The role of Bayesian decision theory in hydrologic design problems is presented both in theory and by example. The theory is applied to an actual flood levee design problem on the Rillito Creek floodplain in Tucson, Arizona. Computer solutions provide a basis for judging the costs of overdesign in the face of uncertainty in the parameters of the flood frequency model (log normal) and for determining the worth of hydrologic data. One conclusion is that decision theoretic analysis looks at the decision situation from the standpoint of the engineer: how can one best decide in the face of limited data and the present knowledge about system behavior?