Tenderness is an important index to evaluate the pork's quality, in this paper a method called three-dimensional diffuse reflectance spectroscopy was proposed to detect pork tenderness. Because pork has a strong scattering impact on light, this method introduced more scattering information of pork samples into spectral analysis of tenderness. Using the special data acquisition system, three-dimensional diffuse reflectance spectra of 64 pork samples were constructed by collecting the emergent light signals of different distances away from the light incident point. And n-way partial least squares (NPLS) regression was applied to establish the calibration model between the pork tenderness and three-dimensional diffuse reflectance spectra which were denoised by wavelet transform. The determination coefficient of model for the calibration set (R2(Cal)) is 0.883 1, and the root mean squared error of calibration (RMSEC) is 3.685 0N. The determination coefficient of model for the prediction set (R2(Pred)) is 0.874 7, and the root mean squared error of prediction (RMSEP) is 3.975 6N. The result indicates that the NPLS model of pork tenderness built by three-dimensional diffuse reflectance spectra has higher calibration accuracy and prediction stability than the traditional diffuse reflectance spectra. Three-dimensional diffuse reflectance spectroscopy can be expected to be a new method to quickly detect the tenderness and the other qualities of pork.