Chronic Pain Through the Prediction Lens

Have you ever wondered why two people with identical scans can have completely different pain experiences, or why pain sometimes persists long after the tissue that first caused it has healed? These are among the most durable puzzles in pain research, and the predictive processing frame offers one of the more productive ways to hold them.

The standard folk model treats pain as a signal — tissue damage in, pain out, roughly proportional. That model was overturned by the pain research community decades ago (Moseley & Butler, 2015) [Supported Finding], but it lingers in popular understanding because it is intuitive. The current, well-supported picture is that pain is a protective output the nervous system produces when it evaluates that a body region requires attention. Nociceptive input from tissue is one input to that evaluation, not a direct readout of it.

Predictive processing extends this picture. In the predictive frame, pain is not merely constructed by the brain — it is actively predicted. The nervous system runs an ongoing model of which body regions are likely to require protection, based on all available information: sensory input, contextual cues, prior experience with similar situations, current stress levels, expectations about outcome. Pain is what surfaces to consciousness when the prediction crosses a threshold.

This explains several observations that the tissue-damage model cannot. It explains why placebo works — a strong prior that a treatment will help genuinely reduces the prediction that this body region needs protection. It explains why chronic pain often outlasts injury — the nervous system's model has learned to predict pain in that region, and the prediction persists even after the original tissue signal has resolved. It explains why anxiety, sleep loss, and social threat all reliably increase pain — each raises the overall prior that protection is required.

What this frame does not do is claim that chronic pain is imagined or that it is the person's fault. That misreading is the most common failure of popularisations of this research. Pain generated by prediction is fully real. The nervous system is doing what it evolved to do. The person is not manufacturing anything. What has changed is that the mechanism is now understood to be centrally driven rather than peripherally driven, and that opens specific therapeutic possibilities.

The practical implications are worth being specific about. Approaches that target the prediction system have shown promise in randomised trials — pain neuroscience education (explaining the mechanism to the patient), graded motor imagery, and specific cognitive-behavioural approaches for chronic pain all show measurable effects. None of them are cures, and none of them replace medical evaluation. They are additional levers the tissue-damage model would not have suggested existed.

A useful way to hold this frame is as an expansion of the pain toolkit, not a replacement of the physical one. Tissue matters. Prediction matters. Both are inputs to what the nervous system ultimately experiences.

For the underlying model, see The Brain Is a Prediction Machine. For a related application, Anxiety as a Predictive-Processing Disorder is the piece to pair with it.