Visuomotor coupling shapes predictive processing in mouse visual cortex


September 6, 2017 - 1:00pm
Northwest Building Room 243
About the Speaker
Alexander Attinger
Speaker Affiliation: 
FMI Basel

Predictive processing theories of brain function are gaining increasing experimental support. An integral component of computation based on predictive coding would be neurons that signal deviations or mismatch between expected sensory input and actual sensory input. In visual cortex of mice, deviations, or mismatches, between actual visual flow and the visual flow predicted from self-movement evokes strong responses. We recently described a circuit mechanism that underlies this computation of a mismatch response in layer 2/3 of mouse visual cortex. In this circuit, a layer 2/3 excitatory neuron balances an excitatory motor-related input against an inhibitory visual input that is mediated by somatostatin interneurons. Hence, this type of layer 2/3 excitatory neuron is activated when the mouse receives less visual input than expected. Conversely, to signal more input than expected a layer 2/3 neuron would need to balance an excitatory visual input against a top-down inhibitory input. Such a neuron would have classical visual responses in a passively observing animal. We have preliminary data that suggest that visual responses in visual cortex are indeed carried by neurons that signal stronger visual flow than predicted. Computationally, these two circuits are symmetric and would signal mismatch of different sign. We speculate that layer 2/3 of cortex functions as a comparator of top-down and feed forward signals.