Some early conclusions were drawn from the stress analyses of bearings assuming simply hydrodynamic lubrication, that is, no interaction between oil film pressure and housing deformation. Negative oil film pressure may arise but its effect on predicted stresses was not systematically assessed; it was argued that the consistency between theoretical predictions and experimental measurements may be improved if negative pressure is taken into account. Tensile tangential stresses were found where the pressure falls to zero, that is, at the location of steepest pressure gradient; their maximum value was predicted at the backing-lining interface. Both flat strip and a bearing shell analyses showed good correlation between the fatigue crack origins in test specimens and their assessments of the maximum tangential stress. However, shear stress and distortion energy maxima were found subsurface. Parametric studies indicated that housing rigidity would have the greatest influence on the magnitude of the tensile stresses that develop at the surface of the lining. These stresses as well as the shear stresses at the lining-backing interface rise with decreasing housing stiffness and increasing lining thickness. The coupled analyses under EHL conditions performed by Bahai and Xu lead to the conclusion that tensile deviatoric tangential stresses are present at the observed crack initiation site during the whole engine cycle while their hydrostatic counterpart dominates the overall stress state in a compressive environment. A more recent, multi-step plain bearing modelling procedure, accounting for all causes of stress development, produced stress and strain cycles with the potential to initiate fatigue damage. Forming and fitting residual stresses were found to have a significant impact on the stress results although strains were affected to a lesser extent. The high plastic deformation predicted by the numerical analyses corresponded to high cycle fatigue; this however was not consistent with fatigue test results and led to the consideration of engine temperature and elastic shakedown as possible reasons for milder stress conditions and the use of simpler elastic analyses. The damage maps based on predicted strain and stress amplitudes were consistent with experimental observations. The effectiveness of sub-modelling technique was tested under accelerated test conditions; the results obtained were in good agreement with those from the full model. The generation of such sub-models can be very useful in damage tolerant analyses since the creation of a full model incorporating a propagating crack would require enormous memory resources and computation times.