The hippocampus constructs a cognitive map from environmental sensory inputs, but the specific conditions that convert a quiescent hippocampal state into an active spatial representation have been unclear. This study asked how visual scenes of varying complexity recruit spatial maps in hippocampal subregions CA1 and CA3, and whether those subregions follow similar or distinct rules when sensory information accumulates.
The authors used a parametric virtual-reality paradigm in rats to vary visual scene complexity systematically while recording simultaneously from CA1 and CA3. The design permitted controlled accumulation of landmarks and manipulation of background textures, enabling direct comparison of how CA1 and CA3 place cells and network oscillations respond as scenes become richer.
CA1 place cells exhibited robust spatial selectivity even under sparse visual conditions. As landmarks were added, CA1 responses updated in a graded, continuous manner: place tuning adjusted progressively with each incremental increase in landmark information. This behavior indicates that CA1 functions as a continuous integrator of landmark inputs, tracking landmark accumulation and maintaining spatial tuning across a range of scene complexities.
By contrast, CA3 remained largely quiescent when landmark information was limited. Recruitment of CA3 place representations was not gradual; instead it was gated nonlinearly. CA3 activation emerged only after the scene reached a certain organization or complexity, consistent with a threshold-like mechanism that prevents engagement until sensory inputs form a coherent ensemble.
Importantly, CA3 recruitment depended on properties of the landmark ensemble beyond mere count. The authors report that ensemble density and configuration—the spatial arrangement and temporal coherence of landmarks—governed CA3 activation. This implies CA3 is sensitive to the integrated structure of a scene, rather than responding only to absolute numbers of cues.
Adding naturalistic background textures to the visual scenes facilitated earlier CA3 activation. Under these more complex, naturalistic conditions CA3 showed stepwise recruitment as scene complexity increased, suggesting that richer contextual information lowers the gating threshold or otherwise promotes the ensemble integration required for CA3 to construct scene representations.
Transitions in single-unit recruitment paralleled changes in local field potential (LFP) coupling. The authors describe a shift from a state of sensory decoupling—characterized by reduced fast-gamma coupling—to a state of global network engagement, marked by increased slow-gamma coupling. These LFP dynamics accompanied the move from sparse, decoupled sensory input toward the ensemble-integrated scenes that recruit CA3.
Together, the findings support a circuit-level framework in which CA1 and CA3 play complementary roles during scene-based map formation. CA1 operates as a continuous landmark integrator, providing immediate and progressive spatial tuning as cues accumulate. CA3 functions as a nonlinearly gated scene constructor, engaging only when landmarks cohere into a sufficiently dense and structured ensemble. The LFP gamma-band shifts provide a network-level correlate for the transition between sensory-driven and ensemble-driven states.
These distinctions have implications for understanding how the hippocampus balances sensitivity to incremental sensory information with stability and pattern completion mechanisms that may require integrated input. The differential rules in CA1 and CA3 could underlie distinct computational roles in navigation, memory encoding, and scene recognition.
This report is a preprint posted on bioRxiv and has not been certified by peer review. The study authors are Eun-Ho Lee and Inah Lee (Seoul National University). Funding sources declared include the National Research Foundation of Korea and Seoul National University Mid-Career Bridging Program. Specific experimental details such as sample sizes, statistical metrics, and full methodological parameters were reported in the original preprint; readers should consult the full manuscript and supplementary materials for those details. As a preprint, conclusions should be interpreted pending peer review and replication.