Sixteen coded compounds were blind-tested at 4 laboratories using the recently described GADD45a-GFP genotoxicity assay. The compounds were chosen to include non-genotoxic compounds as well as weak and strong genotoxins. None of the compounds required metabolic activation in order to exhibit genotoxic effects. The participating laboratories included 2 global pharmaceutical companies, a global consumer goods company and the Gentronix laboratory in Manchester. Each compound was tested 4 times on different days following a protocol previously described. The tests were carried out after a 3-day training period from the parent lab (Manchester). Following the exclusion of data from tests with positive control failures and data series with 'spikes', 92% of assays gave the correct result: non-genotoxins giving negative results and genotoxins giving positive results. There were no randomly distributed problems suggesting that differences between the results from different sites reflected the use of different instruments, procedural differences and operator experience. In naïve operator laboratories the quality of data improved with operator practice. It was concluded that simple clarification of the protocol would provide the level of reliability required for widespread use of the assay in hazard assessment.