无删除的降尺度：区分睡眠中的突触降尺度与类淋巴废物清除

Abstract The dominant paradigm holds that NREM sleep is a window for synaptic pruning — weak synapses are selectively eliminated by microglial phagocytosis to optimize network efficiency, as formalized in the Synaptic Homeostasis Hypothesis (SHY). This paper argues that the paradigm conflates two functionally independent processes: (P1) reversible synaptic downscaling — spine shrinkage, AMPA receptor endocytosis, and proportional strength reduction — and (P2) glymphatic clearance of metabolic waste from the perisynaptic environment. We demonstrate that key experimental evidence supporting pruning-driven synapse elimination can be reinterpreted, without additional ontological entities, under the P1+P2 two-process model. We further propose a falsifiable experimental prediction: blocking glymphatic clearance (via AQP4 inhibition) while leaving downscaling intact will cause downscaled spines to undergo secondary, irreversible degeneration due to waste accumulation — converting reversible downscaling into permanent synapse loss. If confirmed, this would revise the textbook model of “forgetting as deletion” toward “forgetting as downscaled preservation,” with irreversible forgetting as a pathological consequence of clearance failure rather than a default neural operation. The model carries direct implications for early synaptic loss in Alzheimer’s disease, cognitive deficits following sleep deprivation, and theories of memory consolidation.