Abstract: Our work consists in carrying out a survey whose purpose is to focus on the teaching of statistics at the IUT (polytechnics, a two-year course). We aim at explaining the characteristics of the teaching-learning situations specific to the field of statistics, and thus attempt to analyse the status, the effects and the conditions of the implementing of both "real size" surveys performed by the students, and simulations of random experiments designed by the students. Our analysis of statistics in terms of object-subject highlights the fact that its teaching holds in itself a primitive and epistemological breach between data collecting and processing on the one hand, and, on the other hand, an approach based on the calculation of probabilities more or less remote from social practices. It seems therefore that questionnaire-surveys constitute a means of reducing this primitive gap, by allowing the subject to build a learning meaning. Furthermore, we have studied how the constructions of simulations performed by the learners enables to obtain the invariants of the variability present in phenomena whose study is at the core of statistic situations. Our work is based on Gérard Vergnaud's works, as for conceptualization, on Guy Brousseau's theory of educational situations, as well as the educational transposition displayed in Yves Chevallard and Jean-Claude Regnier's works concerning educational statistics and the experimental trials of the learner.
Simultaneously, we have been trying to determine the function of a sandwich course training, as well as the dialectical difference between project and process. In terms of pedagogical engineering, our study displays educational situations, while attempting to point out the various obstacles we have managed to identify; we also show two surveys using statistical implicative analysis (Régis Gras).