The benchmark dose (BMD) is the dose of a substance that is expected to result in a prespecified level of effect, the benchmark response level or BMR. It is a general approach to characterizing dose response, applicable to any toxicant and endpoint. A BMD is conceptually superior to a "no observed adverse effect level" (NOAEL) for this purpose because of being less determined by experimental design, because it is a precisely defined entity, and because its precision can be estimated. Since a BMD is a single number, just as an NOAEL, it is tempting to use the BMD as a straightforward replacement for the NOAEL in the assessment process for calculating allowable daily intakes. However, the level of toxic response at a NOAEL is unknown, while that at a BMD is well defined. Use of the BMD approach potentially adds consistency and objectivity to the process of deriving reference values (RfDs, RfCs, or ADIs) for setting regulatory levels. To take advantage of this, BMRs need to be selected in a consistent way across studies and endpoints. This paper discusses some issues affecting the selection of BMRs, and presents an example of a BMD calculated for the effects of peripubertal exposure to the fungicide vinclozolin.