The feeling that you are controlling your own body and influencing objects around you is known as the sense of agency (SoA). This sensation is essential for everyday functioning and well-being, and it is becoming increasingly important in the design of new human–computer interfaces. As technologies such as virtual reality, rehabilitation tools, and brain–machine interfaces continue to develop, researchers are paying closer attention to how a sense of agency forms—especially in unfamiliar situations where people have not yet learned how their actions produce outcomes.
Traditionally, SoA has often been explained using the “comparator model.” In this framework, the brain predicts what should happen when you perform an action, based on an internal model of how your movements affect the world. A strong sense of agency emerges when those predictions match the sensory feedback you receive.
However, this explanation raises a challenge when people learn new motor skills, such as taking up a sport or regaining everyday abilities after neurological injury. In these situations, outcomes are not yet predictable. Instead, beginners typically try movements first and then observe what happens. This process—known as motor exploration—helps people build and refine the internal model, eventually allowing them to generalize what they have learned to new conditions.
To better understand how SoA develops during early learning, researchers at the University of Tokyo examined how feelings of control change from a “pre-learning” stage, before reliable predictions have formed. They used a motor learning task in which participants wore a specialized data glove and controlled a cursor on a screen through finger movements.
The study included two experiments. In the first, participants learned the hand-to-screen mapping from scratch through trial and error. At different stages of learning, the researchers measured how strongly participants felt they controlled the cursor, including during moments when the cursor’s movement was subtly distorted in space or time.
The results suggested that, before learning, participants judged control mainly by whether their hand movement and the cursor movement were synchronized in time. After sufficient practice, the basis for the sense of agency shifted: participants increasingly felt in control when the cursor followed the mapping they had learned, even when other subtle distortions were introduced. This effect was especially strong among participants who reached higher levels of proficiency.
The second experiment was designed to limit motor exploration. Instead of freely testing movements, participants were asked to imitate predefined gestures that moved the cursor to target locations. In this condition, the researchers did not observe the same strengthening of the sense of agency. The findings indicate that simply memorizing gesture-to-cursor associations is not enough. To develop a robust feeling of control, learners appear to need to actively uncover the underlying rules of the system—for example, understanding that bending a specific finger moves the cursor in a particular direction. The researchers describe this kind of rule-based understanding as a “structural representation,” which seems to emerge through exploration.
Overall, the study offers new insight into how the sense of agency evolves as people acquire new motor skills. By clarifying how the mechanisms behind agency develop in the first place, the findings could help improve future applications in rehabilitation, virtual reality systems, and brain–machine interfaces.