Bond between reinforcing steel and concrete is essential for the combined action of the two materials and the overall resistance of reinforced concrete structures. The bond strength of embedded bars depends on the stress-state of the concrete surrounding the reinforcement. However, existing bond test recommendations typically induce transverse compressive stresses from the support reactions on the bond region, affecting the results and limiting the applicability of the empirical models that are derived. To overcome this parasitic effect, a novel cantilever bond test set-up was developed and validated experimentally. A new specimen geometry with a protruding nib was developed to remove spurious support compressions from the bonded region. The resulting geometry has similarities with that of half-joint beam ends. The longitudinal bars were pulled-out of the full-depth section. Due to the distribution of internal forces, the transverse reinforcement was simultaneously subjected to tensile stresses. The presence of transverse reinforcement limited the opening of longitudinal cracks and the subsequent reduction in bond strength. The new test method leads to a better understanding of the fundamental aspects that underpin bond and anchorage in reinforced concrete. It also allows parameters useful for design and assessment calculations to be evaluated more accurately.