An EEG study on the brain representations in language learning

Abstract

This paper presents an experimental study to understand the key differences in the neural representations when the subject is presented with speech signals of a known and an unknown language and to capture the evolution of neural responses in the brain for a language learning task. In this study, electroencephalography (EEG) signals were recorded while the human subjects listened to a given set of words from English (familiar language) and Japanese (unfamiliar language). The subjects also provided behavioural signals in the form of spoken audio for each input audio stimuli. In order to quantify the representation level differences for the auditory stimuli of two languages, we use a classification approach to discriminate the two languages from the EEG signal recorded during listening phase by designing an off-line classifier. These experiments reveal that the time-frequency features along with phase contain significant language discriminative information. The language discrimination is further confirmed with a second subsequent experiment involving Hindi (the native language of the subjects) and Japanese (unknown language). A detailed analysis is performed on the recorded EEG signals and the audio signals to further understand the language learning processes. A pronunciation rating technique on the spoken audio data confirms the improvement of pronunciation over the course of trials for the Japanese language. Using single trial analysis, we find that the EEG representations also attain a level of consistency indicating a pattern formation. The brain regions responsible for language discrimination and learning are identified based on EEG channel locations and are found to be predominantly in the frontal region.