Experimental Insights into the Prosodic Structure of Ersu (Tibeto-Burman)
Abstract
This study experimentally examines the prosodic structure of Ersu (ISO-639 ers), an endangered Tibeto-Burman language, and empirically tests the hypothesis that Ersu’s prosodic system combines lexical tone in monosyllabic words with stress-like prominence in polysyllabic words (Chirkova et al. 2015). Ersu is proposed as a register tone language with two registers: High (H) and Low (L). In polysyllabic words, the initial syllable retains its tone, while non-initial syllables adopt the tone of the initial syllable. Among the three tone patterns found in di- and tri-syllabic words—H(H)H, L(L)L, and H(H)L—the first two are regular, while the third is exceptional. The exceptional patterns (HL, HHL) are believed to exhibit stress-like features, with the final Low-tone syllable perceived as less prominent, having reduced intensity and shorter duration. This study aims to deepen our understanding of the Ersu prosodic system by empirically testing previous claims using a small corpus of di- and tri-syllabic words in two contexts: citation (phrase-final) and phrasal (embedded within a phrase). It examines three acoustic correlates of prominence—duration, intensity, and fundamental frequency (f0)—to distinguish word-level from phrase-level stress and to investigate their interaction with lexical tone. The analysis draws on field recordings from eight Ersu speakers (3 female, 5 male). Word-list items were transcribed using Praat TextGrids (Boersma & Weenink 2020), and 8,085 acoustic tokens were extracted via a Praat script. Duration and mean intensity were measured once per vowel, while f0 was measured three times per vowel (vowel onset, midpoint, and offset) to capture both pitch height and contour. Due to gender-based differences in mean f0 and f0 range, male and female speakers were analyzed separately. Statistical analysis was conducted using linear mixed-effects models in R (lme4 and lmerTest: Bates et al. 2015; Kuznetsova et al. 2017), with post-hoc pairwise comparisons via the emmeans package (Lenth 2016). Separate models were run for each acoustic parameter. Our results only partially support the original claims. For the HL tone pattern, the first syllable shows significantly greater intensity only in the citation context (estimate = 2.016, t = 2.503, p = 0.0124), while durational differences between syllables are not statistically significant in either context (citation: estimate = -0.003, t = -0.450, p = 0.6528; phrase: estimate = 0.007, t = 1.206, p = 0.2282). For the HH and LL patterns, f0, intensity, and duration do not differ significantly between syllables in either context. At the trisyllabic level, the HHL tone pattern resembles the LLL pattern in exhibiting peak intensity on the initial syllable across all contexts. In terms of duration, all three tone patterns show consistently longer duration on the final syllable in every context examined. This presentation further explores the implications of these findings for the prosodic organization of Ersu and its closely related language Lizu, for which comparable experimental data are available (Chirkova et al. 2023). More broadly, it contributes to the typology of prosodic systems in under-documented Tibeto-Burman languages of Southwest China (e.g. Evans 2008, 2018).
| Origin | Files produced by the author(s) |
|---|---|
| Licence |