Magnetic resonance images of the shoulders of ninety-six asymptomatic individuals were evaluated to determine the prevalence of findings consistent with a tear of the rotator cuff. The scans were reviewed independently by two diagnostic radiologists who are experienced in the interpretation of magnetic resonance images of the shoulder. The over-all prevalence of tears of the rotator cuff in all age-groups was 34 per cent (thirty-three). There were fourteen full-thickness tears (15 per cent) and nineteen partial-thickness tears (20 per cent). The frequency of full-thickness and partial-thickness tears increased significantly with age (p < 0.001 and 0.05, respectively). Twenty-five (54 per cent) of the forty-six individuals who were more than sixty years old had a tear of the rotator cuff: thirteen (28 per cent) had a full-thickness tear and twelve (26 per cent) had a partial-thickness tear. Of the twenty-five individuals who were forty to sixty years old, one (4 per cent) had a full-thickness tear and six (24 per cent) had a partial-thickness tear. Of the twenty-five individuals who were nineteen to thirty-nine years old, none had a full-thickness tear and one (4 per cent) had a partial-thickness tear. Magnetic resonance imaging identified a high prevalence of tears of the rotator cuff in asymptomatic individuals. These tears were increasingly frequent with advancing age and were compatible with normal, painless, functional activity. The results of the present study emphasize the potential hazards of the use of magnetic resonance imaging scans alone as a basis for the determination of operative intervention in the absence of associated clinical findings.