As concerns the long-lasting kinematic vs. static controversy in vowel perception, our claims, based on recent experiments on glides and vowel reduction (carried out by Cathiard et al. and Loevenbruck et al.), will be that kinematics is needed when signals are (i) undersampled or (ii) noisy. Point-light talking faces or sinewave speech typically belong to the first case. Undershot vowels add to the case of undersampling, while cue-tracking in speech in noise is a typical case of CASA, be it a general-purpose or a specific device for speech and/or music.
    Our perceptual results with natural stimuli and with a non-linear biomechanical model of control will be discussed in the framework of a double-component account of vowel gestures. Recalling that the concept can probably be traced back to Sweet, we reset this 2-COMP-VOWEL in a theory formulated within a neural model of goal-directed actions (developped by Jeannerod & Boussaoud). Basically, speech sound mouthing is considered as an outgrowth, an evolutionary *homologue* - say an *exaptation*, via the vocal self monitoring system - departing from the stem function of mouth grasping, itself related to hand capture in primates. The carrier component of vowel mouthing is named *placing*, homologous to the reaching-convey component in mouth "handling". The second component is called *shaping*, homologous to grip formation (preshaping). When both components are fairly synchronous, "steady-state" vowels (typically French and German long vowels) are produced. Glides occur when *shaping* is relaxed asynchronously respective to *placing* changes. Such an asynchrony gives rise to glide-epenthesis in the transition between vowels (French, German), or to diphthongal on/off-glides (in English, Swedish, etc.).
    Of course, like other epenthetic phenomena, glides can be recovered as true phonological controls. But even in this case there is no evidence that motion in the vocalic gesture could offer more than a mere processing benefit for undersampled signals, through a neural *shape-from-motion* mechanism, without any kinematic status of the phonetic long-term representation.