Colloquium with Morgan Sonderegger, McGill University

Title:  New perspectives on speech variability from large-scale studies

Abstract:  I present two studies which aim to understand the structure and sources of variability in speech production, enabled by novel quantitative methods making it easier to study the same phenomenon across many languages (Study 1) or dialects (Study 2). I'll first discuss open-source tools for automatic analysis of speech which enable such large-scale studies by speeding up or replacing manual processing (PolyglotDB, Montreal Forced Aligner). The first study asks how much "the same" phonetic effects vary across 20 languages, and what their distributions can tell us, focusing on effects of vowel height and consonant voicing on F0 ("intrinsic F0 effects"). Consonant-induced effects are larger and more variable than vowel height effects across languages, suggesting a possible explanation for why only the former commonly leads to sound change. The second study, using data from the SPADE project (https://spade.glasgow.ac.uk/), examines variability in English sibilants in 5k speakers from 27 geo-social-ethnic regions to ask: is /s/ more variable than /sh/? On the surface a simple question, the results differ according to the level at which we consider variability.