A method is suggested for modeling uncertainty when there is a lack of information concerning the effect of forest management decisions on tree growth. Dynamic programming is used to investigate the optimality of alternative management strategies. The model is illustrated with an empirical example for a boreal forest region of western Canada. Three tentative conclusions follow: (a) silvicultural strategies to reduce uncertainty or to increase stand growth may not be worth pursuing, at least in northern forests; (b) the discounted cost of ignoring uncertainty may be substantial if taken over the entire forest; and (c) given uncertain forest growth, flexible harvest policies are preferred to a fixed harvest age.