Estimation of the density along with the tensile strength of wood within both the elastic and plastic deformation ranges, represented as modulus of elasticity (MOE) and ultimate tensile stress (UTS), respectively, were performed using near-infrared (NIR) spectroscopy. A partial least squares (PLS) analysis was applied to the measurements of density, MOE, and UTS, and resulted in a high accuracy of prediction, independent of wood species. The correlation coefficient between the NIR spectra and criterion variables, and the regression vector resulting from the PLS analysis, suggested that the characteristic absorption bands were strongly related to the predictability of each property. In the case of softwood, absorption bands due to intra-molecular hydrogen-bonded OH groups in the crystalline regions of cellulose, which are oriented preferentially in a direction parallel to the cellulose chain, might strongly affect the tensile strength of softwood. Hardwoods have much more complex and variable structures than softwoods; therefore, it was supposed that the key factor governing the tensile strength in hardwood would be the interaction between the three principal constituents (i. e., cellulose, hemicellulose, and lignin) of wood.