Theoretical models of airway constriction for sibilants make assumptions which may or may not be justified and which are difficult to extend to pathological cases such as malocclusion. In an attempt to determine the exact shape of the constriction, fast‐setting dental impression material was placed in the mouths of normal English speaking adults, who then attained and held articulatory positions for /s/ and /æ/. The impressions thus formed were then sectioned to obtain successive cross‐sectional areas. There is wide variability among normal speakers in the observed cross‐sectional area patterns, and an attempt will be made to relate these differences to variations in air flow and pressure and the resulting acoustic signal. [Supported by NIH Grant No. 1 R01 DE04267.]