CONTEXT: Isokinetic assessment of shoulder internal-(IR) and external-rotator (ER) strength is commonly used with many different postures (sitting, standing, or supine) and shoulder positions (frontal or scapular plane with 45 degrees or 90 degrees of abduction). OBJECTIVE: To conduct a systematic review to determine the influence of position on the intersession reliability of the assessment of IR and ER isokinetic strength, to identify the most reliable position, and to determine which isokinetic variable appears to be most stable in intersession reliability. EVIDENCE ACQUISITION: A systematic literature search through MEDLINE and Pascal Biomed databases was performed in October 2009. Criteria for inclusion were that studies be written in English or French, describe the isokinetic evaluation methods, and describe statistical analysis. EVIDENCE SYNTHESIS: Sixteen studies meeting the inclusion criteria were included. Variable reliability of ER and IR peak torque (PT) were generally reported for all assessment positions; intraclass correlation coefficients were .44-.98 in the seated position with 45 degrees of shoulder abduction, .09-.77 in the seated position with 90 degrees of shoulder abduction, .86-.99 (coefficient of variation: 7.5-29.8%) in the supine position with 90 degrees of shoulder abduction, .82-.84 in the supine position with 45 degrees of shoulder abduction, and .75-.94 in standing. The ER:IR ratio reliability was low for all positions. CONCLUSIONS: The seated position with 45 degrees of shoulder abduction in the scapular plane seemed the most reliable for IR and ER strength assessment. The standing position or a shoulder posture with 90 degrees of shoulder abduction or in the frontal plane must be used with caution given the low reliability for peak torque. Good reliability of ER and IR PT was generally reported, but ER:IR ratio reliability was low.