It is very important to evaluate the positioning accuracy of NC machine tools, so that the evaluation method is standardized by ISO. It is, however, found from our experiments that two test cycles recommended by ISO gave different results, although only one permissible deviation is provided in the test code. In this study, the positioning accuracy has been precisely investigated, varying the test cycle, the feedback system and the feed rate of a machining center. From the experiments, it is found that the linear cycle gives larger deviation, which increases with the increase of the number of measurements, than the quasi-pilgrim step cycle. The increment of the deviation cannot be neglected in the quasi-pilgrim step cycle or the closed loop controlled system using a linear scale. Therefore, is it necessary to consider the thermal effect as well as the random and systematic errors when evaluating the positioning accuracy. Then, the effectiveness of an alternative method proposed in this paper is confirmed using several machining centers.