Monocular deprivation during a critical period permanently rewires visual cortex — the deficit lives in cortical wiring, not the deprived eye

Hubel and Wiesel sutured one eye of a kitten shut during early development. The open eye's cortical columns permanently annexed the ocular-dominance territory of the deprived eye — even though the deprived eye itself remained healthy. The permanent deficit lives in cortical wiring, not in the eye: "Ocular dominance is established irreversibly early in childhood." Outside that early window the same deprivation does far less harm; inside it, the loss is effectively unrecoverable.

The load-bearing surprise is the relocation of the damage. The naive prior — "the eye is fine, so vision will recover" — is wrong; the clock is in the cortex, not the lens. This is precisely why pediatric surgeons now remove congenital cataracts within weeks of birth and treat amblyopia early or not at all: the therapeutic window is the cortical critical period, and it closes.

This is the biological anchor for the claim that deep networks show the same onset/length-dependent critical periods (claim-deep-nets-have-critical-learning-periods-timed-like-animals), and it carries a further sting once that analogy is taken seriously: a deep net reproduces the phenomenon with no synaptic pruning or neuromodulators at all, suggesting the critical period may be more about information than about pruning chemistry (claim-critical-periods-arise-from-information-plasticity-not-biology). The same Hubel–Wiesel work — the discovery of orientation-selective simple and complex cortical cells — is also the neuroscience the convolutional network was copied from, via Fukushima's S-cells and C-cells (claim-fukushima-1979-neocognitron-first-cnn). Whether the deep-net model has ever run back the other way, predicting a human critical-period timing such as amblyopia's, is open (question-deep-net-critical-period-predicts-human-amblyopia-timing).