Timber as a construction material has been used for millennia, but the research field covering the prediction of the strength of structural timber is still in development. Currently, the common conception is that the determination of strength properties has to be determined for every wood species individually. By combining these strength properties to features that can be measured at the timber (either visually or by machine measurements), strength graded timber can be supplied to the market. Potentially, there are more than 1000 commercially available wood species, the timber of which can be used in structures. The largest amount of these wood species are tropical hardwoods. These wood species are often used when high strength and high durability are required. Nowadays, (tropical) timber is increasingly coming from sustainably managed forests. By application of this method of forest management, the (tropical) forests are preserved and have an economic value for the local population. A result of this approach is that more and more unknown wood species in small quantities are coming on the market, the strength properties of which have to be determined. The present methods for the determination of strength properties of a wood species require extensive testing. An extra problem is that the timber that is tested has to be representative for the timber coming on the market. All future variations in the quality of the timber coming on the market have to be covered. To be able to use the timber in structures, grading rules have to be formulated that are related to the strength properties, determined by tests. For visual grading, features like knots and slope of grain are used. For machine grading, for example, the density and modulus of elasticity are used. For softwoods it has been proven that machine grading is more accurate and gives higher yields in the higher strength classes in comparison with visual grading. For tropical hardwoods, a major problem for visual grading is that the most important feature for the mechanical properties, the slope of grain, is very difficult to measure in practice. For this reason, only one visual grade is defined for tropical hardwoods and optimisation is not possible. A solution for abovementioned problems can be species independent strength grading, where only the influence of the measured features is taken into account, irrespective of the species. To investigate whether this would be possible, the research question dealt with in this thesis was: what are the influencing parameters for the development of species independent strength models, and can they be quantified to ensure safe, economic and sustainable use of softwoods and (tropical) hardwoods in structures ? To answer this question, a database consisting of a large number of test results from bending tests on European softwoods, temperate hardwoods and tropical hardwoods was investigated. This database was built-up in the last ten years in cooperation with the Dutch industry. Based on a literature survey, it was concluded that both the strength and stiffness of clear wood depend on the density of the timber, irrespective of the wood species. The natural variation in test values for both properties strength and stiffness are correlated. As a result, the stiffness is a good predictor of the strength for clear wood. Based on structural mechanics, mathematical models were formulated describing the reduction of strength and stiffness caused by the presence of knots and grain angle deviation. Because the density defines the maximum possible basic strength of the timber species, independent strength grading by visual grading is not possible. For some softwood species and temperate hardwoods, the grow ring width can be a measure for the density. For the majority of hardwood timber, there is no significant correlation. The examination of the visual measurement of the slope of grain has revealed that it is very difficult to accurately estimate the slope of grain for tropical hardwoods before a destructive bending test. As a consequence, the variation in strength properties between test samples from the same wood species can be very large. To determine the strength of timber brought on the market under the same trade name with sufficient safety, a reduction factor has to be applied to the test results. Because it is not known how large the variation can be in the slope of grain for tropical timber brought on the market under the various trade names, it is not possible to determine this reduction factor. By means of machine strength grading it is possible to detect the variation in slope of grain. The reduction of the stiffness (the modulus of elasticity) can be described with the same equation (the well-known Hankinson equation) as the reduction of the bending strength, only with other constant values. Because of this, the modulus of elasticity and the density are parameters that, together, can be used for machine strength grading for timber showing grain angle deviation. The reduction equation describing the strength due to the presence of knots has the same form as the reduction equation describing the stiffness due to the presence of knots, only with other constant values. Because of this, the modulus of elasticity and the density together are also the parameters suited for species independent machine strength grading of timber containing knots. Because the influence of knots and slope of grain on the modulus of elasticity cannot be distinguished from each other in the modulus of elasticity measurement, timber has to be divided into two groups for species independent machine strength grading: timber for which failure is induced by knots and timber for which failure is induced by slope of grain. Therefore, it is necessary to perform a visual assessment, to check for the group containing grain angle deviation that only knots of limited sizes are present in the timber. Furthermore, the visual check has to ensure removal of pieces with features that cannot be detected by machine readings, such as compression failures. A feature like a compression failure causes an unpredictable strength reduction and is therefore not allowed in structural timber. In practice, the modulus of elasticity can be determined in a simple manner by means of vibration measurements. On the basis of mathematical relationships between on the one hand the features knots and slope of grain and on the other hand the density and the modulus of elasticity, it is possible to formulate prediction models of the strength based on the measured density and modulus of elasticity of a piece. The values for the bending strength, the modulus of elasticity and the density for the standardized strength classes are related to timber with a moisture content of 12%. Correction factors have been derived to be able to adjust the test result of timber tested with a different moisture content to this reference moisture content, For structural sizes, no adjustments with regard to the reference sizes are necessary. The shape of the scatter around the prediction lines is theoretically derived on the basis of the distribution of the prediction values. The shape of the scatter turns out to be different for the prediction model for timber containing knots and for the prediction model for timber containing grain angle deviation. A method to derive the shape of the scatter on the basis of available data has been formulated and verified. To actually grade timber, "settings" have to be determined. These are limit values for the prediction values that determine which strength class the timber can be assigned to. The strength values of timber can only be verified on the basis of the properties of a sample that is tested destructively. For small numbers of pieces in a sample, the characteristic values of a strength grade can vary significantly between tested samples. To overcome this problem, a method was developed which takes into account the distribution of the prediction values and the scatter of the prediction model. The characteristic strength value of a strength grade for the required probability can be determined by it, irrespective of the number of pieces in a sample. With the developed prediction models it is possible to perform species independent strength grading. Especially for tropical hardwoods, the assigned strength classes can be determined in a reliable way and the yield in the higher strength classes can be increased. The research results contribute to an economic, safe and sustainable application of timber in structural applications.