Semantic Distance Modulates the N400 Event-Related Potential in Verbal Analogical Reasoning

Semantic Distance Modulates the N400 Event-Related Potential in Verbal Analogical Reasoning Matthew J. Kmiecik (mkmiecik@luc.edu) Department of Psychology Loyola University Chicago 1032 W. Sheridan Rd. Chicago, IL 60626 USA Robert G. Morrison (rmorrison@luc.edu) Department of Psychology, Neuroscience Institute Loyola University Chicago 1032 W. Sheridan Rd. Chicago, IL 60626 USA Four-term verbal analogies have long been used as both a standard measure of intelligence and vocabulary knowledge. According to traditional accounts of analogical processing, to solve this type of problem the reasoner needs to (1) retrieve word meanings from semantic memory, (2) bind words into explicit abstract relations in working memory, and (3) perform a mapping in working memory between corresponding sets of words in the source and target. For instance, to verify the analogy: Abstract Computational accounts have traditionally focused on mapping between structured representations as fundamental to analogical processing. However, a recent connectionist model has been used to argue that structured representations may not be necessary to solve verbal analogies. Green and colleagues (2010) have shown that brain areas associated with analogical mapping become more engaged as semantic distance increases between verbal analogy source and targets. Herein, we had participants verify verbal analogies characterized for semantic distance while we monitored their brain waves using EEG. Our results suggest that the semantic distance between the source and target of a verbal analogy does influence early semantic processing as reflected in the N400 Event-Related Potential. However, successfully differentiating valid and invalid verbal analogies engages areas of prefrontal cortex widely associated with inhibitory processing and the integration of abstract relations in working memory. Thus, it appears that traditional semantic priming alone is likely insufficient to explain the full extent of analogical processing. animal : zoo :: person : house participants may (1) retrieve the meanings of the words animal, zoo, person, and house, (2) bind housed (animal), lived-in (zoo), housed (person), and lived-in (house) (3) and then map lives-in (animal, zoo) to lives-in (person, house) specifically discovering that animal analogically maps to person and zoo maps to house. Several researchers have used this type of approach as embodied in the LISA model (Hummel & Holyoak, 1997; 2003) to account for patterns of verbal analogy performance (Morrison et al., 2004; Michael Vendetti & Knowlton & Holyoak, 2012). In contrast, recent connectionist models of analogy (Leech, Mareschal & Cooper, 2008) have proposed that four-term verbal analogies may be solved without the use of structured relations via a mechanism utilizing guided pattern completion in semantic memory. Contrary to previous accounts of analogical priming (Spellman, Holyoak, & Morrison, 2001), Leech and colleagues argue that this mechanism of analogy could occur automatically without the use of explicitly represented relations and analogical mapping. In addition to many experimental studies (see Holyoak & Hummel, 2008) the former traditional relationally explicit approach is supported by findings showing that solving verbal analogies engages anterior regions of the PFC (Bunge et al., 2005; Green et al., 2010) frequently associated with processing abstract information (e.g., Christoff et al., 2009; Nikitin & Morrison, 2011; Volle et Keywords: analogy, semantic distance, EEG, N400 Introduction Analogical reasoning is fundamental to the way that humans learn and reason in day-to-day life.. Likewise, analogies have long been considered to be a core component of analytic intelligence (Spearman, 1927) and of great importance in learning and discovery (Holyoak & Thagard, 1995). For nearly a century, researchers in cognitive science have developed theories and computational models to offer potential mechanisms for analogical processing (French, 2002). More recently patient-based (Morrison et al., 2004; Krawczyk et al., 2008) and functional neuroimaging studies (e.g., Bunge et al. 2005; Bunge et al. 2009; Green et al. 2010; Krawczyk et al., 2010; Volle et al., 2010; Watson & Chatterjee, 2012) have begun to identify a network of brain areas, particularly the prefrontal cortex (PFC), essential for analogical reasoning.