In French, the syllable is an early and robust prelexical and segmental unit during reading acquisition. But is syllable segmentation possible when no reliable statistical and distributional information is available? We proposed that the sonority might be another reliable source for syllable segmentation. We studied whether dyslexic children compared with chronological-age matched and reading-level matched controls are sensitive to universal sonority-related markedness within the syllable boundary. We used an adapted version of the illusory conjunction paradigm. The statistical and distributional properties were null or quasinull around and within the syllable boundary, while sonorityrelated markedness was manipulated along a continuum from legal, unmarked clusters to illegal, marked ones. Our results show that syllable segmentation strategies primarily depend on sonority-related markedness. As markedness decreases, children better detect the syllable boundaries, while as markedness increases, children promote syllable onsets. This is true with dyslexic children and controls. While this universal phonological sensitivity does not depend on reading level, syllable segmentation abilities develop with reading experience.