Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Beyond the Language Network: A Neurocognitive comparison of conversational utterance initiations
Poster Session A, Wednesday, September 30, 11:00 am - 1:00 pm, Wangari Maathai
Caroline Arvidsson1, Ekaterina Torubarova2, Ture Berg1, André Pereira2, Julia Uddén1; 1Stockholm University, 2KTH Royal Institute of Technology
Human conversation is fundamentally interactive: interlocutors continuously adapt their behavior to one another by signaling attention, monitoring understanding, coordinating speaker transitions, and negotiating the conversational floor in real time. In conversation, speakers can initiate either a backchannel (a brief verbal feedback cue such as “uh-huh” or “mhm” that does not claim the conversational floor) or a turn (an utterance through which the speaker takes the floor). Whereas backchannels are typically drawn from a small and highly frequent inventory of expressions and primarily signal attention or understanding, turns are longer and more linguistically complex. Accordingly, a stronger recruitment of processes subserving the planning, maintenance, and updating of the upcoming utterance is expected to be observed when comparing the neural correlates of a turn initiation (TI) to those of a backchannel initiation (BI). What remains less clear, however, is whether there are processes that are unique to BIs relative to TIs. As these processes depend on real-time interpersonal coordination, they are not captured by standard non-interactive paradigms. Here, we used our own open Neuroengage conversational fMRI dataset to investigate the neurocognitive mechanisms underlying BIs and TIs. Participants (N=28) engaged in three unscripted 10-minute conversations with an experimenter (total 14 hours of conversation). To isolate BI and TI processes, events were modelled as 600-ms windows immediately preceding speech onset, based on the reported minimum latency of speech preparation (Indefrey & Levelt, 2004). This pre-speech window was selected as planning, monitoring, and action selection processes were expected to converge during this period, while minimizing the influence of articulation. Our general linear model included six conversational regressors: BI, TI, production (participant speech), comprehension (experimenter speech), silence, and fixation (a fixation cross displayed between conversations). The contrast BI > TI revealed two significant clusters (pFWE < 0.05): one in the right dorsal temporo-parietal junction (rTPJ), and one in the ventromedial prefrontal cortex. While both of these regions are key nodes in the theory of mind network, previous research on the rTPJ suggests that it is selectively recruited in attributing mental states. Moreover, the contrast TI > BI inter alia revealed a significant cluster in the deeper portions of the right middle frontal gyrus, a cluster in the left post/precentral cortex covering regions implicated in articulation during speech, and a bilateral cluster in the supplementary motor area. Hence, we provide the first evidence suggesting that different types of utterances rely on dissociable mechanisms. Specifically, our results suggest that BI is a unique process connected to monitoring the interlocutor’s mental and communicative state, evaluating whether feedback is needed, and selecting appropriate signals of engagement. In light of parallel findings from our lab showing that speaker engagement modulates rTPJ activity during turn production, we discuss the rTPJ as potentially contributing to monitoring mutual engagement and understanding among interlocutors, highlighting the cooperative nature of conversation. Overall, our findings suggest that conversation recruits neural systems extending beyond the classical perisylvian language network, underscoring the importance of studying language in natural conversation.
Topic Areas: Language Production, Meaning: Discourse and Pragmatics