A clinical study was done to determine how precisely ophthalmologists perform retinoscopy. Ten subjects were examined by five ophthalmologists on two separate occasions. Statistical analysis of the results showed that the physicians' precision was unaffected by patients' sex or type of refractive error; the physicians in general agreed in their estimates of the magnitudes of the refractive errors, but their precision varied significantly; cylinder power was measured more precisely than sphere power; and right eyes were measured more precisely than left eyes. Quantitative estimates are given for the variability of replicate retinoscopic measurements. Suggestions are made for techniques of clinical retinoscopy which may minimize certain errors.