Conscious visual perception is proposed to arise from the selective synchronization of functionally specialized but widely distributed cortical areas. It has been suggested that different frequency bands index distinct canonical computations. Here, we probed visual perception on a fine-grained temporal scale to study the oscillatory dynamics supporting prefrontal-dependent sensory processing. We tested whether a predictive context that was embedded in a rapid visual stream modulated the perception of a subsequent near-threshold target. The rapid stream was presented either rhythmically at 10 Hz, to entrain parietooccipital alpha oscillations, or arrhythmically. We identified a 2-to 4-Hz delta signature that modulated posterior alpha activity and behavior during predictive trials. Importantly, deltamediated top-down control diminished the behavioral effects of bottom-up alpha entrainment. Simultaneous source-reconstructed EEG and cross-frequency directionality analyses revealed that this delta activity originated from prefrontal areas and modulated posterior alpha power. Taken together, this study presents converging behavioral and electrophysiological evidence for frontal deltamediated top-down control of posterior alpha activity, selectively facilitating visual perception.top-down control | directional cross-frequency coupling | prefrontal cortex | alpha oscillations | phase-amplitude coupling V isual perception is flexible, selective, and rapidly integrates sensory evidence with endogenous high-level expectations and predictions (1, 2). It has been suggested that rhythmic brain activity constitutes a key mechanism to coordinate information flow in the human cerebral cortex by transiently forming task-relevant largescale networks (1). However, it is currently unclear how contextual information is dynamically integrated to support visual perception. Numerous studies have shown that visual perception critically depends on prestimulus alpha-band (8-12 Hz) activity (3-7). The gating-by-inhibition hypothesis postulates that alpha serves as a mechanism to route information to task-relevant cortical sites (8) but might also be under top-down control (6, 7). However, it is currently unclear which cortical regions and mechanisms mediate the directed top-down control of alpha oscillations (2). It has been suggested that slow-frequency activity in the delta range (<5 Hz) might reflect a mechanism for endogenous attentional selection and predictions (9, 10). In particular, endogenous low-frequency entrainment is thought to reflect a substrate of top-down processing (11)(12)(13)(14). Importantly, endogenous entrainment does not require rhythmicity in the input stream but reflects an intrinsic mechanism to enable predictions (15). Several studies have demonstrated that visual perception cycles as a function of the alpha phase but only a few reports have demonstrated that multiple rhythms modulate behavior on a fine-grained temporal scale (5,(16)(17)(18)(19).At present, it is uncertain how different temporal scales interact t...

Speech brain-computer interfaces (BCIs) have the potential to restore rapid communication to people with paralysis by decoding neural activity evoked by attempted speaking movements into text or sound. Early demonstrations, while promising, have not yet achieved accuracies high enough for communication of unconstrainted sentences from a large vocabulary. Here, we demonstrate the first speech-to-text BCI that records spiking activity from intracortical microelectrode arrays. Enabled by these high-resolution recordings, our study participant, who can no longer speak intelligibly due amyotrophic lateral sclerosis (ALS), achieved a 9.1% word error rate on a 50 word vocabulary (2.7 times fewer errors than the prior state of the art speech BCI2) and a 23.8% word error rate on a 125,000 word vocabulary (the first successful demonstration of large-vocabulary decoding). Our BCI decoded speech at 62 words per minute, which is 3.4 times faster than the prior record for any kind of BCI and begins to approach the speed of natural conversation (160 words per minute). Finally, we highlight two aspects of the neural code for speech that are encouraging for speech BCIs: spatially intermixed tuning to speech articulators that makes accurate decoding possible from only a small region of cortex, and a detailed articulatory representation of phonemes that persists years after paralysis. These results show a feasible path forward for using intracortical speech BCIs to restore rapid communication to people with paralysis who can no longer speak.