Optical see-through head-mounted displays (HMDs) are less commonly used because they are difficult to accurately calibrate. In this article, we report a user study to compare the accuracy of 4 variants of the SPAAM calibration method. Among the 4 variants, Stylus-marker calibration, where the user aligns a crosshair projected in the HMD with a tracked stylus tip, achieved the most accurate result. A decomposition and analysis of the calibration matrices from the trials is performed and the characteristics of the computed calibration matrices are examined. A physiological engineering point of view is also discussed to explain why calibrating optical see-through HMD is so difficult for users.