Problem-solving is a central component of mathematical activity but frequently leads to failure, regardless of school level or country. In France, recent national assessments reveal that at Grade 4, it is the domain that induces the largest performance gap based on families' socio-economic status. Despite forty years of research, problem-solving remains a persistent source of difficulty and a major research area for both researchers and practitioners.

Previous work has progressively identified the comprehension of problem statements as the main source of difficulty, and thus a crucial lever for intervention to improve student performance in problem-solving. Until now, pedagogical interventions have addressed this aspect indirectly or in a very limited manner.

In our work, we sought to act directly and globally on the comprehension of problem statements and tested the hypothesis that prioritizing comprehension activity before introducing numerical data should improve student performance. To achieve this, we implemented a protocol over three successive years and three grade levels (Grade 3, Grade 4, and Grade 5). This involved two groups: an experimental group and an active control group. Both groups used the same problem bank, following a classic learning method with a pre-test, training, and post-test. Teachers in the experimental group received specific training on our novel approach and were supported throughout the intervention period. This group implemented the protocol with amplified comprehension strategies, while teachers in the active control group maintained their conventional learning approach.

The implemented intervention shows results at all levels, and more particularly among the youngest students and those whose initial comprehension level was the weakest.

The protocol tested a pedagogical innovation grounded in psychological research data, while simultaneously developing a collaborative and respectful approach to teachers' professional expertise. This participatory research involved 103 teachers, 60 of whom were part of the experimental protocol, and engaged several thousand students over the three years. This new approach, though initially challenging for both teachers and students, was co-constructed through close collaboration with the educators. Structured visits, exchanges, and regulation periods, combined with specific measurement tools (grids), enabled the identification of effective professional gestures and classroom practices and ensured adherence to the protocol. Over these three years of implementing this collaborative and participatory methodology, the identified needs and effective practices naturally led to the creation of training programs for teachers and other educational trainers, facilitating the dissemination of this approach.

In the future, the sustainability of the approach in trained classes and its long-term effects on students remain to be studied.

Keywords: Word-problem solving, intervention, comprehension, calculation, learning, primary school.