Expert systems of a certain kind rely essentially upon the availability of a method for handling uncertainty. These systems cannot be conceived without a decision being firstly made about the choice of this method. Obviously this is true for all expert systems using empiricaJ knowledge which in itself is not absolutely certain. As an example, we could mention a medical expert system which draws conclusions from the observed symptoms about whether or not a certain disease is present. All conclusions of this type unavoidably contain an amount of uncertainly. The rules which lead to these conclusions should not be confused with logical rules and must not be treated in the same way.