Nicole E. Neef, Robert Trampel, Lauren Lu, Daniel Hänelt, Pierre-Louis Bazin, Robert Turner, Nikolaus Weiskopf, and Angela D. Friederici

Speaking depends on dynamically organized firing patterns of specific large-scale neural populations in sensorimotor and other brain regions. Conventional in vivo fMRI resolves the activity in these brain areas typically with a resolution of about 3 mm. Given that the primary motor and somatosensory cortices are about 3 mm thick, partial volume effects are inevitable, making it hard to discriminate functional activity in each area.

Here, we used the high magnetic field strength of 7 T and a high quality phased-array RF receive coil to acquire speech-induced and finger-movement-induced BOLD signal from the primary motor and somatosensory cortices with an isotropic resolution of (0.75 mm)³. Within a slow event-related design, 10 healthy participants were asked to perform voiced speaking, voiceless speaking and to imagine speaking, and also to perform isochronous movements of the right index finger and the left index finger, and to imagine moving the right index finger. While actual motion is assumed to involve neural activity across all cortical layers, imagining movement, which does not require output from cortical layer 5, should primarily involve neural populations in superficial layers. In a cortical-depth dependent analysis, the cortex was segmented and automatically contoured into 4 equidistant laminae. This allowed us not only to discriminate BOLD activity in the primary motor cortex (in the precentral gyrus) and the primary somatosensory cortex (in the postcentral gyrus), but also to track BOLD signals from a deep layer adjacent to the white matter, two middle layers, and a superficial layer adjacent to the cortical surface.

In the primary motor cortex, speech and finger movements resulted in a positive BOLD signal that gradually increased from deep to superficial layers, for both actual and imagined moving. By contrast, in the somatosensory cortex, only actual moving induced positive BOLD signals. In the somatosensory cortex, imagined finger tapping induced no modulation of the BOLD signal, but imagined speaking resulted in a negative BOLD signal, especially in superficial layers.

Our results show for the first time that cortical depth-dependent differences can be studied during speech production with in-vivo imaging techniques in human. Furthermore, observed differences in BOLD activity in the primary motor and somatosensory cortex during imagined speech, as compared with imagined finger tapping, may reflect feedforward and feedback control mechanisms for automatized movements (such as speaking) that differ from those involved in less familiar movements (such as finger tapping), which could require a stronger interplay of motor and somatosensory brain hubs.