The approach presented optimizes investment in timber production when the manager is faced with uncertainties both in future product markets and in the response of stands to management actions, and when the objective is to maximize total discounted expected returns over an unending time horizon. The uncertainties are formulated in a Markovian decision process with the state of each stand described by average tree size, stocking level, and market condition. Solution of a functional equation gives a "stationary" management policy in which the optimal action to apply to any stand depends only on the observed state and not on the decision time, past states, or past actions. The approach is applied to young growth Douglas-fir. Forest Sci. 21:109-122.