The scatter in notch fatigue lifetimes of CMSX4 at 650˚C and 725˚C in air and vacuum and with Rene N5 in air at 650˚C is compared under the same (plastic) notch strain range levels. Rene N5 shows shorter lifetimes under equivalent conditions and always exhibits multiple initiation sites. The role of interdendritic porosity in initiating fatigue in both alloys is identified, and the number of initiation sites is found to directly affect fatigue life. In air in CMSX4 and Rene N5, subsurface pores initiate fatigue, and this is believed to be due to the repeated in-filling of surface pores by oxidation product, reducing their associated stress concentration and effectively deactivating them as a fatigue initiation site. Tests in vacuum support this hypothesis as cracks do initiate at surface porosity under vacuum conditions. An attempt to evaluate initiating porosity distributions, has indicated a correlation between total area of initiating porosity and fatigue lifetime, which to some extent may allow for crack coalescence behaviour. A deterministic fracture mechanics based model has been proposed to allow for the effect of pore shape, size and position in determining subsequent fatigue life (and hence scatter). The predictions of the model have been assessed using full factorial design of experiments, assessing the effects of variability in pore shape, size and distance below the notch root, as well as the materials parameters (crack growth laws and K max ) used in the lifing calculation. The model successfully explains some, but not all of the observed scatter in lifetimes.