The aim of this retrospective study was to determine the diagnostic value of preoperative ultrasonography for the characterisation of size and location of the involved tendons. in 332 consecutive patients who underwent surgery, all preoperative ultrasonographic reports were reviewed and ultrasound (US) and surgical findings were compared. Ultrasound criteria for cuff tears were complete nonvisualisation of the cuff tendons or localised absence and focal discontinuity. In all but 12 cases, US diagnoses corresponded with intraoperative findings (sensitivity 98%, confidence interval 95.1-99.3; specificity 93%, CI 85.7-97.1; accuracy 97%, CI 93.8-98.1). Size and location of the tear were correctly predicted in 69 of 96 cases (accuracy 87%, sensitivity 89%, specificity 87%). US demonstrated less extensive tears than observed at surgery in 18%. Ultrasonography was highly accurate and sensitive for detecting rotator cuff tears, but seems to be a method of limited value for evaluation of the size of cuff tears, in particular, for the detection of small tears.