Abstract
We have proposed that the border ownership selectivity that is observed in many neurons in early visual cortex is the result of specific network connectivity (Craft et. al., J. Neurophysiol. 97:4310–26, 2007). According to our model, border ownership cells (B-cells) have reciprocal connections with grouping cells (G-cells) in another level of cortex. A G-cell integrates inputs from B-cells and modulates their firing rates. When a pair of neurons is activated by edges that are part of one figure (the contours are bound), the neurons receive reentrant inputs from the same G-cell. When the contours are part of separate figures, the inputs come from different G-cells. Thus the model predicts that coherence will increase between pairs with bound edges in their receptive fields. Specifically, because pairs that both prefer the bound configuration are connected to the same G-cell, they should display enhanced coherence with binding compared to other pairs. In the model, attention is inserted at a G-cell and propagates back to the B-cells, modulating their firing rates for further processing. Thus as with binding, the model predicts that coherence will increase with attention, especially for pairs that both prefer the bound configuration. We tested these predictions by recording local field potentials and single-unit spiking activity from two electrodes separated by long distances (3–10 mm) in macaque areas V1 and V2. We found that binding increases coherence at around 20 Hz, especially for pairs that both prefer the bound configuration. Selective attention also increases coherence in the same range, though the effect is more subtle. As with binding, the increase in coherence with attention is more pronounced for pairs that both prefer the bound configuration. These results support the network connectivity proposed by the model. The coherence difference (although not gamma) may play a role in coding binding and selective attention.
NIH R01-EY02966, NIH R01-EY16281, NIH T32-EY07143-13.