The Reference Series method of Shacham et al. is modified to enable prediction of normal melting point temperatures (T-m) of pure substances within experimental error level. A homologous series for which large amount and high precision T-m data are available is used as a reference series. To predict T-m for a target series quantitative property-property relationships (QPPRs) are derived to represent the predicted T-m values of the target series in terms of the T-m values of the reference series. Two QPPRs are necessary in order to match the odd and even carbon number (n(C)) oscillations of T-m in the low n(C) region. In the high n(C) region, the QPPR is adjusted to represent correctly the asymptotic behavior of T-m. It is shown that the method is very useful for consistency analysis of T-m data and enables a reliable prediction of T-m in both the low n(C) and the high n(C) regions. (c) 2013 American Institute of Chemical Engineers AIChE J, 59: 3730-3740, 2013