When the brain makes reality fit its predictions

When the brain makes reality fit its predictions

A predictive processing account of self-sabotage, hallucination, and “controlled explosions”

The question of self-sabotage

Why do we so often behave in ways that seem to work against our own interests? Why do we pull away from people we long to be close to, undermine opportunities we care about, hear a hostile voice when no one is there, or suddenly go numb just when life matters most? These experiences are often described as irrational, self-defeating, or pathological. But what if they are not mistakes at all? What if they are the brain’s attempt to make sense of a world – and a body – that no longer matches what it has learned to expect?

A predictive brain

We tend to imagine that the brain simply takes in information from the world and builds an accurate picture of reality. However, contemporary accounts of the brain, particularly within predictive processing and free energy frameworks (Friston, 2010; Clark, 2016), offer a different view. The brain is not a passive recorder but an active prediction system. At any given moment, it is generating a best guess about what is happening (what we are seeing, hearing, and feeling) and comparing that guess to incoming signals. These signals arise both from the external environment (exteroception) and from within the body (interoception). From this perspective, perception is not a direct read-out of the world, but a constructed model: a prediction about what would need to be present for current sensations to make sense.

In this sense, all perception can be understood as a form of “controlled hallucination” (Seth, 2021): a model generated by the brain and continually shaped by sensory input. Most of the time, this system works remarkably well. Predictions are constantly updated in light of new information, allowing our experience to remain closely aligned with the world around us. Yet what we perceive is never a pure copy of reality. It is always a construction – the brain’s best attempt to fit its expectations to its sensory signals.

When expectations and experience diverge

Because perception is constructed in this way, the brain is continually comparing what it expects with what it senses. When these align, experience feels coherent and stable. When they do not, the brain encounters what is known as prediction error – a mismatch between expectation and experience. This idea of prediction error sits at the heart of predictive processing models of perception and action. This mismatch can arise because the external world does not match what was predicted, or because the internal state of the body feels unclear, unexplained, or out of keeping with expectation. In either case, the brain is faced with something that does not make sense and must be resolved.

There are two broad ways in which the brain can reduce this mismatch. One is to update its predictions so that they better fit what is being sensed. The other is to alter perception, bodily state, or behaviour so that experience comes to fit what was predicted. In other words, we can either change our predictions to fit reality, or change our experience of reality to fit our predictions. Both routes reduce surprise and restore a sense of predictability.

“We can either change our predictions to fit reality, or change our experience of reality to fit our predictions”

When old predictions remain

Our predictions are shaped by our past, particularly by repeated and emotionally significant experiences. If earlier life involved threat, unpredictability, criticism, or emotional neglect, the brain may come to expect these conditions. These expectations are not merely abstract beliefs, but deeply embodied patterns that shape how we perceive the world, other people, and ourselves. Because they were once adaptive, helping us to navigate difficult environments, they can become highly automated and resistant to change.

Later in life, circumstances may be very different. Relationships may be safer, opportunities more available, and the environment more stable. Yet if the brain continues to predict threat, rejection, or instability, the present can feel strangely incongruent. It is not that the world is unsafe, but that it does not match what is expected. This creates mismatch, and once again the brain must resolve it.

Behaviour as prediction fulfilment

It is here that many forms of self-sabotage begin to make sense. If the brain predicts rejection, a person may withdraw or behave in ways that create distance until the world once again feels rejecting. If it predicts failure, they may avoid or undermine progress until that expectation is confirmed. If it predicts instability, they may generate chaos until things feel familiar again.

From the outside, these behaviours appear irrational or self-defeating. From the perspective of the brain, however, they are reducing mismatch by making the world more predictable. What looks like self-sabotage may in fact be an attempt to restore coherence between expectation and experience.

In Controlled Explosions in Mental Health (Heriot-Maitland, 2026), I describe these patterns as “controlled explosions” – actions that may create harm, but in doing so avert a deeper, more uncertain threat.

Perception on a continuum

To understand hallucinations, we need to return to the nature of perception itself. If all perception is constructed – a prediction shaped by sensory input – then the difference between everyday perception and hallucination is not absolute, but one of degree. Most of the time, our predictions are tightly constrained by incoming sensory signals, which continually update and correct them. However, under certain conditions this balance can shift. When predictions are particularly strong (shaped by past experience, emotional salience, or urgent motivational states) and when sensory input is weak, ambiguous, or overwhelmed by internal signals, perception may be driven more by prediction than by external data.

