Tongue and vocal tract models that are derived from physiological data are usually based on two-dimensional data, such as lateral X-rays or tracking of midsagittal pellets. Models based on the acoustic wave are also two-dimensional in that they typically assume a uniform cross-sectional shape for the tongue and vocal tract. Cross-sectional tongue shape is not uniform, which adds tremendous complexity to tongue behavior. This paper offers three propositions essential for a three-dimensional model of the tongue: First, extrapolations from two-dimensional lengthwise tongue measurements are not an accurate representation of three-dimensional behaviors. Second, cross-sectional tongue shape can differ considerably at different positions along its length. Third, stabilization of the tongue tip against the palate, as occurs during alveolar consonants, facilitates production of tongue shapes that are different from those produced by the free-standing tongue, as occurs during some vowels. Both anterior and posterior tongue shapes are modified by palatal stabilization, but in different ways. Therefore, a model of tongue shape is proposed, which divides the tongue into “functional segments” in the sagittal and coronal planes, and explains three-dimensional tongue shapes and movements based on ultrasound, MRI, electropalatography, and X-ray microbeam measurements.