The prediction of wool base of greasy wool in the form of miniature cores from digitized near-infrared reflectance spectra is demonstrated. This secondary calibra tion method combines measurements of reflectance at several wavelengths in a multiple linear regression equation for a prediction of wool base. A variety of methods of selecting suitable wavelengths for prediction was employed. Suitable wavelengths could not be selected from considering the spectra of the components, or from the statistical properties of the spectral data. A combination of an exhaustive search and the forward stepwise addition method applied over the spectral range gave the best selection of wavelengths. For the population of 42 samples studied, the error of the calibration is similar in magnitude to that of the conventional gravimetric test.