OpenNovelty.org

From movement to cognitive maps: recurrent neural networks reveal how locomotor development shapes hippocampal spatial coding

ICLR 2026 Conference SubmissionAnonymous Authors
OpenReview Score: 6.5Download Report PDF
recurrent neural networkspatial representationshippocampusdevelopmentlocomotionrats
Abstract:

The hippocampus contains neurons whose firing correlates with an animal's location and orientation in space. Collectively, these neurons are held to support a cognitive map of the environment, enabling the recall of and navigation to specific locations. Although recent studies have characterised the timelines of spatial neuron development, no unifying mechanistic model has yet been proposed. Moreover, the processes driving the emergence of spatial representations in the hippocampus remain unclear (Tan et al., 2017). Here, we combine computational analysis of postnatal locomotor development with a recurrent neural network (RNN) model of hippocampal function to demonstrate how changes in movement statistics -- and the resulting sensory experiences -- shape the formation of spatial tuning. First, we identify distinct developmental stages in rat locomotion during open-field exploration using published experimental data. Then, we train shallow RNNs to predict upcoming visual stimuli from concurrent visual and vestibular inputs, exposing them to trajectories that reflect progressively maturing locomotor patterns. Our findings reveal that these changing movement statistics drive the sequential emergence of spatially tuned units, mirroring the developmental timeline observed in rats. The models generate testable predictions about how spatial tuning properties mature -- predictions we confirm through analysis of hippocampal recordings. Critically, we demonstrate that replicating the specific statistics of developmental locomotion -- rather than merely accelerating sensory change -- is essential for the emergence of an allocentric spatial representation. These results establish a mechanistic link between embodied sensorimotor experience and the ontogeny of hippocampal spatial neurons, with significant implications for neurodevelopmental research and predictive models of navigational brain circuits.

Disclaimer
This report is AI-GENERATED using Large Language Models and WisPaper (A scholar search engine). It analyzes academic papers' tasks and contributions against retrieved prior work. While this system identifies POTENTIAL overlaps and novel directions, ITS COVERAGE IS NOT EXHAUSTIVE AND JUDGMENTS ARE APPROXIMATE. These results are intended to assist human reviewers and SHOULD NOT be relied upon as a definitive verdict on novelty.
NOTE that some papers exist in multiple, slightly different versions (e.g., with different titles or URLs). The system may retrieve several versions of the same underlying work. The current automated pipeline does not reliably align or distinguish these cases, so human reviewers will need to disambiguate them manually.
If you have any questions, please contact: mingzhang23@m.fudan.edu.cn

Overview

Overall Novelty Assessment

The paper proposes a computational framework linking postnatal locomotor development to hippocampal spatial tuning emergence via recurrent neural networks trained on predictive coding. It occupies the 'Recurrent Neural Network Models of Spatial Tuning Emergence' leaf, which currently contains only this work among 28 papers across the taxonomy. This leaf sits within the broader 'Computational and Mechanistic Models' branch, which includes two leaves total. The sparse population of this specific modeling approach suggests the paper addresses a relatively underexplored mechanistic niche within the developmental spatial coding literature.

The taxonomy reveals substantial empirical work on developmental trajectories (three leaves covering postnatal emergence, spatial behavior ontogeny, and human infant studies) and diverse mechanistic branches examining self-motion signals, learning-dependent plasticity, and subcortical regulation. The paper's closest conceptual neighbors appear in 'Experience-Dependent Sculpting of Hippocampal Circuits' (two papers) and 'Vestibular and Otolith Contributions' (one paper), both emphasizing sensorimotor experience. However, the RNN-based predictive coding approach distinguishes this work from geometric constraint models or empirical developmental characterizations, positioning it at the intersection of computational neuroscience and developmental systems.

Among 16 candidates examined across three contributions, the analysis identified one refutable pair for the directional selectivity prediction (6 candidates examined, 1 refutable). The locomotor development classification contribution showed no refutations across 10 candidates, while the core RNN modeling contribution was not matched against any candidates. The limited search scope (top-K semantic retrieval) means these statistics reflect a focused sample rather than exhaustive coverage. The directional selectivity finding appears to have some prior empirical grounding, whereas the computational classification of locomotor stages and the RNN framework itself show less direct overlap within the examined literature.

Given the restricted search scale and the paper's position in a sparsely populated taxonomy leaf, the work appears to occupy a distinctive methodological space combining developmental locomotor analysis with predictive RNN modeling. The analysis does not cover potential overlaps in broader machine learning or theoretical neuroscience literatures outside the hippocampal spatial coding domain. The contribution-level statistics suggest varying degrees of novelty, with the core modeling framework showing minimal direct precedent among examined candidates.

