How do the discrete phonological units of the lexicon map onto continuous-time articulatory gestures and continuous-time auditory signals? The distinctive feature of [vm'ce] in syllable-coda position in English raises these questions with a vengeance. For minimal pairs like buzz/bus, clamber/clamper, tens/tense, etc, most measurable time intervals associated with the first syllable of these words are affected by the value of [voice]. Several of the rules in the traditional standard phonologies of English and many so-called 'phonetic implementation rules' serve to account for the various large and small temporal effects associated with the feature. We show that a very simple model for the English voicing contrast can be proposed that may account for these effects only if this phonological feature is phonetically defined as a velocity perturbation of a periodic dynamical system for English syllables. We summarize some evidence for the generalization that localized speaking-rate changes characterize a change in voicing. Then we suggest a general mathematical form for this dynamic effect that requires only a few parameters. This model implements the voicing feature as a perturbing forcing function for an underlying syllable oscillator.