Current upper-limb prostheses are often rejected as a result of high activation forces that are needed for control. Another reason for rejection is the absence of appropriate proprioceptive feedback. In this study, a pneumatic control system was designed that decreased the activation forces and where proprioceptive feedback was provided. The system was placed on the back of the user and was actuated by shoulder movements. The design has been evaluated on two aspects, namely the sensitivity, determined by the Just Noticeable Difference (JND) and the Weber Fraction (WF), and the displacement (re-)production accuracy (e.g. opening/closing of the prosthetic hand). Ten healthy, right-handed male subjects performed both experiments. The sensitivity was assessed by a reminder task where two forces were compared. Four reference force levels were used: 2, 4, 6, and 8N. This resulted in WF values of: 18% for 2N, 4% for 4N, 3% for 6N, and 2% for 8N. The results for the three higher forces agreed with literature. To evaluate the displacement accuracy the Absolute Displacement Error (ADE), the Relative Displacement Error (RDE), and the Displacement Variability (DV) were measured for three different reference displacements: 5mm, 10mm, and 20mm. The results show that the control of the system is accurate enough for object manipulation and the relative error was lower than 1% for each reference displacement. This design may be the next step in controlling prosthetic hands.