L’objectif de cette étude est de mieux comprendre comment l’enfant construit progressivement l’inconsistance du système lexical français. Une étude des erreurs commises par des élèves du CE1 à la 3^e dans une dictée de mots isolés a été réalisée. Les mots présentés variaient en longueur syllabique, consistance orthographique et fréquence lexicale. Pour chaque catégorie de mots, les trajectoires développementales des erreurs phonologiquement plausibles (PP) et non phonologiquement plausibles (NPP) ont été analysées. Cette étude montre que (1) même si les taux d’erreurs diminuent progressivement, ils restent globalement élevés du CE1 au CM1, ainsi qu’au collège pour les erreurs PP ; (2) les différents types d’erreurs ne suivent pas tous la même évolution ; (3) leur occurrence varie en fonction du niveau scolaire et des caractéristiques des mots. Cette étude invite les chercheurs à mieux prendre en compte ces facteurs psycholinguistiques dans les modèles et/ou études impliquant un traitement orthographique.

The aim of this study was to better understand how the child gradually acquires the orthographic inconsistency in French. Children's spelling mistakes in Grades 2 to 9 were investigated from a word dictation task. Words varied in syllabic length (1, 2, or 3 syllables), spelling consistency (consistent vs inconsistent) and lexical frequency (rare vs frequent). Children were expected to write five words for each experimental category of words (e.g., consistent and frequent word of one syllable), that is 60 words. For each category of words, phonologically plausible (PP) and not phonologically plausible (NPP) errors were analyzed from a developmental point of view. According to Martinet and Valdois (1999), PP errors were distinguished in strict phonologically plausible errors based on French phonographemic correspondences (PPS) and phonologically plausible errors using spellings which transgress the French phonographemic correspondence rules (invented spellings labeled PPL). The results show that (1) even if error rates gradually decreased, they remain overall high from Grades 2 to 9 for PP errors; (2) PP vs NPP errors did not follow the same pattern: NPP errors regularly decreased while PP errors were steady between 2nd and 4th grades before they dropped from Grade 5 to Grade 9; (3) their occurrence varied according to grade level and word characteristics with more NPP errors for younger children and more PP errors for undergraduates; (4) strict phonologically plausible error (PPS) rates were always higher than invented spellings that disappeared beyond grade 5, while PPS errors persisted until grade 9 (24%) for inconsistent and rare words. These results agreed with an implicit statistical assimilation of language graphotactic regularities, in addition to the formal teaching of spelling. This study shows that attention should be paid to these psycholinguistic variables in models and/or studies involving orthographic processing.