We are immersed in a rich and complex visual environment. Foveal animals such as primates interact with this environment primarily by shifting their gaze to objects of interest, an act colloquially known as “looking”. Gaze shifts involve a sensory-to-motor transformation of the visual world into a movement command that redirects the line of sight. Several brain regions involved in the gaze control network enable this transformation by representing sensory, cognitive, and movement-related information in the same population of neurons. For example, neurons in the intermediate layers of the superior colliculus, a critical gaze control node in the midbrain, fire action potentials both in response to the onset of a visual target, during cognitive processing such as target selection and decision-making, and during an eye movement to that target. However, it is not known how this multiplexed dynamics of sensorimotor integration is read out by a downstream decoder. From the perspective of the decoder, when does sensory and cognitive processing end and movement programming begin?

We asked whether sensorimotor integration occurs in sequential stages or is implemented in parallel. Our approach was to 1) study the role of gaze fixation in shaping sensorimotor activity, 2) chart the time course of movement preparation by removing inhibitory gating on the network, and, 3) look for signatures of aforementioned multiplexing in neuronal population dynamics. Influential theories have proposed that sensory and movement processing occur in a cascade implemented by different subsets of neurons, and neural activity must accumulate to a pre-determined level in order to actuate the gaze shift. In contrast, we found that sensorimotor information is represented in a flexible continuum of neurons. Furthermore, ongoing low-frequency activity during the transformation has motor potential - latent ability to produce an eye movement - that is revealed under the right conditions. Together, these results suggest that the brain performs sensorimotor integration in parallel. Finally, the temporal structure of population activity evolved differently for sensory and movement-related processes, providing a novel substrate by which downstream neurons may discriminate between the two in order to decide when to generate a movement.