Brain-machine interface learning is facilitated by specific patterning of distributed cortical feedback

Topographic representations of the peripheral sensory organs are a prominent feature of primary sensory areas in the cerebral cortex1-3. In particular, the whisker representation in the primary somatosensory cortex of rodents is composed of spatially distinct "barrel" columns, each corresponding to a different whisker on the snout4. Although the relationship between the sensory coding properties of neurons and their position in the barrel map has been extensively studied5-8, the functional role of cortical maps remains unclear9. We hypothesize that the body map in the primary somatosensory cortex is a framework for the integration of sensory information into motor control. First, we trained head-fixed mice in a cortical closed-loop brain-machine interface task where learning necessitates the integration of sensory feedback. Second, we show that in this task a biomimetic, topographic organization of the sensory feedback is required for learning. Finally, we show that enhanced performance in the biomimetic feedback condition correlates with improved motor control. Overall our findings support the view that cortical maps are necessary for optimal cortical sensorimotor integration. They should therefore be fully considered when designing direct cortical feedback for brain-machine interfaces with clinical applications10.