In the early 1990s, researchers studying nonhuman primates discovered a population of brain cells they called "mirror neurons." These specialized brain cells fired when the animal performed an action and also when the animal observed another performing that same action. The discovery of mirror neurons seemed to support the idea that we understand the actions of others by internally simulating them in our own minds. But if that were the case, how would anyone be able to imagine doing something they had never done themselves? How could anyone understand Superman flying, for example, when no human has ever been able to fly?
A new study from Caltech researchers finds the first evidence for mirror-like behavior in individual human neurons and suggests that "mirroring" is context dependent and more intricate than previously believed. The scientists found evidence for mirror activity only in a higher-level brain region that encodes for planning, and intention is critical—mirror activity does not happen automatically and can be suppressed depending on the task at hand.
The work informs the development of better brain–machine interfaces (BMIs) that can decode thoughts into robotic actions.
The research was carried out in the laboratory of Richard Andersen, the James G. Boswell Professor of Neuroscience and Leadership Chair and director of the T&C Chen Brain-Machine Interface Center at the Tianqiao and Chrissy Chen Institute for Neuroscience at Caltech. The findings are reported in a paper appearing in the journal Cell on August 20.
"The way that mirror neurons were originally defined, they should respond exactly the same way when observing and executing a particular action, a congruence between observation and execution," Andersen says. "But our findings show that it's not that simple, and behavioral context matters. Mirroring doesn't happen automatically."
For a decade, the Andersen lab has partnered with tetraplegic individuals to develop BMIs that can decode an individual's intended movements and translate them into commands for robotic arms and other apparatuses. The BMIs connect to two specific regions of the brain, called the posterior parietal cortex (PPC) and the motor cortex (MC), through tiny arrays of electrodes that measure neural activity. The PPC encodes for high-level thinking, such as intentions and the transformation of visual information into a motor plan, whereas the MC is more primitive and sends signals to muscles to execute movement.
In this study, led by postdoctoral scholar Vasiliki (Celia) Bougou, the researchers worked with two tetraplegic individuals to utilize the electrode arrays implanted in their PPCs and MCs to measure the activity of millions of neurons and look for mirror activity.
"In the 1950s, a psychologist proposed that our brains are automatically and constantly imagining all possible actions and behaviors of objects within our line of sight, giving us templates for understanding the world," says Bougou, the study's first author. "But this seems like an extremely inefficient use of our brainpower. So, we aimed to measure what is actually occurring in the brain when we observe the world."
In experimental tasks, the participants first observed an animated hand holding a cube and performing different actions: sliding the cube, lifting it, or rotating it. In a separate task, participants were asked to attempt performing the same actions themselves. For each distinct action, neurons in the PPC lit up in corresponding constellations of neural activity. These neurons had mirror-like properties, with patterns of neural activity that looked similar whether the participants were observing the action or attempting to perform it themselves.
Participants were then asked to attempt the sliding, lifting, and rotating actions while being shown the animations of a different action than the one they were instructed to perform. The researchers found that the observed action was not encoded in the PPC or the MC, suggesting the brain can ignore or screen out visual information that is not relevant to the task. However, when participants performed the same task but were also asked to report what they had observed, making the observed action behaviorally relevant, neurons in the PPC encoded both the attempted and observed actions whereas those in the MC encoded only the attempted action.
The findings show that mirror activity was found in the PPC while participants were observing actions, but it was not present in the MC. Importantly, mirroring did not happen automatically—only when the observed actions were task relevant did the PPC neurons encode those actions.
"The observation of an action is not hardwired to the performance of the action in the motor areas," Bougou says. "We are capable of building internal models of other people's actions in our brains that are aligned with our own motor plans, but this is not automatic. This is the first time that such a dissociation between action and observation has been recorded."
The work is critically important for the development of brain–machine interfaces. For example, the findings suggest if a participant were driving a car with their thoughts alone, observing a different action in their visual field should not interfere with their intended way of driving the car. The research also makes the case for implanting BMIs in high-level brain regions like the PPC that have more flexibility and can be dynamically restructured.
The Andersen lab is exploring how the PPC builds internal mental models in order to improve BMIs. "We have an idea that this area builds internal models not only for actions but for the whole physical world," Andersen says. "For example, when you're playing catch, you understand what you're observing—the other player, where the ball is, how it's coming toward you—and plan your own actions in response. We believe these three processes, creating internal models of ourselves, others, and the world, could all be happening in this brain region and have overlapping neural representations."
The paper is titled "Hierarchical and context-dependent encoding of actions in human posterior parietal and motor cortex." In addition to Bougou and Andersen, Caltech co-authors are senior postdoctoral scholar Jorge Gamez de Leon and Kelsie Pejsa, lab and clinical studies manager. Additional co-authors are Emily R. Rosario of Casa Colina Hospital and Center for Healthcare in Pomona, CA; Dr. Charles Liu of Casa Colina and the Keck School of Medicine of USC; and Dr. Ausaf Bari of the David Geffen School of Medicine at UCLA. Funding was provided by the National Eye Institute, the Tianqiao and Chrissy Chen Brain-Machine Interface Center at Caltech, the Swartz Foundation, and the James G. Boswell Foundation.