Canalization on a Living Substrate

Repeated stress carves two kinds of damage into the brain simultaneously — one into the circuit's wiring, one into the biological tissue itself — and this paper argues that understanding both together is what explains why some people get stuck.

The core idea is simple but underappreciated. When a neural circuit fires repeatedly, two things happen in parallel. First, Hebbian plasticity deepens the attractor — the pattern becomes easier to fall into and harder to escape. Second, the biological substrate accumulates wear: oxidative damage, DNA breaks, protein aggregation, and loss of neurotrophic support (especially BDNF). Healthy brains counteract both: sleep-dependent synaptic renormalization pulls biased weights back toward baseline, while cellular repair machinery patches the molecular damage. The paper's claim is that pathological "canalization" — the progressive narrowing of a system's behavioral repertoire into a rigid, maladaptive groove — happens when one or both of these protective layers fail.

To test this, the authors build a minimal model: K neural populations competing via lateral inhibition (a winner-take-all circuit), with two slow variables riding on top of the fast dynamics. The first slow variable is the self-excitation weight, shaped by Hebbian plasticity and pulled back by a renormalization term (standing in for sleep). The second is substrate health, degraded by activity and restored by repair, which directly controls how much noise each population generates. This is the key mechanistic link: a damaged population becomes dynamically rigid — it can't explore, can't escape, can't compete. The model is then stressed (one population gets strong external drive), and the four combinations of high/low renormalization × high/low repair are compared.

The four regimes are qualitatively distinct, not just quantitatively different. Healthy systems recover full diversity after stress. Plasticity-control failure alone produces permanent canalization — the stress attractor is too deep to escape even with intact substrate. Repair failure alone produces something subtler and arguably more insidious: the weights recover (the attractor looks normal), but the substrate stays damaged, so the system is sensitized — a weak trigger later produces a disproportionately large and prolonged response. Combined failure gives both: permanent canalization on a progressively degrading substrate. The phase diagram shows a clean separation, with renormalization rate controlling whether canalization occurs and repair rate controlling substrate fragility.

The clinical translation is careful. Recurrent depression maps onto the combined-failure regime: each episode deepens the low-valence attractor via insufficient renormalization (disrupted sleep, weakened prefrontal control) while also damaging the substrate (stress-induced BDNF reduction, HPA dysregulation), so each subsequent episode requires less provocation — exactly the kindling pattern documented in the epidemiological literature. PTSD maps onto plasticity failure: the threat attractor is Hebbian-deepened during the traumatic event, and failed extinction (itself a form of renormalization) keeps it dominant. Musician's dystonia maps onto circuit-specific overtraining: the same dual failure in a motor circuit rather than an affective one. The model also generates concrete predictions — reduced EEG complexity in canalized states, exaggerated cue-triggered replay in sensitized systems, and a specific interaction between BDNF Val66Met genotype and stress sensitization — that are testable with existing neuroimaging and biomarker methods.