Sequences for the production of explanatory texts have been offered in medium classes four times in the same year along the following lines: science session, individual writing of a first version of a text to explain the phenomenon studied, phase of exchange between partners, individual revision of the text by the author. During the exchange-phase, pupils are divided into two groups: students in the first group read the texts of their partners and send them a written criticism, participants in the second group read and comment on the criticisms addressed to them. The aim of this scheme is to separate a number of language mental activities in a collaborative review scheme: on the one hand, the use of advice and suggestions enabling the scriptor to go beyond what he is able to do on his own; on the other hand, lemprunt in the texts read and the elaboration of a clearer representation of the problems that arise, thanks to the activity of formulating and reformulating critical emitters. An analysis of the texts produced in semantic proposals and the statistical processing of the quantified data thus obtained were supplemented by a qualitative analysis of the relationship between the content of the criticisms and the revised texts, both in the authors and in the critical receptors. These analyses point to a more relevant revision (i.e. changes leading to more real, consistent, topical, gender-sensitive, reader-oriented and fuller text elements) in students emitting criticism than in critical receptors. They also show that progress in the competence to produce and revise its own text goes hand in hand with increasing expertise in criticising the texts of others. This research thus highlights the role of critical activity as a conceptualisation in collaborative review situations. It also illustrates the fertility of a research method based on the construction of educational situations developed for research purposes to shed light on problems arising from normal practice.