The task of enumerating the number of 3-D cube arrays is used to evaluate and develop student’s spatial ability. This research investigates the roles of spatial orientation and visualization of sixth graders in cube enumeration. The material included two types of cube arrays – low regularity and high regularity. We tested two independent variables in low regularity arrays – number and distribution of hidden cubes. The former had four levels (i.e., 4, 5, 6, or 7) of hidden cubes. The latter were divided into two axes and three axes according to the degrees of the hidden cubes. High regularity arrays were divided into outside intact and non-intact types depending on whether the appearance is integral or not. Paper-and-pencil questionnaire was group-administered to collect the hit rate of 204 students. Individual interviews were also conducted to gather data on hit rate, response time, difficulty rating, and strategies from 40 students. Results showed that high regularity arrays yielded better performance than low regularity arrays, and two axes condition yielded better performance than three axes condition. The effects of the number of hidden cubes are not as steady. There are no significantly differences in hit rate and response time between the intact and non-intact groups. Current literature (Battista et al., 1996, 1998, & 1999; Ben-Chaim et al., 1985) has claimed that students’ difficulty in cubes enumeration is due to deficiencies of mental imagery or orthogonal coordination. However, we argue that visualization is more influential than spatial orientation.