For a single d electron in a trigonal environment, the complete ligand-field Hamiltonian, including the interactions with the upper states of the cubic splitting, has been diagonalized in order to re-examine critically the theory of the EPR spectra of various Ti3+ compounds. The compounds for which there are adequate experimental data to justify a detailed discussion of the parametric theory of the electronic structure are titanium in corundum, cesium titanium alum, titanium acetylacetonate, and titanium-doped aluminum trichloride. Only for the first of these is it possible to fit the magnetic properties with crystal-field parameters. In each of the others, covalency is qualitatively indicated, but it is not possible from the present experimental data to give good numerical values either for the trigonal components of the effective potential or for the parameters describing covalency. We suspect that similar uncertainty exists in many of the published interpretations of iron-group compounds.