Lecture | SMILE - Experimental Linguistics series
Formants are better predictors of vowel markedness than features
- Date
- Thursday 23 April 2026
- Time
- Serie
- SMILE - Experimental Linguistics series
- Address
-
Herta Mohr
Witte Singel 27A
2311 BG Leiden - Room
- 0.04
Abstract
Accounting for which sound patterns occur commonly within and across languages is one of the central goals of phonological theory. In constraint-based phonology, these tendencies are typically captured by a set of markedness constraints referring to properties of phonological representations. One of the remaining debates concerns how phonetically detailed these representations should be. Should markedness constraints refer to broad phonological features defined in the same way across all languages or to the specific phonetic realizations of sounds in particular languages? In this presentation, I focus on the phonological representations of vowels, asking whether language-specific formant-based representations are better predictors of vowel markedness than universal feature-based representations. As a testing ground, I use a subset of the multilingual speech corpus DoReCo including 30 languages. I use the Dispersion Theory (Liljencrants & Lindlom 1972) as a general theory of vowel markedness. This theory states that vowels that are more distinct should be more frequent and accounts for the high cross-linguistic frequency of cardinal vowels [a, i, u]. I show that a notion of dispersion based on formants provides a better account of vowel type and token frequency in the corpus than a notion of dispersion based on features. These results provide evidence for phonetically detailed representations in phonology.