This work researches regularities in ordinary teachers’ practices in mathematics classes. It focuses on the correction of exercises involving the inverse function, which represents a didactical obstacle. The transition between the « Terminale » and the first year at the IUT is the selected environment for this study. We state that the notion of inverse function has been a didactical obstacle in France since the curricula reform in 2000 and that we are in the presence of an institutional didactical gap. A didactical de-transposition would certainly be appropriate in this instance. The methodology used is not to be found neither in Education Sciences nor in mathematics didactics publications. It combines a quantitative study (questionnaire) and a qualitative study (class observations and teacher interviews) by searching for variable networks and consequently by the selection of certain individuals. This has been made possible because of new developments in the CHIC software. Teachers’ stated practices are studied in order to understand their personal relationship to the inverse function and to exercise corrections. A model of correction episodes with three variables helped to analyse the teachers’ actual practices: the teacher’s accompanying speech, the knowledge status and the written solutions. The regularities found (related to the networks) correspond to the teacher’s action dimensions. The results show in the observed teachers’ manner of correcting these exercises (involving a notion which is an obstacle) that certain teachers bring to the class the missing content; others, getting around the obstacle, move the difficulty to another mathematical property.