| Volumetric MRI data has been collected during the sustained production of American English and Japanese vowels. Six native speakers (5 male, 1 female) produced nine steady-state American English vowels, and six native speakers (5 male, 1 female) produced the Japanese five vowel inventory. New methods have been developed for identifying the vocal tract airway within such data, resulting in a characterization of cross-sectional area that preserves shape information. Results have been tested through comparison of transfer function spectra derived from these characterizations with those obtained from recorded productions from the same speakers.;In addition to reporting standard area function results, a novel form of parameterization is described that can be used to estimate the vocal tract shape associated with a target vowel from just four weights. Because it provides a mapping between some offset along the tract and the corresponding cross-sectional shape, this 'shape-based' approach can be viewed as a superset of existing area function predictors, since area is readily obtained from polygonal shape by triangulation methods. |