This study examined eight previously published ear prediction methods by Welcker, Gerasimov, Fedosyutkin and Nainys, and Broadbent and Mathews. Computed tomography scans of 78 living adults (n(1)) did not support any of these previously published recommendations. Free earlobes were found to accompany protruding supramastoid crests (Pearson's χ² < 0.05); and ear length [l] and width [w] differed by sex (p < 0.05), correlated with age (r = 0.38[l]; 0.32[w]), and correlated with facial height (r = 0.37[l]; 0.30[w]). New regression equations (for ear length and width) were generated using these variables in several samples and, where possible, cross-validated using independent data (n(1) = 78, n(2) = 2190, n(3) = 1328, n(4) = 1010, and n(5) = 47). As a result of these analyses, four valid and tested methods for ear prediction were identified, but large degrees of error continue to make accurate prediction of the ear, from the skull, problematic.