Preparatory activity and the optimal-subspace hypothesis

When an instructed movement is delayed, motor and premotor cortex show sustained preparatory activity that predicts the upcoming movement and correlates with reaction time. The optimal-subspace hypothesis interprets this as the system steering its population state into a subspace of initial conditions from which the movement-generating dynamics will unroll correctly: preparation is setting the starting point, and reaction time reflects how close the state has come to a good initial condition when the go cue arrives. This dynamical reading explains why preparatory tuning need not match movement-period tuning — the two are different phases of one trajectory, not two separate codes.

A key refinement is the split into output-null and output-potent subspaces. Preparatory activity can be large yet cause no movement because it lies in output-null dimensions that the downstream readout cancels, whereas movement is triggered when the state enters output-potent dimensions that drive the muscles. For BCI this offers a route to decode intent before overt movement and to anticipate timing, but the same overlap of preparatory and movement subspaces that the brain manages carefully is a hazard for decoders, which can mistake preparation for execution if they ignore this structure.