You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster B54, Wednesday, November 8, 3:00 – 4:15 pm, Harborview and Loch Raven Ballrooms

Neural encoding of T3 sandhi in Mandarin Chinese speakers in speech production

Caicai Zhang1,2, Xunan Huang1, Stephen Politzer-Ahles1, Jie Zhang3, Gang Peng1,2;1The Hong Kong Polytechnic University, 2Shenzhen Institutes of Advanced Technology, 3The University of Kansas

Mandarin Tone 3 (T3) sandhi refers to the substitution of T3 with Tone 2 (T2) when followed by another T3 syllable in speech production. For instance, the phrase “mai3 ma3” ("to buy a horse") is ultimately pronounced “mai2 ma3”. T3 sandhi is generally believed to be subserved by a phonological computation mechanism in which the sandhi pattern is applied according to the phonological context regardless of whether it is a real word or not [Zhang, C., Xia, Q., and Peng, G. 2015. Mandarin third tone sandhi requires more effortful phonological encoding in speech production: Evidence from an ERP study. Journal of Neurolinguistics, 33, 149-162]. However, it is unclear whether high-frequency and low-frequency T3 sandhi words are produced via the same computation mechanism. It is possible that both are subserved by the phonological computation mechanism, such that both high- and low- frequency words would elicit similar neural activities during the encoding prior to articulation. Alternatively, it is possible that high-frequency words may be stored in the lexicon in their post-sandhi forms which can be directly accessed for speech production, whereas low-frequency words would rely more on phonological computation that changes T3 to T2 online. On the other hand, such differences should not be observed on words that do not involve T3 sandhi. To adjudicate between these computation-based and storage-based accounts, we examined event-related potentials (ERPs) elicited during the covert production of high- and low-frequency sandhi-undergoing and non-sandhi-undergoing words in Mandarin Chinese. 17 Mandarin-speaking subjects participated in the experiment. Eighty high-frequency and eighty low-frequency disyllabic words were selected, comprising equal numbers of T3+T3 (sandhi-undergoing) and non-T3+T3 (non-sandhi-undergoing) words. The stimuli were recorded and presented as two isolated syllables in citation form with a brief pause in between. Subjects were asked to covertly produce the two isolated syllables as a single word for which tone sandhi would apply, followed by overt production when prompted later. To ensure that the onset of covert production is aligned to the appearance of a visual sign, a second task—loudness judgment, cued by a different visual symbol—was randomly intermixed with the production task in a single block. ERPs were time-locked to the onset of the visual sign for overt production. We observed significantly more negative amplitude elicited by sandhi-undergoing words than by non-sandhi-undergoing words in the time-window of the N200 (250-350 ms). Besides, there was a significant interaction effect between tone sandhi and frequency in the time-window of a late negativity (LN; 450-800 ms). In this late time-window, low-frequency non-sandhi-undergoing words elicited significantly more negative LN amplitude than high-frequency non-sandhi-undergoing words, whereas no frequency effect was found in sandhi-undergoing words. These results suggest that the encoding of sandhi-undergoing and non-sandhi-undergoing words might be subserved by different neural mechanisms, the former primarily engaging phonological computation irrespective of the lexical frequency, and the latter primarily engaging lexical processes. These findings have implications for understanding how phonological knowledge such as tone sandhi is represented in the brain, and encoded in speech production prior to articulation.

Topic Area: Phonology and Phonological Working Memory

Back to Poster Schedule