Timber harvest decision is one of the most important topics of forest economics. Martin Faustmann presented in 1849 the first "correct model" for determining the optimal time to harvest a forest stand. The Faustmann model builds on a set of restrictive assumptions that are far from realistic. During the past four decades the Faustmann model has been extended substantially. One important extension is the inclusion of non-timber benefits. Another is the recognition of uncertainty, especially the adoption of the adaptive optimization framework to determine the optimal time to harvest a stand under conditions of uncertainty. Currently available economic models of forest harvest decisions can be used to determine the optimal time to harvest a forest stand in a variety of special cases, but their ability to describe a typical harvest decision problem remains unsatisfactory. To improve the decision models, researchers must pay more attention to the fact that forests usually are managed for multiple purposes and under conditions of economic, biological, and ecological uncertainties. Therefore, non-timber benefits and uncertainties need to be considered simultaneously, which often implies that the decisions for different stands are interdependent. The information needed for applying the decision models also requires much more research. A particularly important, yet difficult, matter is the rational expectations timber price process.