Multi-axial stress conditions exist in almost all engineering applications, but design procedures, in most cases, assume uniaxial stress conditions. This assumption is not adequate for orthotropic materials such as timber. Initial failure mode in lumber is usually tension perpendicular to the grain or axial tension around natural defects such as knots. Strength properties of lumber have been obtained from clear-wood specimens and in-grade testing of full-scale lumber whereas failure criteria are usually those originally developed for composite materials. Some of the existing failure criteria are relatively easy to use, but prove valid only under special orthotropic conditions. Some studies assume that the tension and compression strengths to be equal, which is not the case for wood. Other criteria are usually based on quadric surfaces, in which certain constraints are taken strictly from geometrical considerations to achieve a closed failure envelope. Since these conditions are not defined from a physical stand point, some parts of the resulting curve leads to errors in failure predictions. These quadric criteria also require evaluation of interaction coefficients for bi-axial stress conditions. Evaluation of these interaction coefficients demands extensive experimental testing, which limits their practical application. Some proposals have been made to deal with the interaction approach either by evaluating the interaction factor in terms of uniaxial strength or by proposing a given value in the absence of experimental data. The proposed paper reviews existing failure criteria. Their relevance to timber structural members is discussed and validated against experimental data. Statistical analyses are performed to assess their adequacy.