This paper describes the application of self-sensing control to a cardiac assist device. We propose to improve the pumping performance of diseased or weakened hearts by applying direct cardiac compression using artificial muscle. This particular application imposes strict limitations on size, weight and system complexity, therefore employing self-sensing could offer advantages over separate sensors and actuators. Many electromagnetic actuators produce a back-e.m.f. proportional to velocity. Using a simple system model, it is possible to separate this back-e.m.f. from the supply voltage, thus the actuator velocity can be self-sensed. Furthermore, using a more detailed model, it also is possible to self-sense the force being applied. Experimental results are presented for linear moving-coil actuators and miniature d.c. motors. Estimation of position has been performed by numerical integration of self-sensed velocity, and shown to compare favourably to data from displacement sensors. Force estimation has also been shown to closely agree with data from a load cell. Combined force and position control has been implemented, without using sensors. Unfortunately, since self-sensed position is derived by integrating velocity, the estimated position can suffer from drifting. An automatic re-calibration scheme is proposed for the cardiac assist application.