What Fires When You Watch Someone Move? A Finding That Outran Its Evidence

Researchers in Parma recording from individual neurons in the premotor cortex of macaque monkeys found cells that fired when the animal grasped an object and also when it observed a researcher grasping one, which was unexpected because the region was understood as motor rather than sensory. The property was reported in the early 1990s, replicated within the laboratory, and extended to other regions and to actions defined by their goal rather than by the specific movement. The core observation is solid and has been confirmed, including in a small number of recordings from human patients undergoing surgery for unrelated reasons, which is the only direct single-cell evidence in humans that exists.

The interpretations that followed were considerably broader than the finding:

• •That the cells explain how we understand the intentions of others
• •That they are the basis of empathy
• •That their dysfunction explains autism, which became a widely repeated claim
• •That they underlie language evolution, imitation and culture
• •That they explain the appeal of fiction, dance and sport
• •That they constitute the most important discovery in neuroscience in decades, which was stated in print

Several gaps separate the observation from the conclusions. Firing when observing an action shows a correlation with observation and does not show that the firing causes understanding, which would require demonstrating that disrupting the cells impairs understanding, and that demonstration is difficult and has not been made convincingly. Human evidence is overwhelmingly indirect, coming from imaging that measures activity across regions containing millions of cells rather than individual neurons with the required property. The property may be learned through associating one's own actions with their visual appearance rather than being an innate mechanism, which several researchers argue and which would change what it explains. And the autism claim has been tested repeatedly with results that do not support it, which has not stopped its circulation.

Setting aside the largest claims, several related findings are solid and worth knowing. Watching an action does activate motor regions of the observer's brain, which is well established by imaging even if the cellular basis is unresolved, and the activation is stronger for actions the observer can perform, so a trained dancer watching their own style shows more than an untrained viewer does. Observing an action speeds up performing the same action and slows performing a different one, which is a robust behavioural effect. Imitation is important in human development and is unusually elaborate in our species. Understanding what somebody intends draws on several systems including regions associated with reasoning about other minds, which are distinct from motor regions. The picture that emerges is of several contributing mechanisms rather than one cell type carrying everything.

The case is now used as a teaching example about how findings propagate. A genuine, interesting and limited result acquired an explanatory scope far beyond its evidence, partly because it appeared to solve several hard problems at once, partly because the name given to the cells was vivid and did explanatory work by itself, and partly because the popular accounts were enthusiastic and the qualifications were left in the technical literature. Critical reviews began appearing in the late 2000s and the field has become considerably more careful, with researchers who made the original discovery also publishing more restrained assessments. The underlying question of how one animal understands another's action remains open and is being addressed by work that does not depend on these particular cells carrying the whole explanation.

Cells firing both when a monkey grasps and when it watches grasping were found in the 1990s, and the observation is solid. Claims that they explain empathy, language, culture and autism went far beyond it, since firing during observation shows correlation rather than cause. Human evidence is almost entirely indirect, and the autism claim has been tested and not supported.