Parallel and serial operations in primary visual cortex (V1) of alert primates
D. Max Snodderly, University of Texas at Austin
As a starting point, one can conceive of primary visual cortex as performing serial operations within multiple parallel pathways that gradually diverge to more specialized cortical areas. In fact, evidence is emerging that the organization of the cortex is massively parallel, with many more pathways than we fully appreciate. This capacity for parallel processing may be what makes biological intelligence so much more powerful than present-day machine intelligence. As an example, we have described a combination of serial and parallel processing in the motion-selective pathways of V1. Two pathways, from layer 4B and from layer 6 go to area MT, and a third, from layer 3, goes to V2. Each of these has a different distribution of receptive field sizes that suggests different roles. The pathways through MT include both large and small receptive fields that are appropriate to contribute to analysis of self-motion and object-motion in the dorsal cortical stream of visual processing. The pathway through V2 originates in the smallest receptive fields of V1 and it is appropriate for sensing small motions within objects (such as smiling or frowning faces) or motions of small objects within the ventral cortical stream. Serial processing within the parallel pathways of V1 was originally thought to be manifest in an increase in receptive field size as information was relayed from the input layers to other cortical layers. However, we have found, and others have confirmed, that the upper layers of V1 that send outputs to other cortical regions have classical receptive fields that are among the smallest in V1. This pattern emphasizes the role of intracortical inhibition in some steps of serial processing. Another contributor to cortical microcircuits in V1 is top-down input from numerous brain regions that terminates preferentially in layer 1. Dendrites of layer 2 cells are especially well placed to receive these inputs, and we have characterized the physiological properties of layer 2 cells. We find that layer 2 cells are not direction selective, and have less specific stimulus requirements than the layer 3 cells immediately below them. We suggest that layer 2 may be an important substrate for modulatory influences from other brain regions. The results described in this abstract have been obtained in collaboration with Prof. Moshe Gur of the Technion. The work was supported in part by NIH and by the US-Israel Binational Science Foundation.
