Perceptual Priors Update Contextual Feedback Processing in V1.

Abstract

Contextual information and prior knowledge facilitate perceptual processing, improving our recognition of even distorted or obstructed visual inputs. As a result, neuronal processing elicited by identical sensory inputs varies depending on the context in which we encounter those inputs. This modulation is in line with predictive processing accounts of vision, which suggest that higher brain areas use internal models of the world to interpret sensory inputs. Cortical feedback signals encoding predictions about those inputs are propagated back down to sensory areas. As such, acquiring knowledge should enhance the internal models that we use to resolve sensory ambiguities, and feedback signals should encode more accurate estimates of sensory inputs. We investigated how knowledge updates contextual feedback processing in V1 by first generating Mooney images, ambiguous two-tone images which are difficult to recognize without prior knowledge of the image content. Across two behavioral experiments and one 3T functional magnetic resonance imaging (fMRI) experiment, participants acquired knowledge either related to the general Mooney image category or Mooney image-specific information. During fMRI, we used partially occluded Mooney images to investigate if contextual feedback signals in early visual areas are modulated after acquiring a high-level interpretation of the images. We show that general information about image categories is sufficient to improve recognition of ambiguous images. We also show that perceptual priors containing image-specific information modulate contextual feedback processing in the early visual areas, in response to previously ambiguous images.