This paper describes an approach for the judgmental assessment of instream flow levels appropriate for local fish protection. The context is that when a hydroelectric plant holds back water, a high level behind the dam increases power potential but leads to low water flows downstream. Modeling stream conditions is often too complex and in practice choices are made by small groups of experts and interested parties who know the stream environment. Usually a few test flows are made by dam releases and results observed and recorded downstream. Depending on fish types, habitat, spawning strategies, predation and so forth, it's not obvious how much flow is better and how to set reasonable values. In this paper Steve Railsback and I explain a detailed methodology we developed for the Electric Power Research Institute which makes use of high-level versions of fish population models. The stream photo above is by Steve Railsback.