It is consistently found in wet ball milling, under the conditions normally used in production mills, that the larger sieve size intervals of the product size distributions do not contain as much material as predicted by the first-order mill simulation models. Four methods of correcting for this effect were investigated. Method 1 uses a hypothetical exit size classifier, Method 2 considers different axial flow velocities for different sizes, Method 3 uses modified specific rates of breakage and Method 4 allows for non-first order grinding kinetics. The methods were tested on literature data from mills of 0.56, 0.82 and 3.35 m in diameter, wet grinding the same phosphate ores in open circuit. The four methods were equally capable of empirically correcting for the effect but the characteristic parameters of the fitting equations for Methods 1, 2 and 3 had to be varied with the fineness of grinding, that is, the apparent effective specific rates of breakage were not constant. It was concluded that the effect is primarily due to non-first order grinding kinetics, but the correction procedure in this case is complex, whereas the other methods are simple to apply. (C) 2004 Published by Elsevier B.V.