Fifteen asymptomatic volunteers and 300 patients with suspected lesions of the rotator cuff were examined with T2*- and proton density-weighted gradient echo (T2*-Rho-GEI) MRI sequences (500-700 ms/30 ms/30 degrees) with a 0.5 T (GE-CGR) system and an adapted surface coil. For the patients, the findings were compared with those of arthrography in 130 cases and of surgery in 42 cases. For the operated patients, T2*-Rho-GEI and arthrography allowed correctly diagnosing 31 complete ruptures (CR). However, surgical exploration of the ruptures showed that T2*-Rho-GEI showed the actual extent of the rupture and the degree of tendon retraction better than arthrography. T2*-Rho-GEI also provides information about the thickness and trophicity of the remaining cuff and on local muscle trophicity. T2*-Rho-GEI allows directly evaluating the tendon of the long head of the biceps both in its extra-articular and its intra-articular parts. In 3 cases of complete cuff rupture, MRI allowed detecting 3 associated lesions of the anterior labrum, which had not been suspected on the clinical examination and had not been demonstrated by conventional arthrography. Lastly, out of 12 cases of deep incomplete rupture (n = 12) correctly diagnosed with MRI, 5 had been undetected on arthrography. Similarly, 8 cases of inflammatory alterations suspected with T2*-Rho-GEI and confirmed by surgery had not been recognized with arthrography. For the non-operated patients, T2*-Rho-GEI and arthrography had corresponding results for the diagnosis of CR (8 cases), of DIR (6 cases) and of intact cuff (6 cases).(ABSTRACT TRUNCATED AT 250 WORDS)