OBJECTIVE: To assess the intra- and interobserver reproducibility of musculoskeletal ultrasonography (US) among rheumatologists in detecting destructive and inflammatory shoulder abnormalities in patients with rheumatoid arthritis (RA) and to determine the overall agreement between the US and magnetic resonance imaging (MRI). METHODS: Fourteen observers examined 5 patients in 2 rounds independently and blindly of each other. US results were compared with MRI. Overall agreements of all findings, of positive findings on MRI, as well as intra- and interobserver reliabilities were calculated. RESULTS: Overall agreement between US and MRI was seen in 79 % with regard to humeral head erosions (HHE), in 64 % with regard to posterior recess synovitis (PRS), in 31 % with regard to axillary recess synovitis (ARS), in 64 % with regard to bursitis, in 50% with regard to biceps tenosynovitis (BT), and in 84 % for complete cuff tear (CCT). Intra-observer and interobserver kappa was 0.69 and 0.43 for HHE, 0.29 and 0.49 for PRS, 0.57 and 1.00 for ARS, -0.17 and 0.51 for bursitis, 0.17 and 0.46 for BT and 0.52 and 0.6 for CCT, respectively. The intra-observer and interobserver kappa for power Doppler (PD) was 0.90 and 0.70 for glenohumeral signals and 0.60 and 0.51 for bursal signals, respectively. CONCLUSION: US is a reliable imaging technique for most shoulder pathology in RA especially with regard to PD. Standardization of scanning technique and definitions of particular lesions may further enhance the reliability of US investigation of the shoulder