The results of fatigue tests are characterized by much scatter. Such scatter is further increased if data from different test series are combined to derive, for instance, characteristic values for individual types of welded joints used in codes. Characteristic values are normally applied to the design of fatigue-resistant ship and offshore structures in connection with the nominal stress approach using S-N curves. More advanced approaches such as the hot-spot stress approach and the notch stress approach are applied to an increasing extent. Such approaches explicitly consider certain influence factors and allow the scatter of these factors to be treated individually. This way, probably even the total uncertainty can be reduced. After reviewing the different approaches used for fatigue strength assessment, the sources of scatter are addressed and assigned to factors considered in the different approaches. Based on published data of fatigue tests and imperfections observed in real structures, an attempt is made to quantify the uncertainties of the different factors and to draw conclusions for their individual consideration in the approaches mentioned.