This study highlights the possibility of developing a high performance grading process based on. the analysis of acoustic vibrations in the frequency domain. The uniqueness of the method is the direct use of the spectrum as predictive variables to estimate modulus of elasticity and modulus of rupture. It is not necessary to compute the bending stiffness or other analytical predictors to evaluate the bending strength. All the information of the mechanical state of the material ís contained in the time or frequency response of the beam. A powerful statistical method, known as partial least squares regression (PLS), was used. PLS is related to both principal component regression (PCR) and multi linear regression (MLR) and can be thought of as occupying a middle ground between them. The use of PLS regression directly on the vibration spectrum permits a very good estimation of the variables of interest (MOE, bending strength, and density). When the linear model of calibration is achieved by PLS, the on-line grading process is reduced to a minimum: a vibration transducer, an acquisition card, and a computer performing only fast fourier transform and multiplying matrices. Therefore, the time for grading a piece of wood is very short, and the cost of grading is low.