The interclass correlation coefficient (ICC 2,1) was 0.97 (95% CI 0.87 to 0.99). The standard error of the measurement was 0.1 cm. Each participant

was seated on a chair with the cervical spine in a neutral position. Participants were asked to flex the affected shoulder to two angles (60° and 90°), either with or without real-time visual feedback. The order of the two angles and the two feedback conditions were randomised by drawing a sealed envelope from a box. Participants were instructed to lift the Sirolimus upper limb being tested slowly with the elbow extended, the forearm and wrist in a neutral position, and a loose fist, and to hold the position for 5 sec at the flexion angle of 60° or 90°. A universal goniometer was used

to determine the flexion angle, and find more a horizontal target bar was positioned at each angle in the sagittal plane. The shoulder level and scapular movement in the lateral and posterior view were recorded on two video cameras connected to a personal computer. The computer screen was positioned at the participant’s eye level and turned on when real-time visual feedback was required. Before the shoulder flexion, the principal investigator placed the scapula in the normal position (vertebral STK38 border parallel with spine spacing at approximately 7 cm, scapula positioned between T2 and T7 and flat on the posterior rib cage). The subject was asked to observe the scapular motion through the computer monitor (Figure 4). If shoulder depression, tilting, or winging were observed during shoulder flexion, the investigator encouraged the subject to protract and elevate the

scapula. Participants practised using the visual feedback to maintain the scapula in a normal position for 15 min. The shoulder flexion task was performed three times. A 3-min rest period was allowed between trials to minimise fatigue. The primary measure in the study was muscle activity in the scapular upward rotators. Surface electromyographic data were collected from the upper and lower trapezius and serratus anterior, using a standard data acquisition systema. Preparation of the electrode sites involved shaving and cleaning the skin with rubbing alcohol (Cram et al 1998). Disposable silver/silver chloride surface electrodesb were positioned at an inter-electrode distance of 2 cm. The reference electrode was attached to the styloid process of the ulna of the upper limb being tested.