Our seemingly effortless ability to reach out and swat a fly or grab a cup belies the sophisticated neural computations at work in our nervous system. It has long been recognized that, before moving, we somehow prepare neural activity such that, when called upon, the desired movement unfolds. But the goals of movement preparation and the underlying neural mechanisms remain poorly understood. I will describe some of our recent electrophysiological investigations of how premotor cortex prepares and helps execute movements. Our results suggest that the brain is attempting to optimize preparatory neural activity  and can delay movement until this activity is sufficiently accurate . We note that the spiking activity during motor preparation exhibits dynamics beyond that driven by external stimulation, presumably reflecting the extensive recurrence of neural circuitry. We have been investigating methods for capturing the dynamics from (96 channel) simultaneous neural recordings using low-dimensional non-linear dynamical systems models, with underlying recurrent structure and stochastic point-process output. I will present recent work using latent variable methods that simultaneously estimate the system parameters and the trial-by-trial dynamical trajectories , and recent evidence that low-dimensional trajectories may be able to better predict the reaction time of the upcoming movement. Characterizing these dynamics may reveal important features of neural computation, and may be useful for further increasing the performance of neural prostheses [4,5].
 Churchland MM, Yu BM, Ryu SI, Santhanam G and Shenoy KV
J. Neurosci. 26:3697-3712 (2006).
 Churchland MM and Shenoy KV J. Neurophysiol. 97:348-359 (2007).
 Yu BM, Afshar A, Santhanam G, Ryu SI, Shenoy KV and Sahani M NIPS 18:1545-1552 (2006).
 Santhanam G, Ryu SI, Yu BM, Afshar A and Shenoy KV Nature 442:195-198 (2006).
 Yu BM, Kemere C, Santhanam G, Afshar A, Ryu SI, Meng TH, Sahani M and Shenoy KV J. Neurophysiol. 97:3763-3780 2007).