Radial velocity curves for U Sco. The published radial velocity data from Johnston & Kulkarni (1992) and Duerbeck et al. (1993b) are plotted together with accurate phasing. The eclipse is at zero phase. The circles, squares, and diamonds are for data from 1990, 1991, and 1992, respectively. The upper plot is for the absorption lines attributed to the companion star, while the lower plot is for emission lines attributed to the white dwarf. The typical error bars are 40 km s−1, so the large observed scatter implies systematic problems. More problems are indicated by the emission and absorption lines having significantly different average velocities. We conclude that the published data are not reliable measures of the orbital velocities of either star, and so the published mass determinations cannot be reliable.