cortex 126 (2020) 83e106 Available online at www.sciencedirect.com ScienceDirect Journal homepage: www.elsevier.com/locate/cortex Research Report Dissociating activation and integration of discourse referents: Evidence from ERPs and oscillations * Cas W. Coopmans a,b, and Mante S. Nieuwland a,c a Max Planck Institute for Psycholinguistics, Nijmegen, the Netherlands b Centre for Language Studies, Radboud University, Nijmegen, the Netherlands c Donders Institute for Brain, Cognition and Behaviour, Radboud University, Nijmegen, the Netherlands article info abstract Article history: A key challenge in understanding stories and conversations is the comprehension of Received 13 June 2019 ‘anaphora’, words that refer back to previously mentioned words or concepts (‘anteced- Reviewed 22 July 2019 ents’). In psycholinguistic theories, anaphor comprehension involves the initial activation Revised 30 August 2019 of the antecedent and its subsequent integration into the unfolding representation of the Accepted 20 December 2019 narrated event. A recent proposal suggests that these processes draw upon the brain’s Action editor Sonja Kotz recognition memory and language networks, respectively, and may be dissociable in pat- Published online 28 January 2020 terns of neural oscillatory synchronization (Nieuwland & Martin, 2017). We addressed this proposal in an electroencephalogram (EEG) study with pre-registered data acquisition and Keywords: analyses, using event-related potentials and neural oscillations. Dutch participants read Reference resolution two-sentence mini stories containing proper names, which were repeated or new (ease of Anaphora activation) and semantically coherent or incoherent with the preceding discourse (ease of Proper names integration). Repeated names elicited smaller N400 and Late Positive Component ampli- Theta tude than new names, and also an increase in theta-band (4e7 Hz) synchronization, which Gamma was largest around 240e450 ms after name onset. Discourse-coherent names elicited an increase in gamma-band (60e80 Hz) synchronization compared to discourse-incoherent names. This effect was largest around 690e1000 ms after name onset and exploratory beamformer analysis suggested a left frontal source. We argue that the initial activation and subsequent discourse-level integration of referents can be dissociated with event- related EEG activity, and are associated with respectively theta- and gamma-band activ- ity. These findings further establish the link between memory and language through neural oscillations. © 2020 Elsevier Ltd. All rights reserved. * Corresponding author. Max Planck Institute for Psycholinguistics, Wundtlaan 1, 6525, XD Nijmegen, the Netherlands. E-mail address: [email protected] (C.W. Coopmans). https://doi.org/10.1016/j.cortex.2019.12.028 0010-9452/© 2020 Elsevier Ltd. All rights reserved. 84 cortex 126 (2020) 83e106 novel, John-he). People have difficulty activating a unique 1. Introduction antecedent when the content of an anaphor does not distinguish between antecedents (e.g., John in a story about Discourse comprehension involves the creation of a coherent two Johns) or does not match any antecedent (e.g., John in a representation of what a story or conversation is about. A key story about David and Peter), in which case they must update challenge in this endeavor is the comprehension of ‘anaphora’, their mental model of the discourse with a new referent (e.g., words that refer back to previously mentioned words or con- Burkhardt, 2006; Gordon & Hendrick, 1998; Haviland & Clark, cepts (‘antecedents’), and which are ubiquitous in cohesive text 1974; Kennison, Fernandez, & Bowers, 2009; Nieuwland, and dialogue (Halliday & Hasan, 1976). Anaphor comprehension 2014). Ease of activation chiefly depends on content overlap is thought to involve the initial activation of the antecedent and of the anaphor with the intended referent relative to other its subsequent integration with the discourse representation antecedents (McElree, 2000, 2006; McElree et al., 2003). (e.g., Almor & Nair, 2007; Garnham, 2001; Garrod & Sanford, Anaphoric reference thus involves a form of recognition 1994; Gernsbacher, 1989; McKoon & Ratcliff, 1980; Sanford, memory, i.e., the mnemonic processes for distinguishing old Garrod, Lucas, & Henderson, 1983; Sturt, 2003). The neural from new information. implementation of these processes is largely unknown, but a Once a referent has been activated, the antecedent rep- recent proposal suggests that they draw upon the brain’s resentation can be integrated into the unfolding representa- recognition memory and language networks, respectively, and tion of the described event (Almor & Nair, 2007; Garrod & may be dissociable in patterns of neural oscillatory synchroni- Sanford, 