This article reports three experiments investigating the use of analogies in spelling acquisition. French children spelled pseudowords to dictation, some of which were phonological neighbours of words with uncommon endings (e.g., /daby/derived from "début" /deby/). A more frequent use of these uncommon graphemes in neighbour pseudowords than in control pseudowords was taken as evidence for spelling by analogy. In Experiment 1, an analogy effect was observed in Grades 3 to 5. Younger children did not use analogies, but they were also unable to spell most reference words. Experiments 2 and 3 introduced a reference word learning phase prior to the pseudoword dictation task. An analogy effect was found in second graders (Experiment 2) and even in first graders (Experiment 3) when children knew how to spell most reference words. Comparable use of analogies was observed in children with comparable lexical knowledge independently of their grade level or alphabetic skills. The results suggest that children establish specific orthographic knowledge from the beginning of literacy acquisition and use this knowledge to generate new word spellings as soon as it is available.