Generalizability of Current Models of Vowel Duration

Abstract

Abstract This paper assesses the generalizability and scope of the three main models of vowel duration: the incompressibility, multiplicative, and additive models. Combinations of vowel tensity with speaking rate, postvocalic consonantal voicing with rate, and tensity with voicing were studied in a database consisting of vowels in a CVC context in sentence frames. Combinations of four pairs of long and short vowels in four different postvocalic consonantal conditions were spoken by 7 speakers of American English across a wide variation in rate. Acoustic analysis of vowel durations revealed only partial support for each of the three models, as none of them adequately captured the combined influences of these three factors. Further, there were observed instances of combinations of effects that were not predicted by any of the models, even instances that were the exact opposite of predictions of the models. A general approach of representing and studying duration behavior is suggested by using simple regression analysis and making slope comparisons of the effects of various factors. This approach provides a framework for describing all of the outcomes in the study, representing all of the three existing models of duration and accounting for outcomes not predicted by any of the models. Within this approach, attaching special importance to interpretations of incompressibility, constant ratios and facilitation is shown to be unnecessary.