PROTOTYPE THEORY
A theory that linguistic representation in the mind (particularly the representation of lexical concepts) takes the form of idealised or highly typical examples to which entities in the real world can be compared for ‘goodness of fit’.
Prototype Theory evolved as a way of dealing with the problem of how weform and apply lexical categories. The American psychologist Eleanor Rosch devised a set of categories (FURNITURE, FRUIT, VEGETABLE, BIRD etc.) and asked subjects to rate around 50 examples of each according to how good an example of the category they felt them to be. The results were strikingly consistent, especially for items rated as good examples. Rosch formed the conclusion that we allocate real-world entities to categories by comparing them to a central exemplar, one which possesses all of the features associated with the category (for BIRD, robin, for VEHICLE, car). The closer an entity is to this prototype, the more confident we are about including it in the category.
There are a number of problems with Prototype Theory:
Rosch’s findings may have simply reflected frequency– with subjects choosing the most common examples of a category.
It was assumed that prototypes were chosen because they possessed a bundle of features which were characteristic of the category as a whole. However, this certainly did not apply to VEGETABLES. The prototypical vegetable was a pea but a carrot (different in size, shape, colour) was rated highly.
Prototypes may be culturally determined. When the Rosch experiment was repeated with subjects of Latin American origin, using Spanish words, very different results were obtained. This raises the question of whether prototypes are to some extent determined by native language, local culture or physical environment.
Later experimenters asked subjects to rate odd and even numbers for typicality and found that they were prepared to do so, despite the fact that the task was patently nonsensical. Thus, 2 was rated as a ‘prototypical’ even number and 806 as a not very typical one. It was therefore argued that we need to distinguish between a conceptual core (‘an even number is divisible by two’) and an identification function (how easy it is to recognise a real-world example as a member of a category). This accords with research which suggests that infants begin by classifying the real world in terms of characteristic features (an UNCLE is somebody who gives presents) and only later recognise defining features (an UNCLE is a parent’s brother).
Rosch’s findings were based upon a limited set of categories. It is possible to demonstrate a prototype effect for categories of concrete noun, but is not so easy to do so for other parts of speech. Nevertheless, Prototype Theory has proved influential and has been extended to other domains. For example, one suggested solution to the problem of phonological representation is that we store in our minds a prototype of each phoneme, an idealised form against which can be matched the many variants that we encounter in connected speech.
Research into how infants form conceptual categories like DOG has suggested that they may establish central points of reference in the form of prototypes. A provisional prototype might be based upon the first exemplar of a word which they encounter or upon the exemplar most referred to by adults. The infant then applies the word to other entities which appear to share a ‘goodness of fit’ with the prototype.
An alternative to Prototype Theory is provided by recent exemplar models. These suggest that, instead of classifying real world entities in relation to a single best example, we store multiple traces of all the exemplars that we have met.
See also: Concept formation, Fuzzy boundary, Pattern recognition
Further reading: Aitchison (2003: Chap. 5); Harley (2001: 288–91); Keil (1987)
