We consider joint position tracking of a desired reference trajectory for a n-dof robotic manipulator, utilizing only joint position measurements. The proposed controller is structurally and computationally simple and achieves prescribed transient and steady state performance bounds on the joint position tracking error. Knowledge of the actual robot dynamics is not required. Furthermore, no approximating structures (i.e., neural networks, fuzzy systems) are utilized to compensate for the absence of such knowledge. The performance of the controller designed is demonstrated via simulations and experiments conducted on a KUKA LWR4+ robotic manipulator.