Neural population dynamics during vocal behavior

PABLO TOSTADO MARCOS; EZEQUIEL M. ARNEODO; VIKASH GILJA; TIMOTHY Q GENTNER

Chicago

Encuentro; Society for Neuroscience annual meeting; 2021

Society for Neuroscience

The increasing capacity of neuroscientists to record the electrophysiological activity of large numbers of individual neurons permits the possibility of understanding brain dynamics that may appear imperceptible in individual unit responses. This is of particular interest when studying complex natural behaviors, like vocalization. To better understand the neural mechanisms that enable complex motor-vocal behavior, we recorded the simultaneous activity of 145 single neurons in the robust nucleus of the arcopallium (RA), a forebrain motor region, of male zebra finches engaged in unconstrained vocal behavior. We apply latent factor-based models to gain insight into the temporal dynamics of neural populations that drive vocal production. The characteristic stereotypy of zebra finch songs allows us to record high numbers of unconstrained, yet highly similar vocalizations in free-behaving birds. We find that the spiking activity across populations of individual RA neurons is well described by a pattern of smooth trajectories in the latent neural space during stereotyped singing (Figure 1). Moreover, the dynamics of the low-dimensional state-space of RA neural trajectories track single notes categorically and reflect more subtle variability across different renditions of the bird’s song. Similar observations are reported in populations of cortical neurons of the primary motor cortex (M1) of macaques during movement generation (Gallego et al. 2018). Our results point to common principles that may underlie the encoding of complex natural motor actions in specific brain regions or circuits.