Some of the most exciting possibilities for dielectric elastomer artificial muscles consist of biologically inspired networks of smart actuators working towards common goals. However, the creation of these networks will only be realised once intelligence and feedback can be fully distributed throughout an artificial muscle device. Here we show that dielectric elastomer artificial muscles can be built with intrinsic sensor, control, and driver circuitry, bringing them closer in capability to their natural analogues. This was achieved by exploiting the piezoresistive behaviour of the actuator's highly compliant electrodes using what we have called the dielectric elastomer switch. We developed suitable switching material using carbon loaded silicone grease and experimentally demonstrated the primitives required for self-sensing actuators and digital computation, namely compliant electromechanical NAND gates and oscillator circuits. We anticipate that dielectric elastomer switches will reduce the need for bulky and rigid external circuitry as well as provide the simple distributed intelligence required for soft, biologically inspired networks of actuators. Examples include many-degree-of-freedom robotic hearts, intestines, and manipulators; wearable assistive devices; smart sensor skins and fabrics; and ultimately new types of artificial muscle embedded, electromechanical computers. © 2010 Springer-Verlag.