Single-well reactive tracer tests, such as the push-pull test are useful tools for characterising in-situ bioattenuation processes in contaminated aquifers. However, the analytical models that are used to interpret push-pull data may be over-simplified, and potentially overlook important processes responsible for the frequent discrepancy between predicted and observed results obtained from push-pull tests. In this study, the limitations underlying the push-pull test methodology were investigated and were supported with results from a push-pull test conducted in a sulphate-reducing aquifer contaminated by crude oil. Poor (<7%) mass recovery was achieved when the push-pull test was performed in a fast-flowing aquifer, preventing a quantifiable reaction rate to be determined. Breakthrough curve data were unexplainable using simplified analytical models, but exhibited trends analogous with tests conducted by others, when >20% mass recoveries were achieved. Push-pull test data collected from sulphate-reducing aquifers indicate that the assumption of a well-mixed batch reactor system is incorrect and that reaction rates obtained from push-pull tests in such systems may be affected by the extraction regime implemented. Evidence of microbial respiration of the reactive tracer was provided by stable sulphur isotope analysis, from which an isotope fractionation factor of +9.9 +/- 8.1%. was estimated. The stable isotope data support the argument that reaction rates calculated using push-pull tests are not uniformly distributed in space and time and are likely to be influenced by heterogeneities in the flow field. (C) 2003 Elsevier Ltd. All rights reserved.