A new study published in Science Advances provides evidence that special brain cells called mirror neurons track the continuous, physical movements of actions, rather than just the final goal. The findings suggest that the brain uses a dynamic, shared network of neurons to understand what others are doing by matching their movements to our own internal blueprints.
The premotor cortex is a region near the front of the brain that helps plan and coordinate physical movements. Within this region, certain brain cells fire both when an animal performs an action and when it watches someone else perform the same action. These special cells are known as mirror neurons.
Since their discovery, experts have debated exactly what kind of information these cells represent. Some scientists propose that mirror neurons focus on the abstract goal of an action, such as grasping an object to eat it or move it. This perspective suggests that the brain is interested in the final outcome rather than the physical effort required to get there.
Vassilis Raos, professor of physiology at the University of Crete Medical School and co-author of the study, noted that past research offered an inconsistent view of how these cells match actions. “At the same time, the ‘goal’ these neurons are claimed to encode has been loosely defined, and there is limited direct evidence about precisely what they represent,” Raos said. “We therefore revisited an important possibility that had not been adequately tested: that mirror-neuron activity reflects detailed kinematics, meaning how a movement unfolds through space and time, during both action execution and observation.”
Raos explained that this idea gained traction after his team’s previous work showed mirror neurons respond to actions that do not involve objects. “This possibility became especially plausible after our previous study showed that a large majority of mirror neurons also respond to the observation of intransitive, or non-object-directed, actions, not only to transitive, object-directed actions,” he said.
Other scientists propose that these cells encode the specific static grip used, like a tight pinch or a wide, whole-hand grab. Another possibility is that mirror neurons track action kinematics. Kinematics refers to the continuous physical details of a movement as it travels through space and time. This includes the speed, the direction of the arm, and the changing shape of the hand as it reaches for an object.
Previous studies often examined the activity of individual mirror neurons in isolation. The authors of the new study took a different approach by looking at populations of neurons. They wanted to see if the overall pattern of brain activity changes dynamically as an action unfolds, rather than just signaling a single, static concept like a final goal.
“Based on a proposal we put forward in a previous study, we expected that the apparent mismatch between execution and observation at the level of individual mirror neurons would be compatible with a systematic correspondence at the population level,” Raos said. “We also predicted that mirror-neuron activity would be modulated by movement kinematics, and our findings confirmed both predictions.”
To get a complete picture of the brain region, the researchers also decided to look at non-mirror neurons. By including cells that only fire during physical movement, they could determine if the ability to track movement details is unique to mirror neurons or a broader feature of the entire motor region.
To test these ideas, the researchers recorded electrical activity from 433 neurons in the premotor cortex of two macaque monkeys. The monkeys performed a reaching and grasping task designed to test different hand shapes. In the execution portion of the experiment, the monkeys reached for and grasped four different three-dimensional objects.
These objects included a sphere, a cylinder, a ring, and a small cube. Each object required the monkey to use a specific grip. The sphere prompted a whole-hand grab, while the small cube required an advanced precision grip using just the tips of the index finger and thumb. The monkeys were trained to grasp these objects consistently and smoothly.
In the observation portion of the experiment, the monkeys simply sat still. They watched a human experimenter perform the exact same reaching and grasping actions with the same set of objects. The researchers used a camera system to ensure the monkeys were looking at the action.
Out of the total recorded cells, 285 fired during both the execution task and the observation task. The researchers classified these as mirror neurons. Another 148 cells fired only when the monkey actually performed the action. The scientists classified these as non-mirror neurons.
For the main statistical analyses, the team focused on narrowed datasets of 240 mirror neurons and 129 non-mirror neurons that had enough recorded trials for accurate mathematical testing. In separate sessions, the researchers used a three-dimensional motion capture system to record the physical movements of the monkeys and the human experimenter. They tracked specific points on the index finger, thumb, and wrist. This equipment allowed them to measure the precise speed, position, and changing distance between the fingers over the course of the action.
The scientists found that information about the type of grip was spread broadly across the population of mirror neurons. Instead of a single neuron maintaining a constant signal throughout the action, different neurons became active at different stages of the movement. This shifting activity indicates that the brain uses a dynamically changing ensemble of cells to represent an action as it happens.
