Neural signatures of engagement and event segmentation during story listening in background noise

Speech in everyday life is often masked by background noise, making comprehension effortful. Characterizing brain activity patterns when individuals listen to masked speech can help clarify the mechanisms underlying such effort. However, most previous research has focused on neural activity related to short, disconnected sentences that little resemble the more continuous, story-like spoken speech individuals typically encounter. In the current study, we used functional magnetic resonance imaging (fMRI) in humans (both sexes) to investigate how neural signatures of story listening change in the presence of masking by 12-talker babble noise. We show that, as speech masking increases, spatial and temporal activation patterns in auditory regions become more idiosyncratic to each listener. In contrast, spatial (and to some extent temporal) activity patterns in brain networks linked to effort (e.g. cinguloopercular network involving the anterior insula and anterior cingulate) are more similar across listeners when speech is highly masked and less intelligible, suggesting shared neural processes. Moreover, at times during stories when one meaningful event ended and another began, neural activation increased over extensive regions in frontal, parietal, and medial cortices. This event-boundary response appeared little affected by background noise, suggesting that listeners process meaningful units and, in turn, the gist in naturalistic, continuous speech even when it is masked somewhat by background noise. Overall, the current data may indicate that people stay engaged and cognitive processes associated with naturalistic speech processing remain intact under moderate levels of background noise, whereas auditory processing becomes more idiosyncratic to each listener.