Traces of navigation as knowledge for simulating information retrieval by navigation in hypertext document

Richness of current websites or hypermedia, the continuous evolution of them, can cause complexity in the structure of these hypermedia and some difficulties in navigation. From their early first design but also during their different evolution, it may be useful to test accessibility of offered information. The use of testing robots - complexes to be elaborated, should be a solution to carry out frequent simulations of website information accessibility. Such a solution requires a model of user behavior that is the subject of the work presented here. A specificity of our approach is that it combines empirical and theoretical approaches of modelling in order to obtain a realistic model. We reached a proposal which take into account traces of navigation previously gathered on the website to test. The proposal of a model, with limited but clear conditions of application, is by its own a contribution to this domain.