Assuming that a phonetic transcription of a speech event is to at least a first approximation its phonetic description, the segmental nature of the transcription may be understood to assert that, if two or more features shift their values “between” two segments, then these shifts occur simultaneously. Physical data indicate that articulatory shifts thought to occur in synchrony may be out of phase by amounts that are not negligible as compared with the durations ascribed to entire segments. Whether such phase variations are subject to the speaker's control in any direct way, or whether they reflect certain physiological constraints on the speech apparatus, is still to be settled to everyone's satisfaction. Moreover the same question is appropriately raised with respect to shifts that are apparently very precisely synchronized, i.e., involving tolerances of a few tens of milliseconds. An interesting case involves the precise timing relation between supraglottal and laryngeal gestures in stop‐vowel and vowel‐stop sequences. In the former this timing relation is variable, yielding distinct stop categories, and this variation is presumably under speaker control. In the latter there appears to be less variability in the timing of closure and extinction of glottal tone. Instead, there is significant variability in vowel duration, ascribable to properties of the following stop. The present paper presents data reporting on the durational relations among prevocalic stops, vowels, and post‐vocalic stops in one‐ and two‐syllable nonsense words. The utterances measured were produced by a single speaker under conditions designed to minimize variability, so that obtained measures of variation might serve as baseline values to be compared with segment duration data from less restricted speech samples.