In symptomatic subjects, the reliability of tests for shoulder joint range of motion has yet to be determined. For this reason, inter-rater and intra-rater agreement trials were undertaken to ascertain the reliability of visual estimation, goniometry, still photography, "stand and reach" and hand behind back reach for six different shoulder movements. Intra-class correlation coefficients (Rho) were derived by using a random effects model. For flexion, abduction and external rotation fair to good reliability was demonstrated for both trials using visual estimation (Inter-rater Rho = 0.57-0.70; Intra-rater Rho = 0.59-0.67), goniometry (Inter-rater Rho = 0.64-0.69; Intra-rater Rho = 0.53-0.65) and still photography (Inter-rater Rho = 0.62-0.73; Intra-rater Rho = 0.56-0.61). The tests had standard errors of measurement of between 14 and 25 degrees (inter-rater trial) and 11 and 23 degrees (intra-rater trial).