The Ferster lab will attempt to answer two critical questions about cortical circuitry raised by the proposed model of visual recognition. The first is whether cells in the cerebral cortex are capable of performing the required MAX operation. For this part of the project, we will use complex cells of area V1 of the cat visual cortex. Much is known about the receptive field properties of these neurons, including the relative positional variance of their responses. They can therefore serve as a quantitatively approachable model system for studying the MAX operation and its implementation in the brain. Simple combinations of two-bar stimuli will be presented and the degree to which the output of the cell conforms to the MAX operation will be measured. We will study both the evoked changes in membrane potential (using intracellular recording) and the evoked changes in spike rate (using extracellular recording). Preliminary experiments indicate that a subset of cells do indeed perform the MAX operation. The second question to be studied will be how the cortical circuitry generates the MAX operation. Four possible circuit models have been studied so far, relying in turn on feedforward and feedback shunting inhibition, threshold nonlinearities, and spiking feedback. Experiments designed to distinguish among these models will include measurements of shunting inhibition evoked by the bar stimuli in complex cells, and measurements of the time course with which the MAX-like output of a cell develops during the course of the response to flashed stimuli. Finally, the nature of the inputs to cortical cells will be explored using neural networks. When applied to responses of complex cells evoked by flashing bar stimuli, the networks could potentially reveal the receptive field properties of the underlying inputs. We hope that the study of these questions will extend scope of the work to the specific cellular mechanisms that serve cognitive function.
Showing the most recent 10 out of 12 publications