Vowel intelligibility is important in communication through lipreading. We studied articulation as one source of vowel confusion. An adult female was filmed frontally at 18 fps as she spoke a random list of ten vowels (/i, ɪ, ɛ, æ, a, ɔ, ᴜ, ᴜ, ʌ, ɝ/) and six diphthongs (/aɪ, eɪ, ɔɪ, aᴜ, oᴜ, jᴜ/) in /b/‐V‐/b/ context (five tokens each). This silent film was shown to twenty normally hearing adults for visual vowel identification. Confusion matrices indicated distinction between rounded (back) vowels, spread (front) vowels, and most diphthongs. High intelligibilities for /i, a, ɔɪ, aᴜ, oᴜ, jᴜ/ and low intelligibilities for /ɪ, ɛ, æ, ɝ/ were obtained, as in several previous studies. Not all tokens of a given vowel were equally intelligible. To examine the relation between articulation and vowel intelligibility, the film images were studied frame‐by‐frame. Measurements were made of frontal articulatory parameters presumed to be important for visual identification of vowels (vertical and horizontal mouth opening, lip thicknesses, teeth visibility, and jaw displacement). Data suggest that highly intelligible tokens are characterized by extreme articulatory changes and quasi steady‐state segments. Tokens which evoke similar response distributions (regardless of intended vowel) exhibited similar patterns of articulatory change. The results suggest articulatory rules for optimizing communication of vowels through lipreading. [Work supported by NINCDS grant NS03856.]