This article demonstrates that science cannot be a "neutral arbiter" for triggering precautionary measures, because both making and warranting findings of risk require non-scientific decisions. Making a risk finding requires decisions about the meaning of "risk of harm," about the meaning of any modifiers for that predicate, and about the degree of confidence asserted for the finding as a whole. Determining that the available scientific evidence warrants a finding of risk requires decisions about acceptable degrees of various types of uncertaintynamely, conceptual uncertainty, measurement uncertainty, sampling uncertainty, modeling uncertainty, and causal uncertainty. This article illustrates these decisions using examples from the food safety law of the United States, recent animal feed cases in the European Community, and Appellate Body decisions in WTO trade disputes. Finding a risk that triggers precautions cannot be a purely scientific act, notwithstanding the myth that a "value-neutral" science can do so.