Abstract: Near infrared spectroscopy combined with pattern recognition methods was used to discriminate the unadulterated and adulterated honey samples. Various crude honey samples from different area in China were collected, and the adulterated honey were prepared according to typical adulteration method, adulteration substance and construction in the market. FT-NIR spectrometer was used to measure the trans-reflectance spectra of honey. The differentiation models for adulteration of honey were constructed by four kinds of pattern recognition methods, including partial least squares discriminate analysis (PLS-DA), soft independent modeling of class analogy (SIMCA), error back propagation network (BP-ANN), least-squares support vector machine (LS-SVM). The results showed that four methods could all correctly differentiate honey samples that were adulterated with high fructose syrup and fructose-plus-glucose solutions. For the adulteration of high fructose syrup, the classification accuracy of calibration set was above 95%, and the classification accuracy of prediction set was above 87%. For the adulteration of fructose-plus-glucose solutions, the classification accuracy of both calibration set was above 93%, and the classification accuracy of prediction set was above 84%. Compared with the four kinds of models, it was found that LS-SVM had the best results, the classification accuracy for both calibration set and prediction set were 100% for two kinds of adulteration. The present study indicated that the fast and accurate differentiation of the adulteration of honey by NIR spectra was feasible.