Abstract
The Binocular Neural Mechanism: Gnostic and Population Coding L. Zhao, B. Farell. Institute for Sensory Research, Syracuse University Purpose: Stereo thresholds usually increase rapidly as the pedestal disparity on which thresholds are measured grows. Recently, however, Farell (ARVO, 2001) found a dip in the disparity increment function for narrow bandwidth patterns: Thresholds are lowest off the horopter. Here we investigate possible neural mechanisms underlying this phenomenon.
Method: A full column of 11248 complex cells was used in the simulation (16 frequencies × 19 orientations × 37 disparity phase angles). Cells were modeled after the energy units of Ohzawa & Freeman. Stimuli were gabors; disparity was a shift in carrier phase between the eyes. In the simulation information is integrated hierarchically, first over neurons with the same orientation and frequency but different preferred disparities, then across neurons sharing only frequency tuning, and finally across different frequencies.
Results: We found that, contrary to current thinking, psychophysically measured sensitivity to changes in disparity may reflect the most active neurons (gnostic coding) rather than the neurons whose responses vary the most. Neurons showing the greatest variation can be tuned to any disparity, not just near the horopter. However, the envelope of the most active neurons can account for the psychophysical measurements. The slope of this envelope is shallow at the horopter, reaches a maximum at a phase shift of about ±45°, and then falls again. This gives the dipper.
Conclusion: Using the complex cells found within a full cortical column, we find a basis for the dipper function for disparity increment thresholds. The dipper follows the envelope of neural activity. We also find that each complex cell can be regarded as a gnostic unit (aka grandmother cell). This results from the high correlation between the neurons' phase difference and the gabor phase difference. Population coding is also supported, for psychophysical performance follows the envelope of neural activation.
Supported by NEI Grant EY12286