Thermal fatigue cracks have been found in austenitic pipe work in many pressurised water reactors, caused by thermal cycling due to the passage of water at different temperatures along the pipe inner surface. The rates of crack initiation and growth for this situation are not well understood because of the stochastic nature of the temperature fluctuations. Therefore, large allowances must be made when assessing the integrity of this pipe work to this failure mechanism. Improved assessment of crack initiation and growth could enable increased plant availability, and better safety cases. A programme of work has been completed consisting of fatigue tests on thick 304L butt-welded pipe specimens, and accompanying predictions of crack initiation and growth. In each test, uniform thermal cycles were generated using a water jet on a small area of the pipe. The magnitude of the cycles differed between the tests. Crack initiation and growth were monitored using a dye penetrant technique, applied to the pipe inner and outer surfaces, together with destructive examination. Crack initiation predictions were made using fatigue data derived from mechanical fatigue tests on the same material as in the pipe specimens. Good predictions were made using a strain-life endurance curve at a temperature corresponding to the average temperature of the metal surface during the thermal cycle. Crack growth predictions were based on an inelastic finite-element model accounting for cyclic hardening, and an enhanced R5 procedure (1) with crack closure taken into account. A linear elastic fracture mechanics definition of a Paris law for crack growth was used, and plastic redistribution effects were included. Predictions were good for all of the experimental scenarios carried out. A further experimental and analytical programme is in hand using the same experimental arrangements, concerning variable amplitude thermal loading.