Abstract
Instead of acquiring speech data in diving experiments involving many difficulties, synthetic vowels under 15 combinations of sound velocity c and density ρof the breathing gas were prepared for a listening test of vowel identification. Error rates and confusion matrices of the listeners' responses are discussed with respect to c and p. The main points of the results are as follows: 1) Although, the intelligibility of speech in a helium environment is greatly improved by restoring the distortion by c (an upward shift of formants proportional to c), the additional restoration of the distortion by p (non-linearity of the upward shift) becomes indispensable at a depth of 500m more. 2) Uni-directional confusion of the vowels pronounced in a helium atmosphere is also explained by the non linear upward shift of the formants.