CoChat

Space is a latent sequence: A theory of the hippocampus

Open access

The DOI matches its Crossref record; 64 of the 64 references it cites were found in a registry or an index; 44 of the 72 sources its claims cite were checked for support.

Abstract

Fascinating phenomena such as landmark vector cells and splitter cells are frequently discovered in the hippocampus. Without a unifying principle, each experiment seemingly uncovers new anomalies or coding types. Here, we provide a unifying principle that the mental representation of space is an emergent property of latent higher-order sequence learning. Treating space as a sequence resolves numerous phenomena and suggests that the place field mapping methodology that interprets sequential neuronal responses in Euclidean terms might itself be a source of anomalies. Our model, clone-structured causal graph (CSCG), employs higher-order graph scaffolding to learn latent representations by mapping aliased egocentric sensory inputs to unique contexts. Learning to compress sequential and episodic experiences using CSCGs yields allocentric cognitive maps that are suitable for planning, introspection, consolidation, and abstraction. By explicating the role of Euclidean place field mapping and demonstrating how latent sequential representations unify myriad observed phenomena, our work positions the hippocampus in a sequence-centric paradigm, challenging the prevailing space-centric view.

Overview

CoChat Summary

TL;DR

The paper proposes that hippocampal spatial representations emerge from latent higher-order sequence learning, not from an explicit Euclidean map. Its clone-structured causal graph (CSCG) learns allocentric cognitive maps from aliased egocentric sensory-motor sequences and reproduces many place-cell remapping phenomena, suggesting Euclidean place-field mapping may itself generate apparent anomalies.

Key findings

  • CSCG learns latent topologies from aliased egocentric observations in 2D and 3D simulated environments without Euclidean assumptions; in a uniform room it discovered four clones per spatial location, corresponding to four possible headings.
  • It can stitch temporally disjoint but spatially overlapping episodes into a global map, represent multiple nonoverlapping maps in one model, and reuse a learned transition graph as a schema to infer a shortest path through an unvisited room, including an elevated platform accessed via a ramp.
  • CSCG reproduces diverse hippocampal phenomena: geometry-dependent remapping (e.g., clone 314 anchored to the top-left corner; clones 161 and 748 split in elongated rooms), cue rotation, barrier disruption, landmark vector remapping, linear-track distance coding on an 18-step track, direction-sensitive place fields, event-specific rate remapping/lap cells (three-lap training vs. four-lap test), lack of remapping with locked doors, place field repetition in identical rooms, hairpin-maze direction-dependent repetition, and place field expansion with room size (side lengths 7, 9, and 11).
  • A testable prediction: place field expansion depends on the uniqueness of visual context rather than the rate of visual change; a checkerboard room produced more expanded place fields than a random-pattern room.

Methods

  • Model: CSCG, an action-augmented cloned hidden Markov model with an overcomplete latent space. Multiple latent states (“clones”) can emit the same observation, allowing sequential contexts to be split or merged. Learning uses expectation-maximization with forward-backward message passing, followed by Viterbi EM to consolidate superpositions; a pseudocount smoothing parameter is used.
  • Continuous observations: K-means is run on training data; clones are assigned to centroids proportional to cluster priors; data are vector-quantized; then the action-conditional transition tensor is learned as in the discrete CSCG. For transfer, the transition matrix is fixed while the emission matrix is learned.
  • Simulations: agents perform random walks in 2D/3D simulated environments with egocentric actions and aliased observations. Place fields are computed as average clone activation per ground-truth spatial location over trials and time steps. No animal data are analyzed.

Limitations

  • Stated: the authors selected a subset of results they judged representative; they did not test remapping with new objects; replay-based planning and schemas were only briefly touched; animals develop place fields with considerably less experience than required; current models are learned using random walks; graph modifications for environmental changes were manually made by the experimenter; the lap-running experiment required multiple random initializations; and neuronal implementation is not addressed.
  • Stated: the work focuses on computational and algorithmic aspects of sequence learning, not on actual implementation in neuronal networks; biological plausibility is argued but not implemented.
  • Evident from the text: all reported results are simulation/modeling studies, not empirical hippocampal recordings; comparisons to experimental phenomena are qualitative reproductions rather than direct neural-data fits; active learning, reward mechanisms, and automatic dynamic graph modification remain untested.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1126/sciadv.adm8470

