In a repository for high-level nuclear waste in bedrock that may carry water, the waste will eventually come in contact with water. Radionuclides will dissolve in the water and migrate away from the repository. In crystalline rock in Sweden, the waters at repository depths are reducing. Many of the important radionuclides, e.g., neptunium, uranium, and technetium, have very low solubilities under these conditions (parts per billion levels). The solubility will considerably limit the transport of these species. If by some means the conditions were to change from reducing to oxidizing, the solubility of these species would increase very much, in some cases by 4 to 6 orders of magnitude. Under such circumstances, these nuclides would escape much faster. We have investigated one possible way in which the redox conditions might change, i.e., spent fuel has a considerable alpha activity, which may radiolyze water and produce oxidizing agents such as hydrogen peroxide. The hydrogen peroxide will make the water oxidizing. The compensating factor is ferrous iron in the bedrock. In the investigation of the interaction of these two species, a con-ceptional and a mathematical model is developed describing the movement of the redox front downstream of a repository. A sample calculation based on minimum measured ferrous iron contents in the bedrock and computed (conservatively on the high side) hydrogen peroxide production shows that the redox front could move several tens of metres downstream in the million-year perspective. The rate of radiolysis would decrease considerably if the spent fuel is not wetted to the high degree assumed in the calculations. The results in the sample calculation should be seen as maximum values for the type of repository considered