The Predictive Processing Loop

Have you ever wanted a single mental picture that ties together the various pieces of predictive processing you have been reading about? This piece is that schematic — a six-step loop that describes what a nervous system is doing, moment to moment, in the predictive frame.

We hold this schematic as a working synthesis of well-supported findings across the literature [Supported Finding]. It is a teaching diagram, not a formal model. Its purpose is to make the moving parts visible in a way that supports further reading.

**Step one: prior.** The nervous system arrives at any given moment carrying an accumulated model of what is likely to happen. This model is built from every prior experience, weighted by frequency, recency, and salience. It runs continuously below awareness.

**Step two: prediction.** The prior generates a specific forecast about the next moment — what the sensory data will look like, what the body will feel, what actions are likely to be needed. The forecast is not conscious. It arrives already shaped by everything that came before.

**Step three: sensory data.** The world sends signals. Photons, air pressure, chemical gradients, internal signals from the body's own tissues. These arrive at the nervous system's sensory surfaces and get processed upward through the hierarchy.

**Step four: comparison.** The prediction meets the sensory data. Where they match, the system moves on quietly — the model was right, no updating is needed. Where they disagree, the system generates a prediction error signal proportional to the discrepancy.

**Step five: choice — update or act.** When prediction error arrives, the system has two ways to resolve it. It can update the prior to better fit the incoming data. Or it can act on the world so that future data matches the current prior. Both moves reduce prediction error. Only the first one produces learning.

**Step six: consequences.** The updated model, or the world-changing action, produces a new moment. The loop restarts. This is happening thousands of times a second across various sensory and cognitive systems.

What the schematic makes visible is that perception, learning, and action are not three separate things. They are aspects of the same loop, run in slightly different modes. What appears as perceiving is mostly steps one through four. What appears as learning is the update pathway from step five. What appears as motivated action is the act pathway from step five. All of them are the same underlying machinery.

A few practical implications follow from the diagram. First, most of the work is invisible. Steps one through three happen below awareness. What consciousness usually gets is a smoothed-out version of step four — the resulting perception, with the underlying prediction and correction hidden. Second, the leverage point is step five. Whether the nervous system updates its model or acts to reinstate the old one is where meaningful change happens. Third, everything upstream — how a nervous system is resourced, how safe it feels, how many resources it has for tolerating prediction error — shapes whether the update path or the act path gets chosen.

This is one useful map of what a nervous system is doing, moment to moment. It is not the only map. It is a productive one.

For each of the six pieces, follow the concepts back to their own articles: The Brain Is a Prediction Machine, Priors — The Brain's Working Assumptions, Prediction Error — The Currency of Learning, and Active Inference — Acting to Change Predictions.