The relationship between lower level linguistic components (phonetics/phonology) and higher level linguistic components (syntax, discourse) is studied from the experimental analysis of the temporal organization of the three extreme vowels [i], [a], and [u]. D. Klatt's (1976) model, based on American English, is examined in order to know how it could be adapted to Brazilian Portuguese. Following this comparison, rules are proposed to predict sound-segment duration of Brazilian Portuguese in continuous speech. Data for this investigation were obtained from one speaker from Rio de Janeiro who read a text for children containing over one thousand words.