Abstract
This study questions the hypothesis according to which Twitter has become a privileged space to study, measure, even «predict» public opinion. Especially, according to several academic works, «opinion lead-ers» important presence on the network could explain its «predictive» character. To discuss this hypothesis, we created a panel of 1228 individ-uals and we followed their activity for 4 months. The main result is that the majority of the participants, although strongly endowed with cultural capital, speak only very rarely politics: only a minority can be assimilated to «opinion leaders»