Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Speaking against expectations: Cognitive control over context-driven conflict in language production
Poster Session F, Friday, October 2, 2:45 - 4:45 pm, Wangari Maathai
Xiaochen Zheng1, Syanah Wynn2, Bryant Jongkees1, Sander Nieuwenhuis1; 1Leiden University, The Netherlands, 2Johannes Gutenberg University of Mainz, Germany
Language production unfolds in context, allowing speakers to anticipate upcoming information and prepare responses efficiently. Yet contextual expectations can also become a source of interference when they conflict with communicative goals. Imagine ordering a coffee and expecting the barista to ask, “What size?”, only to hear, “We just ran out of milk.” Instead of producing the prepared response (“large”), you must suppress it and adapt your utterance (“never mind”). Such situations illustrate internally generated interference: conflict arising not from external distractors, but from context-driven expectations during speaking. Cognitive control in language production has traditionally been studied using external interference paradigms, such as picture–word interference. Less is known about how speakers resolve conflicts between internally generated contextual predictions and intended speech output. Here, we tested whether midfrontal theta—a neural signature of domain-general cognitive control—indexes such internally generated conflict during speech production. EEG was recorded from 40 participants while they read sentence contexts followed by picture naming targets that were congruent, incongruent, or neutral with respect to the preceding context. Behaviorally, naming responses were faster for congruent and slower for incongruent pictures relative to neutral, reflecting facilitation and interference driven by contextual expectations. Predictable sentence contexts elicited stronger alpha–beta desynchronization, consistent with contextual integration and anticipatory processing, although these effects were smaller than previously reported, suggesting that context reliability modulates predictive engagement. Critically, incongruent picture targets elicited enhanced midfrontal theta activity relative to congruent and neutral conditions, indicating increased cognitive control demands when context-driven representations conflicted with speaking goals. To further characterize the representational basis of this conflict, we used a multimodal embedding model (CLIP) to quantify semantic distance between contextual and target representations. Embedding distances predicted naming latencies beyond experimental condition labels and tracked trial-by-trial theta modulation, suggesting that continuous representational conflict contributes to behavioral and neural variation. Together, these findings demonstrate that interference in speaking arises not only from external distractors but also from internally generated contextual expectations. Resolving such conflicts recruits domain-general control mechanisms, highlighting midfrontal theta as a neural marker of cognitive control when contextual predictions and speaking goals diverge.
Topic Areas: Control, Selection, and Executive Processes, Language Production