1994; Garrod & Terras, 2000; McKoon & Ratcliff, zation (Nieuwland & Martin, 2017). To test this hypothesis, the 1989; Nieuwland & Martin, 2017; Sanford et al., 1983; Sturt, current electroencephalogram (EEG) study investigated the 2003). Semantic integration takes place regardless of comprehension of anaphoric proper names embedded in two- whether the expression is anaphoric or non-anaphoric, and sentence mini stories such as John and Peter are the best players is facilitated when the unfolding meaning of the sentence is in the football team. The top scorer of the team was John with thirty coherent with respect to the preceding context and consis- goals in total. To investigate ease of activation, we manipulated tent with what can be expected based on world knowledge whether the referent in the second sentence (‘John’) had or had (e.g., Camblin, Gordon, & Swaab, 2007; Graesser, Millis, & not been mentioned before (i.e., was anaphoric or non- Zwaan, 1997; Hagoort, Baggio, & Willems, 2009; Hagoort & anaphoric). To investigate ease of integration, we manipulated Van Berkum, 2007; Kintsch & Van Dijk, 1978; Menenti, whether the referent was semantically coherent or incoherent Petersson, Scheeringa, & Hagoort, 2008; Nieuwland & Van with the discourse context (top scorer ‘John’ had been intro- Berkum, 2006a; Van Berkum, Hagoort, & Brown, 1999; duced as one of the best or worst players). Using a pre-registered Zwaan & Rapp, 2006). protocol for high-resolution EEG data collection, preprocessing The theoretical distinction between activation and inte- and analysis, we examined whether referent activation and gration thus suggests a difference not only in timing (i.e., integration can be dissociated by event-related potentials (ERPs) activation precedes integration) but also in the underlying and by activity in the theta and gamma frequency bands. processes (i.e., integration, but not activation, is thought to involve combinatorial processes that compute higher-level 1.1. Referent activation and integration meaning from individual word meanings; e.g., Cook & O’Brien, 2014; Kintsch, 1988). Consistent with this distinc- Psycholinguistic theories of anaphor comprehension recog- tion in timing, initial activation processes can be observed nize the importance of memory representations and pro- in earlier reading time measures than subsequent contex- cesses in forming and maintaining referential relationships tual integration processes (e.g., Garrod & Terras, 2000; Sturt, (e.g., Ariel, 2001; Garnham, 2001; Gernsbacher, 1989; Gerrig & 2003; see also; Staub, 2015). In the current EEG study, we McKoon, 1998; MacDonald & MacWhinney, 1990; McElree, attempted to tease apart the processes underlying activa- Foraker, & Dyer, 2003; Myers & O’Brien, 1998; Sanford & tion and integration using ERPs and oscillatory responses, Garrod, 1989, 2005). Anaphora activate antecedents from which are complementary electrophysiological measures the discourse representation, which entails a memory rep- that provide quantitative and qualitative information about resentation of described referents. These activated anteced- online comprehension (e.g., Bastiaansen, Mazaheri, & ents are then integrated into the unfolding representation of Jensen, 2013; Lewis, Wang, & Bastiaansen, 2015). the narrated event.1 Activation appears to be a memory- based process that uses relatively superficial information 1.2. ERP correlates of referent activation and integration (e.g., lexical and semantic overlap or associations, syntactic gender) as cues to potential antecedents in memory (Gerrig & ERPs are voltage fluctuations that are associated with a McKoon, 1998; Martin, 2016; McElree, 2000, 2006; McElree specific event (e.g., onset of a stimulus), and which consist et al., 2003). This involves linking two concepts and recog- of components whose amplitude, polarity, scalp distribu- nizing the anaphor as an instantiation of the antecedent tion, latency and/or shape can be used to inform theories of despite potential differences in linguistic form (e.g., book- cognitive processing. The ERP components that are most relevant to the current study are the N400 and the Late Positive Component (LPC), which have been associated with 1 The distinction between referent activation and integration is similar to that between ‘bonding and resolution’ (Garrod & memory processing and discourse comprehension (for re- & & Sanford, 1994; Garrod & Terras, 2000; Sanford et al., 1983) and views, see Kutas Federmeier, 2011; Rugg Curran, 2007; ‘recovery and integration’ (McKoon & Ratcliff, 1980). Van Berkum, 2012; Van Petten & Luka, 2012). cortex 126 (2020) 83e106 85 The N400 component is a negative deflection that peaks Camblin, & Gordon, 2004). This ‘repeated name penalty’ between 300 and 500