Recent studies have found that stringer-to-cross-girder connections in riveted railway bridges are susceptible to fatigue cracking. Fatigue damage in these connections is caused by secondary, deformation induced effects. These effects are difficult to interpret in terms of a single, applied stress descriptor, which is customarily used in an S-N assessment. In order to address this problem, and, following the work published recently by the authors (Imam et al. 2006(Imam et al. , 2007, in this paper the results of a global-local finite element analysis of a riveted railway bridge are used within the context of the theory of critical distances (TCD). Here, using the TCD in the way proposed by Taylor (Bellett et al. 2005), fatigue damage is shown to converge upon mesh refinement. Comparisons of the TCD-based method with its more traditional, detail-specific S-N counterpart, reveal that the latter can underestimate fatigue damage in some cases by a factor of 3.5.