We look at the process of defuzzification used in fuzzy logic controllers. We provide a parametrized formulation for this procedure based upon the use of level sets which we call generalized level set defuzzification. We show that the commonly used defuzzification procedures, mean of maxima and center of area, are special cases of this general procedure and are simply distinguished by the choice of the parameter. We provide a simple algorithm for adaptive choice of the best defuzzification procedure.