Previous studies conducted in French demonstrated that syllables are used as prelexical and segmental units in silent reading in skilled readers. However, most of these studies focused on the importance of statistical and distributional properties to account for the syllable effects. Our aim was to explore how skilled readers segment printed (pseudo)words when no reliable statistical cues were available around – and within – the syllable boundary. We were interested in how the sonority – a universal feature-like phonological element – might be a reliable source for syllable segmentation. We tested 164 native French-speaking adults with no bilingualism. We used pseudowords in two versions of the classical illusory conjunction paradigm: one classical presentation and one consisted of playing a white noise while the pseudoword displayed. We selected disyllabic pseudowords with bigrams, trigrams and syllables with (quasi) null frequencies in initial positions (first three letters including the syllable boundary). We manipulated five sonority profiles (i.e., high-fall, low-fall, plateau, low-rise, high-rise) within the syllable boundaries along a continuum from legal to illegal clusters. Our results show that French adults were sensitive to the optimality of the sonority profiles to locate the syllable boundary in both experiments; as illegality decreased, French adults better detected the syllable boundaries. We also found that the (quasi) absence of statistical and distributional properties around and within the syllable boundary did not prevent from segmenting syllabically. Moreover, the white noise which aimed at disrupting the possible retrieval of the acoustic-phonetic properties of the sonority did not influence the sonority-based segmentation strategies. Our results also suggest that segmentation strategies of printed stimuli do not only depend on a strict statistical and distributional analysis but also on a universal phonological sensitivity based on sonority.