AUTHOR=Eqlimi Ehsan , Bockstael Annelies , De Coensel Bert , Schönwiesner Marc , Talsma Durk , Botteldooren Dick 

TITLE=EEG Correlates of Learning From Speech Presented in Environmental Noise

JOURNAL=Frontiers in Psychology

VOLUME=Volume 11 - 2020

YEAR=2020

URL=https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2020.01850

DOI=10.3389/fpsyg.2020.01850

ISSN=1664-1078

ABSTRACT=How the human brain retains relevant vocal information while suppressing irrelevant sounds is one of the ongoing challenges in cognitive neuroscience. Studying the underlying mechanisms of this ability can be used to identify whether a person is distracted during listening to a target speech especially in a learning context. This paper investigates the neural correlates of learning from speech presented in a noisy environment using an ecologically valid learning context and electroencephalography (EEG). To this end, the following listening tasks were performed while 64-channels EEG signals were recorded: (1) attentive listening to the lectures in background sound, (2) attentive listening to the background sound presented alone, and (3) inattentive listening to the background sound. For the first task, thirteen five-minute lectures embedded in different types of realistic background noise were presented to participants who were asked to focus on the lectures. As background noise, multi-talker babble, continuous highway, and fluctuating traffic sounds were used. After the second task, a written exam was taken to quantify the amount of information the participants have acquired and retained from the lectures. In addition to various power spectra-based EEG features in different frequency bands, the peak frequency and long-range temporal correlations of alpha-band oscillations were estimated. To reduce these dimensions, a principal component analysis (PCA) was applied on the different listening conditions resulting the feature combinations that discriminate most between listening conditions and persons. Linear mixed-effect modelling was used to explain the origin of extracted principal components, showing their dependence on listening condition and type of background sound. Following this unsupervised step, a supervised analysis was performed to explain the link between the exam results and the EEG-PC scores using both linear fixed and mixed-effect modelling. Results suggest that the ability to learn from speech presented in environmental noise can be predicted by the several components over the specific brain regions better than by knowing the background noise type. These components were linked to deterioration in attention, speech envelope following, decreased focusing during listening, cognitive prediction error, and specific inhibition mechanisms.