Bottom-Up Processing

Abstract

Any form of information processing that is initiated, guided, and determined by input and that proceeds in sequential stages, with each stage coming closer to a final interpretation than the last, as in computational theories of vision that proceed from raw sensory data to more abstract cognitive operations. A clear example of bottom-up processing is provided by feature detection theory. This form of processing allows the possibility of learning from experience during the processing procedure. Bottom-up theories of perception are theories according to which low-level sensory features (1) of a stimulus are first recognized and then built up, with the help of memory and existing schemata, into higher-order perceptions. The term was introduced in 1975 by the US psychologists Donald A(rthur) Norman (born 1935) and David E(verett) Rumelhart (1942–2011). Also called data-driven processing. See also analysis by synthesis. Compare top-down processing.