Neural population dynamics, manifolds, BCI learning, and representational drift in sensorimotor cortex
Seventeen papers, 2012–2026, every one filed against the same ten-section standard — so any two can be compared field by field.
📖 Browse online — https://howardbot.github.io/marmoset-hatlab-guide/
The site offers three views (library, timeline, reading path), filtering by subfield, lab lineage, subjects and priority, full-text search, deep-linkable papers, and light/dark themes.
Every paper is filed against the same ten sections, so any two can be compared field by field:
Metadata → Subjects → Recording & task → Central question → Methods → Key findings → Key concepts → Limitations → Place in the collection → Reading parameters
Field definitions are in TEMPLATE.md. Cross-paper synthesis is in ANALYSIS.md.
- 17 papers, 2012–2026
- Species: 13 macaque (34 individuals) · 2 mouse (8 individuals) · 2 human (4 individuals)
- Venue: 14 peer-reviewed · 3 preprints
- Labs: 16 across 25 institutions
Ordered by dependency rather than date:
-
Lay the foundation — Kaufman 2014 · Elsayed 2016
Build the intuition and the geometric vocabulary for output-null / orthogonal subspaces. Half the confusion in later papers disappears once these two are in place. -
Follow output-null outward — Perich 2018 · Semedo 2019
Kaufman's gate stops being just a gate. In Perich it becomes the substrate that carries motor learning; in Semedo the same geometry routes signals between cortical areas. This is the single thread that runs furthest through the collection. -
Define the manifold — Gallego 2018
The canonical source for 'neural manifold' and 'neural modes'. If you read only one paper to get the intuition, read this one. -
Ask whether the manifold is causal — Sadtler 2014 · Golub 2018 · Oby 2019
Three papers that must be read together. The conclusion moves from 'can't be learned' to 'can be learned, given days rather than hours.' -
Ask whether it is stable — Gallego 2020 · Rule 2020 · Rule & O'Leary 2022 · Clarke 2026
The stability-versus-drift tension, from both sides, plus the mechanism and the missing timescale. -
See whether it is useful — Degenhart 2020
If the manifold theory is right, it should be able to fix a real device. This is that test. -
Move to humans — Dekleva 2024 · Bougou 2025
Whether the framework built in monkeys survives contact with human intracortical recordings. -
Question all of it — Gao & Ganguli 2017 · Zach 2012
Is the low dimensionality you measured a property of the brain or of your task? And what did the single-neuron era already know?
Neural manifold & dimensionality
- Gallego 2018 — Cortical population activity within a preserved neural manifold underlies multiple motor behaviors ·
Essential
BCI learning
- Zach 2012 — Single Neurons in M1 and Premotor Cortex Directly Reflect Behavioral Interference ·
Background - Sadtler 2014 — Neural constraints on learning ·
Essential - Golub 2018 — Learning by neural reassociation ·
Essential - Oby 2019 — New neural activity patterns emerge with long-term learning ·
Essential
Motor preparation & output-null
- Kaufman 2014 — Cortical activity in the null space: permitting preparation without movement ·
Essential - Elsayed 2016 — Reorganization between preparatory and movement population responses in motor cortex ·
Essential - Perich 2018 — A Neural Population Mechanism for Rapid Learning ·
Recommended
Representational drift & stability
- Gallego 2020 — Long-term stability of cortical population dynamics underlying consistent behavior ·
Essential - Degenhart 2020 — Stabilization of a brain-computer interface via the alignment of low-dimensional spaces of neural activity ·
Recommended - Rule 2020 — Stable task information from an unstable neural population ·
Recommended - Rule & O'Leary 2022 — Self-healing codes: How stable neural populations can track continually reconfiguring neural representations ·
Recommended - Clarke 2026 — Neural modes in motor cortex cycle over fast timescales ·
Recommended
Inter-areal communication
- Semedo 2019 — Cortical Areas Interact through a Communication Subspace ·
Essential
Human iBCI & clinical
- Dekleva 2024 — Motor cortex retains and reorients neural dynamics during motor imagery ·
Essential - Bougou 2025 — Hierarchical and Context-Dependent Encoding of Actions in Human Posterior Parietal and Motor Cortex ·
Recommended
Theory & measurement
- Gao & Ganguli 2017 — A theory of multineuronal dimensionality, dynamics and measurement ·
Advanced
| Paper | Published | Lab | Institution | Subfield | Subjects | Brain areas | Priority | Difficulty |
