A study was undertaken to investigate the precision of plain radiographs in the assessment of the width of radiolucent lines and to define parameters for more accurate measurement. A metal-backed glenoid component was inserted into fourteen cadaveric scapulae; the component had a radiolucent spacer at the central post to provide a gap with a known width at the component-bone interface. The specimens were mounted in a custom-designed jig, and initial radiographs were made with the glenoid in neutral version; sequential radiographs then were made, at 5-degree intervals, with the glenoid in 0 to 40 degrees of anteversion and retroversion. Four independent observers with various levels of experience measured the width of the radiolucent lines with use of digital microcalipers. Osteometric analysis demonstrated that normal glenoid version ranged from 3 degrees of anteversion to 13 degrees of retroversion; these values were similar to those reported in previous studies. Radiographic analysis showed that accurate measurement of the width of the gap was dependent on the position of the glenoid. The measured widths of the radiolucent lines were significantly smaller than the known width of the gap when retroversion was 10 degrees or more and when anteversion was 15 degrees or more (p = 0.05). Radiolucent lines were not consistently observed on radiographs that were made with the glenoid in more than 20 degrees of anteversion and retroversion. An analysis of interobserver error showed close agreement among the measurements made by the different observers when the glenoid was in 0 and 5 degrees of rotation, with decreased agreement when the glenoid was rotated more than 10 degrees from neutral.CLINICAL RELEVANCE: Inaccurate positioning of the patient and anatomical variation in glenoid version may explain the variability in the reported onsets, progressions, and frequencies of radiographic loosening of glenoid components. The findings of the present study also may help to explain the poor association between clinical and radiographic findings reported for patients who have pain at the site of a total shoulder prosthesis. Radiographs made within 10 degrees of neutral should allow accurate assessment of radiolucent lines about the glenoid.