Abstract: | In this report, we propose a model of spelling acquisition inspired by statistical learning and a frame-like model of orthographic representations: the fuzzy representation model. To provide an initial test of the model’s predictions for inconsistent words, 107 French-speaking children in Grades 1 to 3 were asked to spell words ending with a silent letter. Half of the words ended with the frequent silent t and half with the less frequent silent d. As predicted, children accurately spelled more t- than d-words. Most errors were omissions and substitutions of the silent letter. Consistent with statistical learning, there was some evidence that children used the preceding orthographic context when making substitution errors. The proportion of omissions, however, was not consistent with the statistical properties of French. These findings are discussed in light of the fuzzy representation model, whereby inconsistencies in words are more likely to be underspecified, if represented at all. |