Recurrent plasticity mechanisms for perceptual decisions in the human brain

Learning and experience are critical for making successful decisions in the face of inherently ambiguous and noisy information. Yet, the human brain computations that mediate this perceptual learning skill remain highly debated, as fMRI at standard resolution does not allow us to discern whether learning alters sensory encoding or top-down influences. Here, we capitalize on the sub-millimetre resolution of ultra-high field imaging to interrogate the finer-scale computations that mediate perceptual learning in the human brain. Combining 7T laminar imaging with orientation discrimination training, we demonstrate learning-dependent changes in superficial V1 layers, suggesting that training alters read-out rather than input signals in the visual cortex. Further, training enhances feedforward connectivity between superficial V1 layers and middle layers of posterior parietal cortex. Our findings propose that the brain learns to translate sensory information to perceptual decisions via recurrent processing within visual cortex and enhanced connectivity from sensory to decision-related areas.