Because the physics that underpin many minerals processing operations is often complex, many researchers find it convenient to give empirical descriptions of their experimental observations. However, such empirical correlations are often given in a dimensionally inconsistent form which means that their utility is limited. In this article, the steps that must be taken to develop a correlation that has the required number of degrees of freedom that are necessary and sufficient to adequately characterise a system and retain dimensional consistency are presented. This methodology is supported by four examples of the rigorous development of empirical correlations that may be of interest to minerals processing engineers. (c) 2007 Published by Elsevier Ltd.