Written French is known for causing much trouble, even to native speakers. As students start university to gain professional and academic knowledge, more and more have to attend Written French classes to improve their spelling skills. Most of the written French inflectional morphology is silent (Dubois, 1965 ; Catach, 1986), making mistakes difficult to get rid of. In order to correct those young adults’ grammatical spelling skills, teachers won’t teach everything again, but rather need to know what went wrong. Various types of knowledge lead to a successful agreement. One can access their declarative knowledge (Anderson, 1983) to apply the rule, implying one understands it and its metalinguistic words. One can also trigger procedural knowledge which is more automatic but which sometimes lead to mistakes. Researchs in psychology show that such mistake can be explained by the attraction of another word to make an agreement (proximity agreement : Largy, 2001; Negro, Chanquoy, 2000) or by the frequency of a word with of without the ending flexion, or even by the existence of a homophonic word belonging to a different grammatical category. Those mistakes are considered as « experts’ mistakes » as they are triggered to avoid a working memory overload, not because they reflect a lack of linguistic knowledge (Fayol, Pacton, 2005; Baddeley 1986). This paper aims at presenting an evaluating tool able to spot the reasons why French students make agreement mistakes. Those reasons range from serious lack of acquisition to punctual mistakes caused by specific linguistic features of the word or specific attentional features of the writing situation. Our study case focuses on two classes of 20 students each. One class is taught Agronomic Engineering while the other French Modern Literature, 1st year of Bachelor degree. Their evaluation shows that it would be more relevant to establish a scale of expertise according to the type of errors they make, rather than according to their level of education and even to the type of curriculum they experience. The results also confirm that it is possible for one to make a correct agreement without being able to verbalise the rule, and vice-versa (Gaux, Gombert, 1999), a fact that should be considered when teaching to this specific public. Anderson, J. R., 1983, The architecture of cognition, Cambridge: Harvard University Press. Baddeley A. D. — (1986) Working memory, Londres, Oxford University Press. Catach N., 1986, L'orthographe française, Paris, Nathan. Dubois J., 1965, Grammaire structurale du Français : Nom et pronom, Paris : Larousse. Fayol M., Pacton S., 2005, « L’accord du participe passé : entre compétition de procédures et récupération en mémoire », Langue française, vol. 151, pp. 59-73. Gaux C., Gombert J.-E., 1999, « La conscience syntaxique chez les préadolescents : question de méthodes », L'année psychologique, vol. 99, n°1, p. 45-74. Largy P., 2001, « La révision des accords nominal et verbal chez l’enfant », L’année psychologie, vol. 101, n°2, p.221-245. Negro I., Chanquoy L., 2000, « Étude des erreurs d'accord sujet-verbe au présent et à l’imparfait. Analyse comparative entre des collégiens et des adultes », L’année psychologique, vol. 100, n°2, p. 209-240.