Recent success in animal studies in which neural signals are used to control machines suggests that an effective neuromotor prosthesis can be tested in humans. Cyberkinetics (CKI) has received FDA approval for a feasibility trial with an investigational neural prosthesis device in 5 patients who have severe motor impairment. There are three central questions which this trial seeks to address: 1) can intracortical neural signals be recorded chronically from humans?, 2) can a filter that maps neural activity into an output signal be built?, and 3) what is the spatiotemporal resolution of the resulting output signal? In our previous work, neural-to-movement decoders by having awake, able-bodied macaques generate complex arm movements. Because paralyzed individuals cannot provide such arm movements, we have decided to use a proxy signal, namely the position of a visual stimulus on the screen guiding an imagined movement. In assessing the utility of a neuromotor prosthetic, we must provide the paralyzed individual with a feedback signal. We have chosen a two-dimensional cursor because (i) point-and-click interfaces are standard among desktop computers, (ii) our previous research has already demonstrated that accurate 2D signals can be reconstructed from motor cortical waveforms, and (iii) 2D signals are a useful starting point to assess the spatiotemporal resolution of a direct neural output signal. We have already shown that a 2D cursor signal can be used to drive a robotic arm in near real time (Serruya et al., SFN, 2001). Our protocol calls for having patients acquire targets on sequentially finer grids. Once the degree of control and ease of use of the output signal is characterized, additional work on the construction and use of more complex signals can be begun.