The researchers then compared the patterns of brain activity when the monkeys performed the action to when they merely observed it. At the level of individual brain cells, the firing patterns rarely matched perfectly between doing and watching. However, when the researchers analyzed the activity of the whole population of cells together, a partially shared neural structure emerged.
This shared structure means the patterns of brain activity during observation overlapped significantly with the patterns during execution. The overlapping activity was most prominent during the middle of the hand movement and at the end of the grasping motion. The neural population tends to align most strongly when the action reaches its most complex physical stages.
The population of mirror neurons also strongly tracked the continuous kinematics of the grasps. The neural activity systematically matched the speed, wrist position, and finger placement as the hand moved toward the object. By looking at mathematical models of the brain activity, the researchers could successfully predict the physical movement of the hand.
This prediction worked in reverse as well. The researchers could use the physical motion tracking data to accurately predict the patterns of neural activity. This two-way relationship was present for both the monkeys’ own movements and the observed movements of the human experimenter.
“Our findings show that mirror neurons really do ‘mirror,’ but this correspondence becomes clearest when we examine patterns of activity across populations of neurons, rather than expecting individual neurons to respond identically during action execution and observation,” Raos told PsyPost. “These populations carry detailed information about how an action unfolds over time.”
He noted that tracking movement does not mean the brain ignores the final goal. “Importantly, sensitivity to movement kinematics does not contradict the idea that mirror neurons encode action goals,” Raos explained. “Research on movement kinematics suggests that intentions and goals are expressed in the movement itself, so the brain may infer what another person is trying to achieve by reading how their movement is performed.”
By reading these ongoing physical details, the brain can decipher deeper intentions. “Our findings suggest that mirror neurons have the properties needed to support this process, bridging the gap between how an action is performed and why it is performed,” Raos said.
The scientists discovered that non-mirror neurons also showed a strong relationship with the physical kinematics of the grasps. Much like the mirror neurons, this group of cells tracked the ongoing physical details of the monkeys’ movements. This finding provides evidence that continuously tracking movement structure is a broad feature shared across different types of cells in the premotor cortex.
The overlap in brain activity between watching and doing was not completely symmetrical. Statistical models trained on the brain activity during observation could successfully predict the monkeys’ physical movements during execution. Models trained on the execution data did not generalize as well to predict the observed human movements. This asymmetry suggests that doing an action involves extra, specific neural processes beyond those shared with simply watching an action.
“Our findings should not be interpreted as showing that mirror neurons are merely analyzers of movement kinematics, or that their role is limited to detecting movement parameters,” Raos cautioned. “Our study identifies information that is systematically present in their activity, but it does not establish their full functional or causal role in understanding others.”
Instead, Raos suggests that this physical tracking is part of a larger system. “Kinematic information may contribute to a broader motor process that links observed actions to an individual’s own motor repertoire, helping bridge the gap between the movements we see and the motor acts and goals they express,” he said.
Several limitations should be noted regarding this research. The recordings were taken from a specific brain area in macaque monkeys performing highly controlled and repetitive actions. “Because our experiment focused on controlled grasping actions, further research will be needed to understand how this process operates during richer and more natural social behavior,” Raos noted.
The scientists recorded the neurons in separate sessions and mathematically combined them later to estimate the overall population activity. Because of this, they did not capture how all the individual brain cells interacted with each other at the exact same moment.
The experimental design also did not completely separate the visual shape of the object from the physical grip used to grab it. The monkeys always used the specific assigned grip for each specific object. This setup makes it difficult to completely rule out the possibility that the brain cells were partially responding to the visual identity of the objects rather than just the kinematics of the hand.
“In this study, we focused primarily on what information is encoded in premotor neuronal activity,” Raos said. “Our next goal is to investigate when and how these representations emerge, particularly how their temporal dynamics depend on the context in which an action occurs.”
To address the object versus grip limitation, the team plans to adjust future experiments. “We would also like to disentangle the contributions of the observed action and the object involved by manipulating them independently, allowing us to determine how each shapes observation-driven activity,” Raos said.
The study, “Dynamic population coding of kinematic structure across executed and observed actions in primate premotor cortex,” was authored by Konstantinos Chatzimichail, Christos Paschalidis, Eleftheria Tzamali, Vassilis Papadourakis, and Vassilis Raos.