This study investigates how statistical orthographic knowledge—how frequently a phonological unit is written in a language—influences phonological awareness tasks. In Experiment 1, 144 children with varying reading levels and 61 undergraduate French-speaking students completed a syllable-counting task for consonant-vowel-consonant (CVC) and consonant-consonant-vowel-consonant (CCVC) acoustic monosyllabic pseudowords. The number of syllables of the pseudowords was ambiguous (one or two) since some of their ending rhymes could be frequently spelled with an “e” (e.g., [-uʃ]), whereas others are frequently spelled without an “e” (e.g., [-is]). The results showed that (1) pseudowords tended to be more frequently counted as one-syllable sequences when they had a high probability of not be spelled with an “e” rather than when they had a low probability of not be spelled with an “e, ” and that (2) this effect was more pronounced in literate adults as compared to children with lower reading levels. Experiment 2, involving prereaders, confirmed that children without formal literacy instruction showed little influence of orthographic information on syllable counts. These findings suggest that phonological responses are shaped by statistical relationships between phonological and orthographic representations, which emerge gradually through learning to read.