This paper is about causality, defined as a specific relation between eventuality on an event-state chain, called a ‘causal chain'. Though causality is a dynamic temporal relation, the expression of causality in discourse contrasts with temporal discourses, in which temporal order between eventualities is parallel to the sequential order of linguistic segments in discourse. Causal discourses are backward, introducing first the effect and second the cause. This property is used to analyze possible causal discourses with and without explicitated causal links by means of connectives (in French parce que, donc, et), and mainly their causal and inferential uses. Finally, the paper tries to answer why causal relations in discourse are used to convey explanation and argumentation. Explanation is the discourse relation corresponding to causal relations in the world, whereas argumentation is a special use of causal relation in discourse, implying causality between states belonging to two causal chains.