Poster Presentation

Search Abstracts | Symposia | Slide Sessions | Poster Sessions

Predicting Wh-in-Situ Questions from Prosody in Mandarin: Evidence from ERPs

Poster Session E, Friday, October 2, 11:00 am - 1:00 pm, Wangari Maathai

Leticia Pablos Robles¹, Hailin Hao², Yang Yang³, Libo Geng⁴, Lisa Cheng¹; ¹Leiden University, ²University of Southern California (USC), ³Guangdong University of Foreign Studies, ⁴Jiangsu Normal University (JSNU)

Mandarin wh-in-situ questions provide a critical test case for predictive language processing because interrogative meaning cannot be inferred from early morphosyntactic wh-marking. Unlike English, where clause type is signaled sentence-initially (What did the child eat?), Mandarin preserves the wh-phrase in situ, delaying overt interrogative information until the object position (The child ate what?). Previous work showed that Mandarin readers do not reliably predict wh-questions in the absence of overt syntactic, contextual, or prosodic information (Pablos Robles et al., 2025), whereas prosodic cues alone can support early anticipation of interrogative structure in auditory comprehension (Gryllia et al., 2020). We investigated whether prosodic information preceding the wh-phrase is sufficient to trigger anticipatory processing during online spoken sentence comprehension. Native Mandarin speakers in China (n = 17) participated in an auditory within-subjects ERP experiment using 64-channel EEG. Participants heard control declaratives and wh-questions, as well as their prosodically cross-spliced counterparts (36 per condition). Cross-splicing was introduced at the direct-object onset, creating prosodic incongruence while preserving identical segmental material prior to the critical region, resulting in 4 conditions: (D|D), (Q|Q), (D|Q) and (Q|D), where “|” indicates the position of the cross-splice. Stimuli differed only in the object position, where declaratives contained a noun phrase júzi (“orange(s)”), and questions contained the wh-word shénme (“what”). Because prior work identified verbal tonal information as predictive of upcoming clause type (Gryllia et al., 2020), only verbs with Tone 1 and 4 were included. Participants were asked to listen carefully to the 144 target sentences, presented randomized together with 144 fillers, and to answer a simple arithmetic question and a comprehension question during 1/3 of the experiment. ERP analyses revealed robust sensitivity to prosodic mismatch from very early on. Relative to their canonical controls, both cross-spliced conditions elicited significant effects over centro-anterior electrodes within a 0–700 ms window, consistent with previous findings on prosodic incongruences (Mietz et al., 2008) and clause-type expectation failures (Astésano et al., 2004; Paulmann et al., 2012). In addition, the Q|Q versus D|Q comparison produced a centro-parietal negativity between 0–400 ms, compatible with an N400-like response associated with lexico-syntactic integration (Kutas & Hillyard, 1980) and syntax-prosody mismatches (Steinhauer et al., 1999; Friederici et al., 2007; Mietz et al., 2008; Bögels et al., 2011). Tone-specific analyses further showed differential contributions of Tone 1 and Tone 4 across conditions, suggesting that tonal and prosodic information jointly participate in anticipatory parsing. These findings demonstrate that Mandarin listeners exploit prosodic information preceding the wh-phrase to predict upcoming interrogative structure, as demonstrated by ERP effects shown immediately upon hearing the onset of the critical region. The results provide neurophysiological evidence that prosody can support incremental prediction during processing of wh-in-situ dependencies and contribute to rapid clause-type computation in spoken Mandarin.

Topic Areas: Prosody,