Abstract: Path integration (PI) is a crucial spatial navigation process that involves continuously integrating self-motion signals—such as vestibular, self-motion, and visual flow cues—to track one's position and orientation relative to a start location. Against the backdrop of global population aging, a key research question is whether changes in PI behavior and its neural substrates can serve as predictive biomarkers for early neurodegenerative conditions, particularly Alzheimer's disease (AD). Impairments in path integration (PI) ability have been demonstrated across cognitively normal older adults, individuals at-risk for AD, and patients with mild cognitive impairment (MCI) or AD, using a variety of experimental paradigms. The triangle completion task serves as a fundamental tool for assessing core PI ability. With advancements in virtual reality (VR) technology, this classic paradigm has evolved into more interactive VR-based tasks, such as the virtual supermarket task, Apple Games, and path estimation tasks. These paradigms differ significantly in the types of cues available to participants—ranging from body-based cues to visual cues—and in their environmental presentation modes, which include real walking, desktop-based VR, and immersive VR.
Utilizing these varied paradigms, distinct patterns of PI decline across the aging spectrum have been identified. In normal aging, PI impairment is primarily characterized by deficits in distance estimation when relying on a single sensory modality. This specific deficit is also evident in at-risk stages for AD. However, both at-risk individuals and those with prodromal AD exhibit significantly increased angular errors during PI. Notably, these pre-clinical groups largely retain the ability to compensate for their deficits when stable environmental cues are available. In contrast, pathological aging (MCI/AD) is characterized by a comprehensive deterioration in PI performance, involving severe inaccuracies in both distance and angular computations, as well as a significantly diminished capacity to utilize environmental cues for compensation. It is important to note that VR paradigms, by eliminating authentic body-based cues inherent in real walking, might accentuate the observed PI deficits. Furthermore, variability in the criteria used to define “at-risk” populations across studies complicates the interpretation of results and the understanding of pathological progression.
To elucidate the neural mechanisms underlying the aforementioned behavioral differences, this study begins by analyzing how the multi-level neural hierarchy operates synergistically during path integration. Compared to other navigation functions, PI underscores the synergistic operation of a multi-level neural hierarchy in processing and integrating self-motion information. At the cellular level, grid cells generate and update spatial representations by integrating information from speed and head-direction cells to continuously track the displacement vector. Place cells support this process by providing positional feedback and error correction. At the brain network level, the initial input and integration of self-motion cues rely on the vestibular system, posterior cingulate cortex (PCC), and retrosplenial cortex (RSC). The core computation for PI is dependent on the hippocampus and entorhinal cortex (EC), a process modulated by rhythmic input from the medial septum. The medial prefrontal cortex (mPFC) serves as an auxiliary region, collaborating with the medial temporal lobe to facilitate precise spatial updating and positioning based on self-motion.
The decline in PI ability is a common feature of both normal and pathological aging, closely linked to the vulnerability of its supporting core brain regions to the aging process. In normal aging, a key neural correlate of PI decline is impaired grid cell-based representations resulting from age-related degradation of the entorhinal-hippocampal circuit. In individuals at risk for AD, the RSC supports compensatory navigation when rich environmental cues are available, and reduced functional connectivity within relevant networks also contributes uniquely to PI deficits. In contrast, those with pathological aging exhibit structural atrophy and neuropathological changes that correspond to the widespread impairment in PI. Current research faces important limitations: the inability to clearly quantify differences between grid cell impairment due to AD pathology versus normal aging, and the lack of causal evidence regarding compensatory mechanisms mediated by posterior cortical regions. These gaps constrain a deeper understanding of the neural mechanisms underlying PI decline in aging and hinder its translation into clinically useful biomarkers.

Key words: path integration, spatial navigation, aging, Alzheimer's disease, neural mechanisms