Bottom Up and Top Down Dynamics in the Visual Cortex

Cortical processing of sensory information integrates not only bottom up inputs from the sensory surface but also top down information about internal state, including attention, expectation and memory. We propose that top down 'learning' signals modulate responses in early cortical areas in accord with predictions about bottom up inputs. Thus, we have shown that orientation-specific responses of V1 neurons in alert monkeys can be modulated according to Bayesian estimates of stimulus location. Using multiple electrode recording and optical imaging in alert monkeys combined with computational models, we are now examining other predictive top down influences, including those due to location based attention and feature based expectation.