Poster D20, Friday, August 17, 4:45 – 6:30 pm, Room 2000AB
Measuring the N400 during naturalistic conversation: An EEG hyperscanning study
Caitriona Douglas1, Antoine Tremblay1, Aaron Newman1;1Dalhousie University
While much work has investigated brain activity in different language production and perception contexts, practical constraints have meant little is known about brain activity during natural conversation. Yet, conversation is the most fundamental mode of language use — it is how we first learn language, and represents a significant proportion of most people’s daily language use. A recent technical advance is ‘hyperscanning’, in which neuroimaging data is acquired from two or more individuals simultaneously. However, most of this work has looked over relatively long time scales, and there is little or no evidence that it is possible to obtain event-related potentials (ERPs) time-locked to individual words during a natural conversation. As an initial exploration of the feasibility of conducting world-level ERP research using hyperscanning, the aim of our experiment was to determine if we could obtain ERPs time-locked to the onset of individual words produced during a conversation between two people. Specifically, we investigated whether we could replicate the modulation of the N400 ERP component by lexical frequency. The N400 is commonly associated with lexical access, and prior studies have consistently reported larger N400s in response to low than high frequency words (e.g., Rugg, 1990, doi:10.3758/BF03197126, Van Petten & Kutas, 1990, doi:10.3758/BF03197127). We created a set of six scripted dialogues that included a number of low and high frequency nouns balanced for word length, number of syllables and morphemes, and orthographic and phonological neighbourhood size and frequency. Subjects (native English speakers) were run in pairs, with each individual reading one part of the dialogue (without being able to see the lines of their interlocutor). The target words for the N400 analysis were not distinguished in any way from other words in the scripts, and participants were instructed to read the scripts silently beforehand, and then read them as naturally as possible to their interlocutor during the experiment. EEG was recorded from both participants simultaneously using a 64 channel amplifier (32 channels/subject, including EEG, EOG, and facial EMG) with a ground splitter that ensured appropriate referencing for each individual; as well participants’ speech was simultaneously recorded in a separate channel to the EEG data file, ensuring accurate time-locking. Following data collection, the onset of each target word was identified in the audio file and used to create event markers for ERP analysis. EEG data were processed separately for each individual, including manual artifact removal and ICA artifact correction. ERPs were time-locked to the onset target words that each individual heard (not to words they produced themselves). The results demonstrated a clear negativity from approximately 250-550 ms that was largest over midline central-parietal electrodes, consistent with past characterizations of the N400. Moreover, this negativity was larger for low than high frequency words, as predicted. These results provide compelling initial evidence that it is possible to obtain ERPs to individual words in a conversational context using hyperscanning, opening the door to an exciting range of future possibilities for neurolinguistic research.
Topic Area: Perception: Auditory