Synthese Library Volume 188, 1987, pp 115-141

Shifts in Selective Visual Attention: Towards the Underlying Neural Circuitry

* Final gross prices may vary according to local VAT.

Get Access

Abstract

A number of psychophysical studies concerning the detection, localization and recognition of objects in the visual field have suggested a two-stage theory of human visual perception. The first stage is the “preattentive” mode, in which simple features are processed rapidly and in parallel over the entire visual field. In the second, “attentive” mode, a specialized processing focus, usually called the focus of attention, is directed to particular locations in the visual field. The analysis of complex forms and the recognition of objects are associated with this second stage.1 The computational justification for such a hypothesis comes from the realization that while it is possible to imagine specific algorithms performing tasks such as shape analysis and recognition at specific locations, it is difficult to imagine these algorithms operating in parallel over the whole visual scene, since such an approach will quickly lead to a combinatorial explosion in terms of required computational resources.2 This is essentially the major critique of Minsky and Papert to a universal application of perceptrons in visual perception.3 Taken together, these empirical and theoretical studies suggest that beyond a certain preprocessing stage, the analysis of visual information proceeds in a sequence of operations, each one applied to a selected location (or locations).