Analytic study of the Tadoma method: further experiments with inexperienced observers

Results are presented for two studies of speechreading through the Tadoma method using normal subjects with simulated deafness and blindness. In the first study, subjects received training on the identification of consonant and vowel stimuli through Tadoma. In posttraining tests, an average score of 73% was obtained on a set of 24 consonants presented in CV syllables, and an average score of 82% was obtained on a set of 15 vowels and diphthongs presented in /g/-V-/d/ syllables. An analysis of the confusion matrices derived from the identification tests in terms of of various articulatory/phonological features indicated that the features place, voicing, frication, and round were well-perceived for consonants and round, tense, vertical lip separation, and low, for vowels. In the second study, subjects received training through Tadoma on comprehension of sentences composed of words from a 43-item vocabulary. In tests for which sentence stimuli were repeated until a correct response was obtained, the subjects identified an average of 30% of the words correctly on the first presentation and required an average of roughly four presentations for complete identification of the stimulus.