Many researchers have proposed different rules of combining conjunctive or disjunctive clauses: for example, instead of taking the minimum or the maximum of the membership functions, they take the arithmetic or the geometric mean. These rules are arbitrary, and there are lots of them. It is possible, if we have enough training data, i.e. conditions and class assignments by the experts, to train our system so that it chooses the best rule that fits the way of reasoning of the expert that did the classification.

Another disadvantage of the rules discussed earlier is that they give the same importance to all factors that are to be combined. For example, it is possible that the role of soil depth or rock permeability is not of the same importance to soil erosion as the role of slope. This issue can be resolved if we do not insist on all membership functions taking values between 0 and 1.