The application of infrared spectroscopy for quantitative microanalysis of silica is hindered by one severe difficulty. There is an almost complete spectral overlap in the region around wavelength 800 cm−1, where the various forms of silica have their main absorption band. Multivariate analysis with the use of plsr techniques has been applied to mixtures of α-quartz and fumed amorphous silica, by the calibration of the wavelength region 900-600 cm−1 at 10-cm−1 intervals. The two silica components could be determined simultaneously in concentrations relevant for occupational hygiene control work. The standard deviation of the method is ±1.8%, whereas the relative prediction error is better than ±5.2%. The plsr multivariate calibration techniques are found to be more robust and more accurate than a combined method of infrared spectroscopy and x-ray diffraction for the quantitative determination of silica mixtures. Thus, multivariate calibration offers a rapid, low-cost, and reliable method for the determination of silica polymorphs in low concentrations by infrared spectroscopy, excluding the use of x-ray diffraction. The ability to extract qualitative spectral information from plsr calibrations is demonstrated, and the principle of matrix specific calibration sets is discussed.