Mineral concentration data could easily be generated by near infrared reflectance spectroscopy (NIRS) while determining quality parameters in forage samples. However, which minerals can be analyzed and why they can be determined has not been documented adequately. Therefore, NIRS spectra were collected on 200 samples of crested wheatgrass (Agropyron spp.), 203 tall fescue (Festuca arundinacea Schreb.), and 59 alfalfa (Medicago sativa L.) hays. Concentrations of Na, K, Ca, P, Mg, Fe, Mn, Cu, and Zn, as determined by atomic absorption, and calculated values of the Ca/P, K/Mg, and K/(Ca+Mg) ratios were regressed against reflectance values measured in 2-nm increments between 1100 and 2500 nm for each sample. Approximately one half of the samples in each forage set was used to develop the calibration equations, while the other half was used to validate the equations. The coefficients of variation [CV=(standard error of analysis÷the mean) X100] generally ranged from 10 to 20% for K, Mg, Ca, and P concentrations in each forage type. The Ca/P ratio in alfalfa was determined with a CV of 18%. The CV values of other minerals and mineral ratios generally exceeded 20%. Chlorophyll and some inorganic salts and organic-acid salts of Ca, Mg, and K were scanned with NIRS for wavelength comparisons with those wavelengths used to determine mineral concentrations in forages. Some of the wavelengths used in the equations for Ca, K, and Mg were related to peaks and changes in slope observed in chlorophyll and organic-acid salts of Ca, K, and Mg, suggesting that NIRS is indirectly measuring these minerals by their association with organic molecules. Accurate use of NIRS to determine mineral cation composition in forages appears limited to certain major minerals (Ca, P, K, and Mg).