Expert reading skills are mandatory in our modern societies as reading represent an important contribution to scholar and social integration. Expert reading should rely on reading by sight automation, which is specifically based on direct lexical access from an orthographic form. Interestingly, reading efficiency can be predicted by several reading related tasks, among which: rapid automatized naming (RAN). The lexical access hypothesis proposes that RAN and reading are linked because of the smoothness of lexical and phonological encoding processing stages, and that RAN performed in a discrete format might be an indicator of reading by sight development. In this thesis, we aim at formally test the hypothesis assuming that discrete RAN is an indicator of reading by sight strategy using EEG recordings during RAN tasks. We also strive to disentangle the respective contributions of visual and lexico-phonological processes through reading instruction.