While speech production primarily conveys linguistic information via acoustics, it has been shown that speech motor control uses all available sensory inputs - auditory, somatosensory, tactile and even visual - to plan and control the evolution of the state of the orofacial system over time. It has also been suggested that somatosensory feedback could be major during on-line speech production, because of its latency, which is shorter (13 to 55ms in EEG and MEG studies) than the one of the auditory feedback (above 100ms in EEG studies). This study investigates the mechanisms underlying somatosensory based tongue motor control in speech production by combining EMG recordings in response to a tongue-stretch perturbation during vowel production and simulations of tongue movements with a 3D biomechanical model. An experimental study was designed to apply tongue-stretch perturbations with a robotic device during the steady-state production of French vowel /i/. EMG was recorded from the anterior part of the mouth floor using a uni-polar surface electrode. EMG signal from this site reflects in large part the activation of the anterior genioglossus muscle, which plays a major role in the control of tongue blade in front/high vowels. EMG data recorded in this task were compared to those recorded in a voluntary reaction task during the same steady-state vowel production and with a resting task. We found in all the three tasks an increase of muscle activation in response to tongue-stretch perturbation with a latency close to 50ms. The magnitude of this response was smaller in the resting task than in both speech production tasks, but it was not reliably influenced by the voluntary reaction. This suggests that the observed EMG response is driven by a reflex mechanism which magnitude is task-dependent. In a previous study, using Electromagnetic Articulography, we recorded tongue kinematics under the same conditions, which enabled us to associate a posteriori EMG activations and tongue movements. Finally, a follow-up study using the same perturbation paradigm during the production of French vowel /e/ with and without auditory masking strongly suggested that these EMG responses were primarily induced by somatosensory feedback. Our group has developed a 3D biomechanical model of the tongue of an adult male subject (called “Reference subject”). It includes an accurate account of the 3D anatomical implementation of the muscles, and of the mechanical properties of tongue tissue based on fresh cadaver’s tongue data. It also incorporates a 3D muscle model that generates stress and modulates muscle tissue’s elasticity in response to EMG-like activation. The model was evaluated and validated based on its capacity to faithfully reproduce 3D tongue postures measured during the production of French phonemes by the Reference subject. We propose here to use this biomechanical model to simulate, from various tongue postures, a step-like increase in the activation of the anterior-genioglossus, in order to study tissue time response. These two combined studies enabled us to disentangle in the timing of tongue responses to stretch perturbation the respective influences of neural feedback delays and time response of tongue tissue.