Abstract
During the past 25 years, there have been attempts to develop speechreading aids that pro vide information about nonvisible articulatory gestures. Five experiments were conducted to de ter mine whether increasing the amount of visible articulatory information can affect speech recognition, and whether such aids might be beneficial. The experiments involved videofluoroscopy, which allows placement of the tongue body, lips, teeth, mandible, and often velum to be observed in a naturally-occurring sequence and time course during speech. Subjects were asked to speechread videofluoroscopic and video images. The results suggest that making visible typically obscured supralaryngeal articulator activity by means of videofluoroscopy does not enhance speechreading performance. Normal hearing and cochlear implant users always performed better with the video than videofluoroscopy medium. Subjects performed as well when the tongue was not visible in the videofluoroscopic records as when it was visible. Training did not affect subjects' ability to recognize speech presented with videofluoroscopy. The results pose a challenge to direct-realist theory and motor theory. Both theories posit that perceivers must recover information about articulatory gestures in order to ab stract linguistic information from the speech signal.

Back to NCVS-sponsored Research Articles