Past experience has shown that stringer-to-cross-girder connections in riveted railway bridges are susceptible to fatigue cracking. This fatigue damage is caused by secondary stresses, which develop in the different components of the connection. For this reason, more detailed analysis techniques are needed to capture this type of behaviour. In this paper, a finite element (FE) model of a typical riveted railway bridge is developed by incorporating the detailed local geometry of a stringer-to-cross-girder connection into the global bridge model. Before the development of this model, benchmark FE studies are carried out on a double-lap joint and the results are presented in terms of stress concentration factors and stress gradients. Further verification studies are carried out on a local bridge connection FE model, in terms of its rotational stiffness. After this investigation, a refined FE model of the bridge is analysed under the passage of a freight train. Principal stress histories at different components of the connection are obtained, which are then combined with the plain material S–N curve, in order to identify the most fatigue-critical locations of the connection. These are identified as being the rivet holes and, in some cases, the angle fillet. By considering different rivet clamping stresses and different rivet defect scenarios it is found that the most damaging effects are caused by the presence of clearance between the rivet shank and the hole, and the loss of a rivet. The rivet clamping stress is also found to affect fatigue damage considerably.