An experiment was carried out to compare the time course of the acquisition of two basic spelling mechanisms in Spanish, a shallow system, and French, a deep system. The first was lexical. It relies on the orthographic lexicon, a hypothetical structure containing the orthographic representations of words accessible for word spelling. To evaluate its contribution the participants were asked to spell words of high and low frequency containing phonemes which can take different graphemic values. The second mechanism relies on sub-lexical processes. Its contribution was evaluated asking the participants to spell words containing consistent phoneme-to-grapheme translation pairs which were identical in Spanish and French. Two contrasting predictions were considered, one derived from the Orthographic Depth Hypothesis (ODH, Frost, 2005) and the other from the Self-Teaching Hypothesis (STH, Share, 2004). According to the ODH, the orthographic lexicon should develop more rapidly in French than in Spanish because Spanish spellers can rely on phoneme-to-grapheme translation mechanisms to spell most words, meaning that they do not need to resort to the orthographic lexicon. In contrast, the STH suggests that effective identification of a word gradually generates its orthographic representation. The results revealed that both spelling mechanisms develop far faster in Spanish than in French. The fact that word frequency effects appeared earlier in Spanish than in French, indicating that the orthographic lexicon incorporates words more rapidly in a shallow than in a deep system, is clearly incompatible with the ODH and easier to handle in the context of the STH.