Speaker: Amy Li (MIT)
Title: Inventory crowding and within-context variability of speech sounds
Time: Monday, November 9, 5pm-6:30pm
Location: 32-D831
Abstract: Many authors have speculated that more crowded inventories tolerate less variability (Martinet, 1952, p. 4; Flege, 1989, and references in pp. 127-8; Cortés Kandler, 2026, and references therein; Zhang & Chodroff, 2026, and references therein). For example, speakers of a language that contrasts /i, a, u/ may vary the exact realization of each of their vowels more than speakers of a language with /i, e, a, o, u/. However, no mechanism explaining the possible effect has been suggested, and empirical evidence is mixed. In response, I propose the “probabilistic dispersion theory of contrast” (augmenting the “dispersion theory of contrast”, Flemming, 2002), which optimizes inventories of speech sounds to balance the probability of misperception with the difficulty associated with reducing variability. The central problem for crowded inventories is high probability of confusion. This model claims that tolerance for confusion is upper bounded in all languages and predicts that languages that permit maximal confusion exhibit the hypothesized relationship. Three new case studies comparing within-context vowel variability across localized differences in inventory crowding find some evidence for more crowded inventories having less variability, supporting the proposed theory.