From this perspective, hallucinations are not a fundamentally different kind of perception, but rather perception in which prediction has taken the lead. This account is also consistent with compassion-focused formulations of voice-hearing, where perceptual experience can emerge from underlying threat-based motivational states (Heriot-Maitland & Bell, 2025).

“Hallucinations are not a fundamentally different kind of perception, but rather perception in which prediction has taken the lead”

This becomes especially apparent in situations of loss, threat, or unmet need. In grief, for example, the brain may continue to predict the ongoing presence of a loved one – their voice, their responsiveness, their relational availability. When the world no longer provides the sensory signals to confirm this, a profound mismatch arises. One way the brain may respond is by generating perceptual experience that fits the prediction, such as sensing their presence, hearing their voice, or encountering them vividly in dreams.

Dreaming provides perhaps the clearest illustration of this process. During sleep, the brain is largely cut off from sensory input from the external world. With little incoming data to constrain its activity, experience becomes almost entirely shaped by internally generated predictions. In this sense, dreaming reveals what perception looks like when prediction runs largely unconstrained. Everyday perception, hallucination, and dreaming can therefore be understood as points along a continuum, differing not in kind but in the degree to which sensory input shapes and corrects the brain’s predictions.

When the body leads perception

Perception is shaped not only by the external world, but also by the internal state of the body. Sometimes the mismatch that drives experience begins internally. The body may be activated (tense, fearful, ashamed, or shut down) without a clear external cause. This creates a form of internal prediction error, a state that demands explanation. In such situations, the brain may effectively ask: “If I feel like this, what must be happening?” It may then generate a perception that fits the bodily state.

In this way, it may be that bodily fear can give rise to hearing a threatening voice. The perception (in this case, hearing a voice) provides orientation, meaning, and direction for action, helping to organise an otherwise confusing internal state. In this way, perception and bodily state can be understood as part of a dynamic, self-reinforcing system, where threat activation and perceptual experience organise one another (Heriot-Maitland & Bell, 2025).

Making sense of controlled explosions

Seen in this light, it becomes less helpful to ask whether such experiences are real or unreal, and more helpful to ask what they are doing. What mismatch might they be resolving? What are they organising, explaining, or moving us towards? This shift in perspective can be applied not only to hallucinations, but also to self-criticism, dissociation, anxiety, and many other patterns of experience.

From this standpoint, what I have described as “controlled explosions” (Heriot-Maitland, 2026) can be understood as one set of strategies the brain uses to manage mismatch. Whether behavioural, perceptual, or bodily, these patterns may serve a common function: reducing uncertainty by bringing experience into line with familiar predictions. They can be harmful, certainly, but they are often patterned, contained, and predictable forms of harm, emerging as part of a trade-off in which the alternative may feel more overwhelming or less controllable.

Returning with compassion

If these patterns are driven by earlier predictions, then change is unlikely to come simply from trying to stop them. Instead, it requires understanding where they came from. Many of these predictive templates were formed in earlier stages of life, when we were more dependent, more vulnerable, and more exposed to particular kinds of threat. Although those conditions may no longer be present, the predictions shaped by them can continue to operate. In this sense, we are not going back for its own sake, but because the brain is still predicting from there.

Engaging with these earlier patterns requires courage, but also compassion. What we encounter is not failure or dysfunction, but adaptation – a system that learned, under pressure, how to survive. Compassion creates the conditions under which these patterns can begin to change. It allows the brain to gradually update its predictions, learning that closeness can be safe, that needs can be met, that vulnerability can be tolerated, and that the present is not the past.

Over time, new experience reshapes old models. Self-sabotage, hallucination, and emotional overwhelm are no longer seen simply as signs that something has gone wrong, but as expressions of a predictive brain attempting to maintain coherence and survival. If these patterns were learned, they can be relearned. If they were predicted, they can be updated. And if they were built to protect us, they can be met, and gradually transformed, with compassion.

References

Friston, K. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience.

Clark, A. (2016). Surfing Uncertainty: Prediction, Action, and the Embodied Mind.

Seth, A. (2021). Being You: A New Science of Consciousness.

Heriot-Maitland, C. (2026). Controlled Explosions in Mental Health: A Compassionate Guide to Understanding Why Our Brains Self-Sabotage, Self-Criticise, and Self-Harm.

Heriot-Maitland, C., & Bell, T. (2025). Compassion-focused chairwork for voice-hearing relationships, body triggers and motivational states. Psychology and Psychotherapy: Theory, Research and Practice.

← Back to Stories

← Back to Research, Writing & Ideas