Clover-like object vector representations support human spatial cognition

Marcia Bécu, Ivan Markel Krasovec, Pearl Saldanha, Jonathan R. Whitlock, Christian F. DoellerbioRxiv 2025 · DOI

Abstract

Vector-based spatial coding has been demonstrated in rodents, yet its neural basis in humans and its contribution to navigation remain poorly understood. Using a novel spatial updating task, we show that egocentric directional signals peak when objects are located behind the navigator, while distance signals emerge only when objects are out of view, suggesting a mnemonic role for vision-independent spatial mapping. Allocentric signals form a clover-shaped, four-axis pattern aligned to visual features, with improved navigation accuracy along its axes. Parallel rodent recordings further show the clover pattern arises from averaged activity of allocentric vectorial neurons, suggesting conserved cross-species mechanisms. Together, our findings uncover vector-based representations in the human brain, potentially serving as a neural reference axis to anchor objects to internal maps and support navigation.

The Challenge of Mapping Human Spatial Vectors

How does the human brain track the location of objects once they are out of sight? While rodent studies have identified "object-vector cells"—neurons that fire based on the distance and direction to a specific landmark—observing these mechanisms in humans has been difficult. Standard fMRI techniques often struggle to detect these signals because if every direction were represented equally, the signals would cancel each other out at a macroscopic level. This study investigates whether human spatial representations are non-uniform, prioritizing specific directions to support navigation.

Methodology and Experimental Design

To bridge the gap between rodent electrophysiology and human neuroimaging, researchers employed a cross-species approach. In the human component, 47 participants performed a spatial updating task within a virtual reality (VR) environment. Participants navigated an arena containing a single column, which would eventually disappear. After being passively moved to a new location, they were required to point back to the invisible object's position, forcing a reliance on internal vector representations.

Figure 2: Experimental setup including the VR environment, task flow, and definitions of egocentric vs. allocentric responses.

High-resolution 7T fMRI was used to monitor brain activity in regions like the Retrosplenial Cortex (RSC) and Parahippocampal Cortex (PHC). Simultaneously, single-unit recordings were taken from the RSC of rats foraging in a similar environment to validate the neural patterns across species.

Discovery of the "Clover" Neural Code

The study revealed that allocentric (world-centered) directional signals in the RSC and PHC are not uniform. Instead, they form a distinct four-axis "clover" pattern, where neural activity peaks at four orthogonal angles relative to the object. Interestingly, egocentric (self-centered) distance signals only emerged when the object was out of view, suggesting these regions act as a "mnemonic buffer" that maintains a spatial record only when visual input is absent.

Furthermore, egocentric directional signals peaked when the object was directly behind the navigator. This "behind bias" suggests the brain allocates more resources to representing space that is currently outside the immediate visual field.

Functional Impact on Navigation

These neural patterns are not merely incidental; they have a direct impact on behavioral performance. The researchers found that participants were significantly more accurate at pointing to the target when their trajectory aligned with one of the four axes of the clover pattern.

Figure 6: Comparison of pointing error (in degrees) between egocentric and allocentric conditions.

As shown in the behavioral data, navigation error was lower on the clover axes, confirming that this neural structure serves as a functional reference frame. The discovery of a similar population-level clover pattern in rodents suggests that this is a fundamental, conserved mechanism for mammalian navigation.

Conclusion and Future Directions

This research uncovers a previously unknown "clover-shaped" neural code that anchors objects to internal maps. While the four-axis symmetry is clear, its exact origin—and why it adopts a four-axis rather than a hexagonal structure—remains a subject for future study. These findings suggest that vectorial codes may support broader cognitive functions, potentially mapping abstract concepts or episodic memories in addition to physical space.

Summary updated

Abstract

Vector-based spatial coding has been demonstrated in rodents, yet its neural basis in humans and its contribution to navigation remain poorly understood. Using a novel spatial updating task, we show that egocentric directional signals peak when objects are located behind the navigator, while distance signals emerge only when objects are out of view, suggesting a mnemonic role for vision-independent spatial mapping. Allocentric signals form a clover-shaped, four-axis pattern aligned to visual features, with improved navigation accuracy along its axes. Parallel rodent recordings further show the clover pattern arises from averaged activity of allocentric vectorial neurons, suggesting conserved cross-species mechanisms. Together, our findings uncover vector-based representations in the human brain, potentially serving as a neural reference axis to anchor objects to internal maps and support navigation.
BibTeX
@misc{becu2025clover,
  author = {Marcia Bécu and Ivan Markel Krasovec and Pearl Saldanha and Jonathan R. Whitlock and Christian F. Doeller},
  title = {{Clover-like object vector representations support human spatial cognition}},
  year = {2025},
  doi = {10.1101/2025.04.18.649400},
  howpublished = {bioRxiv},
}
Figure