The purpose of this study was to evaluate the technical quality of 3D coronary MR angiography (CMRA) with navigator echo and the consistency of image interpretation in repeated imaging sessions. Fourteen subjects underwent CMRA, 10 of whom were imaged twice. The coronary arteries (96 segments) were analyzed twice for hemodynamically significant stenoses. Signal-to-noise and contrast-to-noise ratios varied considerably between the two imagings. Fat saturation was poor or satisfactory in 37%; in 15% of the slabs, the severity of artifacts was moderate; and the overall quality was good to excellent in only 42% of the imagings. The intraobserver reproducibility was good (kappa = 0.54), but the consistency of interpretation for repeated CMRA was only satisfactory (kappa = 0.43). Sensitivities of 84 and 84% and specificities of 70 and 62% were obtained for the two readings. Although the reproducibility of image reading is good, 3D CMRA with navigator echo provides only fair technical consistency, and the frequently compromised image quality impairs the clinical utility of this technique.