The moisture content of samples affects the accuracy of predictions made using near infrared reflectance spectroscopy (NIRS). One way to prevent the problem is to develop a repeatability file, which summarizes variation in the spectra of a single sample resulting from variation in conditions at the time of scanning (Murray, 1993). Another method is to re-read the spectra of all samples on a second occasion when their moisture content has been deliberately altered (Murray, 1993). The present study, using straws with known voluntary intakes, compares these methods with a simple technique that compensates the spectra for moisture content before NIRS calibration (or prediction) begins.