10 Computational EEG Analysis for Hyperscanning …
221
Hereby, more advanced data analysis methods are required to describe the dynamic
interaction across different brains.
The very first hyperlink study on social interaction was also conducted using
fMRI. Montague and colleagues reported a significant correlation in supplementary
motor areas between two persons playing a deception game [39]. Their further and
more influential work revealed an fMRI-based neuro-link between two players in
a trust game that could predict the intention to trust [31]. Speech, as a dominant
social interaction type, has also been studied using fMRI-based hyperscanning. The
dynamic neural coupling between a speaker and a listener was shown to have different patterns in different brain regions: while the auditory related regions exhibited
a delayed coupling of the listener as compared to the speaker, the frontal regions
revealed a leading coupling by the listener possibly responsible for speech anticipation [61]. Furthermore, it has been suggested that interactive experience and skills
play enabling roles in social cognitive functions [7, 8]. Hyperlink hereby provides a
promising new perspective for probing the interactive nature of our brain.
The first EEG hyperscanning study for social interaction by Babiloni and colleagues involved sets of four individuals playing Tressette, a bridge-like game
(Babiloni et al. [1]. The portability of EEG has enabled researchers to extend the
exploration into more realistic conditions. To date, researchers have recorded multiperson EEGs in scenarios representing the majority of our social interaction activities,
ranging from simple motor actions to complex activities such as making a conversation, performing music, playing games, etc. [2]. As different participants may have
different roles, the information flow among participants may have different directions,
with some participants leading the others. The methods for characterizing hyperlinks
can be summarized into the following three categories: (1) undirectional hyperlink
methods; (2) directional hyperlink methods; (3) machine-learning methods.
10.3.1 Undirectional Hyperlink Methods
This category of methods is straightforwardly derived from the inter-subject correlational methods as used for describing social perception. The fundamental hypothesis
underlying most of the correlational methods is that different brains have similar and
synchronized time courses of neural activities, therefore suitable for the cases when
participants have balanced roles. Pairwise Pearson’s correlation can be calculated
for EEG signals filtered at different frequency bands. Inter-brain correlation of the
amplitudes of theta and alpha over right temporal-parietal junction (TPJ), the amplitudes of alpha and beta over frontal regions, have been reported to be associated with
the understanding of others’ intention and high-level cooperative strategies [29, 58].
Phase-locking analysis is another type of method that has been frequently employed.
Phase-locking analysis focuses on the circular correlation of the phases of the neural
oscillations, usually termed as the phase-locking value (PLV), as follows:
221
Hereby, more advanced data analysis methods are required to describe the dynamic
interaction across different brains.
The very first hyperlink study on social interaction was also conducted using
fMRI. Montague and colleagues reported a significant correlation in supplementary
motor areas between two persons playing a deception game [39]. Their further and
more influential work revealed an fMRI-based neuro-link between two players in
a trust game that could predict the intention to trust [31]. Speech, as a dominant
social interaction type, has also been studied using fMRI-based hyperscanning. The
dynamic neural coupling between a speaker and a listener was shown to have different patterns in different brain regions: while the auditory related regions exhibited
a delayed coupling of the listener as compared to the speaker, the frontal regions
revealed a leading coupling by the listener possibly responsible for speech anticipation [61]. Furthermore, it has been suggested that interactive experience and skills
play enabling roles in social cognitive functions [7, 8]. Hyperlink hereby provides a
promising new perspective for probing the interactive nature of our brain.
The first EEG hyperscanning study for social interaction by Babiloni and colleagues involved sets of four individuals playing Tressette, a bridge-like game
(Babiloni et al. [1]. The portability of EEG has enabled researchers to extend the
exploration into more realistic conditions. To date, researchers have recorded multiperson EEGs in scenarios representing the majority of our social interaction activities,
ranging from simple motor actions to complex activities such as making a conversation, performing music, playing games, etc. [2]. As different participants may have
different roles, the information flow among participants may have different directions,
with some participants leading the others. The methods for characterizing hyperlinks
can be summarized into the following three categories: (1) undirectional hyperlink
methods; (2) directional hyperlink methods; (3) machine-learning methods.
10.3.1 Undirectional Hyperlink Methods
This category of methods is straightforwardly derived from the inter-subject correlational methods as used for describing social perception. The fundamental hypothesis
underlying most of the correlational methods is that different brains have similar and
synchronized time courses of neural activities, therefore suitable for the cases when
participants have balanced roles. Pairwise Pearson’s correlation can be calculated
for EEG signals filtered at different frequency bands. Inter-brain correlation of the
amplitudes of theta and alpha over right temporal-parietal junction (TPJ), the amplitudes of alpha and beta over frontal regions, have been reported to be associated with
the understanding of others’ intention and high-level cooperative strategies [29, 58].
Phase-locking analysis is another type of method that has been frequently employed.
Phase-locking analysis focuses on the circular correlation of the phases of the neural
oscillations, usually termed as the phase-locking value (PLV), as follows: