In current theories of word reading the structure and operations of the phonological buffer are quite underspecified. We investigated this issue by running a reading aloud experiment in Italian. We adopted a priming paradigm, with three-syllabic words as primes and targets and we jointly manipulated two effects ascribed to the stage of phonological and phonetic encoding, that is stress priming and syllable frequency. Target words varying for the frequency of their initial syllable were preceded by words congruent or incongruent for the stress pattern. The results showed an interaction between syllable frequency and stress prime, with the stress congruency effect larger for the targets with low-frequency first syllable. This result suggests that, in reading aloud, stress assignment and syllable computation have a tight time dynamics in the phonological output buffer, and that the process at the level of phonology-to-phonetic interface operates interactively.