Simulations based on the Ricker and Beverton–Holt stock–recruitment models illustrate the difficulties with developing information about the effectiveness of habitat restoration efforts from the relation between measurements of habitat and populations of anadromous fishes. The relation between the two is confounded by density-dependent mortality and variable density-independent mortality and is masked by measurement errors. The simulations are considered in terms of populations of fall-run (ocean-type) chinook salmon Oncorhynchus tshawytscha from the Sacramento–San Joaquin river system of central California, where major federal and state efforts are underway to restore anadromous fish populations, as well as brackish and freshwater ecosystems. The simulations show that to implement effective adaptive management of salmon habitat, these efforts must move beyond a trial and error approach in which efforts to restore salmon habitat will be evaluated by population responses. A more promising alternative is evaluating restoration efforts by identifying and testing hypotheses about the mechanisms or processes that relate the restoration actions to populations.