Possibilistic networks are important tools for modelling and reasoning, especially in the presence of imprecise and/or uncertain information. These graphical models have been successfully used in several real applications. Since their construction by experts is complex and time consuming, several researchers have tried to learn them from data. In this paper, we try to present and discuss relevant state-of-the-art works related to learning possibilis-tic networks structure from data. In fact, we give an overview of methods that have already been proposed in this context and limitations of each one of them towards recent researches developed in possibility theory framework. We also present two learning possibilistic networks parameters methods.