The analysis of the results of the French national evaluation in 1989 has been the starting-point of the observations we present in this paper. A leading idea in the questionnaires of the initial evaluation was to design easy questions in various fields of elementary mathematics. In a first part (S1) we recall the major goals of that evaluation and we point out the most important questions which arise from it. In order to study its efficiency, we followed the evolution in mathematics of more than 500 students (age about 11) belonging to 22 different classes during the first year in the secondary school. The initial analysis of individual results (S2.1.1) points out two kinds of errors: lacks in knowledge and failures. The initial situations of the 22 classes are quite different and the evolutions do not appear as linked to the homogeneity of the classes. They are connected with the more or less successful nature of the  »pair teacher-class »(S2.1.3). Among the mathematical fields, we observe contrasted evolutions: achievement already at the beginning of the year in the simplest numerical treatments, improvement in the more complex numerical treatments (with decimals) and the geometrical drawing, almost stationnary process in the learning of problem solving and the use of geometrical terminology, drop of knowledges about areas and perimeters (S2.1.4). The predictive power of the initial evaluation is quite impressive by comparison with the simplicity of the submitted tasks (S2.2). We analyse also (S3) the way the teachers see the organisation of their classroom instruction in geometry. During the year, we collected a corpus consisting in interview with 9 teachers, their propositions for evaluation and their choices of documents to use in the classroom. Concerning the choices of the content topics to be taught, the teachers are quite unanimous; but the analysis of the already mentionned corpus shows great differences in the way the teachers take into account the variety in the levels of presentation of the informations at the entrance and at the exit in a mathematical treatment. Some teachers ignore this variety, some others use it systematically as shown in the presentation of the Van Hiele’s levels.