A neural architecture for imagined and overt speech motor dynamics

Mental imagery is a hallmark of human cognition, yet its neural mechanisms remain poorly understood. Speech imagery (the internal simulation of speech without overt articulation) has been proposed to partially share substrates with speech production, but its spatiotemporal dynamics remain controversial. Here, we leveraged high-resolution electrocorticography to investigate shared and modality-specific coding of articulatory kinematic trajectories during speech imagery and articulation. Linear modeling revealed robust articulatory kinematic trajectories encoding in frontoparietal cortex across modalities. Supramodal populations across middle premotor, subcentral and postcentral-supramarginal regions exhibited spatiotemporal stability during integrative planning. In contrast, modality-specific populations for speech imagery and articulation were somatotopically interleaved in primary sensorimotor cortex, revealing a distinct spatiotemporal organization. We further developed a generalized decoding framework that achieved high prediction accuracy for speech imagery (median 80.4%; chance 16.7%), comparable to speech articulation (78.3%). These findings uncover a somato-cognitive organization linking supramodal planning with modality-specific representations, informing imagery-based brain-computer interfaces.