The two types of imperfection in knowledge have not had the same importance in the interest of scientists. Pascal and Fermat tackled uncertainty through probability theory from 17^th century. However, this theory neither allows subjective belief to be dealt with, as it had been supposed for a long time, nor allows the problem of imprecise and uncertain knowledge to be solved.

This problem has only been taken into account since 1965, when L. A. Zadeh, professor at the University of California at Berkeley, internationally known for its work on system theory, introduced the notion of ``fuzzy set''. Note that the term ``fuzzy set'' is usually used instead of ``fuzzy subset''.

Zadeh started from the ideas of:

• partial membership to a class,
• categories with poorly defined boundaries,
• gradual change from a situation to another.

He generalized the classical set theory, by allowing intermediate situations between the whole and nothing. The developments of this concept provide means of representing and handling imperfectly described knowledge. They build an interface between data described symbolically (with words) and numerically (with figures).

Fuzzy logic allows a reasoning on such knowledge. The possibility theory which was introduced in 1978 by L.A. Zadeh, constitutes a framework making it possible to deal with non-probabilistic concepts of uncertainty. Through the notion of fuzzy sets, it allows, using the same formalism, imprecision and uncertainty to be dealt with.