Intracortical cursor control
Intracortical cursor control is the canonical motor-BCI task: a person with paralysis moves a two-dimensional cursor (and often issues a discrete click) using spiking activity recorded from motor cortex by an implanted microelectrode array. It is the workhorse benchmark of the field because the task is quantifiable, the required kinematics are low-dimensional, and it generalizes directly to typing, computer access, and communication. The signal chain reads binned spike counts or threshold crossings, extracts a velocity (or velocity-plus-position) intention, and drives the cursor in closed loop while the user watches and corrects.
Performance depends far more on the closed-loop training strategy than on the raw information in the neurons. Decoders that model the user's moment-to-moment intent (rotating training kinematics toward the target, zeroing velocity during holds) and that are retrained inside the loop dramatically outperform decoders fit only to open-loop, offline data. The best modern intracortical cursor systems reach typing rates and pointing throughputs competitive with, and in some studies exceeding, other assistive interfaces for the target population.
The cursor task is deceptively simple: because the user sees the cursor and continuously corrects it, a mediocre decoder can look adequate in a slow point-to-point task yet fail badly on fast, complex trajectories where feedback cannot rescue every error.