Taxonomy

Core-task Taxonomy Papers
28
3
Claimed Contributions
16
Contribution Candidate Papers Compared
1
Refutable Paper

Research Landscape Overview

Core task: developmental emergence of hippocampal spatial representations from locomotor experience. The field investigates how spatial coding in the hippocampus arises through experience and maturation, spanning computational modeling, empirical developmental studies, and mechanistic analyses of sensory and network contributions. Computational and mechanistic models explore how recurrent neural networks and self-organizing principles can generate place cells and grid cells from movement statistics, with works like Euclidean Geometry Sculpts[3] and Self-motion Spatial Metric[4] emphasizing geometric constraints and self-motion signals. Empirical developmental trajectories track the maturation of spatial neurons and navigation behavior across postnatal development, as seen in studies such as Development Sculpts Hippocampus[1] and Preweanling Cognitive Map[7]. Additional branches examine vestibular and self-motion inputs, learning-dependent plasticity in task-specific contexts, subcortical-cortical network interactions, cross-species comparisons including primate work like Primate Free Navigation[8], and methodological frameworks for recording and analyzing spatial coding. Particularly active lines of work contrast bottom-up sensory-driven models with top-down network-level accounts of spatial tuning emergence. Some studies focus on how locomotor action sequences and movement statistics directly shape hippocampal representations, as in Locomotor Action Sequences[5], while others emphasize the role of cortical and subcortical circuits in sculpting these codes through recurrent dynamics and plasticity. Movement to Cognitive Maps[0] sits within the computational modeling branch, specifically addressing how recurrent neural network architectures can learn spatial tuning from locomotor experience. Its emphasis on mechanistic emergence from movement statistics aligns closely with Self-motion Spatial Metric[4] and Euclidean Geometry Sculpts[3], which similarly propose that geometric and kinematic features of exploration drive the formation of cognitive maps. Compared to purely empirical developmental studies like Development Sculpts Hippocampus[1], Movement to Cognitive Maps[0] offers a normative computational account of the principles underlying observed maturational trajectories.

Claimed Contributions

Computational classification of postnatal locomotor development stages in rats

The authors introduce a computational method to identify distinct developmental stages in rat locomotion (crawl, walk, run) by analyzing movement statistics from experimental data. This classification provides the foundation for generating synthetic trajectories used to train RNNs at each developmental stage.

10 retrieved papers
RNN model demonstrating how locomotor development drives spatial cell emergence

The authors develop a recurrent neural network trained on visual prediction tasks using trajectories corresponding to developmental locomotion stages. The model recapitulates the sequential emergence of spatially tuned units mirroring biological developmental timelines observed in rats.

0 retrieved papers
Novel prediction of directional selectivity emergence in place cells validated experimentally

The model predicts that directional selectivity in hippocampus emerges primarily through conjunctive place-direction coding rather than pure head direction cells. The authors confirm this previously unreported developmental increase in conjunctive cells through analysis of CA1 recordings.

6 retrieved papers
Can Refute

Core Task Comparisons

Comparisons with papers in the same taxonomy category

Within the taxonomy built over the current TopK core-task papers, the original paper is assigned to a leaf with no direct siblings and no cousin branches under the same grandparent topic. In this retrieved landscape, it appears structurally isolated, which is one partial signal of novelty, but still constrained by search coverage and taxonomy granularity.

Contribution Analysis

Detailed comparisons for each claimed contribution

Contribution

Computational classification of postnatal locomotor development stages in rats

The authors introduce a computational method to identify distinct developmental stages in rat locomotion (crawl, walk, run) by analyzing movement statistics from experimental data. This classification provides the foundation for generating synthetic trajectories used to train RNNs at each developmental stage.

Contribution

RNN model demonstrating how locomotor development drives spatial cell emergence

The authors develop a recurrent neural network trained on visual prediction tasks using trajectories corresponding to developmental locomotion stages. The model recapitulates the sequential emergence of spatially tuned units mirroring biological developmental timelines observed in rats.

Contribution

Novel prediction of directional selectivity emergence in place cells validated experimentally

The model predicts that directional selectivity in hippocampus emerges primarily through conjunctive place-direction coding rather than pure head direction cells. The authors confirm this previously unreported developmental increase in conjunctive cells through analysis of CA1 recordings.

Table of Contents

From movement to cognitive maps: recurrent neural networks reveal how locomotor development shapes hippocampal spatial coding | Novelty Validation