Nonlinear models have frequently been used to characterize dose-response data obtained from biological assays. The effect of a bioactive agent is observed and the model allows prediction of the dose required to obtain the observed effect (the 'inverse prediction'). The precision of this estimate is important in potency determination. Here, a general method is presented for calculating the inverse confidence intervals for estimates of dose potencies obtained from nonlinear models often used to describe these tests. The approach is demonstrated with application of data sets to the negative exponential and four-parameter logistic regression models. Necessary theory is presented and followed by detailed discussion in which estimation strategies are explained and intermediate quantities calculated.