Dynamic Thresholds and Spatial Boundaries Coordinate Memory Association and Discrimination Across Scales.

The ability to connect related experiences while preserving their individual identities is a central requirement of adaptive behavior. Engram overlap offers a cellular substrate for memory linking, but overlap must be regulated: too little prevents integration, whereas too much promotes generalization, interference, and confusion. Here, we organize evidence across molecular, cellular, glial, circuit, and systems levels around a common association-discrimination problem. We distinguish prospective linking, in which temporally adjacent experiences are biased toward shared neuronal allocation during a transient excitability window, from retrospective linking, in which memories separated in time are integrated through replay and cross-regional coordination. Within this framework, CREB- and CCR5-associated excitability states and inhibitory microcircuits set dynamic thresholds that regulate the timing and extent of neuronal recruitment; dendritic compartmentalization and branch-specific plasticity help preserve event-specific information within shared neurons; and representational drift may provide temporal decorrelation across repeated experiences. We further evaluate emerging evidence that learning-associated astrocyte ensembles, perisynaptic processes, calcium dynamics, and metabolic coupling help set the spatial boundaries of plasticity, while emphasizing that a direct role in defining engram boundaries remains a testable hypothesis. Together, these mechanisms form a Memory Engram Association Network that links neural organization to behavioral outcomes including integration, discrimination, generalization, and interference. This framework identifies experimentally separable mechanisms and shows how their disruption may contribute to age- and disease-related cognitive dysfunction. As this evidence is predominantly rodent-derived, we review the human literature on generalization, interference, and integration separately and frame the clinical extensions, including neurorehabilitation, as predictions awaiting human test.