Despite being one of the earliest areas of cortex ever explored, the motor cortex is a notoriously difficult brain region to make sense of. The history of its study has been marred by numerous debates and dead ends. One of the biggest, the ‘kinetics vs. kinematics’ debate, paved the way for the modern approach to studying the motor system.
The debate started in favor of kinetics, the area of physics that deals with the causes of motion. In the 1960s, Edward Evarts applied a new technology to the motor cortex: He used an electrode to eavesdrop on an animal’s motor cortical neurons while the animal was moving. Specifically, he recorded from the motor cortex of monkeys as they moved a bar back and forth with their wrist. By doing so, he found neurons whose activity correlated with the amount of force — a kinetic variable — needed to generate the movement.
In the years that followed, however, scientists found this relationship inconsistent. During more complex movements in particular, the amount of force that a muscle produces changes as the joints and muscles around it do. This makes the chain of influence that runs from neural activity to muscle activity to force less interpretable and the kinetic theory less viable. Moreover, plenty of neurons seemed to care very little about force at all.
Enter Apostolos Georgopoulos, a Greek-born professor of neuroscience at Johns Hopkins University. Based on his doubts that the motor cortex could be understood through isolated movements, Georgopoulos abandoned the simple wrist-flick task of Evarts and recorded from neurons as animals completed full arm reaches.
He also doubted that neurons were signaling such straightforward things as the force. So rather than asking what neurons were saying about the muscles, Georgopoulos asked what they were saying about the movement. In more than a third of the neurons in the motor cortex, he found a very clear and simple relationship between neural activity and the direction the arm was moving. The finding of this ‘direction tuning’ implied that the motor cortex cared more about kinematics than kinetics. Kinematics are descriptive features of motion, defined without regard to the forces that generate them. Kinematic variables therefore indicate the desired outcome for the arm, but not the instructions for how to create it.
Moving from a model where the motor cortex encodes kinetics to one where it encodes kinematics shifts the distribution of labor in the motor system. A kinetic variable, such as force, is only a few small calculations removed from the actual level of muscle activity needed to enact it — transformations potentially performed by neurons in the spinal cord. Because kinematic values only define where the arm should be in space, however, they pose a larger challenge to the rest of the motor system. The burden is thus on areas downstream of the motor cortex to take a set of desired locations external to the body and turn it into patterns of muscle activity. Georgopoulos became an adamant and unceasing defender of this kinematic take on the motor cortex for decades to come.
Based on this finding, Georgopoulos also changed the way he looked at the data. Instead of asking what each neuron encoded, it would be more sensible to consider what all the neurons — the entire population — are saying about the movement. Using the information he had about direction tuning in individual cells, Georgopoulos calculated a ‘population vector’ — essentially an arrow that points in the direction of movement that the neural population is encoding. This calculation works by allowing each neuron to vote for its preferred direction of movement. But this is no perfect democracy, because not all votes are weighted equally. Instead the weight of a neuron’s vote comes from how active the neuron is. In this way, the neurons collectively indicate the desired movement direction.
This population-level approach to the data proved quite powerful — perhaps too powerful. After this work, several studies emerged showing other information that could be read out from the motor cortex if the whole population was taken into account: finger movement, arm speed, muscle activity, force, position and even sensory information about the visual cues indicating where and when to move. Direction may have been one of the original variables decoded in this way, but it was far from unique. Finding both kinetic and kinematic variables (and a host of other information) in the activity was a blow against Georgopoulos’ theory that kinematics is special. Ironically, it was one of his own contributions — a focus on the population — that undermined him.
These findings revealed a possibility that was always lurking just under the surface: There is no single value that neurons in the motor cortex ‘code for.’ It’s not kinematics or kinetics; it’s both, and more, and neither. In many ways, this fact was visible all along. It could be seen in the neurons that didn’t show neat responses to force or direction, or in the neurons that showed big changes in their responses with small changes in the experiment, or simply in the decades-long volley of researchers finding evidence for one side and then the other over and over again.
By some accounts, the field was led astray because it blindly followed the path laid by other scientists: The study of sensory systems that inspired Georgopoulos was a poor model for how to understand movement. The debate over ‘What does the motor cortex encode?’ was unresolved not because the question is hard, but because it was the wrong one to ask from the start. The motor system doesn’t need to track movement parameters, it just needs to produce movements.
That is why today researchers interested in the motor cortex are taking a different approach. Rather than trying to determine what activity in the motor cortex means, they think more about how it gets produced. Specifically, this new view puts the emphasis on the fact that the motor cortex is a dynamical system — that the neurons within it interact in a way that makes them capable of producing complex patterns of activity over time. Because of these interactions between its neurons, the motor cortex has the ability to take in short, simple inputs and produce elaborate and extended outputs in return. What makes this so useful is that it means another brain region could decide where the arm should be and send that information to the motor cortex, and the motor cortex would produce the full trajectory of neural activity needed to make the arm get there.
What’s more, this ‘dynamical systems’ view has the potential to explain why attempts to make sense of the motor cortex have been confounded. If we think of these neurons as part of a larger machinery — where some parts are guiding muscle movements in the moment, but others are planning for the next step — the diversity and malleability of their responses are more expected.
This essay is adapted from Lindsay’s book Models of the Mind: How Physics, Engineering and Mathematics Have Shaped Our Understanding of the Brain.