Investigation by Raman spectroscopy of samples from different geological settings shows that the occurrence of TiO2 polymorphs other than rutile can hardly be predicted, and furthermore, the occurrence of anatase is more widespread than previously thought. Metamorphic pressure and temperature, together with whole rock chemistry, control the occurrence of anatase, whereas variation of mineral assemblage characteristics and/or fluid occurrence or composition takes influence on anatase trace element characteristics and re-equilibration of relict rutiles. Evaluation of trace element contents obtained by electron microprobe in anatase, brookite, and rutile shows that these vary significantly between the three TiO2 phases. Therefore, on the one hand, an appropriation to source rock type according to Nb and Cr contents, but as well application of thermometry on the basis of Zr contents, would lead to erroneous results if no phase specification is done beforehand. For the elements Cr, V, Fe, and Nb, variation between the polymorphs is systematic and can be used for discrimination on the basis of a linear discriminant analysis. Using phase group means and coefficients of linear discriminants obtained from a compilation of analyses from samples with well-defined phase information together with prior probabilities of groupings from a natural sample compilation, one is able to calculate phase grouping probabilities of any TiO2 analysis containing at least the critical elements Cr, V, Fe, and Nb. An application of this calculation shows that for the appropriation to the phase rutile, a correct-classification rate of 99.5% is obtained. Hence, phase specification by trace elements proves to be a valuable tool besides Raman spectroscopy.