Sleep-dependent memory consolidation has been extensively studied over the pasttwo decades, yet a critical construct validity problem remains unaddressed: existingresearch conflates memory retention with crystallized intelligence (Gc). This paperargues that Gc, as defined within the Cattell-Horn-Carroll (CHC) framework, refers tothe flexible application of accumulated, integrated knowledge systems developed overyears, not short-term post-learning retention measurable over days or weeks. Weconduct a systematic review of four research pathways—REM sleep and memoryintegration, the default mode network (DMN) and sleep, DMN and Gc, and sleep andGc—and demonstrate that while individual links exist, no integrated REM → DMN→ Gc pathway has been empirically established. We identify three methodologicalweaknesses in the current literature: (1) brain lesion evidence for REM dispensabilityconfounded by neural plasticity, (2) conflation of memory retention with knowledgecrystallization as a construct, and (3) insufficient observational resolution of fMRIBOLD signals for adjudicating DMN function during REM. We propose the GarbageCollection (GC) Hypothesis, drawing an analogy from computer science: DMN"suppression" observed during REM may represent an active maintenance mode—pruning redundant connections, consolidating hub structures, and reorganizingsemantic network topology—rather than functional inactivity. This hypothesispredicts specific anomalous signatures in neural oscillation patterns (coherencedissolution, atypical frequency bursts, phase relationship inversions) that conventionaltemplate-based analyses would systematically miss. We argue that network theoryapproaches—semantic networks, complex network topology, and brain connectomeanalysis—offer a path toward formally defining the "crystallization" processindependent of current observational limitations.
Ryuhei ISHIBASHI (Thu,) studied this question.