A pattern of activation in the cortex (red dots) induces a pattern of synchronously expanding waves (black circles). Synchronous activation by the expanding waves causes some cortical connections to have temporary increases in gain, via fast acting Hebbian learning. This means that subsequent activation by a translated copy of the original pattern of activation (blue dots connected into a triangle) will be enhanced.

How does the brain maintain object constancy in the face of a constantly changing visual world? This problem is often posed in terms of invariant object feature learning. Instead, we propose that the cortex deals with invariance by storing object information in maps with a specific topological property: as an object is transformed, its representation doesn't change but translates to a different part of the map. Neuro-dynamic mechanisms broadcast the representation across the entire map, allowing the initial learning to be generalized over a set of object transformations. The cortical mechanisms by which stored representations are broadcast involve synchronous activation, travelling waves and timing-dependent learning rules. From a few prior presentations, learning can be generalized over all instances of the mental object. For topographical maps such as primary visual cortex, the system provides a possible mechanism for simultaneously representing identity and variation of objects whose features change through time.