Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Linking Neural Dynamics of Linguistic Prediction to Behavioral Decision-Making: Evidence for N400-Drift-Diffusion Relationship
Poster Session F, Friday, October 2, 2:45 - 4:45 pm, Wangari Maathai
Anna Tumanyants1, Benjamin Straube1, Yifei He1; 1Philipps University Marburg, Germany
Prediction is central to language comprehension, and several neural markers of predictive processing have been identified through EEG research. The classic N400 component has been linked to predictive processing not only in terms of amplitude modulation but also onset latency: several studies reported its earlier emergence at around 200 ms (hereafter ‘eN400’) in highly predictive contexts, hinting at a qualitatively different process within this time window (e.g., Lau et al., 2013; Luka & Van Petten, 2014). However, in linguistic tasks, these indices are typically interpreted in isolation from the decision-making behavior, leaving unclear how distinct neural markers translate into response-related processes. Here, with a novel priming paradigm, we combined EEG with drift-diffusion modeling (DDM) to explore how prediction-specific and semantic priming facilitation mechanisms relate to decision behavior. Participants (N=29) completed a lexical decision task following either one prime (single condition) or three semantically related primes (cumulative condition), manipulating prediction strength while holding semantic relatedness constant. We focused on mean ERP amplitudes time-locked to the target word onset for both 100–300 ms (eN400) and 300–500 ms (N400) time windows and ran a 2 × 2 repeated-measures ANOVA with Condition (single vs. cumulative) and Target Type (related vs. unrelated target). We further conducted ERP-informed DDM analysis in which single-trial eN400 and N400 amplitudes were entered into the DDM models as regressors of interest. For ERPs, we found that the cumulative condition elicited early prediction-sensitive activity (eN400) alongside classic N400 effects. Both components showed a Condition × Target Type interaction: in the cumulative condition, unrelated targets elicited more negative amplitudes than related targets, while no such difference was observed in the single condition. ERP-informed DDM results revealed that eN400 and N400 showed a directional divergence: N400 primarily showed a positive trend with drift rate (P(v > 0) = .829), but not with boundary separation (P(a > 0) = .522). In contrast, increasing eN400 amplitude led to reduced boundary separation (P(a < 0) = .968) and a negative trend of drift rate (P(v < 0) = .788). These results point to distinct relationships between neural signals and decision behavior. Our findings show that predictive context modulates neural signals of semantic priming both at early and classic N400 time windows. Moreover, the early prediction signals (eN400) are preferentially associated with response caution, while later semantic processing (N400) relates primarily to evidence quality, consistent with a distinction between prediction-specific and broader semantic facilitation mechanisms. This differential relationship between ERP components and DDM parameters suggests that earlier neural prediction signals may set the decision threshold before later semantic processing drives evidence accumulation, pointing to a time-dependent influence of predictive processing on decision behavior. Methodologically, as this pattern could not be captured by classical ERP or behavioral analysis alone, our findings demonstrate the added value of linking single-trial neural signals to computational models of behavior.
Topic Areas: Meaning: Lexical Semantics, Computational Approaches