Adequate reliability studies of knee flexion and extension are currently available for isokinetic measurements, but not for the shoulder joint. For this reason, this study examines the variability (%) in the determination of peak torque (PT) and the angle at peak torque (AP) in the test-retest procedure. Differentiation is made between the various types of work (concentric, isometric, eccentric) and the degrees of freedom in the shoulder joint (flexion/extension, abduction/adduction, external rotation/internal rotation). The results show a variability of PT for concentric measurements of 15.0%-19.0% for all degrees of freedom; for flexion/extension and abduction/adduction in isometric and eccentric work, the variability is 17.5%-25.3%. External and internal rotation show generally greater deviation (29.0%-35.3%), except in concentric work. The angle at peak torque can only be reproduced to a limited degree (25.1%-41.1% variability). It is concluded that lower reproducibility can be attained for the shoulder joint compared to the knee joint, depending on the degree of freedom and exercise form tested. It appears meaningful to cite only ranges for the angle at peak torque.