We propose an adaptive control and an adaptive neural network control (composed of two RBF neural components and one adaptive component) for tendon-driven robotic mechanisms with elastic tendons. These controllers can be applied to serial or parallel tendon-driven manipulators having linear or non-linear elastic tendons. We begin by proving the stability of the adaptive control system for our mechanism, and then we prove the stability of the adaptive neural network system and report on the results of numerical simulations and experimental results performed using a 2-DOF tendon-driven mechanism having six elastic tendons. (C) 2003 Elsevier Ltd. All rights reserved.