Acoustic cues in the identification of vowel sequences

Abstract

We have observed that at 30‐ and 60‐ms durations some permuted orders of 4 synthetic vowels /i, æ, a, u/ were much more identifiable than others using a single sequence method [Kerivan, Alfonso, and Bornstein, J. Acoust. Soc. Am. Suppl. 1 65, S113(A) (1979)]. Also, significant improvements in identification of difficult sequences occurred when silent gaps were placed between selected vowels resulting in a rhythmic grouping of the sequence. These findings have been correlated to the various acoustic parameters used to synthesize the vowels. The results of these multiple correlations show that the identification of vowels in sequence follows the laws of pitch continuity established by numerous investigators using tonal sequences. The more monotonic pitch continuity displayed by adjacent first formants the better is sequence identification. A similar but weaker correlation was found between sequence identification and pitch proximity of fundamental frequencies. An interaction was also found between pitch and loudness of adjacent fundamental and formant frequencies. The relative intensity differences for all first formants were within 3 dB; however, those for fundamental frequencies were as disparate as 11 dB. It appears that the ear primarily tracks the first formant of each vowel complex in determining vowel identity in the sequence. [Work supported by an NINCDS NIH grant awarded to Haskins Laboratories and the University of Connecticut Research Foundation.]