Pneumatic artificial muscles (PAMs) are high power-toweight ratio actuators with considerable potential in biomimetic robotics and orthotics due to their similarities with human skeletal muscle. However, precise position control is difficult to achieve due to the highly nonlinear nature of the actuators and the pneumatic systems driving them. A wide range of nonlinear controllers have been proposed to-date, but none have been shown to be entirely satisfactory, and are often optimised for only one region of the PAM's travel. In this paper, a gain-scheduled position controller is designed that aims to achieve equal tracking performance across the entire travel of the PAM. Selected scheduling variables include actuator displacement and error direction, with controller gains defined by 'normalisation curves'determined by data from open-loop characterisation tests. The experimental system consists of a Festo PAM, a pair of on-off valves driven by pulse-width modulated signals, and sensors for PAM displacement and pressure. Controller performance is tested using several dynamic position tracking tests, and the results are compared to an equivalent linear controller. The gain-scheduled approach successfully counteracts the differing inflation / deflation dynamics of the system, showing improved tracking performance over the linear controller with considerably less variability due to operating region.