The use of near infrared reflectance spectroscopy (NIRS) to predict minerals concentration (K, Na, Ca, Mg, Fe) in straw samples was investigated in this study. A total of 222 straw samples were collected in rural area of most provinces in China. Two types of straw samples were prepared, directly cut specimens and oven-dried, milled specimens. The spectra of two kinds of samples were employed to correlate with minerals concentration. Different spectral pre-treatments and regression methods were trialled to optimize the calibration. Coefficient of determination in prediction (R(v)(2)) and standard error of prediction (SEP) were 0.69, 0.54. 0.73, 0.79, 0.41 and 3.77 mg g(-1), 0.69 mg g(-1), 0.58 mg g(-1), 0.31 mg g(-1), 0.11 mg g(-1) for directly cut straw; 0.85, 0.70. 0.82, 0.85, 0.63 and 2.35 mg g(-1), 1.46 mg g(-1), 0.47 mg g(-1), 0.27 mg g(-1), 0.13 mg g(-1) for dried milled samples, respectively. (c) 2008 Elsevier Ltd. All rights reserved.