The generation of prosody in text-to-speech synthesizers usually requires the production of an adequate fundamental frequency movement on the stressed syllables of the sentence. These pitch variations are supposed to correspond, among other factors, to some hierarchy which organize stress groups, and which can be equated to the syntactic structure of the sentence. With this hypothesis, the pitch changes located on the stressed syllables function as acoustic cues to pre-analyze (parse) the speech continuum as it is produced by the synthesizer. This paper deals with one part of this process: the automatic positioning of stress syllables from text in Italian, which is not graphically represented (except in final position), and for which no simple stress rule is known.