A key element of the biomechanical design of trees is their ability to generate large mechanical stresses in wood at the stem periphery. This function is necessary for the tree to control the orientation of its axes, and therefore to grow in height, maintain its branches at an optimal angle or achieve adaptive reorientations (Fournier et al., 2013). This “maturation stress” appears in wood fibres during their cellular maturation when their secondary cell wall is thickening. In hardwoods, the stress asymmetry is generated by the formation of specific fibres with a very high tensile growth stress on the upper side of the inclined axis. The resulting tension wood has microstructural features highly different from the normal wood. Tension wood in almost all temperate species has a specific gelatinous cell wall layer, called G-layer, that is not lignified and highly mesoporous (Chang et al., 2015). It has been shown recently, at the macroscopic scale, that cellulose tension appears during maturation and is synchronous with the development of this specific layer (Clair et al., 2011). Tropical species, like simarouba, are able to generate lignified tension wood fibres. The mechanisms of stress generation in tension wood fibres are not yet well known. The aim of our study is to understand these mechanisms and to know the spatial and temporal kinetics of the different cell wall layers stiffening during maturation, in parallel with tensile stress generation. Contact-Resonance Atomic Force Microscopy, CR-AFM, was used here to address this question on embedded samples (Arnould and Arinero, 2015) using dual resonance frequency tracking (DRFT, Rodriguez et al., 2007). Change in the contact modulus between, and within, each layer of the cell wall has been measured on several radial lines of developing fibres at different steps. In order to understand the obtained results, and to estimate the sensitivity of the AFM indentation-like technique to the different cell wall components stiffness, an elastic anisotropic indentation model was used (Vlassak et al., 2003; Jäger et al., 2011). Finally, these mechanical measurements have been compared to topochemical data obtained on the same fibres.