Yuka Tatsumi

All Publications

• Studying mandarin tone sandhi in interaction. PloS one Pelzl, E., Tatsumi, Y., Cui, Z., Viswanathan, N., Olmstead, A. J. 2026; 21 (8): e0356597

  Abstract

  The Mandarin low tone (T3) undergoes an alteration, third tone sandhi, when followed by another T3. The resulting F0 is superficially the same as that of the rising tone (T2) and has the potential to induce confusion. We investigated how Mandarin speakers adjust their speech production to overcome potential ambiguities induced by third tone sandhi during a structured interactive task. Ten pairs of L1 Chinese participants completed an interactive phrase matching task. Participants were shown displays with Chinese phrases (surname+title). One participant read an indicated phrase which the other selected from their display. There were two conditions. In the no sandhi condition, the title did not induce sandhi, and should result in distinct F0 patterns for the surnames. In the sandhi condition, the title induced third tone sandhi, and should result in homophonous surnames. Task performance showed clear evidence of sandhi-induced ambiguity. Examination of accuracy and tone acoustics suggested that pairs reduced sandhi-induced ambiguity primarily by lowering the F0 of T3 towards the end of its trajectory. We interpret this as avoidance of sandhi application. This study shows that even highly structured interactive tasks have potential to reveal new insights into how speakers and listeners navigate communicative ambiguities caused by tone sandhi. This constitutes an important first step to evaluating how phenomena such as tone sandhi generalize to everyday interactive communication.

  View details for DOI 10.1371/journal.pone.0356597

  View details for PubMedID 42616737

  View details for PubMedCentralID PMC13489460

• Neural mechanisms of face cue predictability and the integration of facial and acoustic cues in native- and nonnative-accented words BRAIN AND LANGUAGE Lei, D., Tatsumi, Y., Hsieh, E., Giorio, C., van Hell, J. G. 2025; 269: 105621

  Abstract

  This study examined the integration of face cue and native- and nonnative-accented English speech by manipulating the face cue predictability of a speaker's accent (predictable: one accent (American-accent or Chinese-accented), not predictable: two accents (American-accented and Chinese-accented). Monolingual listeners were first familiarized with each speaker's number and type of accent(s). Then, they completed an EEG-recorded auditory go/no-go animal decision task where the no-go items were critical words and nonwords. Listeners saw a face cue (speaker image) before speech onset and concurrently with the speech. Pre-speech ERP results revealed that listeners processed face cues differently based on face cue predictability. Post-speech ERP analyses revealed N400 lexicality effects for native-accented speech, and face cue predictability effects for nonnative-accented speech. No N400 effects were found for an audio-only experiment. This indicates that the monolingual listeners integrate face cues and auditory cues during real-time nonnative-accented speech processing, but not during native-accented speech processing.

  View details for DOI 10.1016/j.bandl.2025.105621

  View details for Web of Science ID 001541211700001

  View details for PubMedID 40716179

• In-Context Learning Boosts Speech Recognition via Human-like Adaptation to Speakers and Language Varieties Roll, N., Graham, C., Tatsumi, Y., Kim Tien Nguyen, Sumner, M., Jurafsky, D. edited by Christodoulopoulos, C., Chakraborty, T., Rose, C., Peng ASSOC COMPUTATIONAL LINGUISTICS-ACL. 2025: 4412-4426

  View details for Web of Science ID 001805531300219