A review of experimental data on free embedded cantilever walls is presented. Normalization of limit equilibrium depths of embedment and maximum bending moment was successfully carried out, but it was found that very little data were in existence. In contrast, many methods of limit equilibrium design were found to be in use. Subsequently, new experiments have been performed, using a small-scale model of a free embedded cantilever wall. Further experimental data have been obtained on maximum bending moments, depths of embedment at failure, and also normal and shear stresses mobilized between soil and wall. New comparative analyses have been performed between these data and predictions by current design methods. The adequacy of these methods has been assessed in the light of the new data. A special model wall facility was built for these experiments. It comprised a test tank, a small-scale model wall, and instrumentation for measuring normal and shear stresses between soil and wall. New types of boundary normal stress and boundary shear stress transducers were developed, using the Hall effect principle. A re-assessment of the phenomenon of cell action factor has also been carried out. Previous design criteria for diaphragm-type boundary normal stress transducers did not predict accurately values of cell action factor measured in this thesis. Accurate predictions of cell action factor were obtained only by a design criterion considering the influence of soil modulus, which was determined in the triaxial test using local strain devices.