Water quality classification for stream has been major tool for water quality management in Korea. This paper examines the application of the fuzzy inference mechanism to develop a fuzzy expert system for proper determination of WQCS from uncertain and imprecise ecological information. This study proposes a rule matrix composed of seven water quality grades, toxicity of water and rarity of cases. From this rule matrix, 30 rules for WQCS determination are generated. From the comparison of performance of the fuzzy expert system and the conventional expert system for the determination of class, toxicity, and rarity, it seems that the smoothly varying curves of WQCS determination from the fuzzy expert system represent our real-world experience more realistically than stepwise curves from the conventional expert system.