Best-Fit Principle

last changed in 2002

The Best-Fit Principle is the greatest mystery of linguistics, and maybe of all cognitive science. It is the core of pragmatics, the set of processes by which we apply our knowledge (including grammar) to particular cases. It says that if we want to understand a particular piece of experience E, we look in our knowledge for the stored concept which best fits E, and assume that E isa this concept. Having made this initial link we can then apply default inheritance and guess vast amounts of information, so the link is vital to the whole of our understanding.

In relation to spoken language, we hear a sound and decide it is an example of /b/, or we hear a more complex pattern of sounds and decide it was an example of the word breakfast; but how do we do it? It can't be that we take our input experience and look through our knowledge for an exact match, because our experiences are often distorted in one way or another - e.g. you can recognise an example of breakfast even if the speaker is drunk, has a cold, etc. So we must be looking for the best match, given everything else we know about the input - e.g. what the speaker is likely to be talking about. But does that mean that we must consider all possible interpretations in order to pick the best? That can't be so; but if we don't do that, what do we do - and how do we do it so quickly? (For discussion see Hudson 1990 pp. 45-52.)

The most likely solution to the problem is spreading activation: the best fit is with the most active concept - the one that receives most activation from the input experience.







This is an article from the Encyclopedia of Word Grammar and English grammar. If you refer to it, please give the url as "".