Predictive coding in visual cortex
The predictive coding framework represents a paradigm shift in neuroscience. Brain processes are traditionally studied as a function of sensory stimulation. In contrast, predictive coding states that the brain continually generates models of the world based on context and information from memory in order to predict sensory input. In terms of brain computation, a predictive model is created in higher cortical areas and communicated to lower sensory areas through feedback connections. Empirical data testing the predictive coding framework in visual cortex is relatively sparse. I will argue that fMRI is a useful tool for investigating predictive feedback and prediction error. To investigate the information content of feedback projections, we have exploited a strategy based on non-stimulated sections in retinotopic regions (apparent motion path, occluded natural scenes, blindfolded subjects). To investigate prediction error, we have used probe stimuli that were presented in a matching or non-matching context (i.e. apparent motion illusion). The results demonstrate expectation-related information content in non-stimulated parts of V1.