Maitre, Jean-PhilippeJean-PhilippeMaitre2023-05-042023-05-042015-08-28http://hdl.handle.net/20.500.12162/6692Despite several approaches and many researches, studies of classroom interactions have mostly taken interest in short time spans of teaching for each class observed (from minutes to a couple of hours). And yet, learning is certainly grounded in longer durations. In our contribution, we will describe the method we are developing to work on larger data set. It is based on the hypothesis that mere analyses of uses of certain terms by the teacher ñ keywords and clues ñ may bear insights about what the teacher supposes the students know. Through video recordings and interviews, we test this hypothesis at first on two hours of teaching of 13 science teachers. 70% of the teachersí suppositions are predicted from the terms they use. The results are encouraging enough to use the method in two exploratory studies. From qualitative and quantitative analyses of use of keywords and clues, we come to grasp teachersí choices regarding the management of the class heterogeneity, and aspects of studentsí comprehension of the keywords. From these further results, we discuss our methodís potential to deal with longer observations of classroom interactions.enType de référence::Communications::Communication scientifique non publiée::Communication orale