In this thesis, we study the role of visual attention when an expert reader acquires new orthographic lexical knowledge. Our contribution is twofold. On the one hand, we develop an original computational, probabilistic model of orthographic learning. Our model, named BRAID-Learn, is an extension of BRAID, a hierarchical probabilistic model of visual word recognition and lexical decision. On the other hand, we gather original experimental data on the evolution of eye movements during incidental learning of new orthographic forms and demonstrate the ability of the model to account for these observations. Our contribution is described in three articles.In the first article, we simulate the length effect as experimentally observed for words in ...