A general theory for the origin of growth stresses in the reaction wood of trees is presented. In both gymnosperms and arboreal dicotyledons, stress is considered to arise from the cellulosic component of the wood. It is suggested that in gymnosperms, cellulose microfibrils behave as helical springs. The helical springs are thought to be laid down in a compressed state and thus exert a compressive force which acts to right or stabilise the tree. In arboreal dicotyledons the cellulose microfibrils are laid down as extended, longitudinally oriented springs and thus exert a tensile force tending to right or stabilise the tree. Contrary to the prevailing view, lignin is not considered to be involved in the generation of growth stress in reaction wood. It is suggested that the sole function of lignin is to cement the cellulosic constituents into a whole, thus ensuring the transmission of stresses through the wood. The low values of lignin often reported from tension wood and especially from gelatinous fibres is seen as a mechanism to facilitate the contraction of microfibils thus maximising longitudinal tensile stress. The high values of lignin in of compression wood is seen as a mechanism to increase the compressive strength of compression wood.