This study investigates the relative timing of vocalic gestures as a function of syllable position, manifested in the traditional distinction between hiatus and diphthong. The hiatus-diphthong distinction has been treated as one in syllable affiliation, therefore differences in their production can be interpreted as phonetic correlates of the syllable. I compare the acoustic realization of Romanian diphthongs [jV] and similar hiatus sequences [iV]. The latter have variable pronunciation, occasionally realized as diphthongs. A contrast is maintained between them, although individual speakers show tendencies toward a merger. Differences between diphthongs and hiatus sequences are found in: the vocalic portion duration, the F2 value at the onset of the vocalic portion, the duration of high energy frication at the release of a preceding [p]. I propose that all of these differences result from differences in articulatory timing between [i]/[j] and V. Inter-gestural timing in [jV] is tightly controlled, producing a short vocalic gesture for [j]. In a hiatus sequence [iV] timing is loose, allowing for variation. An occasional early onset of the V gesture can produce a glide percept. The results are interpreted in the framework of Browman and Goldstein's (2000) Bonding Strength Model.