When an image doesn’t line up, the visual cortex quickly picks one version

The Cold Spring Harbor team looked at a pair of regions in the mouse: V1, the primary visual cortex, and LM, a neighboring area that already assembles more complex features. There are bidirectional connections between them. When activity patterns matched, they persisted longer. When they diverged, the conflict disappeared in less than a second. The authors call this consensus-building.

If one of the regions is temporarily switched off, the other’s behavior changes in the way the model of this loop predicts. An artificial network with the same reciprocal connections behaves similarly: what’s shared remains, the disagreement collapses. So it’s not just that a “higher-level” area dictates. The two areas, as it were, vote each other in.

This hasn’t yet been tested for hearing or in humans. Two mouse areas aren’t the whole cortex. But the principle is elegant. Perception here is neither a photograph nor a permanent compromise. It’s a rapid cleanup: keep what’s consistent, discard what doesn’t match. Illusions and failures may be precisely the places where the vote got stuck or the wrong side won.

Sources:

1. Cold Spring Harbor Laboratory. "When what you see doesn’t make sense, your brain does something remarkable." ScienceDaily (24 September 2026).
2. Javadzadeh M. et al. "Reciprocal connections dynamically build consensus between neocortical areas." Nature Neuroscience (2026). DOI: 10.1038/s41593-026-02437-3.

Illustration: Javadzadeh lab/CSHL