SummaryThe feasibility of near-infrared (NIR) diffuse reflectance spectroscopy scanning from 1100 to 2500nm in the analyses of the amounts of moisture, protein, starch, amylose, and tannin in buckwheat flours was examined. Fifty ground samples comprised of 27 different cultivars harvested in 12 different countries were divided into two sets: 35 samples as a calibration set and 15 samples as a prediction set. The multiple regression equations (MREs) established between the second derivative NIR spectra data and the reference data, which were obtained by chemical analyses of the calibration set, gave multiple correlation coefficients of higher than 0.93 for moisture, protein, and starch, and workable standard error of predictions (SEPs) in relation to the reference standard deviation data. In contrast, the MREs for amylose and tannin were judged as unstable because the SEPs had no difference with their reference standard deviation data. The influential wavelengths were 1925 nm for water (assigned to the combination of the stretching and bending vibrations of hydroxyl [OH]), 2057nm for protein (assigned to the combination of NH and amide II or III), and 2100nm for starch (assigned to the combination of OH and CO). The obtained results indicated that the NIR procedure can be used as a nondestructive analysis method for rapid and simple measurement of the amounts of moisture, protein, and starch in buckwheat flours.