Cites

64 of 64 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1. Clone-structured cognitive graph. (A) Learning cognitive maps from sequential sensory observations is challenging because observations do not identify locations uniquely. (B) The cognitive map learning problem can be understood as learning a latent graph from observations emitted at every node, where two different nodes can emit the same observation. The challenge is to learn context-specific representations that will disambiguate sensory observations in the latent space. The observation D occurs in three different contexts in sequences A → D → E (purple), B → D → F (green), and C → D → G (orange) from the environment, a distinction that is not represented in a first-order Markov model. Two of these contexts (purple and green) correspond to the same latent state, and the third (orange) to a different latent state. Cloning D into multiple latent states allows for flexible merging and splitting of contexts as appropriate. (C) The cloning structure of dynamic Markov coding can be incorporated in an HMM with a structured emission matrix, the cloned HMM. CSCG extends cloned HMMs by including actions. (D) CSCG learns an allocentric map from aliased egocentric local observations in a 2D room with uniform interiors even with long runs of the same observation (i). (ii) Each unique sensation, indexed by color, is attached to a set of latent states (clones) through the emission matrix. Through learning of the transition matrix, these clones learn to represent different temporal contexts of that sensation. (iii) Learned transition graph among clones. Each clone’s color represents the observation it is attached to. The red arrows highlight the observations corresponding to the clones in the top-left corner of the graph. (iv) Clone activations as the agent navigates the room can be used to compute their place fields, which reveal the spatial locations they represent.
Fig. 1. Clone-structured cognitive graph. (A) Learning cognitive maps from sequential sensory observations is challenging because observations do not identify locations uniquely. (B) The cognitive map learning problem can be understood as learning a latent graph from observations emitted at every node, where two different nodes can emit the same observation. The challenge is to learn context-specific representations that will disambiguate sensory observations in the latent space. The observation D occurs in three different contexts in sequences A → D → E (purple), B → D → F (green), and C → D → G (orange) from the environment, a distinction that is not represented in a first-order Markov model. Two of these contexts (purple and green) correspond to the same latent state, and the third (orange) to a different latent state. Cloning D into multiple latent states allows for flexible merging and splitting of contexts as appropriate. (C) The cloning structure of dynamic Markov coding can be incorporated in an HMM with a structured emission matrix, the cloned HMM. CSCG extends cloned HMMs by including actions. (D) CSCG learns an allocentric map from aliased egocentric local observations in a 2D room with uniform interiors even with long runs of the same observation (i). (ii) Each unique sensation, indexed by color, is attached to a set of latent states (clones) through the emission matrix. Through learning of the transition matrix, these clones learn to represent different temporal contexts of that sensation. (iii) Learned transition graph among clones. Each clone’s color represents the observation it is attached to. The red arrows highlight the observations corresponding to the clones in the top-left corner of the graph. (iv) Clone activations as the agent navigates the room can be used to compute their place fields, which reveal the spatial locations they represent.CC BY-NC · PubMed Central
Fig. 2. CSCGs learn diverse latent topologies, transitively stitch them, and transfer structure to new environments. (A) CSCG learning in an example 3D environment. The agent navigates the environment with egocentric actions and gets observations as RGB images. The images are passed through a vector quantizer to obtain cluster indices, which are used as observations for training a CSCG. The learned transition graph reflects the topology of the environment, which is a square room with a plus-shaped barrier in the center. (B) An agent experiences an environment composed of four connected 3D rooms in disjoint sequential episodes. Each episode has only partial coverage of the locations in the composite room. CSCG learning stitches together the disjoint experiences into a coherent global map. (C) An agent experiences multiple sequential episodes sampled from four nonoverlapping 3D rooms. In this case, CSCG learning correctly learns separate maps for each maze. (D) The learned transition graph of a CSCG trained on one environment can be considered a reusable schema. Given partial experience in an unfamiliar environment, a differently colored variation of the four-room environment used in (B), the previously learned joint transition model can be used to identify the agent’s location within it and rapidly navigate around obstacles to find the shortest path to the goal (here, to return to the starting location).
Fig. 2. CSCGs learn diverse latent topologies, transitively stitch them, and transfer structure to new environments. (A) CSCG learning in an example 3D environment. The agent navigates the environment with egocentric actions and gets observations as RGB images. The images are passed through a vector quantizer to obtain cluster indices, which are used as observations for training a CSCG. The learned transition graph reflects the topology of the environment, which is a square room with a plus-shaped barrier in the center. (B) An agent experiences an environment composed of four connected 3D rooms in disjoint sequential episodes. Each episode has only partial coverage of the locations in the composite room. CSCG learning stitches together the disjoint experiences into a coherent global map. (C) An agent experiences multiple sequential episodes sampled from four nonoverlapping 3D rooms. In this case, CSCG learning correctly learns separate maps for each maze. (D) The learned transition graph of a CSCG trained on one environment can be considered a reusable schema. Given partial experience in an unfamiliar environment, a differently colored variation of the four-room environment used in (B), the previously learned joint transition model can be used to identify the agent’s location within it and rapidly navigate around obstacles to find the shortest path to the goal (here, to return to the starting location).CC BY-NC · PubMed Central
