Role of the monkey midbrain in target identification and localisation in naturalistic multisensory environments.

To localise and identify a target as fast and as accurately as possible within a complex multisensory environment requires the precise coordination between the auditory and visual systems, and the programming of a fast eye-head orienting response. However, in environments with many distractors, unisensory information processing is typically noisy, giving rise to large localization errors and long reaction times. We argue that multisensory integration is an essential aspect within the perception-action cycle, as it will reduce spatial-temporal uncertainty about the target, leading to better accuracy and faster responses. An important question is how the brain knows which sensory inputs to integrate, as they belong to a single object, and which not, because they arose from separate objects? A prominent theory proposes that Bayesian reasoning underlies multisensory integration, and many behavioral studies have provided support for this principle. However, how neural populations represent the involved probability distributions is not known. In this project trained, head-unrestrained monkeys generate eye-head gaze shifts to visual, auditory and audiovisual stimuli, that are hidden within a cluttered audiovisual background. Meanwhile we record single-unit activity in the auditory midbrain inferior colliculus (a central hub for acoustic processing), and in the midbrain superior colliculus (a crucial gaze-control center), under a variety of background stimulus distributions. By careful manipulation of the environmental conditions, we correlate the variability and errors in the eye-head movement responses, with the variability in the firing rates of single cells to test the quantitative predictions of the Bayesian hypothesis. Moreover, we will perturb audiovisual integration through local microstimulation in either structure to assess its functional role.