PATTERN RECOGNITION (also PATTERN MATCHING)
The establishment of a one-to-one match between, on the one hand, a set of features drawn from a stimulus and held temporarily in a sensory store; and, on the other, a stored representation in the mind. Pattern recognition enables us to identify familiar patterns (e.g. letter shapes) with a high degree of automaticity, and to impose patterns upon unfamiliar forms (as when users of the Latin alphabet are exposed over time to signage in the Greek or Russian ones).
Approaches to pattern recognition offer different accounts of how a pattern is stored in long-term memory. They include:
Template matching theories: where the pattern is matched with an exact counterpart in long-term memory. This would appear to entail storing all possible variants of the pattern– a very inefficient solution since, in order to recognise (say) the letter E, one would have to store it not only in all possible fonts but also in all possible sizes. The solution is to assume a two-stage process, where a stimulus is normalised, with non-essential features edited out, before being matched to the template.
Feature analysis theories: where the pattern is broken into constituent parts; and is identified as a combination of those features. For example, a small number of distinctive features (lines, curves etc.) would allow us to identify all the letters of the alphabet. There is evidence that the visual cortex in mammals is so organised as to detect the presence of simple features within a complex pattern.
Prototype theories: where the pattern is compared on a ‘best fit’ basis with an idealised example of the pattern in long-term memory.
See also: Normalisation, Phonological representation, Prototype Theory
Further reading: Anderson (1990: Chap. 3); Kellogg (1995: Chap. 2); Lund (2001: Chap. 5)
