Our aim was to determine the most repeatable three-dimensional measurement of glenoid orientation and to compare it between shoulders with intact and torn rotator cuffs. Our null hypothesis was that glenoid orientation in the scapulae of shoulders with a full-thickness tear of the rotator cuff was the same as that in shoulders with an intact rotator cuff. We studied 24 shoulders in cadavers, 12 with an intact rotator cuff and 12 with a full-thickness tear. Two different observers used a three-dimensional digitising system to measure glenoid orientation in the scapular plane (ie glenoid inclination) using six different techniques. Glenoid version was also measured. The overall precision of the measurements revealed an error of less than 0.6°. Intraobserver reliability (correlation coefficients of 0.990 and 0.984 for each observer) and interobserver reliability (correlation coefficient of 0.985) were highest for measurement of glenoid inclination based on the angle obtained from a line connecting the superior and inferior points of the glenoid and that connecting the most superior point of the glenoid and the most superior point on the body of the scapula. There were no differences in glenoid inclination (p = 0.34) or glenoid version (p = 0.12) in scapulae from shoulders with an intact rotator cuff and those with a full-thickness tear. Abnormal glenoid orientation was not present in shoulders with a torn rotator cuff