Written argumentative discourses were produced by 7 to 14 year-old children in two debate situations: one concerning a scientific issue (“Discours Formel”: DF,) and another concerning an opinion issue (“Discours Naturel”: DN). We had made the following developmental hypothesis: a specific discourse representation would be gradually built up by children in each situation, and would enable them to produce two different discourses, particularly with regard to the implication marks used by the writer. The two debate situations had been set up (...) to be only differentiated by the type of proposed issues. All the remaining situational characteristics had been made identical (social scenario, announced goal, addressee, etc.).Children's verbal products were divided into utterances. Each utterance was coded with regard to whether the following implication marks were present or not: the assuming by the writer of its assertive content; the writer's linguistically expressed reference to him- or herself; the evaluative or prescriptive value of the utterance; and the certainty modalities.The main results were: writer's assuming the assertive content specifies more and more DN, with a clear opposition between the two discourses realised by 13–14 children. While decreasing with age in both situations, the writer's self references are found to only occur in DN from 13–14. Prescriptive and evaluative types of utterance appear only in DN, at all the ages; the first ones decrease and the second ones increase from 7 to 14. At the opposite, DF proves to consist exclusively of constative utterances. At last, certainty modalities mainly occur in DN utterances. Differentiation between the two discourses has so begun by 7–8. But it becomes complete, for all the implication marks, at 13–14. The results observed at 13–14 join those shown by two reference adult groups, which performed the same discourse tasks.Finally, results are discussed with respect to parts of conceptions developed by Grize and by Bronckart. Given the many controls brought to bear in setting up the two discourse situations, children seem really to have been led, by the one representation of the argumentative issue, to produce more and more differentiated discourses. (shrink)