Statistical tools were used to analyze membrane fouling occurring during ultrafiltration (UF) of clarified grape juice. Experimental flux data were correlated to dead-end and crossflow filtration models. The UF experiments were performed at 30 and 40 degrees C for an 8-kDa multitubular ceramic membrane and also at 30 degrees C for a 1-kDa membrane to determine the effect of pore size. ANOVA was used for the statistical analysis of fluxes. The modified Kolmorov-Smirnov and the Shapiro-Wilk tests were used to evaluate the normality of the residuals obtained from each model, thus allowing validation or rejection of initial adjustments given by a simple R-square statistical test. Using the standard error of the estimate and the mean absolute error it was possible to identify the best model for each case. Differences in the permeate flux and fouling mechanisms between both temperatures and pore size were discussed.