A working feedback-based framework for goal-directed spatial memory deficits in Alzheimer's disease: Feedforward and feedback interactions along the RSC–MEC–CA1–RSC axis

Spatial navigation deficits are among the earliest and most clinically significant cognitive impairments in Alzheimer's disease (AD), particularly when navigation depends on goal-directed spatial memory. This review defines goal-directed spatial memory as a task-oriented construct and proposes a network-level framework for interpreting early AD navigation deficits. We synthesize anatomical, physiological, behavioral, and disease-related evidence concerning the retrosplenial cortex (RSC), medial entorhinal cortex (MEC), hippocampal CA1, and CA1-RSC feedback interactions. We propose the RSC-MEC-CA1-RSC axis as a testable working framework based on coordinated feedforward and feedback interactions. Within this system, the RSC integrates behaviorally relevant external cues and supports reference-frame transformation. The MEC contributes to path integration and self–goal relational coding, and CA1 consolidates these signals into functional goal-location representations. These representations can be retrieved, stabilized, and then transmitted back to cortical networks to guide continuous behavioral updating. Notably, we define the CA1-RSC pathway as a functional feedback route that supports iterative information updating, route correction and strategic adjustment. We argue that AD navigation deficits stem from the progressive breakdown of feedforward and feedback interactions across this axis, rather than from isolated dysfunction of single brain regions. The RSC-MEC-CA1-RSC framework offers a circuit-level mechanistic account for spatial disorientation in early AD. As a working model instead of a fully validated canonical circuit, it puts forward a series of testable hypotheses for future research, including cross-regional electrophysiological recordings, projection-specific circuit manipulation, imaging assessments, and differentiated navigation paradigms to examine cue use, path integration, goal retrieval and feedback-dependent updating.