The purpose of this study was to develop new standards for determining the sex of fragmentary human skeletal remains. We measured height, width, and length of the mastoid process in medieval to early modern Japanese skeletons, from the Yuigahama-minami and Hitotsubashi sites, in order to provide a metric standard for the diagnosis of sex using the mastoid process. We calculated discriminant functions based on these measurements; the accuracy of sex classification was over 80% using a single variable, and reached 82-92% with two variables, mastoid height and width. This accuracy is equal to or better than that reported by some previous studies of sex determination using the cranium. However, when we examined intra-and interobserver errors in the mastoid process measurements, we found a high level of errors, and this highlights the difficulty involved in intraobserver repeatability and interobserver reproducibility. Our results imply that, in order to achieve reliable results of sex determination using the mastoid process, the measurement methods need to be carefully determined and executed.