🧠 neuroscience

Primary and Supplementary Motor cortex implement parallel control solutions for rhythmic and discrete arm movements

This study reconciles conflicting views on motor control by demonstrating that the primary motor cortex (M1) and supplementary motor area (SMA) implement parallel, distinct strategies—converging limit-cycle dynamics in M1 versus diverging, movement-specific helical spiral dynamics in SMA—to govern both rhythmic and discrete arm movements.

Original authors: Colins Rodriguez, A., Fuentes-Flores, R., Humphries, M. D.

Published 2026-02-17

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Colins Rodriguez, A., Fuentes-Flores, R., Humphries, M. D.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine your brain's motor cortex as a massive, high-tech control center for your arms. For decades, scientists have been arguing about how this control center handles two very different types of movements: rhythmic movements (like pedaling a bike or swimming) and discrete movements (like throwing a dart or picking up a cup).

The big question was: Does the brain use one master strategy for both, or does it switch to a completely different playbook depending on what you're doing?

This paper says: Yes, it does both. And it does them at the same time, using two different "departments" within the brain.

Here is the breakdown using simple analogies:

The Two Departments: M1 and SMA

Think of the motor cortex as a company with two main departments:

M1 (Primary Motor Cortex): The "Muscle Manager." This is the older, more direct department that talks to your muscles.
SMA (Supplementary Motor Area): The "Strategic Planner." This is a newer department that helps plan complex sequences and timing.

The Experiment: The Virtual Pedal

The researchers asked monkeys to pedal a virtual wheel.

Sometimes they had to pedal just half a turn (a discrete movement, like a quick push).
Sometimes they had to pedal seven full turns (a rhythmic movement, like riding a bike).
They recorded the electrical signals in the monkeys' brains while they did this.

The Discovery: Two Different Solutions

1. The Muscle Manager (M1) uses the "Same Strategy"

The Analogy: Imagine a Ferris Wheel.

Rhythmic Movement: When the monkey pedals 7 times, the brain's activity spins around and around the Ferris Wheel in a perfect circle. This is called a "limit cycle."
Discrete Movement: When the monkey only needs to pedal half a turn, the brain's activity still heads toward that same Ferris Wheel. It starts spinning, but a "stop signal" hits it just after half a turn, and it stops.

The Takeaway: M1 doesn't care if you are doing a short push or a long ride. It has one single "Ferris Wheel" engine. It just decides when to stop the ride. Whether it's a short trip or a long one, the engine is the same.

2. The Strategic Planner (SMA) uses "Different Strategies"

The Analogy: Imagine a Spiral Staircase.

Rhythmic Movement: If you need to go up 7 flights, the brain starts at the very bottom of the spiral and walks all the way up, circling around the center pole seven times.
Discrete Movement: If you only need to go up half a flight, the brain doesn't start at the bottom. Instead, it teleports you to the top of the staircase (or near the top) before you even start moving. You only have to take a tiny step to finish.

The Takeaway: The SMA is smart. It knows before the movement starts whether you need a long ride or a short one.

For a long ride, it starts the plan from the beginning.
For a short ride, it sets the plan up near the finish line so you don't waste time.
Because the starting points are so different, the brain uses two completely different "maps" for these two types of movements.

Why Does This Matter?

This solves a decades-old mystery.

Some scientists thought the brain used one strategy for everything.
Others thought it used two totally different strategies.

The Truth: It uses both, but in different places.

M1 is the "engine" that just runs the same loop (the Ferris Wheel) and stops when told. It's efficient and simple.
SMA is the "navigator" that figures out the timing and the starting point (the Spiral Staircase) based on the goal.

The Evolutionary Twist

The paper suggests this makes sense from an evolutionary perspective.

M1 is an older part of the brain (like the engine of a car). It evolved to handle rhythmic things like walking or running. It just keeps the rhythm going until told to stop.
SMA is a newer part of the brain (like the GPS). As animals evolved to do more complex, precise tasks (like throwing a spear), the brain added this "navigator" layer to tell the old engine exactly when to start and stop, allowing for those precise, one-off movements.

In a Nutshell

Your brain is a brilliant multitasker. When you move your arm, the Strategic Planner (SMA) decides how long the trip is and sets the starting point accordingly (using different maps). Then, it tells the Muscle Manager (M1) to just spin the wheel. Whether the trip is short or long, the wheel spins the same way; the only difference is when the Manager gets the order to "Stop."

This parallel system allows us to be both rhythmic (like a dancer) and precise (like a surgeon) at the same time.

1. Problem Statement

The motor cortex is responsible for generating a vast spectrum of arm movements, ranging from discrete (e.g., throwing a dart, distinct start/stop) to rhythmic (e.g., pedaling, continuous oscillation). A longstanding debate exists regarding how the cortex controls this spectrum:

Same-Control Hypothesis: The motor cortex uses a single, unified control strategy for both movement types. Evidence suggests similarities in Primary Motor Cortex (M1) population dynamics (e.g., rotational dynamics) across both types.
Different-Control Hypothesis: The cortex employs distinct strategies for discrete and rhythmic movements. Evidence suggests differences in the Supplementary Motor Area (SMA), where rhythmic movements show rotational dynamics while discrete movements show linear dynamics.

The core question is whether these opposing views represent mutually exclusive mechanisms or if the brain implements both solutions in parallel across different cortical regions.

