This paper presents a fuzzy diagnosis support system for fatty liver and an algorithm of membership function acquisition in consequent clause in this system. This system shows the possibility of fatty liver using fuzzy inference from blood tests. In constructing this fuzzy system, the expert knowledge must be represented by fuzzy if-then rules. However, the membership function acquisition in the fuzzy if-then rule is time consuming owing to less information about the weight of each clinical fatty liver diagnostic rule. This algorithm calculates the relative positions and widths of membership functions in consequent clause using the probabilities of fuzzy events and the specificities of fuzzy sets in antecedent clause. Furthermore, the membership functions are detected by optimization using learning data. The input of this system is the values of five blood tests: ChE, GOT, GPT, ALP and LAP. The output is the possibility of fatty liver. To evaluate the performance of our system, we have applied it on 59 cases. Our system correctly diagnosed 30 cases as fatty liver, because the possibility of fatty liver was one. These 30 cases were diagnosed as fatty liver by both the ultrasonic imaging and blood test methods. For the remaining 29 cases our system showed the possibility of existence of fatty liver, where ultrasonic imaging diagnosed them as fatty liver, but the blood test method evaluated them as normal. We would expect the usefulness of this system to be evident in fatty liver diagnosis using the blood test data.