Abstract: Based on the near-infrared reflectance spectroscopy （NIRS） method and modified partial least squares （MPLS） regression method, a prediction model for determination of protein content in soybean meal was established using 40 soybean meal feed samples from different origins. The results showed that the R square （RSQ） of the calibration model was 0.916 0, the standard error of calibration （SEC） was 0.150 3, and the standard error of cross validation （SECV） were 0.324 2. These data suggest that the established NIRS method is an effective technology for the quantitative determination of protein content in soybean meal.