The purpose of this study is to analyse the structure of written argumentative texts produced by pupils in grades 3, 5, 7 and 11 in relation to three different tasks: Group A — subjects are assigned a topic question consisting of a single statement (open question); Group B — subjects are given a topic question consisting of both a statement and its opposite (opposite opinions); Group C — subjects are given an initial and a final sentence of a text, which they have to complete, and which refers to opposite opinions (oriented task). Six different types of text organization were evidenced on the basis of the argumentative moves presented in the texts, the linguistic indexes used to express the relationships between the moves, the presence or absence of counter-arguments and the type of argumentation or reasoning used. The organization of the texts range from a simple affirmation of a position (repitition of the proposition or its opposition without any support) to a use of data or reasons to support a position, to others still in which two positions are presented in the body of a unitary structure, developed ‘in parallel’ and interconnected regularly. Our analysis found a well-defined tendency corresponding to the age levels in the types of texts produced by the subjects. The texts produced by the younger group, the pupils in elementary school, contained a fairly simple structural organization, characterized by very few argumentations or a lack of explicitness of the essential moves and a juxtaposing of arguments. The older subjects, in particular those attending secondary school, were able to produce texts which were more complex with regard to the presence and the organization of the argumentative moves, and more coherent and cohesive with respect to their articulation of opinions and use of adequate linguistic indices to analyze t topic. As regards the different types of tasks, when the subjects were presented explicitly with two opposite opinions (though in different terms), they developed both positions in their responses, contrary to the situation in which they produced one position when only one of two opinions was presented.