Simple and complex cells, cortical columns
In January 1962 Hubel and Wiesel divided the fields of 303 cells in the cat's visual cortex into simple and complex: a complex cell answered to a correctly oriented slit, edge or bar anywhere in its field. The cortex turned out to be built of columns of cells sharing one orientation.
Why it matters
The paper proposed a hierarchy: simple cells take their input directly, complex cells from several simple cells in the same column, and so tolerate a shift of the stimulus. It is the scheme the neocognitron of 1980 named as its model.
Recordings in forty cats; 233 of the 303 cells had simple fields. The most effective stimuli were long narrow slits of light, straight borders between areas of different brightness and dark bars; orientation was critical. Four fifths of cells were influenced by both eyes. Columns of common field-axis orientation ran from the surface to the white matter, about 0.5 mm across at the surface. That complex cells receive input from simple ones is offered as a suggestion, not a measurement. The full text was read in a copy of the scan in New York University course readings.