This paper analyzes the processes by which three levels of government analyzed potential risks from a precedent setting environmental application of biotechnology, Advanced Genetic Science's FROSTBAN product. The research is based on regulatory decision dockets, accounts of the case in scientific journals and newspapers, and personal interviews with key scientists, policy analysts, and senior decision makers. The 1.5 hour interviews employed both fixed choice and open-ended questions. Four distinct conceptions of potential hazards arose in the AGS case: toxicity to humans, pathogenicity to plants, ecological risks to non target species, and climate changes. Worst case scenarios were devised in each organization to evaluate the application. The scientific assumptions of these analyses differed substantially. All three organizations requested additional technical data from AGS. Criteria for evaluating such data were not always prepared before performing experiments. Scientists believed the most critical uncertainties related to their individual discipline. Many of the uncertainties identified in the AGS case could not be unambiguously resolved by provision of more data. A framework for interpreting information on potential ecological hazards of introduced organisms is not available from any one of the disciplines that contributed to the AGS review. Future risk analyses could be improved by developing explicit ecological goals and criteria for data interpretation. In the meantime, biological controls should be required for tests of engineered microorganisms to reduce many identified uncertainties.