Sensorimotor dynamics differentiate singing and speaking

Singing and speaking often dissociate clinically--people who stutter can sing fluently, and individuals with aphasia and speech output problems from stroke may express sentences fluently in song--yet the neural mechanisms of this centuries-old clinical phenomenon remain unclear. We recorded intracranial EEG while neurosurgical patients produced matched sentences by singing or speaking, sampling millimeter- and millisecond-scale activity across both hemispheres. During articulation, high-frequency activity (70-150 Hz) lateralized oppositely across behaviors, with right-dominant sensorimotor cortex (SMC) activation for singing and left-dominant activation for speaking. Phase-amplitude coupling revealed that mu-band ([~]10 Hz) phase locally organized HFA in left SMC in a channel-specific, spatially interdigitated architecture preserved across both behaviors. Mu-band synchrony in speaking showed an early left-led pattern, whereas singing exhibited a ramping of synchrony within the left sensorimotor cortex and between the two motor cortices, supporting progressive interhemispheric recruitment. Frequency-domain Granger- Geweke causality revealed that the left primary somatosensory cortex drives both motor cortices at speech onset. In contrast in singing, control over motor cortices relied on both hemispheres. Singing and speaking engage a shared mu-coupled sensorimotor substrate through distinct recruitment dynamics.