The study proposes a rigid-body biomechanical model of the trunk and whole upper limb including scapula and the test of this model with a kinematic method using a six-dimensional (6-D) electromagnetic motion capture (mocap) device. Large unconstrained natural trunk-assisted reaching movements were recorded in 7 healthy subjects. The 3-D positions of anatomical landmarks were measured and then compared to their estimation given by the biomechanical chain fed with joint angles (the direct kinematics). Thus, the prediction errors was attributed to the different joints and to the different simplifications introduced in the model. Large (approx. 4 cm) end-point prediction errors at the level of the hand were reduced (to approx. 2 cm) if translations of the scapula were taken into account. As a whole, the 6-D mocap seems to give accurate results, except for prono-supination. The direct kinematic model could be used as a virtual mannequin for other applications, such as computer animation or clinical and ergonomical evaluations.