Three-dimensional MR images of the vocal tract of three adult males were obtained during steady-state production of the Japanese fricatives /s/, /sh/, and /shu/. MR images of dental crowns containing a small amount of water were obtained using a dental crown plate which was shaped so as to attach tightly to the subject's dental crown by thermoforming which contained a contrast medium for MR imaging. A computer algorithm based on gray level interpolation and threshold operation for air--tissue boundary tracing of arbitrary sections of the vocal tract taken from axial and coronal MR images were proposed. Images of arbitrary sections of the vocal tract were reconstructed from 3-D MR images using gray level interpolation based on the sampling theorem. Gray level thresholds were computed from the gray levels at points on the air--tissue border. The boundary tracing error of the dental crowns from the 3-D MR images was less than 5.7%. The 3-D shape and area function of the vocal tract during production of the Japanese fricatives /s/, /sh/, and /shu/ were obtained.