International Symposium on "Dynamics of Perception". (Okazaki, Japan)
生理学研究所主催 知覚のダイナミクスに関する国際シンポジウム(岡崎)1993.1.27

Columns in the Inferotemporal Cortex:
Machinery for Visual Representation of Objects

Ichiro Fujita
Precursory Research for Embryonic Science and Technology (PRESTO),
Research Development Corporation of Japan, and Frontier Research Program,
RIKEN Institute, Wako, Saitama, 351-01 Japan

 The visual cortical areas of the monkey are organized along 2 major streams: one crucial for object vision and the other involved in spatial vision. The anterior part of the inferotemporal cortex, cytoarchitectonic area TE, lies at the final stage of the object vision pathway. Monkeys with bilaterally lesioned TE cannot distinguish different shapes or recognize objects seen before. TE cells respond to partial features of objects such as particular shape, color, texture or a combination of these. Discharges of individual TE cells on presentation of an object do not signal abstract representation of the object.

 We have recently found that cells responsive to similar object features cluster together within TE (Fujita, Tanaka, Ito and Cheng, 1992). This clustering is vertically elongated in the cortical depth and patchy across the cortical sheet, forming columnar organizations. Cells within a column, however, show differences in sensitivity to small variations of stimuli. For example, a cell responds best to a horizontally striped triangle, while an adjacent cell in the same column prefers a square filled with horizontal stripes. The width of each column is 0.4 mm on average, and one TE is estimated to contain 2000 columns.

 When a monkey sees an object, a particular subset of columns, not the entire TE, may be activated by the image. Different component features of the object activate different columns. Composition of the entire image may thus be represented by the combination of the active columns. I suggest that it is this combination that allows two thousand columns to represent an almost infinite number of objects in the natural world, much like only a few tens of phonemes are enough to express the enormously rich ideas in a spoken language. Heterogeneity of neuronal selectivities to fine parameters within a column could contribute to stable encoding of particular object-features in the presence of changes in illumination or viewing angles.

Fujita, I., Tanaka, K., Ito, M. and Cheng, K. (1992) Columns for visual features of objects
in monkey inferotemporal cortex. Nature 360:343-346.