This is a preprint. It has not undergone peer review. Neuromodulators are classically described as setting the rate of synaptic learning. We develop a framework in which a single eligibility-trace mechanism is controlled along two dissociable channels: noradrenaline and serotonin set the direction of plastic change, through opposite transformations of the trace, while dopamine sets the gain of its consolidation. On one consolidation equation — an induction gate and a ceiling acting on the eligible trace, an autocatalytic maintenance term holding what has been consolidated, and a homeostatic term removing weight without drive — the two channels yield dissociable loci of failure that converge on a common region of phase space: saturation of the consolidated weight, collapse of the effective learning rate toward a floor set by the asymptote of prediction-error precision, and loss of plastic reversibility. The convergence is not an identity. Only the direction channel relaxes prediction-error precision, and only there does the collapse run deep; the gain channel is bounded above that floor by the ceiling on synaptic weight. Direction failures instantiate as opposite regimes — sustained noradrenergic excess locks in threat circuits (anxiety), a sustained serotonergic deficit leaves compulsive circuits undepotentiable (obsessive-compulsive disorder) — while a gain failure is proposed as the aberrant-salience-to-stabilization route to psychosis. A formal cholinergic reset parameter discounts prior precision without depotentiating the circuit that holds it, defining computable conditions for escaping the collapse. The framework yields falsifiable predictions distinguishing the regimes in reversal learning, electrophysiology and neuroimaging, including a dissociation on the depth and recoverability of the effective learning rate. Supporting data and code: https://doi.org/10.17605/OSF.IO/RDHB3. An earlier version of the manuscript (v30) is archived at https://doi.org/10.17605/OSF.IO/S3UQ4.
No takes yet. Share an insight, caveat, or question.
Salvo Montalbano (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: