Ken Miller, Anton Krukowski, Todd Troyer, Nicholas Priebe
Dept. of Physiology,UCSF
The origin of orientation selectivity in primary visual cortex has long been a central, model problem for the understanding of cortical circuitry. Two major approaches have emerged. Beginning with Hubel and Wiesel, many have argued that the selectivity derives simply from the pattern of feedforward inputs from LGN. Others have argued that feedforward input will be weakly tuned, and therefore intracortical circuitry is required to sharpen the tuning: in particular, inhibition between cortical regions of differing orientation preference (e.g. cross-orientation inhibition). An additional problem is raised by contrast invariance: stimuli of increasing luminance contrast cause higher firing rates in the LGN inputs. This should push weaker stimuli above threshold for cortical response; yet the orientation tuning of cortical cells is invariant to changes in contrast. Most previous solutions to this problem have also relied on cross-orientation inhibition. Yet direct measurements of the synaptic inputs to cortical cells argue strongly against the existence of cross-orientation inhibition.
We propose a very simple model of the cortical circuitry that resolves these apparent problems. Careful consideration of the LGN input, assuming a Hubel-Wiesel model of feedforward connectivity, shows that the input has an untuned component - the mean rate or DC component - and a tuned component - the temporal modulation of the rate, or F1 component. Cross-phase inhibition - inhibition between cells of similar orientation but opposite spatial phase - eliminates the DC component, while rendering contrast-invariant the orientation tuning of the F1. To obtain strong response to the preferred orientation, intracortical excitation must amplify an effective stimulus. This is achieved by same-phase excitation - excitation between cells of similar orientation and similar spatial phase. These cortical connectivity patterns - cross-phase inhibition and same-phase excitation - will emerge directly from simple Hebbian rules for development, and are sufficient to account for the major experimental results about orientation tuning.