As already mentioned, uncertainties are pervasive in risk assessment. When uncertainty concerns the magnitude of a physical quantity that can be measured or inferred from assumptions (e.g., ambient concentration), it can often be quantified, as Chapter 9 suggests.

Model uncertainties result from an inability to determine which scientific theory is correct or what assumptions should be used to derive risk estimates. Such uncertainties cannot be quantified on the basis of data. Any expression of probability, whether qualitative (e.g., a scientist's statement that a threshold is likely) or quantitative (e.g., a scientist's statement that there is a 90% probability of a threshold), is likely to be subjective. Subjective quantitative probabilities could be useful in conveying the judgments of individual scientists to risk managers and to the public, but the process of assessing subjective probabilities is difficult and essentially untried in a regulatory context. Substantial disagreement and misunderstanding about the reliability of quantitative probabilities could occur, especially if their basis is not set forth clearly and in detail.

In the face of important model uncertainties, it may still be undesirable to reduce a risk characterization to a single number, or even to a range of numbers intended to portray uncertainty. Instead, EPA should consider giving risk managers risk characterizations that are both qualitative and quantitative and both verbal and mathematical.

If EPA takes this route, quantitative assessments provided to risk managers should be based on the principles selected by EPA. EPA might choose to require that a risk assessment be accompanied by a statement describing alternative assumptions presented to the agency that, although they do not meet the principles selected by EPA for use in the risk characterization, satisfy some lesser test (e.g., plausibility). For example, EPA generally assumes that no threshold exists for carcinogenicity and calculates cancer potency using the linearized multistage model as the default. Commenters to the agency on a specific substance might attempt to show that there is a threshold for that substance on the basis of what is known about its mechanism of action. If the threshold can be demonstrated in a manner that is satisfactory under the agency's risk-assessment principles, the risk characterization would be based on the threshold assumption. If such a demonstration cannot be made, then the risk characterization would be based on the no-threshold assumption; but if the threshold assumption were found to be plausible, the risk manager might be informed of its existence as a plausible assumption, its rationale, and its effect on the risk estimate. In this way, risk assessors would receive both qualitative and quantitative information relevant to characterizing the uncertainty associated with the risk estimate.