Three experiments on beginning readers′ ability to decode vowel graphemes are presented and are used to test an interactive analogy model of reading development. According to this model, children come to school with phonological knowledge about onsets and rimes, and this phonological knowledge plays an important role in setting up orthographic recognition units for words when they begin learning to read. So these orthographic recognition units are linked to phonological segments of the words as well as to their corresponding entries in the mental lexicon. One way of testing this model is to study vowel decoding, as according to the model pronunciations for vowel graphemes should originally be coded as part of the rime. For example, the pronunciation of the vowel grapheme u in the word bug should be initially coded as the rime ug. The nature of this coding can be studied by looking at transfer. Children who have set up an orthographic recognition unit for bug should show transfer to new words that share entire rimes with bug (e.g., rug), but not to words that share vowel graphemes only (e.g., cup). Three experiments examined such contrasts with progressively more complex single-syllable words and progressively older readers. Experiment 1 found that beginning readers only transferred pronunciations corresponding to rimes in words (e.g., bug-rug). Experiment 2 found that as reading progressed, pronunciations for vowel graphemes and onset-vowel units were also transferred (e.g., beak-heap, beak-bean). Experiment 3 supported this finding, but showed that vowel grapheme transfer was restricted to vowel digraphs for 7-year-old readers.