The objective of the present research was to study the potential of NIR diffuse reflectance spectroscopy as a way for nondestructive measurement of the firmness of "Xueqing" pear. NIR spectral data were collected in the spectral region between 800 nm and 2,630 nm, but only the data between 800 nm and 2,500 nm were used for the analysis. Spectral data smoothing of different gaps using Savitsky-Goaly method was discussed both qualitatively and quantitatively. The results showed that the smoothing of NIR original spectra had little influence on spectral data analysis. Statistical models were developed using partial least square method. Models for three kinds of mathematical treatments of spectra (original, first derivative and second derivative) were established in three spectral regions: 800-2,500 nm, 800-1,890 nm and 832-872 nm. The original spectra showed better results than the first derivative spectra and second derivative in all the three spectral regions. And for the three spectral regions, the full wavelength range of 800-2,500 nm offered the best result. The PLS model of original spectra in the range of 800-2,500 nm gave best predictions of pear firmness, with correlation coefficient (r) of 0.869 and root mean square errors of calibration (RMSEC) of 3.88 N for the calibration sample set and with correlation coefficient (r) of 0.840 and root mean square error of prediction (RMSEP) of 4.26 N for the validation sample set. The results of the study show that NIR diffuse reflectance specroscopy can be used to predict the firmness of "Xueqing" pear.