Neural dynamics in the rodent motor cortex enables flexible control of vocal timing

Abstract

Neocortical activity is thought to mediate voluntary control over vocal production, but the underlying neural mechanisms remain unclear. In a highly vocal rodent, the Alstons singing mouse, we investigate neural dynamics in the orofacial motor cortex (OMC), a structure critical for vocal behavior. We first describe neural activity that is modulated by component notes (~ 100 ms), likely representing sensory feedback. At longer timescales, however, OMC neurons exhibit diverse and often persistent premotor firing patterns that stretch or compress with song duration (~ 10 s). Using computational modeling, we demonstrate that such temporal scaling, acting via downstream motor production circuits, can enable vocal flexibility. These results provide a framework for studying hierarchical control circuits, a common design principle across many natural and artificial systems.

Integrator dynamics in the cortico-basal ganglia loop underlie flexible motor timing

“…5b-d). Thus, a large proportion of ALM activity between the cue and the lick, which we refer to as ‘timing dynamics’, demonstrated temporal scaling, similar to observations in other timing tasks and species 9,11,22,24,49,65 .…”

“… a. Schema illustrating two hypothetical cells that encode time differently through ramping activity 49 . Left, a cell with ramping activity encoding relative time, where the ramping speed changes as the lick time varies (i.e., temporal scaling).…”

“…While neural correlates of an integrator have been observed in the frontal cortex and striatum 15,18,22,[35][36][37][38] -two forebrain regions forming a long-range loop -the precise roles of these areas in implementing the integrator dynamics remain unclear. Manipulations of the frontal cortex and striatum affect timing behavior, supporting their causal roles 11,[39][40][41][42][43][44][45][46][47][48][49][50][51][52][53][54] . However, the presence of neural correlates and necessity cannot determine their computational roles 8,55,56 .…”

A Recurrent Neural Circuit Mechanism of Temporal-scaling Equivariant Representation

“…The representation of time information in recurrent circuits was supported by a wide range of neurophysiology experiments in many brain areas, e.g., medial frontal cortex [6], hippocampus [24], striatum [25], and HVC in songbirds [26], etc. The disentangled architecture is also supported by a recent study on vocal motor circuits in a highly vocal rodent [27], where the motor cortex dictates the timing of songs, while the mid-brain or brain stem areas generate music notes (patterns).…”

“…In contrast, the present paper analytically derives a recurrent circuit implementation of TS equivariant representation, and the explicit representation of TS group operators. Moreover, our hand-crafted circuit model generalizes over temporal scales well, which outperforms recent neural circuit models where the trained network cannot generalize (extrapolate) well to temporal scales beyond the range in the training set [11, 27]. In addition, previous studies considered generating a moving neural sequence by adding an anti-symmetric recurrent weight component in the recurrent network (e.g., [45–48]), whose magnitude increases with temporal speed (or TS factor) [37, 47].…”