The accurate prediction of droplet sizes is fundamental in many industrial applications. In order to be able to simulate the evolution of a size distribution, suitable kernels should be used in the population balance. In this paper we make use of four different breakage kernels in order to predict a size distribution and compare the results among them as well as with experimental data. Two breakage kernels are derived from inverse problems and the other two are derived from physical or empirical interpretations. The Sauter mean diameter, a moments-based error and Gaussian shape factors were used to compare between the resulting distributions. (C) 2008 Elsevier Ltd. All rights reserved.