Neural mechanisms of mixed speech and grasp representation in sensorimotor cortices

Objective. Recent brain-machine interface (BMI) studies have challenged traditional views of functional specialization in human motor cortices, suggesting that regions associated with hand movement also support speech. However, the extent of this dual functionality as well as the neural mechanisms underlying it are unclear. This study aims to determine whether mixed speech and grasp representation generalizes across a wide range of cortical regions and if so, to identify the population-level mechanisms supporting such mixed selectivity.Approach. We analyzed intracortical neural activity from seven brain regions (spanning motor, premotor, somatosensory and parietal cortices) while two human participants with tetraplegia completed speech and hand movement tasks. We assessed the population-level encoding of the two tasks using offline neural decoding, functional connectivity and subspace analyses.Main results. Robust grasp decoding was evident across all regions, along with reliable discrete-word decoding during both silent and vocalized speech. Within each region, overlapping neural populations contributed to both tasks, indicating mixed selectivity. However, these populations reconfigured their functional connectivity between tasks, organizing neural activity into non-overlapping speech and grasp subspaces.Significance. Our results indicate that mixed selectivity for speech and grasping generalizes across distributed sensorimotor cortical networks. Additionally, our findings support the development of multi-functional BMIs capable of decoding both speech and grasping from the same implant and highlight ventral premotor area 6r as a novel target.