Abstract The goal of this project is to understand how spatio-temporal patterns of activity across motor cortex initiate different types of voluntary movements. Large distributed ensembles of motor cortical neurons begin modulating their firing rates prior to voluntary movement and are thought to causally generate these movements. However, it is still unresolved why movement is not initiated when similar modulations in single unit motor cortical activity occur during movement planning, imagery, and visual observation of action. The amplitude of local field potential (LFP) oscillations in the beta frequency (15-40 Hz) range is known to attenuate prior to movement onset and is considered a mesoscopic signature of corticospinal excitability. We have recently discovered a sequential pattern of LFP beta attenuation and single unit modulation timing in primary motor cortex that is spatially organized prior to reaching movement onset but not during movement preparation. Our working hypothesis is that such a propagating sequential pattern is necessary to initiate movement. In this project, we will test and extend this hypothesis by demonstrating that such propagating patterns generalize to movement initiation of different behaviors including 2D reaching under different conditions, more complex 3D reach-to-grasp, and tongue protrusion and occur in premotor cortex. We will first demonstrate that propagating sequences in beta attenuation and single unit modulation timing occur during initiation of each of these behaviors along different portions of the somatotopic map of primary motor and premotor cortices. Second, we will provide a causal link between these propagating patterns and movement initiation by applying subthreshold, spatio-temporal patterns of electrical stimulation. We will demonstrate that movement initiation is delayed when patterned stimulation travels against the natural propagating sequence but not when it mimics the natural propagating pattern. Third, we will provide a mechanistic explanation of how these propagating sequences lead to muscle activation that supports movement initiation using patterned stimulus-triggered muscle activity and muscle decoding. To accomplish these aims, four high-density electrode arrays will be chronically implanted in the either the upper limb or orofacial areas of primary motor and premotor cortices from which 100s of single units and LFPs will be simultaneously recorded. A two-link exoskeletal robot and a motion tracking system using a set of fourteen infrared cameras will monitor the kinematics of the arm and hand. A strain gauge will measure tongue force and kinematics of the tongue will be tracked with a novel 3D x-ray fluoroscopy system. Indwelling EMG electrodes will also measure activity from arm, hand, and tongue muscles. A set of classical and novel computational methods will be employed to characterize the spatio-temporal dynamics of motor cortical activity during movement initiation.