Conventional fuzzy controllers are employed for the purpose of designing nonlinear key field controllers. Fuzzy control methods utilise existing qualitative know-how or experience to find favourable key field controllers. However, some empirical knowledge cannot be processed adequately employing known fuzzy controllers. This paper focuses on the introduction of a new class of defuzzification methods differing from conventional methods in being nontranslation-invariant. Reasons are put forth to justify why this method is advantageous in certain applications. A simple version of these new defuzzification rules is presented in the form of the torque method.