. The brain can now reflect, update its generative models, and encode learning from the experience.
How to Support (and Why it Helps):
• Engage in restorative conversations.
• Talk through what helped, and what to try next time.
• Focus on understanding, not blame.
• Gently align student’s beliefs of what happened with what actually happened using concrete information.
Reflection at this stage helps rewire the generative models — strengthening adaptive responses and building a deeper understanding of emotional regulation. It is possible for maladaptive responses like the fight or flight to become embedded in the generative model if they were successful at minimising the prediction error. In the regulated time after the event before the next sleep cycle is the opportunity to try to modify the memory consolidation of this maladaptive response consciously- but in a gentle and supportive way.
Trauma and Neurodiversity
Why Neurodivergent Students Are More Vulnerable to Dysregulation
From a predictive processing perspective, neurodivergent individuals — including those with autism, ADHD, and sensory processing differences — are often operating with much higher baseline levels of prediction error. Their brains may assign greater precision to sensory input, meaning they give more weight to what their senses are telling them, even when the information is noisy, ambiguous, or overwhelming. At the same time, their generative models tend to be narrower or less generalisable — they may not easily form broad, flexible expectations about the world.
The result? Their internal predictions are more often mismatched with what’s happening in real time, leading to frequent and intense prediction errors. This pushes the brain closer to its bandwidth limits more regularly, making dysregulation more likely — and sometimes quicker to emerge.
Peter Vermuelen’s excellent book Autism and the Predictive Brain is an excellent introduction to Predictive Processing theory and how it provides an informative model on how neurodiversity affects cognition.
Why Trauma Lowers Thresholds for Dysregulation
Traumatic experiences often lead the brain to develop strong, narrowly tuned predictive models shaped by fear or threat. When current situations even vaguely resemble those past events, the brain may quickly default to these trauma-based predictions. Precision is often over-weighted toward threat detection, and the system operates with a chronic sense of prediction error. This can cause rapid escalation even in safe environments.
This aligns with recent neuroethological models (Linson, Parr, & Friston, 2020) and predictive frameworks of trauma (Putica & Agathos, 2024), which describe how maladaptive precision weighting and generative model updating contribute to chronic dysregulation.
Understanding this helps us frame trauma behaviour as a mismatch between past and present prediction systems — not as an overreaction but as misfiring survival systems. Trauma-informed support involves creating sensory and relational safety that allows the brain to rebuild trust in the here and now.
This is why predictability is not just helpful but essential for many neurodivergent and traumatised students. Routines, clear transitions, and careful management of change can all help reduce prediction error and support regulation. Importantly, understanding this can shape how we support anyone in crisis — showing just how much our environments, not just internal traits, affect regulation.
The Impact of a Predictive Processing Lens - Personal Perspective
Looking at crisis and behaviour that challenges through a predictive processing lens has transformed how I respond as a professional. Understanding what might be happening in the student’s brain — the clash between internal predictions and overwhelming sensory input — has unlocked a deeper empathy. It’s helped me let go of old narratives like “they’re just being naughty” or “they just want attention.” But it’s also helped me avoid the opposite trap — that of removing all agency from the student. This model doesn’t excuse behaviour; it explains it. And in doing so, it allows us to think clearly and compassionately about the long-term work: how we can help students reshape their generative models over time. That takes patience, strong boundaries, and consistent, supportive relationships. Not as quick fixes — but as the slow, steady work of rewiring belief systems that were built for survival, not safety.
Final Reflections: A Predictive Processing Approach to Support
Understanding dysregulation through predictive processing allows us to act not just as behaviour regulators but as prediction error managers.
• We reduce sensory and relational complexity
• We stabilise patterns and build trust in the present
• We help students rebuild generative models that align with reality
In doing so, we shift from managing meltdowns to fostering cognitive safety. This framework offers a proactive, compassionate, scientifically grounded pathway to support our most vulnerable learners.
I encourage you to get involved in the debate, comment on this post, and subscribe(free) to my substack
REFERENCES
Clark, A. (2023). The experience machine: How our minds predict and shape reality. Pantheon Books.
Friston, K. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138. https://doi.org/10.1038/nrn2787
Hohwy, J. (2013). The predictive mind. Oxford University Press.
Linson, A., Parr, T., & Friston, K. J. (2020). Active inference, stressors, and psychological trauma: A neuroethological model of (mal)adaptive explore-exploit dynamics in ecological context. Behavioural Brain Research, 380, 112421. https://doi.org/10.1016/j.bbr.2019.112421
Putica, A., & Agathos, J. (2024). Reconceptualizing complex posttraumatic stress disorder: A predictive processing framework for mechanisms and intervention. Neuroscience & Biobehavioral Reviews, 164, 105836. https://doi.org/10.1016/j.neubiorev.2024.105836
SEMH Education. (2025, March 23). Understanding the 6 stages of crisis. SEMH Education.
🧠Understanding the 6 Stages of Crisis🌋
17 days ago · 9 likes · 7 comments · SEMH Education
Seth, A. (2021). Being you: A new science of consciousness. Faber & Faber.
Vermeulen, P. (2022). Autism and the predictive brain: The predictive coding model of autism. Routledge.
Wray, A. (2024, May 29). Teachers are prediction error managers. Predictably Correct.
This is a great post. I'm not an educator currently, though I did work as a special ed para for some time, and I'm sending this article to some of my old colleagues. I'm fascinated by predictive processing theory generally which is how I came across this piece. Thank you for writing it.
You might enjoy my writing as well. I write primarily about psychology, neuroscience, and semi-esoteric spirituality. And funnily enough predictive processing has, in one way or another, come up in just about everything I've written so far. It's by far the strongest framework I've ever come across for understanding how I process reality and it's answered so many questions about life I never even realized I had.
It'd be an honor if you subscribed! But either way, I just wanted to thank you for a great read, much appreciated.