Covariation of stop voice onset time across languages: Evidence for a universal constraint on phonetic realization

Abstract

Stop consonant voice onset time (VOT) was examined in a typological survey of over 100 languages. Within broadly defined laryngeal categories (long-lag, short-lag, and lead voicing), VOT means were found to vary extensively. Importantly, the means for members of the same laryngeal series did not vary independently but instead were highly correlated across languages. The strong linear relations identified here cannot be reduced to previously reported ordinal relations, and provide evidence for a uniformity constraint on phonetic realization: within a language, each laryngeal specification must be realized in approximately the same way across stops of different places of articulation.

Voice Onset Time (VOT) at 50: Theoretical and practical issues in measuring voicing distinctions

“…It seems that speakers adopt settings for the glottal gesture that are consistent across stops, resulting in the covariation. Further research has shown that this pattern extends to languages as a whole (Chodroff, Golden, & Wilson, in preparation)…”

The Pacific Expansion: Optimizing Phonetic Transcription of Archival Corpora

“…There is significant interest in adapting or tailoring forced alignment tools and techniques for use with data for low-resource languages, particularly material originating from archives and language documentation projects [13,14,15]. Phonetic annotation of such material allows greater inclusion of understudied languages in research on crosslinguistic phonetic patterns [16,17] and explorations of language-internal patterns of variation and change [18]. However, the limited size and coverage of many language documentation corpora, and level of associated linguistic description, pose challenges for the development of effective speech technologies.…”

Xie, Liu, & Jaeger (2020). Cross-talker generalization during foreign-accented speech perception

“…For example, exposure to repeated instances of a category with atypical cue realization (e.g., /b/s with atypically long or atypically variable voice onset times) can affect how listeners interpret that cue during test (e.g., Clayards et al, 2008;Kraljic & Samuel, 2006;Munson, 2011;Theodore & Monto, 2019). Evidence for the relevant distribution might also be inferred based on implicit knowledge listeners have about the correlational structure of cues across categories (see Chodroff, Golden & Wilson, 2019;Idemaru & Holt, 2011;Kleinschmidt & Jaeger, 2015, Part II). For example, talkers who produce long VOTs for /p/s tend to produce long VOTs for /t/s (e.g., Chodroff & Wilson, 2017;Chodroff et al, 2019), and F0 tends to be higher in voiceless stops than in voiced stop (references in Idemaru & Holt, 2011).…”

“…Evidence for the relevant distribution might also be inferred based on implicit knowledge listeners have about the correlational structure of cues across categories (see Chodroff, Golden & Wilson, 2019;Idemaru & Holt, 2011;Kleinschmidt & Jaeger, 2015, Part II). For example, talkers who produce long VOTs for /p/s tend to produce long VOTs for /t/s (e.g., Chodroff & Wilson, 2017;Chodroff et al, 2019), and F0 tends to be higher in voiceless stops than in voiced stop (references in Idemaru & Holt, 2011). Whether through direct observation or inferences, exposure is expected to facilitate comprehension during test only if the category-cue mappings during exposure lead to sufficiently correct expectations about the actual category-cue mappings during test.…”