Static travel cost models have been only partially successful in modeling the sequential aspects of a recreation decision like the decision to fish. In this paper we describe a dynamic structural model of the decision to visit a recreation site. The model is best described as a dynamic multinomial logit model. By virtue of its dynamic nature, the model avoids the problem of the independence of irrelevant alternatives that afflicts static multinomial logit models. Welfare effects of changes in site quality are easily calculated via dynamic programming.