Cognitive control in thematic role assignment: Evidence from neural oscillations
Tal Ness1, Valerie Langlois2, Wing-Yee Chow3, Colin Phillips1, Jared Novick1, Albert Kim2; 1University Of Maryland, College Park, 2University of Colorado, Boulder, 3University College London
A key function in language comprehension is determining who did what to whom, i.e., assigning thematic roles to an event. Real-time thematic role assignment can generate conflict between multiple representations when syntactic and semantic cues support different interpretations (e.g., ‘…which waitress the customer had served…’). Previous research has indicated that cognitive control engages to resolve such conflict: role reversals increase activity in frontal brain regions associated with cognitive control across linguistic and nonlinguistic domains; and upregulation of cognitive control facilitates the resolution of competing event roles (Thothathiri et al., 2012, 2018). These findings, however, do not address the temporal dynamics of how and when cognitive control influences sentence comprehension. Here, we ask whether neural oscillatory EEG activity in the theta band (4-8 Hz) — which is widely associated with cognitive control in tasks like Stroop and Flanker (e.g., Cavanaugh & Frank, 2014) — can provide a real-time index of cognitive-control engagement during language comprehension. Specifically, we tested whether theta activity increases for role-reversal sentences, which create two competing interpretations, but not for implausible sentences in which syntactic and semantic cues do not conflict. We hypothesized that cognitive control is rapidly recruited during sentence processing, exclusively when conflict arises, and that this manifests in EEG as increased theta-band activity. We conducted a reanalysis of EEG data from Chow, Smith, Lau, and Phillips (2016; Experiment 1; N=24). Participants read two types of semantically implausible sentences and their baseline counterparts: Role-Reversal sentences (e.g., ‘…which waitress the customer had served…’), which were rendered implausible by swapping argument roles from a highly plausible baseline sentence (‘…which customer the waitress had served…’); and Argument-Substitution sentences (e.g., ‘…which realtor the landlord had evicted…’), which were rendered implausible by substituting an argument from a highly plausible baseline sentence (‘…which tenant the landlord had evicted…’). The critical word (‘served’/‘evicted’) was unpredictable and implausible in both sentence types. In Role-Reversal but not Argument-Substitution sentences, conflict arises between the thematic-role assignment supported by world knowledge and that signaled by the sentence’s structure. We computed time-frequency representations of EEG activity during the critical words (5-cycle Morlet wavelets at 30 log-spaced scales 2-80). Role-Reversal sentences elicited increased theta-band power versus baseline, 500-950ms after word onset (cluster-level p = 0.039 in a cluster-based permutation test). No increase in theta-band activity was observed in Argument-Substitution sentences (versus baseline). Notably, the theta effect is functionally distinct from the P600 effects observed in the original ERP analysis, which were observed in both implausible conditions. Our findings provide a new neurophysiological index of cognitive-control engagement during sentence processing: increased theta-band activity. The time-course of theta effects, beginning around 500ms, suggests rapid, reactive cognitive-control engagement in response to conflict. Increased theta-band activity was exclusive to Role-Reversal sentences, which suggests that cognitive control engages specifically when multiple competing interpretations of a sentence are temporarily pursued, and not in implausible sentences involving no conflict, as in Argument Substitution. These patterns show that EEG activity in the theta band can clarify the temporal dynamics of when cognitive-control mechanisms deploy during language processing.
Topic Areas: Control, Selection, and Executive Processes, Methods