Overview

• In visually naive animals, feedforward inputs do not match modular responses, producing inconsistent patterns to the same stimuli.
• Recordings and computational modeling predicted two requirements for reliable perception: greater discriminability of layer 4 inputs and alignment with recurrent layer 2/3 connectivity, both confirmed experimentally.
• After experience, interconnected modules respond to the same features, yielding stable, coherent representations across layers and time.
• Pre-existing modular activity is present before eye opening, which the authors say primes rapid and efficient learning once vision begins.
• The team now aims to pinpoint the synaptic wiring changes that drive this alignment and to test whether the mechanism generalizes beyond vision.