Reduced cost and greater portability of 3D motion tracking technology increases the likelihood of its use in clinical settings to evaluate therapeutic exercise quality in reference to an ideal movement pattern. It must first be determined how accurately healthy individuals can mirror a demonstrated exercise to understand the inherent variability of this approach. PURPOSE: To test the kinematic fidelity between therapeutic shoulder exercises demonstrated by a therapist and mirrored by patients. METHODS: Ten, healthy, young (21.4 ± 0.5 yr, 70.2 ± 12.3 kg, 1.68 ± 0.08 m) men and women with a history of resistance exercise training simulated the roles of therapist and patients. A physical therapy aid performed unloaded forward and lateral raise shoulder exercises while a 2D video camera and 12-camera, 3D motion capture system recorded shoulder kinematics. The 2D video recording of the therapist was played back on a life-size screen to each patient who emulated the therapist’s demonstrated exercise. Upper-body marker data of both therapist and patients were recorded at 250 Hz, smoothed with a 6 Hz low-pass filter, and shoulder joint kinematics were obtained. Custom software was used to time normalize kinematic data, obtain cross-correlations between each patients’ joint positions and the therapist’s as a global measure of agreement, calculate the mean absolute error across the range of motion, and error at peak joint excursion. Shoulder angles were compared in the sagittal plane for forward flexion and frontal plane for lateral flexion exercises. RESULTS: The mean cross-correlation coefficient for the forward raise was r = 0.98, 95% CI [0.96, 1.00], the mean error across the range of motion was 11.9 deg, 95% CI [7.8, 16.0], and error at peak excursion was 7.1 deg, 95% CI [1.6, 12.5]. The mean cross-correlation coefficient for the lateral raise was r = 0.96, 95% CI [0.95, 0.98], the mean error across the range of motion was 14.6 deg, 95% CI [11.6, 17.7], and error at peak excursion was 4.4 deg, 95% CI [0.6, 8.2]. CONCLUSIONS: After time normalization, cross-correlations revealed healthy, young people have an excellent ability to replicate the shape of demonstrated uniaxial shoulder exercises. Yet, a modest amount of absolute position error exists across the range of motion and at peak joint excursion. Supported by NSF Grant 1830597