A methodology for assessing toxic effects of chemicals on fish should be able to begin with an emission rate, an LC₅₀, and a description of the receiving system and generate an estimate of the likelihood of reductions in fish populations. This process is a series of extrapolations, each with an associated variance. The LC₅₀ must be extrapolated from the test species to the species of interest, to life-cycle toxicity, to long-term toxicity in the field, to changes in population size due to direct toxic effects and, finally, to the combined direct and indirect toxic effects. Similarly, the emission rate must be converted into an effective environmental concentration in an imperfectly known hydrologic, chemical, physical, and biological system. In this paper we summarize some data and methods for making these extrapolations and indicate sources of uncertainty in each stage of the analysis.