Summary : The effects of the addressee's degree of conviction on the use of cleft sentences. This study investigates the effects of how a speaker A expresses a belief on the syntactic form chosen by a speaker B to assert a proposition uncompatible with this speaker A's belief. In reference to linguistic and psycho-linguistic analyses of cleft constructions, it is hypothesized that the more certain a speaker A is of the truth of what he says, the more a cleft construction will be used by the speaker B to carry out this belief modification. The results of a first experiment (matching task) show that the subjects systematically match the cleft version of B's reply with the context where the contested belief is expressed by A with a high degree of certainty ; the uncleft version of the same reply, on the contrary, is systematically matched with the context where A appears more hesitant about the truth of what he says. In a second experiment (elicited production task), the same pattern of results is observed, with one exception : when the contested information concerns the agent of an action, the reply always takes the form of an it-cleft sentence, whatever the speaker A's belief might be. However, whatever the status of the contested information, the speaker A's degree of certainty has an effect on the use of complete or elliptical clefts by B as well as of an introductory explicit negative sentence. These results are discussed in terms of facilitating the processing of information by the addressee. Key words : cleft sentence, contrastive function, production.