New three-dimensional (3D) magnetic resonance imaging (MRI) methods for measuring the tendon moment arm were created and were evaluated on the tendon moment arm of the flexor digitorum profundus at the third metacarpophalangeal joint. Using an open magnet MRI system and a hand holder, a series of static images were acquired at four joint angles and analyzed using specially created computer programs. Three methods were evaluated: (1) a 3D tendon excursion method that extended the method of Landsmeer; (2) a 3D geometric method whereby the moment arm was the perpendicular distance between the joint axis of rotation and the tendon path, and (3) a two-dimensional (2D) geometric method whereby single image slices were analyzed. Repeating the imaging and measurement processes, the 3D tendon excursion method was more reproducible (6% variation) than the 3D geometric method (12%), and both were much more reproducible than the 2D geometric method (27%). By having three operators analyze a single set of image data, we found that the precision of the 3D tendon excursion method was much less affected by segmentation error than the 3D geometric method. With the 3D imaging methods, tendon bowstringing and a displacement of the joint center of rotation toward the dorsal side of the hand were evident, leading to as much as a 60% increase in moment arm with joint flexion. Because of the dependence on flexion and variation between subjects, we recommend patient-specific measurements for target applications in functional neuromuscular stimulation interventions and tendon transfer surgery.