Statistical learning has been proposed to underlie the developmental transition during infancy from allophonic to phonemic speech sound perception. Based on this, it can be hypothesized that in dyslexic individuals, core phonemic representation deficits arise from reduced sensitivity to the statistical distribution of sounds. This study aims to investigate (a) whether statistical learning contributes to the construction of phonemic representations in typical readers, and (b) whether deficits in statistical learning underlie dyslexia. Fifty-eight children performed an identification task of a non-native phonetic contrast, before and after exposure to the sounds of the continuum. Our results suggest that the statistical distribution of the presented sounds implicitly enhanced the formation of phonemic representations and that dyslexic readers make less use of the statistical cues embedded in oral language, resulting in less distinct phonemic categories and thus a higher risk for failing to establish robust connections between these and written language.