2. Methodology

The authors combined electrophysiological recordings in non-human primates with Recurrent Neural Network (RNN) modeling to test these hypotheses.

Behavioral Task: Two monkeys performed an arm-pedaling task in a virtual environment. They controlled a pedal to reach a target, performing movements spanning the spectrum from 0.5 cycles (discrete, half-cycle) to 7 cycles (rhythmic).
- Conditions varied by direction (forward/backward), starting position (top/bottom), and number of cycles ( $N_{cycle}$ ).
- Visual cues indicated direction and distance (cycles) 0.5–1.0s before movement onset.
Data Collection: Simultaneous multi-unit recordings were obtained from M1, SMA, and arm muscles (EMG).
RNN Modeling Strategy:
- The authors trained two families of RNNs to predict neural/muscle dynamics under competing hypotheses:
  1. Same-Control RNNs: Trained only on rhythmic conditions ( $N_{cycle} > 1$ ). Discrete movements were generated solely by varying the timing of a "stop" signal.
  2. Different-Control RNNs: Trained on both ends of the spectrum ( $N_{cycle} = 0.5$ and $7$) with an explicit input indicating the movement type (discrete vs. rhythmic) via orthogonal weights.
- M1 Models: Trained to output low-dimensional muscle (EMG) trajectories.
- SMA Models: Trained to output low-dimensional M1 trajectories (since SMA projects to M1). These included a "temporal-context" input to track movement duration, necessary to replicate SMA's helical dynamics.
Analysis Techniques:
- Canonical Correlation Analysis (CCA): To compare similarity between RNN outputs and biological data.
- Limit Cycle Analysis: Quantifying how close neural trajectories approach a stable rotational attractor.
- Demixed PCA (dPCA): To identify if specific neural subspaces encode the "number of cycles" (movement type) during preparation.

3. Key Results

A. Primary Motor Cortex (M1): Same Control Strategy

Convergence to a Limit Cycle: M1 dynamics for both discrete and rhythmic movements converged to the same limit cycle (a stable rotational attractor).
Trajectory Behavior:
- Rhythmic movements: Neural trajectories entered the limit cycle and continued rotating.
- Discrete movements: Neural trajectories approached the limit cycle but were interrupted by a stop signal before completing a full rotation, then receded.
Preparation Phase: During the preparation phase (before movement onset), M1 neural trajectories did not separate based on the upcoming number of cycles. The type of movement (discrete vs. rhythmic) was not linearly decodable from M1 activity before the movement started.
Conclusion: M1 employs a single control strategy where the difference between movement types is determined by when the trajectory is stopped relative to the limit cycle.

B. Supplementary Motor Area (SMA): Different Control Strategies

Helical Dynamics: SMA dynamics followed a helical spiral structure, where the trajectory rotates around a central axis for each cycle performed.
Divergent Initial Conditions:
- Rhythmic movements: Trajectories started near the beginning of the helix (far from the end point).
- Discrete movements: Trajectories started near the end of the helix (equivalent to the last cycle of a rhythmic movement).
Preparation Phase: Unlike M1, SMA neural trajectories diverged significantly during the preparation phase based on the upcoming movement type. The "number of cycles" was explicitly encoded in the neural state before movement onset.
Conclusion: SMA employs distinct control strategies for discrete and rhythmic movements, setting different initial conditions on a shared helical manifold to dictate movement duration.

4. Key Contributions

Resolution of a Scientific Debate: The paper reconciles the "same-control" vs. "different-control" hypotheses by demonstrating that both are true, but implemented in parallel within different motor cortical regions.
Functional Specialization:
- M1 acts as a universal oscillator (limit cycle) that generates the movement pattern; the "stop" signal determines the movement type.
- SMA acts as a higher-order planner that sets the initial conditions (start point on the helix) to define the movement duration and type.
Methodological Advance: The study demonstrates the power of using RNNs to generate specific, testable predictions about neural dynamics (e.g., distance to attractors, separation during preparation) that cannot be distinguished by observing raw neural data alone.
Evolutionary Insight: The findings align with evolutionary theories where M1 (older) handles rhythmic locomotion, while SMA (newer) evolved to provide top-down timing signals for fine, discrete limb control.

5. Significance

Theoretical Impact: This work shifts the understanding of motor control from a "one-size-fits-all" or "completely separate" view to a hierarchical, parallel control architecture. It suggests that the brain solves the rhythmic-discrete spectrum problem by having a lower-level region (M1) handle the dynamics of movement generation, while a higher-level region (SMA) handles the timing and initiation parameters.
Clinical & Engineering Relevance:
- Neuroprosthetics: Understanding that M1 uses a limit cycle for all movements could simplify decoding algorithms for brain-computer interfaces (BCIs), focusing on detecting the "stop" signal rather than distinct movement classes.
- Parkinson's Disease: Since SMA is involved in movement initiation and timing, and M1 in execution, this model helps explain why different symptoms (e.g., freezing of gait vs. difficulty initiating discrete movements) might arise from specific regional dysfunctions.
- Robotics: The proposed architecture (a shared oscillator with variable initial conditions) offers a robust control solution for robots requiring both continuous and discrete motion.

In summary, the paper provides a unified framework where the motor cortex solves the problem of controlling the rhythmic-to-discrete spectrum by running parallel control solutions: M1 uses a unified limit-cycle strategy, while SMA uses distinct strategies to set the initial state of that system.