Abstract Calculations have been performed to estimate the removal rate of fission gas atoms from bubbles due to collisions with energetic fission fragments and recoil cascades. The efficiency of this process was found to be higher than estimated earlier, but is still too low to be responsible for the experimental observations of fission gas bubble destruction during irradiation of oxide fuel. An irradiation experiment to investigate the interaction of fission spikes with free surfaces has enabled a simple theory to be developed which can explain the shrinkage of bubbles and pores by the surface relaxation of a shock wave produced by the passage of a fission fragment. This mechanism occurs in oxides but not carbides because of the faster dispersion of the fission fragment energy and provides the major reason for the difference in gas bubble distributions in oxide and carbide fuel. This process, however, does not remove gas atoms from the bubbles. Since high levels of apparently diffusive fission gas release are observed in oxides, the “effective solubility” of the fission gases required for this release must be sought in phenomena other than the fission spike.