BACKGROUND: Isokinetic assessment of shoulder internal and external rotators is commonly used by clinicians to assess muscle performance and to guide rehabilitation. The reliability of isokinetic assessment is fundamental to track small but clinically relevant changes. OBJECTIVES: We aimed to analyze the absolute and relative reliability of strength imbalance indices such as peak torque ratios (ERconc/IRconc, ERecc/IRecc, ERecc/IRcon, IRecc/ERcon), bilateral concentric and eccentric strength ratios, and to examine the reliability of external rotator and internal rotator peak torque measured using a Biodex((R)) dynamometer in the seated position. DESIGN: Cross-sectional laboratory study. METHODS: Forty-six healthy participants were tested twice with seven days between sessions, at 60 degrees /s and 120 degrees /s concentrically, and 30 degrees /s eccentrically. RESULTS: Low to moderate relative reliability (intraclass correlation coefficient: 0.25-0.81) was found for unilateral and bilateral strength imbalance ratios. High intraclass correlation coefficient values (0.87-0.97) were found for peak torque. Concerning absolute reliability, the standard error of measurement ranged from 9.1 to 25.6% for strength imbalance ratios and from 7.7 to 14.5% for peak torque measurements, and minimal detectable change ranged from 25.2 to 71% for strength imbalance ratios and from 21.3 to 40.2% for peak torque measurements. CONCLUSIONS: The standard error of measurement and minimal detectable change reported in the present study should be taken into account when evaluating the individual longitudinal changes in clinical practice.