We report on developmental vowel dyslexia, a type of dyslexia that selectively affects the reading aloud of vowel letters. We identified this dyslexia in 55 Turkish-readers aged 9-10, and made an in-depth multiple-case analysis of the reading of 17 participants whose vowel dyslexia was relatively selective. These participants made significantly more vowel errors (vowel substitution, omission, migration, and addition) than age-matched controls, and significantly more errors in vowel letters than in consonants. Vowel harmony, a pivotal property of Turkish phonology, was intact and the majority of their vowel errors yielded harmonic responses. The transparent character of Turkish orthography indicates that vowel dyslexia is not related to ambiguity in vowel conversion. The dyslexia did not result from a deficit in the phonological-output stage, as the participants did not make vowel errors in nonword repetition or in repeating words they had read with a vowel error. The locus of the deficit was not in the orthographic-visual-analyzer either, as their same-different decision on words differing in vowels was intact, and so was their written-word comprehension. They made significantly more errors on nonwords than on words, indicating that their deficit was in vowel processing in the sublexical route. Given that their single-vowels conversion was intact, and that they showed an effect of the number of vowels, we conclude that their deficit is in a vowel-specific buffer in the sublexical route. They did not make vowel errors within suffixes, indicating that suffixes are converted as wholes in a separate sublexical sub-route. These results have theoretical implications for the dual-route model: they indicate that the sublexical route converts vowels and consonants separately, that the sublexical route includes a vowel buffer, and a separate morphological conversion route. The results also indicate that types of dyslexia can be detected in transparent languages given detailed error-analysis and dyslexia-relevant stimuli.