Discrimination between different rival models for describing the inhibitory effect of ethanol both on yeast growth and on fermentation was studied for a continuous process of alcoholic fermentation in a tower reactor with recycling of flocculating cells. Models tested include linear, parabolic, hyperbolic, exponential, and generalized nonlinear power-law types. The best expressions were identified under the criteria that all the kinetic parameters should assume acceptable values in a feasible range and should result in the best fit of the experimental data. The kinetic parameters were estimated from steady-state data of several sugar concentrations in feeding stream (S-0 = 160, 170, 180, 190, 200 g/L), constant dilution rate (D = 0.2 h(-1)), recycle ratio (alpha = 13.6), and temperature (T = 30 degrees C). The best model for the yeast growth was of power-law type, whereas for the product formation the best model was of linear type. These models were able to reproduce the trends of the process variables satisfactorily.