The problem with the available cetane indices is that they give significant errors in the prediction of cetane number of a fuel that is very different from the fuels used to derive the index. This stems from the fact that the way physical properties are used in the correlations does not properly describe the fuel chemistry. This can be overcome, to a certain extent, by use of proper combinations of the common physical properties. Our investigations, to base cetane index correlations on sound models, have resulted in several parameters in terms of some combinations of the common physical properties, which can show the degree of branchiness of a given diesel fuel. Through the use of these parameters a predictive equation has been developed and compared to the existing correlations and to the ones that are being developed. The proposed equation provides substantial improvements over the available cetane index correlations. The data base for this study consists of nearly 400 fuels. 70 of these fuels are from an exchange program with multiple engine cetane ratings, 260 are from a CGSB study, and the remainder of the fuels are from various sources with diversified properties.