The reproducibility of the three-dimensional (3D) kinematic aspects of motion coupling patterns during manual mobilizing techniques is still a debatable matter. The present in vitro study analysed segmental 3D motion of the atlanto-axial joint during manual axial rotation mobilization. Twenty fresh frozen human cervical specimens were studied in a test-retest situation with two examiners. The specimens were manually mobilized using three different techniques: (1) a regional mobilization technique of the cervical spine; (2) a segmental mobilization technique of the atlas with manual fixation of the axis; and (3) a segmental mobilization of the atlas on the axis applying a locking technique. Segmental atlanto-axial kinematics was registered with a Zebris CMS-20 ultrasound-based tracking system. The Euclidian norm was used as a representation of overall 3D motion. The results indicated good reproducibility (mean intraclass correlation coefficient, ICC: 0.87). Intraobserver reproducibility was slightly higher (mean ICC: 0.91; range: 0.76-0.99) than interobserver reproducibility (mean ICC: 0.85; range: 0.56-0.98) (P , 0.05). The total range of motion expressed as the Euclidean norm of 3D motion components was a parameter with good reproducibility in the study of segmental kinematics of manual atlanto-axial mobilization. Although previous studies have demonstrated poor inter-rater reliability of manual examination and mobilization of segmental motion components, the results of the present study shed a new and more positive light on the reproducibility of techniques for manual mobilization of the upper cervical spine.