The significance of modeling and measuring various attributes of the Web in part or as a whole is undeniable. In this paper, we consider the application of patterns in browsing behavior of users for predicting access to Web documents. We proposed two models for addressing our specification of the access prediction problem. The first lays out a preliminary statistical approach using observed distributions of interaccess times of individual documents in the collection. To overcome its deficiencies, we adapted a stochastic model for library circulations, i.e., Burrell’s model, that accounts for differences in mean access rates ofWeb documents. We verified the assumptions of this model with experiments performed on a server log of accesses recorded over a six month period. Our results show that the model is reasonably accurate in predicting Web page access probabilities based on the history of accesses.