Male vowel-consonant-vowel formant transitions are analyzed using cepstral techniques. The first three formants are detected and represented as trajectories in the F1-F2 plane. Distinctive formant trajectories are clearly revealed, exhibiting specific trajectory angles at the onset during the initial vowel, during the consonant production, and at the offset during the target vowel. Formant stabilities are also studied. These trajectories are interpreted using a "distinctive region" model (Mrayati, Carre & Guerin, 1988). The acoustic tube is divided longitudinally into regions, the boundaries of which are deduced from the zero-crossings of the sensitivity functions of the uniform tube. The lengths of the regions are equal to a fixed percentage of the "effective" length of the tube. Thus, the command strategy for controlling the model is simple: transversal at specific places. Formant trajectories are generally monotonic for area variations between 1 and 16 cm2, and pseudo-orthogonal behaviour is obtained, i.e., for each of the regions, an increase or a decrease in its cross-sectional area leads to well-defined formant transitions for the formants that are considered. V1CV2 trajectories predicted by the model correspond to those observed on natural speech. The timing control of the model is discussed. A simple synthesis-by-rule technique is deduced using the distinctive region model as a co-articulation model. Good synthesis of V1CV2 sequences is obtained from such a technique. Theoretical consequences on the speech production mechanism are discussed.