The brain doesn't just react to the world: it's constantly guessing what will happen next. This principle is called predictive coding: neurons compare expectations with reality and update their internal picture when an error occurs. But how such a complex system could be wired into the living tissue of the cortex remained unclear.
The authors of the new work have built a bridge between mathematics and biology. They took a model of recursive Gaussian processes (a way to build multi-layer predictions where each layer builds on the previous one) and showed that it naturally performs all the necessary operations: it propagates uncertainty, weighs errors by their importance, and builds a hierarchy of guesses. The key trick is a single shared 'backbone' stochastic process that runs through all layers, keeping them connected.
The researchers then went further and fitted this mathematical construct onto the real microanatomy of the cortex. It turned out that each component of the model (the shared process, the selection mechanism, the update dynamics) has a clear counterpart in known neuronal populations and connections between layers. This is not just an analogy: the scheme predicts specific differences in activity rhythms for bottom-up and top-down signal transmission, which can be tested experimentally. Thus, recursive Gaussian processes become not only a convenient computational tool but also a candidate for a unified language to describe the brain's predictive machinery.
The model predicts that prediction errors traveling from lower to higher cortical layers should be carried by higher-frequency waves than the corrective signals going in the opposite direction.