Fig. 3. CSCG reproduces several place field remapping phenomena. (A) A CSCG was first trained in a small square (SS) room (i). (ii) Place fields were computed in the SS, horizontal rectangle (HR), vertical rectangle (VR), and large square (LS) rooms. Notably, clone 314 maintained a consistent place field anchored to the top-left corner, while clones 161 and 748 exhibited field splitting when the room is elongated along the horizontal and vertical axes, respectively. (iii) The CSCG also replicated directional place fields reported in (27). (B) When trained in a circular room with a cue, the CSCG demonstrated that place fields also rotate when the cue is rotated. For most clones, the place fields disappear when a barrier is introduced. (C) Trained in a rectangular layout with a landmark, the CSCG exhibited place fields with two components in the modified layout: one at the original location and another at a vector displacement from the new landmark location. (D) A CSCG was trained on a linear track using outbound and inbound walks. Place fields were computed using trials with different starting positions for the outbound trajectories and different end positions for the inbound trajectories. Most clones coded distance from the start box, while others were anchored to the end box. (E) A CSCG was trained in a rectangular maze similar to (8), with training trials comprising three laps followed by a reward at the end. During test trials, the reward was shown at the end of four laps. Place fields computed with training trials show that there are different clones that are maximally active for different laps. In test trials, lap three clones are substantially active in both the third and fourth laps, reflecting the absence of a reward in the third lap.
Fig. 3. CSCG reproduces several place field remapping phenomena. (A) A CSCG was first trained in a small square (SS) room (i). (ii) Place fields were computed in the SS, horizontal rectangle (HR), vertical rectangle (VR), and large square (LS) rooms. Notably, clone 314 maintained a consistent place field anchored to the top-left corner, while clones 161 and 748 exhibited field splitting when the room is elongated along the horizontal and vertical axes, respectively. (iii) The CSCG also replicated directional place fields reported in (27). (B) When trained in a circular room with a cue, the CSCG demonstrated that place fields also rotate when the cue is rotated. For most clones, the place fields disappear when a barrier is introduced. (C) Trained in a rectangular layout with a landmark, the CSCG exhibited place fields with two components in the modified layout: one at the original location and another at a vector displacement from the new landmark location. (D) A CSCG was trained on a linear track using outbound and inbound walks. Place fields were computed using trials with different starting positions for the outbound trajectories and different end positions for the inbound trajectories. Most clones coded distance from the start box, while others were anchored to the end box. (E) A CSCG was trained in a rectangular maze similar to (8), with training trials comprising three laps followed by a reward at the end. During test trials, the reward was shown at the end of four laps. Place fields computed with training trials show that there are different clones that are maximally active for different laps. In test trials, lap three clones are substantially active in both the third and fourth laps, reflecting the absence of a reward in the third lap.CC BY-NC · PubMed Central
Fig. 4. CSCG reproduces various observations about place cells such as place field repetition, size, and shape variations. (A) A CSCG trained in a layout with interconnected square rooms showed consistent place fields across settings with locked doors. (B) A CSCG was first trained on a layout comprising two visually identical rooms in the same orientation connected by a corridor. Place fields were computed in a layout where the two rooms were rotated such that their orientations differed by 180. We also considered a second modification, where we introduced an asymmetry in the connection between the two rooms. In all three layouts, we observed place field repetition across the two rooms, in contrast to the findings in (30), where place field repetition was observed only in the same orientation layout. (C) However, when we retrained the CSCG in the layout with asymmetric connectivity, we observed that the place field repetition disappeared. (D) CSCG reproduces direction-dependent place field repetition in a hairpin maze as observed in (30). (E) We trained CSCGs on square rooms with uniform interiors of three different sizes. We observed that place fields at the edge and center of the room elongate and enlarge as the room size increases, while place fields anchored to a corner remain consistent across sizes. (F) A CSCG trained on the checkerboard room has more expanded place fields compared to one trained on a room with a random pattern. (G) In a room elongation experiment, we initially trained a CSCG on a rectangular room with local landmarks on one side. Place fields were computed in both the training room and an elongated room where landmarks remained in the same position. Place fields anchored to the landmarks remained consistent in both layouts, while those farther from the landmarks expanded in the elongated layout.
Fig. 4. CSCG reproduces various observations about place cells such as place field repetition, size, and shape variations. (A) A CSCG trained in a layout with interconnected square rooms showed consistent place fields across settings with locked doors. (B) A CSCG was first trained on a layout comprising two visually identical rooms in the same orientation connected by a corridor. Place fields were computed in a layout where the two rooms were rotated such that their orientations differed by 180. We also considered a second modification, where we introduced an asymmetry in the connection between the two rooms. In all three layouts, we observed place field repetition across the two rooms, in contrast to the findings in (30), where place field repetition was observed only in the same orientation layout. (C) However, when we retrained the CSCG in the layout with asymmetric connectivity, we observed that the place field repetition disappeared. (D) CSCG reproduces direction-dependent place field repetition in a hairpin maze as observed in (30). (E) We trained CSCGs on square rooms with uniform interiors of three different sizes. We observed that place fields at the edge and center of the room elongate and enlarge as the room size increases, while place fields anchored to a corner remain consistent across sizes. (F) A CSCG trained on the checkerboard room has more expanded place fields compared to one trained on a room with a random pattern. (G) In a room elongation experiment, we initially trained a CSCG on a rectangular room with local landmarks on one side. Place fields were computed in both the training room and an elongated room where landmarks remained in the same position. Place fields anchored to the landmarks remained consistent in both layouts, while those farther from the landmarks expanded in the elongated layout.CC BY-NC · PubMed Central

