Abstract
This multidimensional scaling investigation seeks to determine the dimensions underlying the perception of diphthongs in American English and whether such dimensions are radically different from those found in studies utilizing monophthongal vowels. Dissimilarity data were obtained from subjects' paired comparison judgments of 15 English vowels [i e ε æ a o u i at a ju]. INDSCAL analysis of the data revealed that the perceptual structure of the vowels was best modeled in terms of four perceptual dimensions labeled front/back, high/low, low-back onset, and mid/nonmid. Regression analyses demonstrated that the perceptual structure of the vowels was most significantly related to the first two formants of the stimulus vowels. No dimensions relating to dynamic acoustic information (such as direction and/or extent of formant change over time) were found. Separate regression analyses demonstrated that these empirically derived dimensions were significantly related to the distinctive features relating to tongue position and, in particular, that the high/low dimension was better predicted by Ladefoged's multivalued Height feature than by Chomsky and Halle's binary High, Low, and Tense.
Get full access to this article
View all access options for this article.
