Prof. ZHANG Caicai, Department of Chinese and Bilingual Studies

Variation in Real-time Lexical Tone Processing: A Psycholinguistic Perspective. The Third International Conference on Tone and Intonation (TAI 2025). Institute for Phonetics and Speech Processing, Ludwig-Maximilians-Universität München, Herrsching near Munich, Germany, 16-18 May 2025.

Abstract
Variation is ubiquitous in human speech. A speech sound has different acoustic realizations when produced by different speakers or surrounded by different speech sounds, posing a challenge for fast and consistent speech perception in the listener’s brain. Beyond acoustic variation, there are systematic phonological alternations, where the context-specific phonological form must be selected in a super-fast manner prior to articulation in the speaker’s mind. However, how variation is accommodated in speech perception or planned in speech production is not well understood. In this talk, I will focus on variation on a short timescale, i.e., a few hundred milliseconds of real-time speech perception and production, probed via the lens of a highly variable phonological entity – lexical tone. I will present evidence showing that listeners deploy multiple cues, including contextual acoustic cues, population F0 knowledge, and their own vocal F0 cues, to achieve fast and accurate tone perception, especially when the stimuli are variable, ambiguous, or acoustically degraded. On the production side, our studies on Mandarin Third Tone sandhi show that speakers encode the abstract tonal category at an earlier phonological encoding stage and the surface, context-specific tonal form at a later phonetic encoding or motor preparation stage before articulation. These neural processes appear to be largely similar across lexicality (real vs. pseudowords) and word frequency (high vs. low frequency) conditions. These findings have implications for advancing speech perception and production theories. Psycholinguistic methods are instrumental in understanding basic mental processes in real-time lexical tone processing, highlighting that the human speech system is fast, flexible, and multi-pronged.