Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Semantic similarity as partial cue-match: EEG evidence from the fan effect
Poster Session D, Thursday, October 1, 4:30 - 6:30 pm, Wangari Maathai
Philine Link1, Jakub Dotlačil1, Leendert van Maanen2; 1Institute for Language Sciences, Utrecht University, 2Department of Experimental Psychology, Utrecht University
Cue-based retrieval accounts of memory and language processing propose that retrieval requires matching cues to items in memory. Multiple items matching a cue causes inhibitory interference (Lewis & Vasishth, 2005; Van Dyke, 2007). This effect originates in cognitive science: fan effect studies show that retrieval difficulty increases as more facts are associated with a concept (Anderson, 1974; Anderson & Reder, 1999). However, a gap remains between classical fan experiments and cue-based retrieval models based on ACT-R (Anderson & Lebiere, 1998). Fan effect studies primarily examine interference from exact word repetition, whereas cue-based retrieval assumes interference can also arise from partial cue matches. To link the ACT-R model and cue-based retrieval, it is therefore necessary to show that fan effects also emerge when using semantically similar instead of repeated words to create the associative fan. Only one previous study has demonstrated semantic inhibitory interference in a fan paradigm (Link et al., 2026), and only behaviorally. Here, we provide the first evidence that semantically induced fan effects are also reflected in neural processing. We conducted an EEG fan experiment with 50 participants. Participants studied 24 target person-location word pairs. Fan was manipulated by associating each target concept with either two (fan 2) or four (fan 4) semantically related concepts. During retrieval, participants judged whether a presented word pair had been studied (target) or not (foil), while EEG was recorded. Hierarchical Bayesian linear mixed models of the behavioral data revealed longer reaction times and lower accuracy for higher fan conditions, consistent with semantic inhibitory interference. For EEG analyses, we defined time windows and regions of interest based on the Early Effect reported for classical fan effects (Borst et al., 2013). Trial-averaged EEG amplitudes were analyzed using hierarchical Bayesian linear mixed models. For foils, 96% of the posterior distribution for the fan effect was above zero (CI: -0.071, 1.135), indicating greater positivity for higher fan conditions, consistent with Borst et al. (2013). Because our semantically constrained stimuli included lower-frequency and less length-controlled words, we expected effects to emerge later. Shifting the analysis window by 50 ms yielded a positive fan effect for both targets and foils (CI: 0.024, 1.038), with 98% of the posterior distribution above zero. In addition, we applied a novel machine learning approach for detecting cognitive events (Weindel et al., 2024). This analysis revealed differences in the duration of the event interpreted as retrieval, further supporting the idea that semantic similarity modulates retrieval dynamics. Together, these findings suggest that semantically induced interference engages neural mechanisms similar to those underlying the classical fan effect. This supports the view that semantic similarity constitutes partial cue matching and provides crucial evidence for the use of ACT-R-based retrieval models in explaining semantic inhibitory interference in sentence processing.
Topic Areas: Meaning: Lexical Semantics, Computational Approaches