Purpose.To evaluate the Complete Ophthalmic Analysis System (COAS; WaveFront Science) for accuracy, repeatability, and instrument myopia when measuring myopic refractive errors.Methods.We measured the refractive errors of 20 myopic subjects (+0.25 to −10 D sphere; 0 to −1.75 D cylinder) with a COAS, a phoropter, and a Nidek ARK-2000 autorefractor. Measurements were made for right and left eyes, with and without cycloplegia, and data were analyzed for large and small pupils. We used the phoropter refraction as our estimate of the true refractive error, so accuracy was defined as the difference between phoropter refraction and that of the COAS and autorefractor. Differences and means were computed using power vectors, and accuracy was summarized in terms of mean vector and mean spherocylindrical power errors. To assess repeatability, we computed the mean vector deviation for each of five measurements from the mean power vector and computed a coefficient of repeatability. Instrument myopia was defined as the difference between cycloplegic and noncycloplegic refractions for the same eyes.Results.Without cycloplegia, both the COAS and autorefractor had mean power vector errors of 0.3 to 0.4 D. Cycloplegia improved autorefractor accuracy by 0.1 D, but COAS accuracy remained the same. For large pupils, COAS accuracy was best when Zernike mode Z40(primary spherical aberration) was included in the computation of sphere power. COAS repeatability was slightly better than autorefraction repeatability. Mean instrument myopia for the COAS was not significantly different from zero.Conclusions.When measuring myopes, COAS accuracy, repeatability, and instrument myopia were similar to those of the autorefractor. Error margins for both were better than the accuracy of subjective refraction. We conclude that in addition to its capability to measure higher-order aberrations, the COAS can be used as a reliable, accurate autorefractor.