Tactile communication of speech: II. Comparison of two spectral displays in a vowel discrimination task

Vowel discrimination experiments were performed comparing two tactile, Optacon-based, spectral displays: a frequency-amplitude (FA) display and a time-swept (TS) display. The set of vowel pairs tested consisted of all pairs from the set of ten American nondipthongized vowels. One set of materials consisted of natural /b/-V-/t/'s (with each vowel represented by 12 utterances generated by four speakers producing each vowel three times). A second set consisted of synthetic vowels (with each vowel represented by a single waveform). On the average, th score obtained with the second set of materials was substantially better than with the first (92% vs 79%); the score obtained with the TS display was slightly better than with the FA display (87% vs 83%); the feature best differentiated was tenseness, followed by the features high and low, then by round and back, and finally by retroflexion; and the correlation between discrimination performance and the physical parameters duration, amplitude, F1, and F2/F1 (taken singly) was relatively weak. In general, the results of this study, together with those of other studies previously reported, suggest that the two displays studied are far from optimum.