Objectives: (1) To identify any differences in the accuracy of the prediction of the summed widths of unerupted permanent canines and premolars for males vs females and, if differences are identified, (2) to examine the sex differences in the best combination of independent variables and (3) to evaluate the practical accuracy using our models for both sexes, respectively. Materials and Methods: Records for 200 orthodontic patients (100 males and 100 females) with discernible tooth crowding in the permanent dentition were selected. Mesiodistal tooth crown widths, dental arch widths, and basal arch lengths were measured. Predicted sum widths of permanent canines and premolars in both dental arches were calculated using stepwise regression analysis. Results: Among male subjects, the best combinations of independent variables were the widths of the upper central and lateral incisors (U1, U2) and the first molar (U6) for the maxilla (R 2 5 0.615) and the mandible, lower central and lateral incisors (L1, L2), and the first molar (L6) (R 2 5 0.685), whereas in females, the best combinations for prediction were found to be U1, U2, U6, and the maxillary arch width for the maxilla (R 2 5 0.429) and L1, L2, L6, and the basal arch length for the mandible (R 2 5 0.426). Correlation coefficients between each independent and dependent variable in the female subject group were found to be weaker than those for the male subject group. Conclusions: We found the sex differences in the accuracy of prediction and the best combination of independent variables for predicting the summed widths of unerupted canines and premolars. These new models should be clinically useful, providing highly accurate prediction. (Angle Orthod. 2011;81:938-944.)