The shoulder is the most mobile joint of the human body due to bony constraint scarcity and soft tissue function unlocking several degrees of freedom (DOF). Clinical evaluation of the shoulder range of motion (RoM) is often limited to a few monoplanar measurements where each DOF varies independently. The main objective of this study was to provide a method and its experimental approach to assess shoulder 3D RoM with DOF interactions. Sixteen participants performed four series of active arm movements with maximal amplitude consisting in (1) elevations with fixed arm axial rotations (elevation series), (2) axial rotations at different elevations (rotation series), both in five planes of elevation, (3) free arm movements with the instruction to fill the largest volume in space while varying hand orientation (random series), and (4) a combination of elevation and rotation series (overall series). A motion analysis system combined with an upper limb kinematic model was used to estimate the 3D joint kinematics. Thoracohumeral Euler angles with correction were chosen to represent rotations. The angle-time-histories were treated altogether to analyze their 3D interaction. Then, all 3D angular poses were included into a nonconvex hull representing the RoM space accounting for DOF interactions. The effect of series of movements (n = 4) on RoM volumes was tested with a one-way repeated-measures ANOVA followed by Bonferroni posthoc analysis. A normalized 3D RoM space was defined by including 3D poses common to a maximal number of participants into a hull of average volume. A significant effect of the series of movements (p < 0.001) on the volumes of thoracohumeral RoM was found. The overall series measured the largest RoM with an average volume of 3.46 ± 0.89 million cubic degrees. The main difference between the series of movements was due to axial rotation. A normalized RoM hull with average volume was found by encompassing arm poses common to more than 50% of the participants. In general, the results confirmed and characterized the complex 3D interaction of shoulder RoM between the DOF. The combination of elevation and rotation series (overall series) is recommended to fully evaluate shoulder RoM. The normalized 3D RoM hull is expected to provide a reliable reference to evaluate shoulder function in clinical research and for defining physiologic continuous limits in 3D shoulder computer simulation models.