Linear regression models were constructed for chestnut beams of Spanish origin (Asturias, Galicia, Catalonia and Extremadura) using the global modulus of elasticity (MOE g ) and bending strength (MOR), both obtained by destructive tests, as dependent variables, and the results of non-destructive measurements, visual grading parameters and density as independent variables. The variables selected were density, wave velocity, sample length, dynamic modulus of elasticity, maximum knot diameter in relation to height and concentrated knot diameter ratio. Linear regression models were constructed to indirectly estimate the mechanical properties of the beams. Ultrasonic velocity, density and sample length were the best predictors of MOE g (R 2 = 0.740 and 0.734 SE). Regression adjustments for MOR presented low coefficients of determination and high errors. The visual grading parameters of the beams did not play a significant role in the prediction of either MOE g or MOR.