A model of syllable duration is outlined. It takes into account segmental structures, lexical categories, stress and word accent assignment and positional effects within a syntactic frame. The model is used for predicting temporal structures in text reading to be compared with specific utterances. Prediction errors are small enough to sort out speaker specific variations in local emphasis-deemphasis and speech rate. The concept of speech tempo in relation to speech rate is discussed. Other applications of the model are in studies of speaking style and language contrasting patterns, and as a tool for quantifying syllabic stress.