|---|---|---|---|---|---|---|---|---|
| Zach 2012 | 2012-03-12 | Eilon Vaadia Lab | Hebrew University of Jerusalem (ELSC / ICNC) | BCI learning | Cynomolgus macaque (Macaca fascicularis) ×2 | M1 (primary motor cortex); PM (premotor cortex) | Background | 2/5 |
| Kaufman 2014 | 2014-03-01 | Krishna Shenoy Lab | Stanford University | Output-null | Rhesus macaque (Macaca mulatta) ×2 | M1 (primary motor cortex); caudal PMd (dorsal premotor cortex) | Essential | 3/5 |
| Sadtler 2014 | 2014-08-28 | Aaron Batista Lab + Byron Yu Lab | University of Pittsburgh | BCI learning | Rhesus macaque (Macaca mulatta) ×2 | M1 (primary motor cortex, proximal arm region) | Essential | 3/5 |
| Elsayed 2016 | 2016-10-27 | John Cunningham Lab + Mark Churchland Lab | Columbia University (Center for Theoretical Neuroscience, Zuckerman/Kavli) | Output-null | Rhesus macaque (Macaca mulatta) ×2 | M1 (primary motor cortex, hand/arm region); PMd (dorsal premotor cortex) | Essential | 4/5 |
| Gao & Ganguli 2017 | 2017-11-12 | Surya Ganguli Lab | Stanford University | Theory | Rhesus macaque (validation dataset) ×2 | PMd (dorsal premotor cortex); M1 (primary motor cortex) | Advanced | 5/5 |
| Golub 2018 | 2018-04-01 | Byron Yu Lab + Steven Chase Lab | Carnegie Mellon University | BCI learning | Rhesus macaque (Macaca mulatta) ×3 | M1 (primary motor cortex, proximal arm region) | Essential | 3/5 |
| Gallego 2018 | 2018-10-12 | Lee Miller Lab | Northwestern University | Manifold | Rhesus macaque (Macaca mulatta) ×3 | M1 (primary motor cortex, hand region) | Essential | 3/5 |
| Perich 2018 | 2018-11-21 | Lee Miller Lab | Northwestern University | Output-null | Rhesus macaque (Macaca mulatta) ×2 | M1 (primary motor cortex); PMd (dorsal premotor cortex) | Recommended | 4/5 |
| Semedo 2019 | 2019-04-03 | Adam Kohn + Byron Yu + Christian Machens Labs | Carnegie Mellon University | Communication | Macaque ×3 | V1 (primary visual cortex); V2 middle layers (24–37 neurons, mean 29.4) | Essential | 4/5 |
| Oby 2019 | 2019-07-23 | Aaron Batista + Steven Chase + Byron Yu Labs | University of Pittsburgh | BCI learning | Rhesus macaque (Macaca mulatta) ×2 | M1 (primary motor cortex, arm region) | Essential | 3/5 |
| Gallego 2020 | 2020-02-01 | Lee Miller Lab | Northwestern University | Drift | Rhesus macaque (Macaca mulatta) ×6 | PMd (dorsal premotor cortex); M1 (primary motor cortex) | Essential | 3/5 |
| Degenhart 2020 | 2020-07-01 | Byron Yu Lab | Carnegie Mellon University | Drift | Rhesus macaque (Macaca mulatta) ×2 | M1 (primary motor cortex, proximal arm region) | Recommended | 3/5 |
| Rule 2020 | 2020-07-14 | Timothy O'Leary Lab | University of Cambridge | Drift | Mouse (Mus musculus) ×4 | PPC (posterior parietal cortex) | Recommended | 3/5 |
| Rule & O'Leary 2022 | 2022-02-10 | Timothy O'Leary Lab | University of Cambridge (Engineering Department) | Drift | Mouse (modelled dataset) ×4 | PPC (posterior parietal cortex) | Recommended | 4/5 |
| Dekleva 2024 | 2024-04-01 | Jennifer Collinger Lab | University of Pittsburgh | Human iBCI | Human (Homo sapiens) ×2 | Motor cortex (hand and arm areas) | Essential | 3/5 |
| Bougou 2025 | 2025-11-13 | Richard Andersen Lab | California Institute of Technology | Human iBCI | Human (Homo sapiens) ×2 | MC (motor cortex); SPL (superior parietal lobule / posterior parietal cortex) | Recommended | 3/5 |
| Clarke 2026 | 2026-01-26 | Paul Nuyujukian Lab | Stanford University (Bioengineering, Neurosurgery, Wu Tsai Neurosciences Institute) | Drift | Rhesus macaque (Macaca mulatta) ×3 | Motor cortex | Recommended | 4/5 |
data/papers.json single source of truth — every structured field
build.py generator: JSON → cards + README + site
papers/*.md 17 standardised cards
docs/index.html the site (GitHub Pages serves from /docs)
TEMPLATE.md field definitions for the standard
ANALYSIS.md cross-paper synthesis
Append an entry to the papers array in data/papers.json (fields are documented in TEMPLATE.md). That is the only file you need to touch — reverse links are filled in automatically, and everything else is generated.
Committing data/papers.json alone is enough. A GitHub Action validates it, reruns build.py, and commits the regenerated cards, README tables and site — so you can add a paper from the GitHub web editor on any machine, with no local Python.
To preview locally before committing:
python3 build.pyNever hand-edit anything in papers/, docs/ or the generated parts of this README — the next build overwrites them.
The PDFs themselves are copyrighted and are not included in this repository. What is here is the reading notes and the metadata.