The objectives of this study were to study wood property variation in Pinus patula from different physiographic sites, the interrelationship between wood properties and the effects of some forest site factors on wood properties and their within-tree variation. The wood properties considered were density, transverse shrinkage, grain angle and stiffness (modulus of elasticity). Ten trees were taken from each of seventeen sites selected during a previous study of soil ecology and tree growth patterns on the escarpment of Mpumalanga in South Africa. This is a complex physiographic region with several different growth sites within a relatively small geographical area. The key finding of the study was that the effect of radial distance from the pith (i.e. the age of the tree) was highly significant and accounted for most of the variation of the trees sampled. Although the effect of site on the selected wood properties was highly significant, its effect accounted for a relatively small percentage of the total variation in each case. Consequently, solid wood processors must take into account the large variation within a tree when implementing strategies to minimise the downgrade of timber as result of excessive board distortion and/or inferior strength. There were significant correlations between wood density and density slope, radial shrinkage, shrinkage ratio and stiffness and also significant correlations between wood stiffness and density slope and radial shrinkage. Density and grain angle both showed significant correlations with the most important climatic factors. ‘Growth days’ (which is an index reflecting the number of days per year when soil moisture is not a limiting factor for plant growth) was identified as one of the principal growth factors influencing the selected wood properties, because of it’s strong relationships with wood density, density slope, grain angle and stiffness. Regression models were developed to predict wood density and grain angle from growth variables. The results show that those wood properties that correlate with geographic representable site variables can be included in site classification and evaluation systems, to assist in predicting growth in terms of volume yield and wood quality.