The uncontrolled release of radioactive fission products from the primary circuit of a nuclear power reactor, described as ?release source term?, is one of the major criteria for assessing a severe nuclear accident. A large number of tests were carried out under the European Commission (EC) Framework Programmes to improve our understanding of the fission product behaviour. Especially the five integral pile tests that were carried out in the Phebus FP programme lead to an increased understanding of the mechanisms of severe accidents, but also emphasized the need of further separate effect tests to study some individual phenomena in more detail. One of these was the possibility of chemical revaporisation of fission products (especially 137Cs) after their initial deposition inside the primary circuit or containment. This revaporisation can lead to a significant release in the late phase of a severe accident which was previously not taken into account when calculating the release source term. To study this behaviour using realistic samples of fission product deposits, which were obtained from the Phebus test circuit, a lead shielded revaporisation device was designed and set-up inside a glove box. In the device, a compact high temperature furnace is used to heat the samples under a controlled gas flow up to 1273 K with nearly ambient pressure. The gas flow is slow enough to prevent physical resuspension. It consists of hydrogen, steam, air or argon. An on-line gamma spectroscopy system is used to scrutinize the revaporisation behaviour by monitoring the loss of activity for the 137Cs in the sample. In the most recent test sequence, consisting of 3 experiments, samples from the Phebus FPT3 upper vertical line were heated with 2 K/min up to 1273 K under steam, hydrogen or air. The tests showed similar results for the two experiments under oxidising conditions (steam or air) were used as process gas. Revaporisation started for steam at ~810 K respectively for air at ~860 K and progressed for some time with a loss of activity equal to ~1 %/min. More than ~90 % of the 137Cs had revaporised at the end of the experiment. The experiment using reducing conditions (pure hydrogen) showed diverse results. The revaporisation did not start before ~955 K and progressed with a slower loss of activity (~0.7 %/min). The total amount of revaporisation was only 79 %, which was caused by a power outage which the system experienced before the planned heating sequence had finished. Nevertheless, the results show that revaporisation is possible and that a switch in the atmospheric conditions (reducing to oxidising) lowers the threshold temperature which is necessary to trigger revaporisation. For the last experiment a sampling method was introduced to determine the size and chemical composition of the aerosols that form after the process gas is cooled down. Analysis showed a large number of particles in the diameter range 8- 20 nm and EDX revealed Cs, Re and Si to be their main components.