A continuing controversy in motor neuroscience is over what aspects of movement are represented and planned centrally (i.e., in the brain). The virtual trajectory hypothesis claims that since the peripheral neuromuscular system appears to translate central commands into stable postures, the brain need only be concerned with the kinematic aspects of a motor task. If this hypothesis were true, we would expect to find that central commands bear a close resemblance to the task description and that the actual limb trajectory would be distorted by dynamic effects. Some have criticized this hypothesis as untestable because a direct test based on interpreting the symbolic content of neural activity is currently beyond our capabilities and indirect tests based on mechanical behaviour have been thought to require models of muscle impedances--a notoriously difficult proposition given their highly nonlinear and time-varying properties.

I have developed an alternative argument which compares the actual trajectory executed by a subject with the trajectory which they would follow if the inertial contributions to the movement could be cancelled (what I call the attractor trajectory). I have also devised a novel experimental technique to measure attractor trajectories;​ in my experiments, the subject interacts with a robotic arm which supplies interaction forces designed to cancel the inertial effects. This technique depends only on having a reasonable estimate of the inertia of the subject's arm and does not require a model of the muscle impedance, so we now have a convincing means to test the virtual trajectory hypothesis.

In point-to-point movement experiments using this technique, I have disproven the virtual trajectory hypothesis by showing that the brain's plans for unconstrained movements are more complex than the actual arm trajectories;​ the planning process therefore cannot be as simple as the hypothesis suggests, but must incorporate some form of compensation for dynamic effects. However, when I change the speed of the movement or ask the subject to cope with spring-like or coriolis-like forces, I find that the plans are simpler than the actual trajectories and are in fact very similar to the plans for the unconstrained movements. These results indicate that, in the spirit of the virtual trajectory hypothesis, the brain executes a new or unusual task by recycling an already-well-learned plan used in a similar but more common movement.