OBJECTIVE To compare hard-tissue morphology by cephalometric measurements between extraction and non-extraction orthodontic treatment in borderline cases. METHODS The samples consisted of 33 cases selected as borderline cases by 5 orthodontic specialists. They were divided into 21 extraction cases (including 13 four first premolar extraction cases and 8 second premolar extraction cases) and 12 nonextraction cases by checking patients' treatment records. Conventional cephalometric analysis was made to compare hard tissue structures before and after orthodontic treatment and the same comparison was made between two different extraction patterns. RESULTS No statistical difference was found on pretreatment hard-tissue morphology between extraction and non-extraction groups divided from borderline cases. The SNB angle of the four first premolars extraction group was smaller than that of the four second premolars extraction group by (4.0+/-1.3) degrees P<0.01 . The statistical significant differences of the post-treatment hard-tissue cephalometry between extraction and nonextraction and between two different extraction patterns were found to be limited to the items related to the tooth position. CONCLUSIONS Conventional cephalometric analysis cannot differentiate extraction or non-extraction treatment selection from borderline cases. The differences between extraction and non-extraction orthodontic treatments were limited to tooth positions but had no skeletal relation. Extraction of the four second premolars could prevent upper incisors from uprighting as it often follows the four first premolar extraction treatment. Extraction of the four first premolars could change the incisor position much more than extraction of the four second premolars. The changes of four second premolar extraction treatment were much similar to non-extraction treatment.