A time-dependent parameter, the oxidative phosphorylation, P/O ratio, was estimated in a metabolic reaction model of aerobic yeast fermentation by two methods and the accuracies of the estimates were compared. In the first method, P/O was estimated according to on-line physiological state recognition results by a fuzzy inferencing method based on error vectors. The amount of cell growth was then calculated using the metabolic reaction model with the estimated P/O. In the second method, P/O was estimated adaptively in the metabolic reaction model using the pseudo-inverse matrix (least square form for total residual error minimum). When the two methods were compared with respect to the on-line estimation of cell growth and the ease of use on-line, the method based on fuzzy inferencing was better than that using the pseudo-inverse matrix.