The Parietal Cortex in Sensemaking: The Dissociation of Multiple Types of Spatial Information
Figure 7
An illustration of numerosity detection (a) and size detection (b). In both figures, each unit on the hidden layer takes the sum of the visual input units, hence the name “summation units”. For simplicity, we show only three units as the visual inputs and two units on the hidden layer. In representing numerosity, a summation unit takes the sum from each visual input unit uniformly (i.e., with equal connection weights). Thus, the activation of such a unit only responds to numerosity monotonically, and the spatial information is completely discarded. Different summation units have different connection weights from visual inputs, and their combined activation pattern is projected to the final tuned numerosity detectors that are ultimately selective to numerosity. In contrast, the summation units for encoding size (or distance between the two furthest active visual units) must receive nonuniform weights selectively from different spatial locations in the visual inputs (i.e., with unequal connection weights) in order to preserve the spatial information.
