We measured the visible to near-infrared (IR) spectra of 176 synthetic and natural samples of Fe oxides, oxyhydroxides and an oxyhydroxysulfate (here collectively called “Fe oxides”), and of 56 soil samples ranging widely in goethite/hematite and goethite/lepidocrocite ratios. The positions of the second-derivative minima, corresponding to crystal-field bands, varied substantially within each group of the Fe oxide minerals. Because of overlapping band positions, goethite, maghemite and schwertmannite could not be discriminated. Using the positions of the 4T1←6A1, 4T2←6A1, (4E;4A1)←6A1 and the electron pair transition (4T1+4T1)←(6A1+6A1), at least 80% of the pure akaganeite, feroxyhite, ferrihydrite, hematite and lepidocrocite samples could be correctly classified by discriminant functions. In soils containing mixtures of Fe oxides, however, only hematite and magnetite could be unequivocally discriminated from other Fe oxides. The characteristic features of hematite are the lower wavelengths of the 4T1 transition (848–906 nm) and the higher wavelengths of the electron pair transition (521–565 nm) as compared to the other Fe oxides (909–1022 nm and 479–499 nm, resp.). Magnetite could be identified by a unique band at 1500 nm due to Fe(II) to Fe(III) intervalence charge transfer. As the bands of goethite and hematite are well separated, the goethite/hematite ratio of soils not containing other Fe oxides could be reasonably predicted from the amplitude of the second-derivative bands. The detection limit of these 2 minerals in soils was below 5 g kg−1, which is about 1 order of magnitude lower than the detection limit for routine X-ray diffraction (XRD) analysis. This low detection limit, and the little time and effort involved in the measurements, make second-derivative diffuse reflectance spectroscopy a practical means of routinely determining goethite and hematite contents in soils. The identification of other accessory Fe oxide minerals in soils is, however, very restricted.