Quantitative estimates of muscle-tendon loading are inherently important in the field of biomechanics. Researchers have primarily attempted to obtain these estimates using modeling approaches, which rely on assumptions on musculoskeletal coordination and geometry. Tendon loading may be directly measured, but the most reliable approaches are invasive and not suitable for widespread use. Non-invasive approaches exist, but these have primarily measured tendon elasticity, which has a variable nonlinear relationship with tendon loading. We have developed a non-invasive technique for measuring axial tendon stress based on shear wave propagation speed. A tensioned beam model of tendon predicts that wave speed squared will increase linearly with stress, and that stress will be the primary determinant of wave speed at physiological loads. The constant of proportionality between wave speed squared and stress is predicted to be the tendon’s effective density, which is influenced by the surrounding medium. We tested this technique in ex vivo porcine digital flexor tendons, in vivo Achilles, patellar and hamstring tendons, and in situ cadaveric Achilles tendons. Experimentation in ex vivo and in situ tendons showed that wave speed squared is highly correlated with applied stress. The constant of proportionality could be predicted in ex vivo tendons immersed in saline or in air, but the in situ case was found to be more complicated. These experiments also showed that both specimen-specific and group-average calibration approaches could be used to predict stress from wave speed, with specimen-specific approaches being more accurate. Experiments on in vivo tendons showed that tendon wave speeds could be measured during dynamic activities such as walking and running, and that wave speed measurements could be used to obtain valuable information on tendon loading and how it is affected by altered performance of a task. Also included here is an investigation into whether shear wave speed may predict other mechanical properties of tendon in a rabbit model of tendon rupture and repair. Wave speed was found to be a weak predictor of both tendon elasticity and ultimate strength.