The term polyphony is used in the field of discourse analysis and pragmatics in order to describe one important dimension of the organization of discourse, and also of utterances, i.e. the fact that they can express and combine different voices. The notion of polyphony has a strange history. Early on, in the thirties, polyphony was identified as a major dimension of discourse by the Russian linguist Mikhail Bakhtin, in the frame of his dialogical approach to discourse, but compared to the large number of studies on speech acts or implicatures, the body of literature on polyphony is considerably smaller (the notion of polyphony being still absent from most major textbooks and dictionaries). This is probably due to a conjunction of different facts: first, for various political and personal reasons, Bakhtin’s texts were discovered and translated outside Russia only at the beginning of the seventies (see Todorov 1981); second, Bakhtin’s thought is often complex, elliptic and very difficult to translate; third, the paradigms which have dominated anglo-saxon pragmatics during the past twenty years (speech act theory, conversation analysis, relevance theory) ignored Bakhtin’s sociological approach. Given this state of affairs, it comes as no surprise that, although the importance of the polyphonic dimension of discourse is not contested, few linguists have worked on it. The only one to have undertaken recently a systematic elaboration of the notion of polyphony is the French linguist Oswald Ducrot, in the frame of his enunciation theory, but he is more interested in the utterance than in the discourse level. This paper will first present briefly Bakhtin’s and Ducrot’s notions of polyphony and evaluate their contribution to our knowledge of this dimension of discourse, and then, describe the different aspects of the polyphonic organization which have to be taken into account in discourse analysis. Finally, it will present an analysis of the polyphonic organization of a short fragment of a letter.