Decoding hidden goal-directed navigational states and their neuronal representations using a novel labyrinth paradigm and probabilistic modeling framework

Goal-directed navigation involves a sequence of planned actions aimed at achieving long-term goals through reinforcement, but detecting hidden states that support this process and their neuronal substrates remains a fundamental challenge. To address this, we developed a complex labyrinth test that mimics naturalistic foraging and implemented a novel hierarchical probabilistic modeling framework, Cognitive Mapping of Planned Actions with State Spaces (CoMPASS). This framework infers a nested state structure, comprising short-term surveillance and ambulation states (Level 1) and long-term goal-oriented navigational states (Level 2). Using CoMPASS, we show that successful navigation in wild-type mice is marked by increased recruitment of both surveillance and goal-oriented states specifically at decision nodes, revealing how sequential behavioral decisions culminate in long-term goals. In contrast, the humanized App ^SAA mouse model of Alzheimer′s disease (AD) exhibited navigational impairments marked by diminished surveillance during decisions, reduced goal-directed states, and increased navigation stochasticity. Importantly, we show that gamma oscillations in the posterior parietal cortex (PPC), a region involved in spatial navigation planning, encode these CoMPASS behavioral states and their dynamic operating modes linking spatial locations to long-term goals. Our findings provide a novel paradigm for assessing hidden goal-directed navigational states and identify gamma oscillations in the PPC as their neural substrates.