Interpreting vowel “quality”: The dimension of rounding

According to one view, speech perception depends crucially on the articulatory interpretation of the acoustic signal, since this would explain the identification of the acoustically different initial portions of two formant pseudospeech patterns such as, e.g., /di/ and /du/: Listeners “know” that their onsets are produced with very similar tongue gestures. Is this relation between perception and articulation versus acoustic signal typical, or is it instead restricted to certain phonetic units? For vowels it appears that different combinations of tongue and lip articulations yield very similar acoustic patterns and auditory qualities. The 18 “cardinal vowels” of Daniel Jones were presented to several groups of listeners asked to decide which had been produced with lip rounding. Error rates varied significantly over the vowel space, and showed a bias toward reporting front vowels as unrounded and back ones as rounded, a tendency that presumably reflects the relative frequencies across languages of front unrounded, back rounded, front rounded, and back unrounded vowels. Thus, if speech perception is crucially dependent on an articulatory interpretation of an acoustic signal, that interpretation need not be correct. [Work supported by NIH Grant HD-01994 to Haskins Laboratories.]