Test-retest reliability is important for long-term follow-up; however, data on the reliability of isokinetic dynamometry of the shoulder are scarce. Twenty subjects (50% male) were measured; 10 with asymmetrical use of the arms (mean age 27 years) and 10 used their arms symmetrically (mean age 32 years). A Biodex dynamometer (Multi joint system 2) was used. Abduction/adduction and external/internal rotation were measured following a standardized protocol. Performed scheme: two sessions with a 2-week interval, all measurements were done with 60 degrees/second (5 repetitions) and respectively 120 degrees/second and 180 degrees/second (10 repetitions). Differences in the mean peak torques, split for muscle group and gender were significant. Intraclass correlation coefficients ranged from 0.69 to 0.92. This implies good to excellent reliability in research on groups. To determine test-retest reliability of two consecutive individual measurements smallest detectable differences (SDD) were computed and ranged from 21% to 43%. It is questionable whether the SDDs are small enough to detect real changes in muscle strength.