Vowel identification and goodness based on level of formant detail

Abstract

Neel (2004) asked how much time-varying formant detail is needed for vowel identification. In that study, starting with vowel recordings from one male speaker and one female speaker, multiple stimuli were synthesized for each vowel: 1-point (monophthongal with midpoint values), 2-point (linear from onset to offset), 3-point, 5-point, and 11-point. Results suggested that a 3-point model was optimal. The present study partially replicates Neel (2004) but draws from more robust phonetic sources (Jacewicz et al., 2011; Hillenbrand et al., 1995). Eight English monophthongs were chosen for synthesis. 1-, 2-, 3-, and 5-point stimuli were created as described above, and another 1-point stimulus was created with onset values rather than midpoint values. These 40 stimuli were used in two studies (n = 18 for each). First was a vowel identification task, where the ten choices were [hVd] words with the chosen vowels (heed, hid, etc.). Second was a goodness rating task with a 7-point Likert scale, where participants were played a vowel stimulus while being shown the orthographic [hVd] word containing that vowel. The results of neither study showed improvements beyond 2-point stimuli, in contrast to the previous finding of 3-point stimuli as optimal.