Verification

Claims the paper makes

30 Valid · 5 Not valid · 18 Not checked · 148 Own results and interpretations

129 more claims · 13 with their cited sources checked

Create a free account to see all 201 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 48
  • Probably not a substantial claim. 28

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

4 of 4

Citations

Cites

Research works printed but not linked: 5

  1. L. Chrisman AAAI (Citeseer 1992) vol. 1992 pp. 183–188.
  2. V. Sharan S. M. Kakade P. S. Liang G. Valiant “Learning overcomplete hmms” in Advances in Neural Information Processing Systems (2017) pp. 940–949.
  3. J. Pearl Causality (Cambridge Univ. Press ed. 2 2013).
  4. J. Peters D. Janzing B. Schölkopf Elements of Causal Inference: Foundations and Learning Algorithms (The MIT Press 2017).
  5. D. Eaton K. Murphy Artificial Intelligence and Statistics (PMLR 2007) pp. 107–114.

Cited by

See all 27

Metadata

Authors
Raju, R. V., Guntupalli, J. S., Zhou, G., Wendelken, C., Lázaro-Gredilla, M., & George, D.
Affiliations
  • Google (United States) (US)
  • Google DeepMind, Mountain View, CA 94043, USA.
Venue
Science Advances
Published
2024-08-02
Volume
10
Issue
31
Pages
eadm8470-eadm8470
ISSN
2375-2548 (ISSN-L)
OpenAlex
W4401169058
Cited by
25
References
63
Citations by year
2026: 11 · 2025: 8 · 2024: 6
FWCI
6.06
Type
journal article
Language
en
Open access
Licence
cc-by-nc
Locations
  • Science Advances (publishedVersion, cc-by-nc)
  • PubMed Central (publishedVersion, cc-by-nc)
  • PubMed (publishedVersion)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Memory and Neural Mechanisms
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Spatial Cognition and Navigation
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural dynamics and brain function
Keywords
hippocampus, mental representation of space, sequence learning, place fields, episodic memory, memory consolidation, self-reflection, spatial planning
MeSH terms
Animals, Hippocampus, Humans, Learning, Models, Neurological, Neurons, Space Perception

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-01, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-01 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.