Many studies on the structural effects of corrosion in reinforcement have been conducted. However, most of them are based on artificially corroded test specimens. Thus, the knowledge available entails one major uncertainty, i.e. whether the results are reliable enough to be used for naturally corroded structures. The purpose of this study was to develop a test method and carry out experiments on naturally corroded specimens taken from an existing structure to investigate the anchorage capacity. Beam specimens were taken from the edge beams of a bridge at repair. The specimens showed corrosion-induced damage to a varying extent from no sign of corrosion to extensive cracking and spalling of the concrete cover. A four-point bending test indirectly supported by suspension hangers was chosen. The beams were strengthened with transverse reinforcement around the suspension hangers to avoid premature failure. Eight successful tests were carried out; in all these tests, diagonal shear cracks preceded a splitting induced pull-out failure; i.e. anchorage failure was achieved as intended. The results showed around 10 % lower capacity for the corroded specimens than for the reference ones. The average bond stress in the anchorage zone was estimated based on the applied load and available anchorage length. The stress was about 16 % lower in the beams with corrosion cracks, and 9 % lower in the beams with cover spalling compared to the reference specimens; there was also a larger variation among the damaged specimens than for the reference specimens. The results extend our knowledge concerning the structural behaviour of corroded reinforced concrete structures during field conditions.