Abstract The crystallinity of bone mineral represents an established method of measuring heat-induced change and is of importance to fields including material science, clinical science, anthropology and archaeology. A commonly used technique involves the calculation of the Crystallinity Index (CI) using selected peaks from Fourier Transform Infrared (FTIR) absorbance spectra. However, the choice of peaks has to date not been statistically justified. In this study a novel application of statistical techniques to the FTIR spectra of bone samples burned in the range 100 °C–1100 °C has been used to identify 5 new spectral indices of heat-induced crystallinity change. The validity of the new indices was tested by using a statistical classification model (LDA) to predict the burning temperature of a set of 108 bone samples burned between 100 °C and 1100 °C. A correct classification rate (CCR) of 97.2% was obtained when a combination of 4 indices including the CI was used. This was significantly better than the CCR of 66.7% which was obtained when the CI was used on its own.