Information on the strength distribution of timbers and other wood products seems to have become more important for users and producers after revision of the Japan architectural standard in 1998, which emphasizes the performance requirements of structures. Because there is no way other than expensive destructive tests to collect strength data, many researchers have proposed many inspecting methods for predicting strength by nondestructive evaluation. The most popular method for structural timber is the mechanical grading method based on the relation between Young's modulus (E) and strength (σ) with some linear regression models. On the other hand, it is well known that the proof loading test is superior for obtaining information on the lower tail ofσ distribution. If the E distribution of the objective timbers is known approximately, selecting timbers nearest to the projected E values saves timbers for destructive tests. We examined the alternative sampling method using the reported e-σ data sets of Japanese larch square-sawn timber. The simulated results showed that the estimated lower tail of the bending strength distribution by the alternative method was a better approximation of the experimental distribution than that derived from the conventional linear regression model.