BACKGROUND: This pilot study assesses level of agreement between surface and fine-wire electromyography (EMG), in order to establish if surface is as reliable as fine wire in the diagnosis and treatment of abnormal muscle patterning in the shoulder. METHOD: Eighteen participants (11 female) with unstable shoulders were recruited after written consent and ethical approval. Anthropometric information and mean skinfold size for triceps, subscapular, biceps and suprailiac sites were obtained. Triple-stud self-adhesive surface electrodes ('Triode'; Thermo Scientific, Physio Med Services, Glossop, Derbyshire, England) were placed over pectoralis major (PM), latissimus dorsi (LD), anterior deltoid (AD) and infraspinatus (IS) at standardized locations. Participants performed five identical uniplanar standard movements (flexion, abduction, external rotation, extension and cross-body adduction). After a 20-minute rest period, a dual-needle technique for fine-wire insertion was performed and the standard movements were repeated. An experienced examiner in each technique reported if muscle activation patterns differed from agreed normal during any movement and were blinded to the other test results. Sensitivity, specificity and Kappa values for level of agreement between methods were calculated for each muscle according to the method of Altman (1991). RESULTS: Fifteen participants were successfully tested. Sensitivity, specificity and Kappa values between techniques for each muscle were PM (57%, 50%, 0.07), LD (38%, 85%, 0.22), AD (0%, 76%, -0.19) and IS (85%, 75%, 0.6). Only IS demonstrated high sensitivity and specificity and a moderate level of agreement between the two techniques. There was no correlation between skinfold size and agreement levels. CONCLUSION: The use of surface EMG may help to classify types of shoulder instability and recognize abnormal muscle patterns. It may allow physiotherapists to direct specific rehabilitation strategies, avoiding strengthening of inappropriate muscles. It has a reasonable degree of confidence to evaluate IS but may have poor sensitivity in detecting abnormal patterns in PM, LD and AD. Further work is required to see if investigator interpretation may have been a factor for the poor level of agreement.