Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Differential Integration of Factual, Counterfactual, and Possible Information in Discourse Processing
Poster Session E, Friday, October 2, 11:00 am - 1:00 pm, Wangari Maathai
Maxime Tulling1, Shiyu Li1, Loïc Rainville1; 1University of Montreal
[Introduction] Language allows us to describe possibilities that are open (“maybe Mike grabbed the knife”) or contrary to reality (“if only Mike had grabbed the knife”). In factual utterances (“Mike grabbed the knife”), linking “knife” to “Mike” elicits neural responses reflecting discourse model updating (Speer et al., 2007; Tulling et al., 2021; Whitney et al., 2009) and increases accessibility of associated compared to dislocated objects (faster responses in probe-recognition; Glenberg et al., 1987; Morrow et al., 1989; Zwaan & Madden, 2004). However, counterfactual statements appear to block model updating (de Vega et al., 2007; de Vega & Urrutia, 2012), and both counterfactual and open possibilities have been shown to lack electrophysiological signatures of updating (Tulling et al., 2021; Urrutia et al., 2012). This study investigates how these possibilities are integrated into discourse representations and maintained in short-term memory. [Methods] 40 participants (6 excluded) completed a probe-recognition task while reading six murder mysteries in French. Their accuracy, reaction time (RT), and ERPs were recorded. Stories were split into 30 blocks containing 72 key sentences combining three utterance types: FACTUAL, COUNTERFACTUAL, and POSSIBLE, with two object statuses: ASSOCIATED (e.g., ‘took’) or DISLOCATED (e.g., ‘dropped’). There were 72 fillers and 36 distractors. Each target sentence was presented in six timed chunks (e.g., It was | a bottle | of wine | And Mr. Green | maybe grabbed it (target epoch: -0.2-1s) | from the counter), preceded by a context and continuation sentence. After each block, a probe word appeared, e.g., BOUTEILLE ‘bottle’ (probe epoch: -0.2-0.8s), and participants responded if it had appeared. Comprehension questions assured attention. EEG data (64 channels, mastoid-referenced, 0.1–30Hz; baseline: −200–0ms; AutoReject + 150µV artifact threshold) were analyzed in 300–500ms and 500–800ms windows in frontal and parietal regions to capture potential N400, LPC, and sustained frontal negativity effects (Nieuwland et al., 2019). EEG and RTs were analyzed using linear mixed-effects models with association, utterance type, and their interaction as fixed effects, subject and item as random factors, and relevant covariates for reading time and response time. [Results & Discussion]. For target epochs, a main effect of association (F=5.25; p=.022) was observed in the frontal ROI within the 500–800ms window. Factual (p=.034) and Possible (p=.032) utterances showed sustained anterior negativity for the ASSOCIATED compared to DISLOCATED conditions. This may reflect increased discourse integration demands for linked referents. This pattern aligns with RT results where the probe was more accessible (faster RTs) in the ASSOCIATED than DISLOCATED condition (factual p=.008, possible p=.002). In the same window, Possible utterances showed an opposite effect in parietal ROIs (F=3.5, p=.05), with greater negativity for DISLOCATED items (p=.032), possibly reflecting concurrent consideration of multiple discourse alternatives. For Counterfactual utterances, no EEG differences emerged between conditions. However, RTs showed faster responses for the DISLOCATED than ASSOCIATED probes (p=.002), suggesting that, after sentence completion, counterfactual inference reverses the implied state of the referent, increasing accessibility of the actual scenario. We thus show a dissociation between neural and behavioral signatures of discourse integration across factual, possible, and counterfactual utterances.
Topic Areas: Meaning: Discourse and Pragmatics,