In this paper we study the problem of identifying linear and nonlinear feedback mechanisms, or controllers, operating in closed loop. A recently developed identification method for nonlinear systems, the LAVA method, is used for this purpose. Identification data is obtained from inertial sensors, that provide information about the movement of the system, in the form of linear acceleration and angular velocity measurements. This information is different from the information that is available to the controller to be identified, which makes use of unknown internal sensors instead. We provide two examples, a simulated neuromuscular controller in standing human balance, and a lead-filter controlling a physical position servo using a DC motor. Both linear and nonlinear controllers are used in the examples. We show that the LAVA method is able to identify sparse, parsimonious models of the controllers.