Magnetic resonance imaging (MRI) is well suited for studying the three-dimensional (3-D) shape of the vocal tract and tongue. However, when MRI is used for speech production measurements, it does not measure normal speech, as the subject lies in supine position and often has to sustain the articulation artificially due to very long acquisition times. In this chapter, MRI is used to estimate the effects of the sustained articulation. Coarticulation in real-time and sustained articulations are compared in particular. The subject was further positioned facing downwards during one part of the acquisition in order to evaluate the effect of gravitation. Finally, a 3-D tongue model, derived from sustained articulations was tested regarding its ability to replicate real-time midsagittal tongue contours.