Audible-frequency sounds from the human body are an invaluable diagnostic tool. For over 200 years, stethoscopes have been used to listen to these sounds. Despite this, the physics of how stethoscopes work remain poorly understood. While the stethoscope itself is a simple device, its performance depends on how it is coupled to both the patient and the clinician. Existing models do not adequately address these interactions, forcing design choices to be made based on simple heuristics. The aims of this thesis are to provide a theoretical framework for understanding the acoustics of stethoscopes, propose a low order model to simulate the response, and develop an experimental methodology to validate the model. When a stethoscope is pressed against the chest, body sounds induce small perturbations around the equilibrium position of the nonlinear chest-stethoscope system. In this thesis, a lumped-element approach is used to model these perturbations. The resulting models are validated using experiments conducted on a phantom (a laboratory model representing the human chest). Impedance measurements on the phantom and on the human chest allow differences between these systems to be accounted for. The models presented in this thesis capture the trends associated with each of the key design parameters. Minimising the cavity volume maximises the response, while tubing significantly attenuates low frequencies and introduces distorting standing-wave resonances. Using a diaphragm attenuates the response and shifts the resonances to higher frequencies, but also allows smaller air cavities to be used. Holding a stethoscope against the chest sets the equilibrium position of the coupled system and provides a damping-dominated impedance load on the chestpiece. The strong dependence of a stethoscope’s performance on external factors, such as the properties of the chest and the way it is held, makes it difficult to compare sensors objectively. The work presented in this thesis dispels various misconceptions about how stethoscopes work and can be used to inform design choices, ultimately improving the diagnostic capabilities of future stethoscopes.